The author(s) shown below used Federal funds provided by the U.S.
Department of Justice and prepared the following final report:
Document Title:

The North Carolina Highway Traffic Study

Author(s):

William R. Smith ; Donald Tomaskovic-Devey ;
Matthew T. Zingraff ; H. Marcinda Mason ;
Patricia Y. Warren ; Cynthia Pfaff Wright

Document No.:

204021

Date Received:

January 2004

Award Number:

1999–MU–CX–0022

This report has not been published by the U.S. Department of Justice.
To provide better customer service, NCJRS has made this Federallyfunded grant final report available electronically in addition to
traditional paper copies.

Opinions or points of view expressed are those
of the author(s) and do not necessarily reflect
the official position or policies of the U.S.
Department of Justice.

The North Carolina Highway Traffic Study
Final Report to the
National Institute of Justice
U.S. Department of Justice

By
William R. Smith
Donald Tomaskovic-Devey
Matthew T. Zingraff
H. Marcinda Mason
Patricia Y. Warren
Cynthia Pfaff Wright
North Carolina State University
Raleigh, North Carolina
With
Harvey McMurray
C. Robert Fenlon
North Carolina Central University
Durham, North Carolina
July 21, 2003
*This project was supported by Grant No.1999-MU-CX-0022 awarded by the National Institute
of Justice, Office of Justice Programs, U.S. Department of Justice. Points of view in this
document are those of the authors and are solely those of the authors and do not necessarily
represent the official position or policies of the U.S. Department of Justice. Authors are listed
with senior authors first, followed by junior authors (in alphabetical order). The authors would
like to thank the five anonymous readers for their comments on an earlier draft of this report.

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table of Contents
Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Project Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Official Records (Citations, Written Warnings, Searches, Stops) . . . . . . . . . . . . . 4
The Survey of North Carolina Drivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Citizen Focus Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
North Carolina State Highway Patrol Focus Groups . . . . . . . . . . . . . . . . . . . . . . 19
Chapter 1: “Driving While Black” and the North Carolina Highway Study . . . . . . . . . 23
The Work of the North Carolina State Highway Patrol . . . . . . . . . . . . . . . . . . . . 25
The Importance of Process and Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Non-Bias Mechanisms that Could Produce Racial Disparity in Stops . . . . . . . . 37
Deployment and Patrol Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
References for Chapter One . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Chapter 2: The North Carolina State Highway Patrol Data Bases and Evidence of
Racial Disparity at the Aggregate Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Racial Disparity in Geographic Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Theories of Racial Disparity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Methodological Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
The Importance of Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Citation Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Spatial Heterogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Proxy Measures of Citizen Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Command Policy and Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Aggregation Units of Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Stops, Citations and Written Warnings by Race. . . . . . . . . . . . . . . . . . . . . . . . . . 61
Racial Disparity in Districts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Proxy Measures of Citizen Driving Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Nationwide Personal Transportation Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Spatial Heterogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Day and Night Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
What Are the Mechanisms that Generate Positives (and Negatives)? . . . . . . . . 107
Citation Zones as a Causal Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
References for Chapter Two . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Chapter 3: Racial Disparity at the Individual Trooper Level . . . . . . . . . . . . . . . . . . . 119
Goals of Individual Trooper Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Individual Level Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Measuring the Context of the Officer’s Behavior . . . . . . . . . . . . . . . . . . . . . . 124
Identification of Troop Districts with Troopers as Statistical Outliers . . . . . . . 138
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
References for Chapter Three . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
i

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Chapter 4: Searches by the North Carolina State Highway Patrol . . . . . . . . . . . . . . . . 153
Race and Searches by the North Carolina State Highway Patrol . . . . . . . . . . . 154
Troopers Accounts of Search Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Recorded Search Activity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Probable Cause versus Consent Searches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
Hit Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Individual Trooper Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Chapter 5: Citizen Survey Results: Racial Disparity in Self-Reported Stops . . . . . . . . 181
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Sample Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Racial Differences in Stops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Comparison of Driver Survey Estimates of Stops to Official Records. . . . . . . 185
Racial Differences in Reported Stop Experience . . . . . . . . . . . . . . . . . . . . . . . 187
Racial Differences in Reported Driving Behavior . . . . . . . . . . . . . . . . . . . . . . . 191
Racial Differences in Demographic Background . . . . . . . . . . . . . . . . . . . . . . . 195
Modeling Police Stops in a Multivariate Context . . . . . . . . . . . . . . . . . . . . . . . 196
Discussion of Multivariate Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Racial Differences in Stop Experiences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
References for Chapter Five . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Chapter 6: Racial Differences in Trust in the Police . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .212
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Racial Differences in Trust of the Police . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
Racial Differences in Personal and Network Stop Experiences . . . . . . . . . . . . 215
Distrust of Government and Belief in Profiling . . . . . . . . . . . . . . . . . . . . . . . . . 217
Modeling Distrust of the Police in a Multivariate Context . . . . . . . . . . . . . . . . 219
Distrust of Government Officials Other than the Police . . . . . . . . . . . . . . . . . . 221
Belief in Racial Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Belief in Other Forms of Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
General Distrust of the Police . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Distrust of the Local Police . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Distrust of the North Carolina State Highway Patrol . . . . . . . . . . . . . . . . . . . . 232
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
References for Chapter Six . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Chapter 7: Discussion and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Official Record Evidence of Disparity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Criminal Interdiction Team . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Indicators of Racial Disparity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Self-Reported Traffic Violations and Stops . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Attitudes of the General Public . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
Viewpoints of the North Carolina State Highway Patrol Troopers . . . . . . . . . . 247
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
ii
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Appendix A: Baseline Observational Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Studying Driver Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Modified Carousel Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Baseline Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
Validating the Stopwatch Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Correcting for Under-Estimation of Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
References for Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Appendix B: North Carolina State Highway Patrol Focus Groups . . . . . . . . . . . . . . . 274
North Carolina State Highway Patrol Focus Groups . . . . . . . . . . . . . . . . . . . . . 276
Decision to Stop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
Enforcement Decision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Decision to Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
Minorities and Traffic Stops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Steps to Stop Racial Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
Appendix C: Citation Charge versus Reason for Stop . . . . . . . . . . . . . . . . . . . . . . . . . 292
Appendix D: Offense Codes for Citation Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Appendix E: Self-Reported Police Speeding Stops: Results from a North Carolina
Record Check Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
Background Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
Item Under- or Non- Reporting for Sensitive Questions . . . . . . . . . . . . 335
Race and Item Under- or Non- Reporting for Sensitive Questions . . . . 337
Record Check Surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
Study Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
Design of North Carolina Record Check Survey . . . . . . . . . . . . . . . . . . 340
Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
Response Bias in Reports of Police Stops. . . . . . . . . . . . . . . . . . . . . . . 342
Using Reverse Record Check Estimates of Response Bias to
Adjust for Race Differences in Stop Reports. . . . . . . . . . . . . . . 344
Response Bias and Self-Reports of Other Driving Behavior. . . . . . . . . 346
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
Appendix F: Citizen Focus Groups. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .351
Literature Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .352
Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .354
Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .355
Issue One: General Feeling Toward Police . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
Issue Two: Compared Types of Law Enforcement . . . . . . . . . . . . . . . . . . . . . . 358
Issue Three: Police Stops and Fairness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .359
Issue Four: Targeting African American Drivers? . . . . . . . . . . . . . . . . . . . . . . 360
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .362

iii
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Appendix G: General Issues in Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
The Stop Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
Measurement and Classification of Hispanic Drivers . . . . . . . . . . . . . . . . . . . . 374
Missing Stop Records by District . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
Comparison of Stop Written Warning Actions to Written Warning Records . . 381
Missing Citation Records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
Accidents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
Baseline Observational Stud . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
References for Appendix G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400

iv

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Executive Summary
The North Carolina Highway Traffic Study
NIJ Grant Award Number 1999-MU-CX-0022
Project Overview
The North Carolina Highway Traffic Study is a multi-method investigation of the
phenomenon popularly referred to as “driving while black,” or more generically as “racial
profiling” and “racial targeting.” There is widespread belief that African Americans and other
minorities are at increased risk of police stops compared to white drivers. A 1999 Gallup Poll,
for example, found that 56 percent of whites and 77 percent of African Americans believed that
racial profiling exists (Newport 1999). In our own survey of adult, North Carolina-licensed
drivers, 30 percent of whites and 80 percent of African Americans reported that they believed
that African Americans were more likely than other drivers to be pulled over by the police.
“Racial profiling” and “racial targeting” refer to a fairly specific police practice of using race as
an explicit criterion for deciding which cars to stop or search. “Driving while black” is a less
focused term, but summarizes a widespread belief in minority communities that they are singled
out for harsher treatment than are white drivers.
In designing this project we were faced with three associated issues. First, the actual
degree and spatial/organizational distribution of racial disparity in stops are not known and
current methodologies are inadequate for establishing scientifically reasonable estimates of
disparity. This project develops and evaluates a series of alternate methodologies for
1
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

establishing the degree of racial disparity in stops. Second, the political attention to this complex
phenomenon needs to be clarified with theoretical understandings of the various mechanisms
which plausibly could produce racial disparity in police stops. While racial profiling, the
explicit use of race by police as an indicator of potential criminal status, might be one such
mechanism, there is no reason to believe it is the only or even most general mechanism. For
policy, ending explicit racial profiling might do very little to reduce racial disparity in police
stops if other mechanisms produce racial disparity in stops and post-stop outcomes. Indeed, it is
necessary to examine mechanisms that, on their face, are not racially biased, but may in fact
work to produce observed racial disparity in traffic stops. For example, one might expect some
level of observed disparity in stops if police deployment, perhaps in response to calls for service
or accident rates, increases police presence in areas that are disproportionately minority. Here,
minority vehicle stops may be a function of increased patrols and resultant citizen contact.
Third, regardless of the level or dispersion of ethnic disparity in stops, the perception that
“driving while black” places some community members at special risk represents a widespread
threat to the legitimacy of law enforcement.
This project combines demographic analyses, highway observations, surveys of citizens,
focus groups with drivers, and focus groups with North Carolina State Highway Patrol troopers
to develop methodologies to estimate racial disparity in traffic stops, identify plausible
mechanisms producing those disparities, and learn more about the consequences of perceptions
of racial disparity in policing for trust in the police.
This project began in 1999 when the North Carolina State Legislature mandated that the
North Carolina State Highway Patrol (NCSHP), and all state law enforcement agencies, begin to
assemble data on the racial distribution of all vehicular stops initiated by officers. The NCSHP
2
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

agreed to cooperate with us in a more thorough study of traffic stops and outcomes. The
cooperation of the NCSHP allowed us access to a great deal of demographic data on stops as
well as cooperation in organizing focus groups with troopers and collecting observational data
on North Carolina highways. Our project is a joint effort of faculty researchers and graduate
research assistants at North Carolina State University and North Carolina Central University.
The research is intended to answer four basic questions: 1) Do NCSHP troopers stop
minorities, particularly African Americans, on the road at higher rates than they do whites? 2)
Once stopped, do African American citizens and white citizens experience different rates for
citations, written warnings, and searches? 3) What factors might account for highway stops? and
4) How do African Americans and other ethnic minorities experience and respond to traffic
stops? Our goal is to produce informed answers to these questions that can help to shape public
policy, police training, and citizen outreach.
These specific research questions result from the way that we conceptualize the“driving
while black” phenomenon. That is, we view “driving while black” as the result of at least three
processes: police stops of motorists, the decisions that motivate police stops, including racial
and ethnic bias as well as drivers’ behavior, and the interpretations of police stops by the
minority communities.
We have taken substantial care to produce appropriate baseline comparisons
(denominators) for highway racial homogeneity or heterogeneity and driving behaviors.
Baseline denominators have been produced from available records (such as licensed drivers
within counties), calculations to create estimates of “drivers driving” within NCSHP districts (in
other words, it is necessary to estimate the number of drivers driving in a particular district [or
area] who do not reside in the district of observation), direct observation of motor vehicles and
3
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

drivers on the highway, and traffic accident rates. These various baseline denominators allow us
to more completely examine the racial/ethnic differences in stop rates and outcomes in the
contexts of racial composition on the highway, driving behavior, and trooper activity (individual
and organizational).

Official Records (Citations, Written Warnings, Searches, Stops)
Official record information provided by the NCSHP includes data bases on vehicular
stops, citations, written warnings, and searches. Our ability to verify that there were stop records
to match all associated written warning or citation records is limited by the fact that there was no
identifying number linking the different data bases. The lack of such a linkage has implications
for our ability to assess whether, for example, every non-accident related citation has an
associated stop record. Using a restricted list of specific measures across data bases, we were
unable to identify many stop records as matching the citation or written warning records.
Indeed, we found that approximately one in three incidents that could have had a corresponding
stop record did not – in part because stops at checkpoints no not require that a stop form be
completed, calling into question the completeness of the stop record data and raising questions
about why some stop records would not be filed.
As a consequence, we focused much of our analytic attention on the citation, written
warning, and search records. As for the general relationship between citations, written warnings,
and the race of the driver, we found that there was considerable variation in the racial
distribution of these interventions across the types of behaviors that were likely to have resulted
in the citation or written warning incident (such as speeding or unsafe vehicular movement).
Our initial hypothesis was that the more subjectively measured behaviors, such as “driving too
4
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

close” or “failure to yield,” would show greater disparity than the more objectively measured
behaviors, such as vehicular speeding (virtually always measured by radar guns when citations
are issued), having a revoked or expired license, or failure to wear a seat belt. The opposite is
the case: the more objectively measured indicators of violating behaviors in citations are more
often involving African American drivers. This suggests the possible importance of variations in
driver’s behavior as a primary determinant of whether or not someone is cited.
Racial disparity in official written warnings was generally found to be more pronounced
than in citations. However, such disparity is inherently ambiguous. If African Americans are
more likely than whites to receive a written warning for “unsafe movement,” it may be evidence
of more lenient treatment of African Americans (who are receiving a written warning rather than
a citation) or it may be evidence of so-called pre-textual bias: troopers are stopping vehicles and
giving warnings as a pretext to looking over the vehicle for signs of contraband. Since searches
by the NCSHP troopers are very rare, the latter interpretation is doubtful.
The question of the extent of racial disparity in citations and written warnings issued by
the NCSHP is much more complicated than looking at the racial distribution of the stated
reasons for these interventions. Ideally, the statewide incidence of citations for a type of
vehicular behavior needs to be evaluated in terms of what are the likely mechanisms by which
statewide patterns of disparity are generated. The understanding of when and where troopers are
deployed may go a long way to account for racial disparity. For example, if there are relatively
many African Americans living in urban areas of North Carolina, and the NCSHP over-patrol
highways in and near urban areas, the proportion of citations issued to African Americans
statewide may be enlarged.

5
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Somewhat to our surprise, we found empirical evidence to the effect that there is also
racial variation by time of day in the distribution of drivers on the highways of North Carolina.
African Americans are more likely to be driving in the evening and early morning hours relative
to their distribution in the licensed driver population. To the extent that the NCSHP happens to
patrol more at night is some areas could also enhance the racial disparity estimate.
The process of evaluating whether the NCSHP stops and cites African Americans
excessively is made easier if objective measures of drivers’ behaviors are available against
which to compare with citation rates. For example, all else equal, we would expect that the
proportion of African Americans stopped for speeding or other infractions would mirror the
corresponding rate at which these infractions occurred (if 20 percent of the drivers speeding are
African American then 20 percent of the drivers cited for speeding would also be African
American). Toward the goal of establishing a method to measure objectively drivers speeding,
we conducted an observational baseline study at fourteen sites. Each site consisted of between
10 and 15 miles of highway (both directions). We sent a research team to each site for a week to
measure the speeds of motorists passing our research vehicle, which was driving at the speed
limit while traveling these stretches of highway. We estimated the speed of the passing
motorists by using stop watches to measure the time it took the passing vehicle to pass the rear
and front bumper of the researcher’s van. These estimates were found to be within a few miles
per hour of the actual speed (as validated in road tests that we conducted).
We operationalized speeding as driving at or above the speed at which drivers were
generally found to be cited for speeding and found that there were differences in the driving
behaviors by race. Specifically, African Americans were over-represented among those driving
above the “speeding threshold” at which drivers tend to be cited. However, the African
6
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

American over-representation declined above approximately 8 mph over the threshold speed,
indicating that the relationship between vehicular speeding and race is more complicated than we
initially thought. The racial speeding differences that were observed, however, do not
necessarily generalize to all highways in North Carolina, nor should they be used to generalize to
other states. As argued in the report, the relationship between speeding and race is potentially
quite complex, having to do with the types of roads and reasons for driving, among other things.
The observational baseline study was particularly useful in alerting us to the fact that
there is considerable variation in the proportion of drivers who are African American across even
very proximate locales. Even the same highway a few miles apart or two highways intersecting
may have a substantially different racial composition. Subtle differences in the volume of
patrolling could interact with the variations in where African American drivers drive to generate
higher or lower percentages of African American drivers stopped and cited. That is, the percent
of those stopped and cited who are African American may reflect variation in the deployment of
troopers to one highway over another. Unfortunately, we do not have measures at the very
micro-level of where troopers patrol. We only have records of troopers having written a citation
or written warning in what we call county highway areas (segments of highway within about a
third of a county). The unmeasured variation in patrolling, coupled with unmeasured racial
composition of specific segments of highway (smaller than a county highway area), is referred to
as the “spatial heterogeneity” problem. This reduces our ability to make strong claims about the
degree of racial disparity and about the likelihood that racial bias accounts for such disparity.
So far we have been speaking about what is unavailable to us: multiple and objective
measures of driving behavior and of where (precisely) troopers spend their time patrolling. On a
more positive methodological note, we do have measures that can partially substitute for the lack
7
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

of direct measures of the behavior of drivers, and we can make some reasonable assumptions
about patrolling so as to lessen the likelihood of spatial heterogeneity problems. Specifically, we
have measures of the resident driving population living in an area such as a county (100 counties
in North Carolina) or patrol district (fifty-three such districts). We were able to utilize data on
the residency of drivers cited outside of their county of residency to create an estimate of
“drivers driving” in an area (in a nutshell, we estimate the proportion of drivers driving in a
district based on the assumption that their driving outside of their district of residency is racially
proportional to the composition of licensed drivers within their district—see discussion in the
report for details). A third source of data to be used as a baseline against which to compare
citation and written warning rates is the racial composition of drivers in accidents. This third
source of information is particularly useful in that it can be measured at relatively small units of
analysis (the county highway areas alluded to above). We found that we could measure the
proportion of those drivers involved in accidents who are African American at these relatively
small units of analysis. While not an ideal unit of analysis, the county highway areas represent
some degree of control for spatial heterogeneity problems.
Having discussed some of the primary methodological issues, we turn our attention to the
results of the analysis. The analysis is broken down into two parts. The first assesses whether or
not there are any districts which have a high rate of citations of African Americans (end of
Chapter 2). The second describes some models for assessing whether specific troopers have
unduly high numbers of citations of African Americans (Chapter 3).
For the aggregate analysis, in which our goal is to determine whether or not there are any
districts (of the fifty-three NCSHP districts) with unduly high citation rates of African
Americans, we show that there are several such districts (varying by night and day times) that
8
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

qualify as “positive outliers”—the districts “lie” outside a range of measurement that is likely
due to measurement error, and as such, qualifies them for further scrutiny as to possible racial
disparity. At the same time, it must be mentioned that there are even more districts that are
“negative outliers”— in other words, they have fewer citations of African Americans relative to
the prevalence of African Americans in accidents. Thus, some observers might attribute the
pattern of results to measurement error or to factors that were unspecified in the analysis
(uncontrolled factors). It should also be mentioned that the specific districts defined as positive
or negative outliers will vary somewhat depending on which baseline (resident drivers, “drivers
driving,” or accidents) is used. Thus, there is ambiguity as to whether any particular area is
suspect as having unduly high levels of racial disparity.
Such a finding of ambiguity may be disquieting to those who would want to know what
might seem to be a straight answer to a simple question: is a troop racially biased or not? We
interpret our evidence to mean that we can safely rule out widespread and large degrees of racial
disparity in the behaviors of the NCSHP across districts. However, we cannot rule out with
certainty the presence of small degrees of racial disparity. That is, in some districts there may be
some disparity that cannot be accounted for by the deployment patterns of troopers, but our
measures and methods are not adequate to tell conclusively. Some will argue that our models are
not sufficiently fine-tuned to rule out non-bias interpretations. We agree, but we do not have
measures of what those factors might be. If there is racial bias operating, it is most likely of a
“cognitive” sort (defined as bias that does not stem from conscious or overt racial antagonism).
As for the presence of racial bias in the behaviors of individual troopers, we are dealing
with a somewhat different policy issue since evidence of racial disparity may be grounds for
“personnel action”—at least of an investigative sort. That is, if a trooper has a tendency to cite
9
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

more African Americans than he or she “should”—relative to some baseline– then there may be
grounds for looking further as to whether or not the officer is a “positive outlier.” This
investigation would presumably weigh further evidence so as to determine if there are reasons
other than bias for the high number of citations of African Americans. Thus, our research here
has as its goal to show methods that could be used to identify the “outlier” troopers, and not to
claim that any specific trooper is disparate or biased in his or her citation behavior.
We present some regression models at the individual trooper level in which the
dependent variable is the number of African Americans cited in the year 2000. We find that a
control variable for the volume of citations (specifically, the number of whites cited) is a strong
determinant of citing of African Americans. So too is any of a number of contextual measures
(the percent of citations issued by other troopers to African Americans or the percent of drivers
in accidents who are African American). Time of day effects (such as late evening or early
morning) and type of highway (interstate or rural highways) are also found. Approximately 60
to 70 percent of the variance in the citation of African Americans is explained by these factors.
When we try to identify troopers who represent positive and negative outliers relative to
the ordinary least squares model, we find that approximately eighteen are “positive” outliers
(have more citations of African Americans than our model indicates should be “expected”) and
slightly more are “negative” outliers (fewer citations of African Americans). More sophisticated
statistical models indicate only partial overlap in who is a positive or negative outlier. Thus,
some of the “positive outliers” from the ordinary least squares model may be “false positives,” as
some of the negatives may be “false” also. Such findings are unfortunate from an efficiency
point of view, since there are multiple classifications of who has high (or low) levels of citations
of African Americans. However, recall that our purpose is to show that the statistical models can
10
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

be used as part of a more general strategy to assess whether or not racial disparity or bias is
present in the actions of individual troopers rather than to claim that any officer is disparate or
biased. We think we have accomplished that purpose.
One final set of findings based on official records includes the study of the consent and
probable cause searches conducted by the NCSHP. We distinguish between the behavior of the
regular road trooper (who rarely searches a vehicle) from that of the Criminal Interdiction Team
(CIT—whose job it is to stop vehicles, question drivers, and search for contraband such as drugs
or guns). For the regular trooper, almost all of them seem to avoid proactive work toward the
goal of conducting a search. The small volume of consent searches (searches justified based on
some suspicion, yet requiring the permission of the driver to conduct the search) indicates that
the regular trooper is not proactively looking to conduct searches. Among the small number of
troopers who make up the CIT (about thirteen in calender year 2000), searches are more
common place (although the number dropped off in the late 1990s to only a little more than one
vehicular search per day by the year 2000). As for the racial composition of the searches, the
CIT troopers are more likely to search the vehicle of an African American stopped than that of a
white. They did so somewhat inefficiently as late as 1997, but by 2000 the “hit rate”
(successfully finding drugs or other contraband) was higher for an African American driven
vehicle than a white driven vehicle (however, we do not mean to suggest that the high rate of
finding contraband justifies the high search rate of African Americans, as that is a more complex
question, involving utility and civil libertarian concerns beyond to the scope of our project).
As for the mechanism that could account for the disparity in the search rates of African
Americans, it seems to us, based on discussions with CIT troopers, that their use of the
“conversational method” is one that could easily lead to the manifestation of disparity or bias. In
11
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

this method, they question drivers routinely and evaluate such factors as the consistency of the
driver’s answers and the degree of nervousness of the driver (and other vehicle occupants).
Based on what they see and hear, they may decide to ask permission to search the vehicle. We
do not know whether African American drivers are more likely to draw the suspicions of the
troopers because the troopers are “looking harder” for the proper “signs” due to the race of the
driver. There are alternative explanations about which we simply lack sufficient information.
For example, African Americans’ perceptions that the troopers are “looking harder” for some
violation when they are stopped may result in a higher prevalence of “nervous behavior,”
independent of culpability during their stops.
The plausibility that “cognitive bias” may account for the higher search rate of African
American driven vehicles is enhanced by the CIT troopers’ recognition that they use certain
generalizations in their everyday interaction with the public. Decisions must be made on a daily
basis as to whether or not the citizen in front of them poses a threat to them or not. Style of
dress, hair or verbal expressions will all be drawn upon by the trooper in making the decisions.
Some typing is probably necessary for some decisions, such as whether or not to exercise
extreme caution. While we do not question the practical need for generalizing behaviors and
situations, we merely point out that “typing” people may lead to decisions that have racially
biased implications.

The Survey of North Carolina Drivers
The North Carolina Driver Survey was designed to compliment the official statistics
analysis described above. Official law enforcement statistics are accounts of citizen-trooper
encounters provided by the individual trooper and organization, as complete and accurate as they
12
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

may be. Further, official data contain little information about driver behaviors which may
provide an opportunity to examine whether drivers who are stopped actually drive differently
than those who are not stopped. The survey data we collected allows us to collect information
on reported typical driving behaviors that may influence the probability of being stopped and to
capture information about stops conducted by law enforcement agencies across the sate. In the
survey we not only asked North Carolina drivers whether or not they were stopped, but also why
they were stopped, the outcome of the stop, and how they were treated. Overall, the survey was
intended to 1) help develop more inclusive baseline estimates of African American and white
motorists’ differences in driving patterns and driving behavior; and 2) measure African
American and white motorists’ differences in traffic stop experiences and their respective
interpretations of the events.
We conducted a telephone survey for a stratified random sample of current North
Carolina licensed drivers (African American=1,368; white=1,487). The sample was stratified by
race in order to have sufficient sample sizes to compare the experiences of white and African
American drivers. The sampling frame included white and African American drivers who had
applied for or renewed their licenses in the previous six months. Data were collected between
June 22, 2000, and March 20, 2001.
A comparison of our final sample to the actual race-gender-age distribution of licensed
drivers in North Carolina shows that our final sample is quite a good match to the state
distributions. Still, in all four gender-race groups, young adults age 30–39 are under
represented. In most statistical analyses we weight the data to correspond to the known gender
and age distributions of licensed drivers within the two race strata.

13
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

African American drivers are significantly more likely than their white counterparts to
report a traffic stop in North Carolina. The odds of a stop by local police may be twice as high
for African American as they are for white drivers even after controlling for other demographic
statuses and reported driving behavior. Local police are also significantly more likely to stop
African American males relative to African American females, while among whites there is no
gender disparity in stops after controlling for driving behavior.
The estimated racial disparity in stops by the NCSHP is much smaller, but still
statistically significant after controls for driver characteristics and reported driving behavior.
The NCSHP does not stop African American males at higher rates than African American
females net of driving behavior. Among the NCSHP troopers, race is linked to other attributes in
the stop decision. Older whites and whites driving late model cars are less likely to be stopped
than are other whites. African Americans who report more risky driving behaviors are more
likely to be stopped. This suggests that the NCSHP troopers are not simply reacting to the race
of the driver, but perhaps to the combination of race and other status attributes for whites and
race and driving behavior for African Americans.
After the stop, differences in white and African American reported experiences are less
dramatic. African Americans are slightly more likely to have been informed that the stop was
for a more discretionary reason. African Americans are also slightly more likely to report that
they were treated disrespectfully after the stop. There are no racial differences in the distribution
of self-reported citations, written warnings, and verbal warnings. Racial differences in
experiences after the stop are small.
Our telephone survey also addressed citizen trust. We found that distrust of law
enforcement is produced by a combination of negative personal experiences with the police,
14
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

negative experiences of family and friends, belief in police profiling on both racial and nonracial grounds, general distrust of government institutions, and being a minority.
The related problems of racial profiling and trust in the police are not simple ones.
African Americans distrust the police because of their personal experiences and more general
community orientations. Disrespectful interactions are particularly powerful sources of both
distrust in the police and belief in racial profiling. This is not, however, simply a perception
produced by direct experience. On the contrary, negative encounters with the police by family
and friends generate distrust and increase belief in racial profiling. In fact, among African
Americans, disrespectful police treatment or stories of disrespectful police treatment can even
undermine trust in government institutions in general. Belief in racial profiling undermines trust
in the police even among whites.
African Americans are more forgiving of the NCSHP than are whites. African
Americans are more likely to translate negative experiences into distrust of local police forces
than the NCSHP. This may reflect their observations of lower bias or more professional carriage
by NCSHP troopers. Whites, on the other hand, are less discriminating. Any perception of
disrespect or profiling undermines white trust in all types of police. Whites are particularly
influenced by perceptions of non-racial profiling (for example a young driver playing loud
music), perhaps because these are the types of profiling for which they or family members are
thought to be most at risk. Thus while African Americans are more distrustful of the police in
general than are white citizens, whites’ trust in the police seems more vulnerable to recent
experiences and media portrayals.
Citizen trust in police is also influenced by more general dispositions toward trust in
government. This is true for white and African American citizens and for all types of police
15
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

examined. This suggests that the legitimacy of the police in general, and of specific police
forces, is a nested problem. Police legitimacy is undermined by perceived disrespectful
treatment (especially among whites), belief in racial profiling (especially among African
Americans), and belief in other forms of profiling (especially among whites). Where racial
disparity in treatment is lower, as in the NCSHP versus local police, African Americans do not
translate negative experience into reduced trust. Police legitimacy is more vulnerable among
whites. African Americans, however, have a lower level of trust in the police of all types
stemming from their past experiences in, and cultural understanding of American society. Some
of this can be seen in African Americans’ lower trust in government institutions in general, but
most seems to be focused on a specific fear of the police. Distrust of the police among whites is
more strongly tied to distrust of government institutions in general.

Citizen Focus Groups
Citizen focus groups enable us to examine reported experiences of drivers and better
understand the range of feelings about racially motivated traffic enforcement among both
African Americans and whites. Specifically, we used these discussions with citizens to explore
reported and perceived reasons for police stops, the perceived treatment of citizens by NCSHP
troopers as reported by the respondents, their experience with other law enforcement encounters
(local, county, state), how the police-citizen encounter began and unfolded (did the stop result
from a stationary radar unit, passing on a two-lane or four-lane road, or driving side by side), and
what knowledge citizens can report about police-citizen encounters by other community members, friends and relatives. We were very interested in learning something about the themes and
patterns of their police-citizen encounters that might directly inform both policy and practice.
16
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Our focus groups with African American drivers revealed a generally positive evaluation
of the job that the police do. Participants were quick to say that the police had an important and
difficult job and that they were grateful for the good work they do. At the very same time many
of the African American drivers had very little trust in individual police officers. They felt at
risk for harassment and bias based on race and made considered analytic distinctions for each
and every time they were stopped. Some stops were judged fair, typically when the law was
broken and they were treated with respect. In general, law enforcement as an institution was
described as legitimate and reasonable, individual police were suspect, and racial bias was
attributed to “bad apples.” There was, however, some disagreement among African Americans
as to how common the bad apples are.
Stops that were not tied to serious illegal driving behavior—the most common of which
was the “rolling stop sign” pull-over—were considered to be likely instances of racial bias. In
many of these cases African Americans assumed race was the cause of the stop, because they did
not recognize any other legitimate reason. In some cases this assumption was confirmed by the
officer making the stop, such as when reporting that the African American citizen was stopped
for being in the “wrong”—that is to say “white”— neighborhood (and thus out of place). One
young man spoke of the time he was stopped (with his brother), removed from the car, tackled,
and had guns drawn on him for driving in a neighborhood where another African American man
on foot was being pursued by the police. Here, apparently, “young, back male on foot” was
interpreted as “African American male anywhere.”
Lack of respect by the police during legitimate stops were also evaluated by some
African American drivers as likely evidence of racial bias. Lack of respect in the interaction was
interpreted as an indicator of racial bias, and encouraged the suspicion that the pull-over was
17
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

racially motivated as well. Troopers of the NCSHP, in contrast to officers attached to various
local police forces, were singled out as treating drivers with respect and professionally. It was a
clear pattern in the focus groups that African American drivers had less suspicion of the NCSHP
than they did of other police officers. While this evaluation mirrors our findings in the citizen
survey that racial disparity in police stops is lower among the NCSHP than among other law
enforcement agencies in North Carolina, the focus group participants used respectful treatment,
rather than the rate of stops, as the basis for arguing that the NCSHP was better.
In general, African Americans were more likely to perceive racial bias in a stop if the
officer interacted with them in a disrespectful manner or if they were stopped without what they
believed to be legitimate driving infractions. They seemed to be more than willing to
acknowledge their personal responsibility for a “real” violation. What were seen as minor
violations were another story. Here they saw race as the predictor of the stop, not the violation.
This is in contrast with white drivers who tended to see all stops, “real” or otherwise, as
discretionary and idiosyncratic. White drivers talked about “driving while blond” or “driving
while a musician” or “he should have cut me a break.” In many ways white drivers evaluated the
police more harshly than African American drivers did, and were also more likely to generalize
“unnecessary” enforcement across agencies. African American drivers saw many stops as
legitimate and some as potentially racially biased. White drivers saw most stops as illegitimate,
but idiosyncratic.
We also found stark contrasts between African American and white drivers in evaluations
of the “driving while black” phenomenon. African Americans tended to see it as just another
example of general and continuing racial bias. Racial bias in the policing of drivers was seen as a
form of discrimination similar to the other forms of discrimination faced each day. Its existence
18
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

was confirmed by some combination of their own experiences, stories they had heard from
friends and family, media reports on questionable police behavior (Rodney King was often
mentioned), and the existence of general levels of prejudice and discrimination in the society at
large.
White descriptions were considerably simpler and more disturbing. The white focus
groups tended to accept that police targeted African American drivers, but described racial
targeting as at least understandable if not fair and justifiable. Since African Americans were
stereotypically assumed to be more dangerous and thus more culpable, white citizens typically
saw police stops on the basis of race as reasonable. Whites tended to use stereotypes and
statistical discrimination arguments similar to those sometimes used by police to justify racial
targeting. It seemed very easy for the white subjects to collectively justify discrimination in
policing, even though they were quite resistant to taking personal responsibility for their own
police stops. As such, disgruntled white drivers are not natural allies of African American
drivers who fear they are being harassed because of their race.

NCSHP Focus Groups
The purpose of the focus groups with NCSHP troopers was to inquire into the following
domains: What circumstances are considered before, during, and after a stop? What are the
training issues pertinent to highway traffic stops and how are they interpreted by NCSHP
troopers? Is there a perceived reward structure that might influence the behavior of NCSHP
troopers? And, what are the troopers’ interpretations of racial differences in stops made by the
NCSHP?

19
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Six focus groups were conducted in early June, 2001. Due to the racially sensitive nature
of the topic, four focus groups were race-specific (two African American and two white). This
was deemed appropriate in order to best provide a forum where respondents would feel less
restricted, although each group noted that they would feel comfortable speaking in the presence
of fellow troopers. The management and command groups were racially diverse. Our random
selection process did not capture any women troopers (there are few women in the NCSHP
relative to men). Focus groups numbered from six to nine troopers and each lasted
approximately 2 hours. The sessions were facilitated by two members of the research team.
Three other members of the research team observed the sessions shielded by a one-way glass
window.
The central issue, of course, was whether or not extra-legal factors (especially race) serve
as the basis for the disproportionate number of stops involving African American and Latino
drivers. It appears that troopers, for the most part, engage in enforcement patterns they believe
would yield the greatest number of enforcement opportunities. A major determinant in the
decision to make a stop appears to be the “behavior” of the vehicle. The focus on such behavior
seems to vary situationally. The interstate, it is believed, is more likely to yield speeding
violations rather than seatbelt violations. Participants state that it is not possible to know the
race of the driver on the interstate or at night. Rather, they report that their focus is the behavior
of the vehicle. Still, and significant given our interests here, not all troopers were unwilling to
attribute likely traffic violations to specific segments of the community.
Troopers also generally acknowledged that it is easier, because of reasons out of their
control, to do their jobs in some places and not others. Simply, some citizens are more likely to
resist the legality of the troopers’ actions, complain to supervisors, and challenge the citation.
20
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Each of these situations is something that most troopers would like to avoid. Interestingly, but
not surprisingly, it was reported that the NCSHP receives more complaints from whites as
compared to African Americans and Latinos. We pick up the same theme in our citizen focus
groups and our telephone survey. White focus group citizens tended to see any stop as an
unnecessary intrusion and an unproductive use of police time. African American citizens tended
to acknowledge and accept responsibility for stops resulting from clear violations of traffic laws.
This raised an auxiliary issue that compounds the problem of racial profiling: how does
the coupling of expected resistance from a white person—who may also be from a more affluent
segment of society that possibly harbors stereotypes of the minority community (such as
perceived levels of law violations)—impact the level and degree of disparity in stop outcomes?
This coupling, intentionally or unintentionally, may produce deployment or locales of
enforcement that will serve only to increase disparity in traffic stops. It appears that while some
deployment decisions are based on traffic demands (for example, a road with a history of
accidents or fatalities), others are based on areas with a significant “opportunity” factor. That is,
opportunities associated with density of bars or other environmental factors (“where the fishing
is good”). Such decisions are more likely to target low-income people (thus disproportionately
people of color) than their high-income counterparts. This is manifested with the presumption
that they are less likely to challenge the action in court and that the higher income areas are
involved in less overall criminality and disorder.
While there was acknowledgment of the possibility of racial profiling in the NCSHP, it is
generally believed to be an infrequent occurrence today but perhaps was a more frequent
occurrence in recent years past. They attributed the reduction in complaints to the dismantling

21
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

the statewide drug interdiction units and reducing the competitive nature surrounding the
quantity of drugs seized.

22
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Chapter 1 “Driving While Black” and the North Carolina Highway Study
This project is a multi-method investigation of the phenomenon popularly referred to as
“driving while black.” Cognate terms that refer to police behavior include “racial profiling” and
“racial targeting.” The general social problem is the widespread perception that African
Americans and other minorities are at increased risk of police stops compared to white drivers.
A 1999 Gallup Poll found that 56 percent of whites and 77 percent of African Americans
believed that racial profiling is widespread (Gallup 1999). In our own survey of North Carolina
drivers, 30 percent of whites and 80 percent of African Americans reported that they believed
that African Americans were more likely to be pulled over by the police than other drivers.
“Racial profiling” and “racial targeting” refer to fairly specific police practices of using race as
an explicit criterion for deciding which cars to stop or search. “Driving while black” is a less
focused term, but summarizes a widespread belief in minority communities that they are singled
out for harsher treatment than are white drivers.
In designing this project we were faced with three associated issues. First, the actual
degree and spatial/organizational distribution of racial disparity in stops are not known and
current methodologies are inadequate for establishing scientifically reasonable estimates of
disparity. This project develops and evaluates a series of alternate methodologies for
establishing the degree of racial disparity in stops. Second, the political attention to this complex
phenomenon needs to be clarified with theoretical understandings of the various mechanisms
which plausibly could produce racial disparity in police stops. While “racial profiling,” the
explicit use by police of race as an indicator of potential criminal status, might be one such
mechanism, there is no reason to believe it is the only or even most general mechanism. For

23
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

policy, ending explicit racial profiling might do very little to reduce racial bias in police stops if
other racially biased mechanisms produce racial disparity in stops. It is also possible that other,
not racially biased, mechanisms produce the observed racial disparity in stops. For example, if
police deployment in response to accident rates or calls for service increase police presence in
minority neighborhoods this might lead to higher minority automobile stops as a function of
increased patrols and contact. Third, regardless of the level or dispersion of ethnic disparity in
stops, the perception that “driving while black” places some community members at special risk
represents a widespread threat to the legitimacy of law enforcement.
This project combines demographic analyses, highway observations, surveys of citizens,
focus groups with drivers, and focus groups with troopers to develop methodologies to estimate
racial disparity in police stops, identify plausible mechanisms producing those disparities, and
learn about the consequences of perceptions of racial bias in policing for trust in the police. This
project began in 1999 when the North Carolina State Legislature mandated that the North
Carolina State Highway Patrol (NCSHP), and all state law enforcement agencies, collect data on
the racial distribution of all vehicular stops initiated by officers. The NCSHP agreed to
cooperate with us in a more thorough study of traffic stops and outcomes. The National Institute
of Justice funded this research. The cooperation of the NCSHP allowed us access to a great deal
of demographic data on stops as well as cooperation in organizing focus groups with troopers
and collecting observational data on North Carolina highways.
In this introductory chapter we first present a brief review of current approaches to the
problem of racial bias in traffic stops and suggest that simple assumptions about the universality
of police bias or lack thereof are unlikely to capture the reality of the situation. Instead we
advocate that if bias occurs, it occurs in the context of the work different police forces do, their
24
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

organizational practices, the immediate locality where individual officers patrol, and in
interaction with citizens. This is followed by a theoretical elaboration of several mechanisms
that could plausibly produce racial differences in police stops. For example, one mechanism
could be average racial differences in driving behavior. Simple differences in racial composition
of stops do not demonstrate the presence or absence of police bias. Rather, we should suspect
racial bias only after accounting for racial/ethnic differences in driving behavior. On a practical
level, what is most important in this regard is the racial distribution of drivers on the road. Who
is driving and where they drive will vary dramatically from place to place as a function of racial
differences in residence, employment, and driving destinations. We argue that, for policy
purposes, estimates of racial disparity in police stops adjusted for driving behavior are preferable
to simple counts of the racial distribution of police stops. A variety of approaches to establishing
racial disparity are briefly discussed in this chapter. Finally, the chapter introduces the issue of
when and how minorities perceive police bias and the consequences of such perceptions for trust
in law enforcement as an institution.

The Work of the North Carolina State Highway Patrol
Since much of this report focuses on the stop and search decisions of the NCSHP we
begin with a description of the work done by the NCSHP. We wish to emphasize at the outset
that this research did not arise out of a specific lawsuit or other accusation of gross racial bias by
the NCSHP. Instead, this research was developed with the cooperation of the NCSHP who was
willing to take the public risk of external research on the topic of racial bias in policing in order
to both facilitate the research and improve police practice.

25
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Public attention is often focused on the NCSHP when they make a large drug bust or
when a trooper loses her or his life in the performance of duty, but the day-to-day reality of
patrolling for most of the approximately 1,400 NCSHP troopers is one of responding to accidents
and stopping vehicles that are speeding or are otherwise in violation of safety laws, and then
writing citations and warnings. This is occasionally dangerous but seldom glamorous work.
NCSHP troopers generally carry themselves with a great deal of dignity; they keep their
uniforms crisp and clean, wear their hats when approaching vehicles, and endeavor to treat
citizens politely but firmly. Making the highways safe is the primary and predominant function
of the NCSHP in North Carolina. In addition, a small special force of approximately twelve
troopers is assigned primary responsibility for drug interdiction and identification of other types
of contraband. We will discuss this unit, the “Criminal Interdiction Team” in Chapter 4. For the
most part we will be focusing on racial disparity and possible racial bias in the routine day-today activities of the NCSHP.
Because the activity of the NCSHP is primarily oriented toward vehicle-safety-related
stops, about half of which are for speeding, the opportunities and motives for racial bias in police
stops are probably small relative to other police forces that have both broader jurisdiction and
enforcement goals. While troopers have some discretion in the decision to write a ticket, the
information that generates speeding stops is mostly guided by race-blind radar devices. Racial
stereotypes that associate minority status with criminality or drugs might be expected to
encourage racial bias in policing, but most NCSHP troopers do little or no criminal or drug
investigation work. Still, traffic law enforcement is a combination of proactive and reactive
activities. NCSHP troopers receive few 911 or other citizen initiated calls, relative to local law
enforcement, for example. This situation would seem to clearly increase the time available to
26
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

them to choose which citizens they encounter. But it is important to recognize that both citizen
initiated calls for service and citizen initiated driving violations result in a reactive response from
police. It is precisely the duality of traffic enforcement activities that is at the heart of the racial
profiling controversy: Do police stop only those whom they see or only that which they see?
The major work of the NCSHP officer is stopping cars for traffic law violations. In
Figure 1.1, the numbers of “citation events” handled by the NCSHP are presented by month for
the years 1997 through 2000. A citation event is an occasion in which a citation or multiple
citations are written. Several citations can be written at a stop or accident scene (an event).
Since it is important for us to know about the composition of the drivers who are stopped,
warned, cited, and so forth, in this report we will often be examining data on citation events.
Here, and throughout most of the discussion in this and the next two chapters, we will be
focusing on citation events when we refer to “citations.”
In Figure 1.1 we see that the four-year average of citation events is approximately 52,000
per month (horizontal reference line), or approximately thirty-seven per month for each of 1,400
troopers. Each year the NCSHP cites more than .5 million drivers in North Carolina (roughly
500,000 to 650,000, varying by year). The citation rates were generally above the four-year

27
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 1.1 Trend in Citations, 1997–2000

28
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

average (1997–2000) prior to the summer of 1999, and below it afterwards, although for some
months in 2000, the rate does rise to or above the average. At least a portion of the decline after
1997 in the number of citations is probably due to the implementation of the NCSHP’s
implementation of a Total Quality Management (TQM) plan intended to shift troopers’ activities
to focus on their core responsibility—reducing accidents—rather than the volume of citations.
We will discuss TQM more below in our account of the focus groups with NCSHP troopers.
The second, vertical, reference line highlights January, 2000, when Senate Bill 76,
requiring NCSHP troopers to record all traffic stops, went into effect. Senate Bill 76 was
specifically concerned with the possibility of racial bias in police stops among state law
enforcement agencies, particularly in stops that did not result in a citation or written warning.
Across 1999, there is a steep drop in citation activity. This is a period when the NCSHP
patrolling was the lowest in several years (NCSHP Statistics, 2001) (The NCSHP Web site
shows almost one-hundred-thousand fewer patrolling hours in 1999 than in 1998).1 It was also
the period of political discussions about racial bias and the drafting of legislation to require the
NCSHP to collect new stop data and investigate the potential for racially biased policing. After
January, 2000, the frequency of citations seems to become stable around the four-year mean. It
appears that neither Senate Bill 76, nor the political and media attention to “driving while black,”
influenced aggregate citation activity. It should be noted that the number of citation events will
vary with the number of troopers working the highways, and we will have more to say about

1

In 1998, the NCSHP Statistics Web page shows that there were 1,056,049 “preventive
patrolling” hours, compared to only 960,297 hours in 1999. In 2000, there were 1,064,283
preventive patrolling hours. The number of hours investigating collisions in 1999 was also up
from 1998 (276,190 versus 264,972 in 1998). Thus only a relatively small percentage of the
reduction in the number of hours on patrol could be attributed to increased hours investigating
collisions.
29
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

that in subsequent chapters. Another interesting pattern we see from the graph is the seasonal
effect. Here, the numbers of citation events decline during the fall, probably due to an increase in
the number of accidents during that time (see Figure 1.2). Accidents take up a considerable
proportion of the troopers’ time, leaving less shift time for general traffic enforcement that would
result in citations and warnings (written and verbal).
The number of accidents that the NCSHP responded to and noted in accident reports for
1997–2000 is presented in Figure 1.2. There are numerous accidents, and reporting them
constitutes a significant proportion of the workdays of the NCSHP. The four-year average is
approximately 9,900 accidents per month (horizontal reference line) or seven per officer per
month. Looking at this reference line, we see that the accident rates seem to have the same
seasonal cycle over the four years, peaking every fall. As with the graph on citations, the second
reference line (vertical reference line) highlights January, 2000, when Senate Bill 76 was
initiated. Unlike this reference line on Figure 1.1, this line does not mark any change in the
pattern of accident rates. The main effects we see from the graph are seasonal effects.
Many of the stops by the NCSHP do not result in citations, but in written warnings. In
Figure 1.3 we see that the mean four-year average is approximately 24,500 warning events per
month (horizontal reference line). Looking at this reference line, we see that the warning rates
were generally above the four-year average prior to the summer of 1999, the time period
following the implementation of the NCSHP’s TQM program, wherein quality, not quantity of

30
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 1.2 Trends in Accidents, 1997-2000

stops became a focus. Other than a much lower mean, this graph is almost identical to Figure 1.1
for citation events. At least some of the decline after 1997 in the number of warnings is due to
TQM. The second reference line (vertical reference line) highlights January, 2000, when Senate
Bill 76, requiring NCSHP troopers to record their traffic stops, went into effect. This line
immediately follows the lowest point in the number of warning events.

31
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Other than a much lower mean for written warnings, Figure 1.3 is almost identical to
Figure 1.1 (of citation events). Thus, there is no need to repeat the discussion here of the trend.
Note that there are fewer written warnings issued per month than citations. In Figure 1.3, we see
that the mean four-year average is approximately 24,500 warning events per month (horizontal
reference line).

Figure 1.3 Trends in Warnings, 1997–2000

In Figure 1.4, the number of “consent searches” (which here include probable cause searches as
well as consent searches) per month is presented for the years 1997–2000.
32
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

There are far fewer searches conducted per month (sixty-five), of course, than citations
and written warnings. Consequently, the pattern of the trend line is less stable. The volume of
searches seems to have remained low since 1999, unlike the pattern observed for citations and
written warnings. Thus, one could say that there has been a decline in the number of consent and
probable cause searches since the spring of 1999. Another interesting effect we see from the

Figure 1.4 Trends in Searches, 1997–2000

33
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

graph is the seasonal effect, but somewhat different from those observed earlier. Here, the
number of search events increases during the summer months, perhaps reflective of increases in
the volume of traffic on the interstate highways (where many of the searches take place).
In Chapter 5, we analyze survey data which allows us to compare in a preliminary way
the degree of racial disparity and potential racial bias in vehicle stops by the NCSHP and the
many town, city, and county police officers also operating in North Carolina. The work of local
police forces is typically different from the NCSHP. They respond to 911 calls for service more
often, investigate criminal activity more often, typically have more knowledge of residential
areas and individuals in the community, and operate their patrol vehicles at far slower speeds.
With the exception of 911 calls, all of these aspects of local police work increase the potential
for discretionary stops relative to the NCSHP. In addition, previous research has shown that
larger police forces (Gardiner 1969) and state police forces (Mastrofski et al. 1987) tend to have
more professional training. Some reports (see for example, Adams 2000; Parker 2001; Gaines
2002) have tended to show large stop disparities by city and suburban police. Allegations of
racial bias by state police and highway patrol troopers have often focused on vehicle searches,
where discretion is presumably higher and where stereotypes about minority behavior may
encourage searches for contraband. We report in Chapter 5, that even after accounting for racial
differences in self-reported driving behavior, racial disparity in police vehicle stops is reported to
be much higher among stops made by local officers than by members of the NCSHP.
The vast majority of this report is about the NCSHP. We include a contrast with local
police forces in part to emphasize that the analysis of racial bias in policing needs to take into
account the local context and content of actual police work. We develop in this chapter
theoretical tools for thinking about racial bias in police stops that may be more useful in
34
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

explaining bias processes in other settings. Similarly, we develop across this report a series of
methodologies for examining bias processes. We expect that the methodologies will be usefully
generalized to other contexts. We do not think it is reasonable to generalize any of our findings
(except perhaps the findings regarding trust in the police) to other police forces or geographic
contexts. Our main contributions to the study of the “driving while black” phenomenon are
theoretical and methodological. Our substantive conclusions are mostly restricted to the activity
of the NCSHP in the most recent years.

The Importance of Process and Context
During the course of this research we have been struck by the deterministic flavor of
discussions of racial bias in policing. On the one hand, stakeholders who wish to raise awareness
of racial bias in policing and the consequent distrust of police in the minority community, tend to
level blanket charges of police bias with little attention to the source of the bias or to alternative
sources of disparity in stops. All bias is the same and the assumption is if you look hard enough
you will find it everywhere. We call this the lay theory of a racist society. In our focus groups
with African American citizens, we discovered that this lay theory of pervasive societal racism
was generally endorsed. Conversely, police officers and other stakeholders tend to dismiss the
charge of pervasive racism, claiming they don’t know any (or few) racists and that there are
other good explanations for racial disparity in stops, such as not properly accounting for racial
differences in who is actually on the road and breaking the law. White citizens we talked to also
endorsed a more extreme lay theory in which racism was not the point: minority behavior
required more police attention and intervention. Proponents of both lay theories, when pressed,

35
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

often will point to “bad apples”—that is, racist individual officers—as the source of the problem,
although they will disagree as to the pervasiveness of the amount of rot.
From the viewpoint of social science, both lay theories seem quite misdirected. To argue
that racial bias and racism are a unitary, pervasive phenomenon flies in the face of what we
know about racial bias and its consequences. To argue that there is no racial bias anywhere in a
very large organization or across police forces is equally improbable. As we will outline below
there are a number of racially biased mechanisms that may produce racial disparity in traffic
stops. Self conscious, mean-spirited racism, while culturally the most recognizable, is by no
means the most likely. It is also well known by social scientists who study organizational bias
that the degree of decision making discretion and policies that encourage and discourage bias
vary across organizational contexts (Bielby 2000; Reskin 2000). It is simply not plausible to
expect that racial bias will be produced in the same way or with the same intensity everywhere.
Nor is it plausible to expect that it does not exist somewhere. Our discussion of the difference
between the NCSHP and more locally-oriented police forces raised some of these issues. We
will also investigate variation across the fifty-three troop districts within the NCSHP in the
degree of racial disparity and potential racial bias. We think that one of the most important
contributions this project may make is to clarify the variety of bias producing mechanisms we
might encounter and the importance of organizational context and policies for encouraging or
discouraging racial bias. Racial bias in policing is not an all-or-nothing phenomenon. Instead, it
is a series of questions about how and where bias is a problem.

36
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Non-Bias Mechanisms that Could Produce Racial Disparity in Stops
The simple observation of a racial disparity in police stops or searches is not sufficient
evidence to support accusations of racial bias in policing. Conversely, a finding that minority
drivers are stopped or searched by police in numbers roughly proportional to their incidence in
the population cannot be used to rule out the possibility of biased police stops. We define bias in
police stops as disproportionally stopping, citing, or searching minority drivers given their
incidence in the population of offending drivers encountered by police. This definition, assumes
that it is police discretion as to who is stopped (cited, searched) which potentially generates
racial or ethnic bias in the distribution of stops. Since the bias is an interaction between officers’
discretion and the drivers available to stop, a suspicion of racial bias requires us to first develop
estimates of who is at risk to be stopped and where the officers actually patrol before considering
explanations of racially biased policing.
In most places we expect to encounter relatively large average racial and ethnic
differences in driving behavior. The most important reason for these racial differences in driving
behavior is locational. That is, there are very large African American-white, and Hispanic-white
differences in residence patterns, reflecting historical patterns of residential segregation (Massey
and Denton 1993). Although, not much is known about its spatial distribution, there are also
well known patterns of racial/ethnic employment segregation (Tomaskovic-Devey 1993). Since
we know that in most regions of the country the average white and minority citizens live in
segregated neighborhoods and work in different organizations we can be fairly sure that in most
places the racial composition of drivers in different places (neighborhoods, highway segments,
main versus local streets) will be highly variable. Since we have no reason to believe that police
patrol streets randomly and proportional to population, a city or state level ethnic disparity in
37
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

police stops may simply reflect the racial composition of roads that police patrol. This cuts both
ways, observing that African Americans are stopped less than their proportion in the population
does not necessarily imply an absence of racial bias. Rather, African Americans simply may be
driving less in those places where police patrol more.
Where people drive is obviously related to both the locations of where they live and their
likely destinations (such as where they work or where they travel to for shopping). How much
people drive is obviously related to how far they must go to get places they need to be, but also
to other factors which may be related to race. Outside of the South, the United States African
American and Hispanic populations tend to live in more urban areas on average relative to
whites. This simple residential difference will have implications for the degree of reliance on
automobiles and mass transit. In urban areas especially, simple comparisons of those stopped to
minority and majority population proportions may be quite different from the actual racial/ethnic
distribution of drivers on the road.
There may also be racial/ethnic differences in the driving behaviors that are likely to
place one most at risk of a traffic stop. Some recent evidence in New Jersey suggests that on
some highways at some times of day, African Americans are more likely to break the speed limit
at higher speeds than white drivers (Lange, Blackman and Johnson 2001). Our own research at
fourteen sites indicates that African Americans speed disproportionately.2 We know of no other
research at this time that suggests that this is the case.
These preliminary observations must be cautiously examined. The minority population is
slightly younger on average than the white population in many places. Since it is well known
2

Note however that our observational study is of a non-random sample of observed
drivers on four-lane highways, and one should not generalize the findings to all types of
locations.
38
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

that young drivers engage in more risky driving behaviors, racial differences in the age structure
might lead to small average racial differences in risky driving behavior. On the other hand,
because of widespread fear of bias by the police in the minority community, we might expect
minorities to drive more carefully than whites. Since so much of social life is organized or
associated with race, it is not unreasonable to assume that, at least in some localities, there may
be average racial/ethnic differences not only in where and how much people drive, but also how
they drive. There is no reason, however, to assume, a priori, that it is minority drivers who tend
to be worse drivers. It could quite reasonably be white drivers, secure in their privilege to drive
as they wish, who are more frequently the risky drivers.
It is our expectation that racial differences in where people drive and how much they
drive (because of their strong links to residence and income) will be much greater than racial
differences in the use of seatbelts, turn signals, or excessive speed. One of the primary
contributions of this project is to develop a series of methodologies for examining the spatial
distribution, density, and driving behaviors of drivers at risk to be stopped. Another contribution
is that discussions of racial disparity in policing require a good faith effort to account for nondiscriminatory sources of racial disparity in stops associated with driving behavior before
reaching a conclusion that a particular police force is guilty of racial profiling in traffic stops.

39
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Deployment and Patrol Patterns
Police deployment may produce absolutely no racial disparity in police stop decisions,
but a large racial disparity in stops. Deployment refers to where police patrol and concentrate
their activities. Deployment for crime control, for example, tends to be greatest in higher crime
neighborhoods. If higher crime neighborhoods tend to have larger minority communities,
minorities might be stopped for traffic offenses at higher rates, simply because as a group they
have a higher probability of encountering an officer. A similar pattern might hold if state
troopers were deployed and concentrated their efforts in areas for reasons unrelated to race (for
example, areas with higher accident rates) but those areas tended to have more African
Americans routinely traveling these specific areas. This latter example is less plausible on its
face, but the general point is that who gets stopped reflects, in many ways, where the police are
deployed. It is possible for deployment to generate more minority (or majority) stops than their
incidence in the population. Because residential segregation and employment segregation tend
to be powerful forces in most places, we strongly believe that studies of racial bias in policing
must be able to adjust for deployment patterns.
If police are deployed specifically to harass African American drivers, deployment could
be influenced by a bias process at the organizational level. We suspect that it is more likely that
police deployment is intended to reflect public safety goals such as crime prevention or highway
safety. In this project we find that the racial composition of accidents is a useful tool with which
to identify the racial composition of drivers at risk for stops, but the figures must be adjusted for
police deployment by location and time of day to be most useful.

40
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

References for Chapter One
Adams, Jim. 2000. Study: Police Stopped Blacks Twice as Often as Whites. The
Courier-Journal. October 29, 2000.
Bielby, William T. 2000. Minimizing Workplace Gender and Racial Bias.
Contemporary Sociology. 29(1), 120–129.
Gaines, Larry K. 2002. An Analysis of Traffic Stop Data in the City of Riverside.
Report Submitted to the City of Riverside, March 5, 2002.
Gardiner, Mary J. 1969. The Inner City Marketplace: The Need For Law and Order. The
George Washington Law Review. 37(5), Jul 1015–1030.
Gallup Poll. 1999. “Racial Profiling is Seen as Widespread, Particularly Among Young
Black Men.” http:/www.gallup.com/poll/releases/pr991209.asp
Lange, James E., Kenneth O. Blackman, Mark B. Johnson. 2001. Speed Violation Survey
on the New Jersey Turnpike: Final Report. Report Submitted to the Office of Attorney General,
Trenton, New Jersey.
Massey, Douglas S., Nancy A. Denton. 1993. American Apartheid: Segregation and the
Making of the Underclass. Harvard University Press: Cambridge, MA.
Mastrofski, S., D. Ritti, D. Hoffmaster. 1987. Organizational Determinants of Police
Discretion—The Case of Drinking and Driving. Journal of Criminal Justice, 15(5), 387–402.
Parker, Robert N. 2001. Traffic Tickets, Ethnicity and Police Patrol in Riverside, 1998:
Evidence for Racial Profiling in Patterns of Traffic Enforcement. Report Submitted to Press
Enterprise, November 13, 2001.
Reskin, Barbara. 2000. The Proximate Causes of Discrimination. Contemporary Sociology,
29(2), 319–328.

41
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Chapter Two: The North Carolina State Highway Patrol Data Bases and
Evidence of Racial Disparity at the Aggregate Level

Racial Disparity in Geographic Areas
In this chapter we address the question of whether there is an excessive number of stops
and citations of African Americans across the North Carolina State Highway Patrol (NCSHP)
districts. This seemingly simple question is not easily answered. As researchers, our goal is to
uncover indications of possible racial bias, yet we have no data on the motives of individual
troopers who may or may not be biased in their behavior. Overt racism is seldom admitted to,
and thus, evidence of possible racial bias must be gleaned from a statistical analysis of the
behavior of troopers. As we will see, this evidence, while allowing us to rule out some claims
about the prevalence of bias, is equivocal about others.
We separate two forms of NCSHP behavior relevant to the questions of bias: the
deployment of troopers to a district (are some districts with relatively many African Americans
“over” patrolled?) and the behavior of individual troopers toward citizens (Are the behaviors of
individual troopers race neutral?). We take up the first question in this chapter and the second in
the following chapter.
The analysis of race-specific rates of stops/citations across areas that we address in this
chapter is important in two ways. First, these areal or aggregate rates can be used to study
patterns of stop/citation disparity.3 If an NCSHP district, for example, has an unusually high rate
of citations or stops of African Americans, it may indicate organizational practices in that district

3

By “aggregate” rates we simply mean that the number of stops/citations are summed for
an entire geographic area such as a district or a county.
42
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

which need to be corrected. These practices may range from a seemingly benign decision to
patrol one highway more than another because of tradition (“we always patrol there because
there are a lot of speeders there”— and more African Americans happen to drive there, driving
up the rate of African American stops in a district), to policy that possibly involves targeting
highways because of the racial composition (troopers are deployed because a highway is known
to have many African American drivers). In such a case, disparity may be reduced or eliminated
by simply changing deployment patterns, or making deployment patterns more rationally based
(patrol where preventable accidents occur) rather than based on tradition.
A second importance of areal analysis is that, as researchers, we cannot evaluate how
troopers behave individually without knowing about the contexts in which they work. For
example, a trooper working in the predominantly white counties of western North Carolina
would not be expected to have the same rate of stops of African Americans as a trooper in the
counties in eastern North Carolina, where relatively many African Americans live. If we can
ascertain that a district does not have an excess number of African Americans stopped/cited, then
we have a baseline against which to compare the behavior of individual troopers.
In the organization of this report, this chapter explores whether there are any patterns of
excessive stops/citations of African Americans at the area level. Where there are no differences
in stops/citations, or where the range of disparity is small (or within the range of measurement
error), we have, as a result of this comparison, established a baseline against which to judge the
behavior of individual troopers (this latter task we leave for the next chapter). For example, if
we know that 23 percent of the drivers speeding on the highways in an area are African
American, and that 23 percent of the drivers cited for speeding on the highways of an area are
African American, then we have one base rate (later we will discuss others) against which to
43
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

evaluate individual troopers. That is, we can reasonably take the next step to see if there are
specific troopers who have an excessively high percentage of their citations for speeding issued
to African Americans. If an entire geographic area has an excessively high rate of citations of
African Americans, however, then we will need to adjust our estimates to account for this fact
when evaluating individual trooper behavior. (We will address this issue in greater detail in the
next chapter).

Theories of Racial Disparity
Previous research has generally been focused on individual officer bias. As discussed in
the first chapter, individual officers may make decisions (to stop, to issue a citation) that are
racially motivated (active, racial animus) or make decisions that represent cognitive bias
(officers are unaware that they give the benefit of the doubt to whites more than African
Americans). What is called organizational bias or “institutional racism” (alternatively,
“institutionalized racial bias”) also can be at work. The term has several different meanings, and
can include any organizational policy that has an adverse effect on African Americans,
irrespective of the basis or rationale for the policy. We find it helpful to differentiate policies
(organizational decisions) that may have racial bias as its motivational origin from those that do
not have such a motivational source, but which nevertheless have adverse consequences for
African Americans. We argue that there can be three types of such “unconscious” bias at the
organizational level: 1) cognitive bias, 2) irrational incidental bias, and 3) rational incidental bias
(some may prefer the term “disparity” to “bias” here).
As an example of cognitive bias at the organizational level, suppose it were found that
troopers over-patrolled a given stretch of highway because the local sergeant believes that there
44
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

are “numerous bad drivers” on that stretch of highway. But he or she (and presumably others in
the Troop) may have formed that view based in part on the observation that there are
disproportionately high numbers of African Americans usually found driving that specific
highway who contribute to the high number of “bad drivers” (white and African American). In
other words, unconscious bias against African Americans could lead to a labeling process
(labeling an area or stretch of highway) resulting in over-patrolling of that stretch of highway.
“Organizational cognitive bias” is distinguished from individual cognitive bias as the latter refers
to possible bias of an individual trooper.
As an example of irrational incidental bias, suppose it were found that troopers tend to
patrol a given stretch of highway because “it is easier to catch speeders there,” but that same
stretch of highway is where, coincidently, there are a disproportionate number of African
American drivers (we say “coincidently” because we are not implying that troopers are aware of
there being a disproportionate number of African American drivers speeding). Again, stops and
citations of African Americans for the area may be excessively high due to the over-patrolling of
this stretch of highway. Reducing the patrolling of this highway would result in fewer African
Americans stopped/cited and could reduce/eliminate racial disparity in the area. The “over-”
stopping or citing of African Americans, in this case, is not related to any racial bias of a
cognitive sort (nor motivational source, for that matter), and it is easily corrected by limiting the
patrolling of that stretch of highway.
The third type of organizational decision that can generate racial disparity is one that is
not racially motivated, and which nevertheless affects African Americans in an adverse manner
(we assume that more citations and stops are “adverse”), but which is justified on the basis of
rational principles of organizational action. For example, if it were the case that Highway A,
45
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

with disproportionately many African Americans, had substantially more accidents than
Highway B, thereby justifying the allocation of a higher proportion of patrols to Highway A,
then the consequence for African Americans would appear as a form of racial disparity, but one
incidental to the rational allocation of troopers to that highway. We refer to this as “rational
incidental disparity.”
Since we have no data on the motives of organizational actors (organizational decision
makers), and differences in the patterns of NCSHP troopers’ behaviors are often small, we will
most times not be in a position to interpret the data as being more consistent with one
explanation than another. Is disparity due to cognitive bias, irrational incidental bias, or rational
incidental bias? Sometimes, however, we will be able to weigh the evidence that is available to
see whether for specific findings one explanation is more consistent with those findings than
another, or perhaps only more plausible.
As mentioned, we do not have direct measures of the motivations of individual troopers
or of Troop captains or sergeants. As such, as stated above, it will generally be the case that
evidence of racial disparity cannot be interpreted unequivocally as bias. Nevertheless, despite
possible ambiguity, organizations need to respond to the possibility that disparity and possible
discrimination may exist. For example, it is possible that everyday deployment decisions by the
NCSHP, even if made without any conscious implication of race, could have profound impacts
on the racial composition of drivers who are stopped and cited. For example, suppose there are
only two highways in a county and they have equal traffic volume. Because of patterns of
activity between residence, work, and recreation, 30 percent of Highway A’s drivers are African
American, while only 20 percent of Highway B’s drivers are African American. The average for
the county would be 25 percent (the average of 30 and 20). If 25 percent of those cited in the
46
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

county were African American, one could conclude that there is no evidence of racial disparity.
But what if, say, 28.5 percent of those cited were African American? That presumably would be
defined as a racial disparity worthy of further investigation (in other words, 3.5 percent higher
African American citation rate than the base rate).
The evidence of a 3.5 percent difference could be interpreted as follows: at the individual
level, NCSHP troopers (acting alone) make either or both biased decisions to cite African
Americans (racial animus) or racially unconscious decisions to let whites go with unreported
warnings (cognitive bias). But there is another plausible interpretation: deployment can be
unequal. That is, the difference in allocation could be accounted for statistically by deployment
of patrols to highways where African Americans are known to drive, or the deployment of
patrols at times of the day when African Americans are more likely to be driving. If this were
the case, the various forms of organizational bias discussed above might be the appropriate
interpretation. If the excessive deployment of troopers to Highway A is justified by a perception
of the highway as a “problem highway” then organizational cognitive bias may be at work. If it
were found that troopers patrol a highway because “the fishing is better there” (violators are
easier to find), then what we call “irrational incidental bias” may be operative. If there has been
a rash of accidents on Highway A, triggering a deployment decision that 70 percent of the patrol
time is devoted to Highway A, while only 30 percent is spent on Highway B, then perhaps
rational incidental disparity is operating. For the county in question, a citation rate of 28.5
percent for African Americans could be in part a consequence of the deployment decision (which
had been made for non-race-related reasons). Without knowledge of the fact that Highways A
and B differ both in the amount of patrolling and in the proportions of their drivers who are
African Americans, a researcher might mistakenly interpret the 28.5 percent as evidence of
47
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

individual troopers’ racial bias, instead of a consequence of a non-racially motivated deployment
decision.
Suppose the situation were different and the sergeant deployed 70 percent of the
patrolling to Highway A because his or her perception was that there were a lot of African
Americans speeding on Highway A and “those people” needed to be stopped and cited. It would
be unlikely that we, as researchers, would have access to information of such explicit bias of this
sort (here, overt racial animus). Except for some very specific instances that have gained
national attention because of the litigation process, there is seldom “smoking gun” evidence in
racial bias research. Yet, the deployment decision in this hypothetical example was in fact a
racially motivated and biased decision with arguably negative consequences for African
Americans (excessive number of stops and citations compared to what would have been the case
if a racially neutral decision had been made). For these reasons, we have to be cautious in
interpreting the patterns of data on stops and citations. Any evidence of disparity requires
further analysis and scrutiny. At the same time, it is difficult to determine when disparity should
be interpreted as bias.
The real life situations of patrolling and citizen driving are far more complicated than
described above. Troopers of the NCSHP typically have a great deal of discretion concerning
where they patrol in a given hour of a given day, so that even if there was a directive of the first
sergeant to focus patrols on Highway A, there is no guarantee that such a decision would be
followed. There are several reasons for the discretion. So called “micro-managing troopers” has
not been a tradition in highway patrolling, presumably because it would be time consuming to
administer. Someone would essentially have to dispatch troopers and verify that they have not
left a given highway. Patrol cars are generally mobile. From our conversations with NCSHP
48
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

troopers, they perceive themselves as more effective if their vehicles are mobile because they are
more likely to come upon unsuspecting motorists violating the law. Factors such as the “good
fishing” principle may be operating: troopers cite drivers in locations where they have
historically found speeders and other violators to be plentiful.
From the focus group data collected for this project, we learned that troopers go where
violations are plentiful not because they have quotas to fill, but because they have citations and
written warnings to write as evidence of their productivity. Presumably they also go to where
accidents are plentiful. But the two are not always highly correlated. For example, a very busy
interstate highway may have relatively few accidents per vehicular miles driven, yet many
speeders. Since speeding (frequent event) is presumably only moderately correlated with
accidents (rare events), there must be locations where speeding occurs but few accidents (for
example, a straight-away), and other areas where there are many accidents but little speeding
(sharp curve or steep incline). Furthermore, the stopping and citing of drivers may be due to the
fact that an area is within a short range of a major intersection such that drivers speed up as they
drive away from that intersection, violating the speed limit by an excessive amount as they settle
in on a comfortable speed for that highway. Data we have examined as to the location of stops
suggests that troopers have a tendency to stop drivers within a mile or two of major intersections
(of interstate highways). Whether it is because drivers are driving faster there or the troopers
choose these areas for other reasons (an intersection is a convenient turn-around place) is not
known, but we suspect that both citizen behavior and the “fishing principle” are operating.
As stated above, the general problem that we face as researchers is that data on possible
racial motivation is generally not available, only data on stops, citations and accidents. Some
inference need be drawn on the possibly biased nature of organizational responses from the
49
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

patterns of these data. One of the specific problems a patrol organization faces is that decisions
by lieutenants and sergeants that affect deployment of patrol cars are likely to not always be
based on accident data, but rather on partially subjective evaluations of accident patterns. Some
subjectivity is probably necessary, since the type of accidents, their presumed causes and so forth
ought to be taken into consideration. A second problem they face is the fact that the individual
trooper is generally free (and, to some extent, required given staff shortages) to roam a rather
large area while patrolling and the specific locations are chosen at his or her discretion. Thus,
patrolling is unlikely to ever be fully rational, as the trooper’s beliefs and ideas of what are
“good fishing” areas will almost inevitably come into play. Thirdly, when troopers in patrol cars
stop drivers in heavy traffic, they can be inadvertent causes of accidents. Thus, some of the most
heavily trafficked areas may be “under-patrolled” for fear of causing more of a problem than is
necessary. For all these reasons, the correlation of accidents and the volume of patrolling is
likely to be attenuated (less than perfect).

Methodological Issues
Keeping in mind our two goals of studying aggregate units of analysis to inform our
subsequent individual-level analysis and identifying areas where racial disparity is most
pronounced, we next turn our attention to several methodological issues. First, we as researchers
do not have direct measures of routine citizen-driver behavior. As a consequence, we must
examine patterns among variables that stand as “proxies” (approximate measures) for citizen
driver behavior. Some measurement error pertains to these proxy measures. These proxy
measures include the number of resident licensed drivers in a district, an estimate of the number

50
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

of what we call “drivers driving” in a district,4 and the number of individuals in accidents (we
will look at accidents reported by the NCSHP and by other agencies). We also have collected
information at fourteen sites on speeding behavior of drivers. The data from these sites can help
us validate the proxy variables, that is, help us determine how much measurement error there is
across the proxy measures. Appendix A discusses the speeding study of the fourteen sites.
Another methodological problem we have to address is how substantial a statistical
difference is necessary before we consider an area to have an excessive number of African
American stops/citations. When dealing with stops/citations, there are hundreds of thousands of
these events involving the NCSHP in North Carolina each year. It is not a given that a
statistically significant difference is a meaningful difference. Also, differences can be attributed
to measurement error. For example, if 23 percent of the drivers speeding in an area are African
American and 22 percent of the drivers cited for speeding are African American, we must ask if
that difference can be attributed to measurement error (in this case most likely in our estimate of
African Americans speeding) or not (is there actually “reverse discrimination”?). Secondly,
even if the difference is not attributed to measurement error, we must decide whether the
difference is “actionable.” That is, should a policy decision reasonably be based on the observed
differences that we find? These are complex issues, and we offer suggestions as to how to
proceed toward resolving them.

4

More specifically, “drivers driving” refers to a statistical estimate of the racial
composition of the drivers in an area, based on the aggregate racial composition of the area in
which a driver who has been cited resides. This is discussed in greater detail elsewhere in this
report.
51
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

The Importance of Context
As we have indicated above, to assess the presence and extent of racial disparity in an
organization such as a highway patrol, it is useful to distinguish methodologically between
individual trooper behavior and the context or environment in which the trooper works. We
claim that without understanding the context of the trooper’s work, one cannot evaluate
statistical evidence that measures the individual trooper’s behavior. In the example discussed
above, a trooper working in the predominantly white, western counties of North Carolina would
be expected, all else equal, to have a lower proportion of African Americans among his or her
citations than a trooper patrolling in the predominantly urban counties, which in North Carolina
have relatively large African American populations. But variation in context occurs within
relatively small areas, such as a county. For example, a patrol of a highway near the
predominantly white suburb of Cary in Wake County would be expected to yield a relatively low
percentage of African Americans issued citations, compared to a highway on the east side of
Wake County, outside of a predominantly African American residential section of the city of
Raleigh.
Very “local” factors may also be relevant to the racial make up of a troopers’s citations.
The presence of a textile plant with predominantly African American employees would
presumably have an influence on the percentage of citations issued to African Americans on a
highway near the plant. Similarly, the presence of a bar where African Americans recreate and
consume alcohol would affect the “driving under the influence” rate for African Americans on
the highway near the location of the bar. Organizational factors, such as a sergeant’s decision to
“crack down” on DUI driving on a highway where the “African American bar” is located, also
can affect the rate of DUI citations (and arrests), generally, and the African American rates,
52
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

specifically, for the district. Thus, the volume and often type of behavior that police observe or
come into contact with will vary, depending on where the patrolling occurs.
Another set of factors can influence the rates of stops/citations of African Americans.
Our analysis below reveals that the representativeness of African Americans on the highways is
different at different times of the day, and different from that of whites. Our initial evidence of
this was surprising to us, and so we looked for confirming evidence in the literature and found
that the 1995 National Transportation Survey confirms our findings that African Americans are
disproportionately to be found on the highway in the evening and nighttime hours (relative to
white drivers). Also, we discovered that the NCSHP does not work the highways evenly across
the hours of the day, and that the number of patrols vary across the state. Essentially, there is
more patrolling at night in the more urban or heavily driven areas (especially where the
interstates pass through) and less patrolling in the rural areas. The possibility of a mismatch
between when drivers are on the highway and when troopers are on the highway needs be
addressed. The issues are complicated because it is conceivable that some decisions about
deployment are themselves not racially neutral. For example, if there is an excessive number of
African Americans issued citations because there is an “excessive amount” of patrolling at night,
a reasonable question is “why?” The answer may or may not be racially neutral. We will
discuss this possibility again further below.

Citation Zones
One further issue will be developed which we did not anticipate when we began this
research and which seems to have a profound impact on the rate at which drivers are stopped. It
is essentially similar to the “the speed trap.” We prefer to use the term citation zone as we
53
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

imply by that term something more than a trooper hiding his or her patrol vehicle behind a bush
or in a ditch to catch speeders, but rather use the term generically to refer to an area of a highway
where many citations are issued. Also, we imply by the term “citation zone” that other behaviors
are looked for besides speeding, such as failures to stop, yield, weaving or other unsafe driving,
or perhaps simply expired automobile tags. By using the term citation zone, we want to capture
the idea that patrols tend to work certain stretches of a highway (a specific mile or several miles
of highway), presumably because “the fishing is good there” — in other words, there are many
speeders or other violators, and the highway design may be conducive to making stops (for
example, the highway has broad shoulders to safely pull over vehicles). If these citation zones
account statistically for a large proportion of the stops/citations, then very local conditions and
circumstances are crucial to understanding the stop/citation rate of an area. For example, we
notice that there tends to be a high rate of citation activity near major intersections, indicative of
heavy patrolling of those areas. Why the high rate? The most probable explanation is that as
cars accelerate coming from an intersection they often exceed the speed limit. One might even
say it is a natural tendency to speed up as one enters onto a highway from another highway.
Also, as one leaves a city or town where one has been driving more slowly — in accordance
with slower speed limits — one may tend to accelerate to a speed well above the limit (the vast
majority of drivers on major highways speed at some level). Perhaps the “liberating” effect of
the “open road” represents a form of psychological “release.” Troopers may have learned that
these are good locations to pick up speeders (and collaterally, other violators) because “the
fishing is good” and the landscape is suitable for a stop. The racial distribution of stops in a
district can be influenced by the choice of where to look for drivers. If the citation zone is near a
residential area that is mostly African American, or it is near a factory where African Americans
54
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

make up a high percentage of the labor force, the stop rate of African Americans will be high for
that district or county.
Our point about what we call the citation zone is that they can unduly “tip” the statistics
in the larger geographic area in which they are embedded. That is, an area which statistically
may not show evidence of racial disparity, may do so based on one “citation zone” that happens
to be located where African Americans frequently drive (alternatively, the speed trap is located
there for racially motivated reasons). In either case, the location of the speed trap is an important
part of understanding the explanatory mechanisms by which racial disparity evidence is
generated.

Spatial Heterogeneity
At a more abstract level, one of the predominant issues that we have to address in this
chapter is that of “spatial heterogeneity” (also known as geographic heterogeneity) – a fancy
term for the simple mismatch between when and where patrolling occurs and when/where
driving occurs as these phenomena are measured across geographic areas. If troopers patrol
more frequently at night in one district than troopers in another district, and African Americans
drive more at night in both districts, the former will have a higher percentage of African
American stops/citations than the latter. If troopers choose citation zones near areas with
relatively high numbers of African American drivers, the percent of stops involving African
Americans for the whole district may be affected. In both instances, race may or may not have
something to do with the deployment decision, and we must consider that possibility in the
analysis and interpretation of our data.

55
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Proxy Measures of Citizen Behavior
One of the key issues to the study of racial disparity is that there is seldom any direct —
much less error-free — measure of the behavior of the citizens driving on the highways where
the troopers patrol. If one wants to know if the troopers issue an excessive number of citations
to African Americans in an area, enforcement must be gaged relative to the prevalence of
violating behavior of African Americans where the troopers are patrolling. Depending on the
county, area of the county, and local features on a highway, the presence of African Americans,
as well as their violating behaviors, may vary. The most difficult issue faced by researchers is to
assess the adequacy of measures to approximate the behavior of citizens so as to have a basis for
comparison. For example, if one knows that 25 percent of the “threshold speeders” on a
highway are African American, then we would expect that 25 percent of the citations for
speeding would be written to African Americans. (A “threshold speeder” could be defined as a
speeder who is traveling at a speed that usually triggers a stop and citation by the NCSHP on that
specific highway.) Lacking a measure of the threshold speeding behavior of drivers, a researcher
must look for alternative measures. Such measures could seem rather crude, such as estimates of
the proportion of drivers in an area who are African American or white. For example, census
data, or data from the state Division of Motor Vehicles on the addresses of licensed drivers could
be used to estimate what percent of the eligible drivers in an area are African American. Under
the assumption that whites and African Americans violate traffic laws to the same extent and
degree, and do so where the NCSHP troopers can see them, such measures would constitute
adequate proxy measures of behavior (further assumptions could be made about out of state
drivers or drivers from other areas within the state). If, however, there is variation in the
56
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

distribution of traffic violating behavior across races and across the locations where the patrols
occur, or variation in the levels of travel from other areas of the state or out of state, then these
resident-based proxy measures might prove to be inadequate.

Command Policy and Procedures
In addition to examining our data at the aggregate level to measure characteristics of the
contexts in which troopers work, we also study at the aggregate level to gain information on the
nature of the local command policies and procedures as they impact the racial composition of
drivers stopped, cited or warned. Some sergeants may promote the issuing of citations, while
others promote the idea of deterring through presence on the highway (particularly highways
where accidents in general or specific types of accidents occur). The implications for race are
not obvious, but if a push for a high volume of citations happens to occur on highways used
disproportionately by African Americans, then that district may appear to have an excessively
high rate of such citations. Yet, the statistical disparity is but an artifact of a possibly arbitrary
decision at to where to patrol intensively. Of course it is also possible that the decision to patrol
more intensively is itself racially motivated. We will also discuss that possibility below.
Other command policies or practices in place may have the effect of lowering the rate of
citations of African Americans.5 For example, troopers have reported to us that citations are
reviewed each week to monitor the racial distribution of that activity. Presumably the intent is to
identify some number that represents “too many” and a trooper who is found to cite consistently
a high percentage of African Americans would be called upon to account for this. If this
5

It would be naive to assume that it has been “business as usual” for the NCSHP since
implementation of the special stop data collection. From our conversations with troopers, we
know that they are well aware of the public and legal scrutiny to which their behavior is subject.
57
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

managerial decision has a real impact, it might lead troopers to be cautious in their personal
citation rates of African Americans. We can only imagine what strategies a trooper would
devise to avoid any possible reprimand. For example, a trooper may “look the other way” for
some African American drivers so as not to come close to the acceptable proportion of African
American citations. Or he or she may avoid citations of African Americans early in the week so
as to have the option open for him or her to cite African Americans in the latter part of the week,
should a particularly serious offense be observed. We do not know about these possible
individual adaptations. However, it should be mentioned that we are unlikely to find any trooper
with extremely high rates of citations of African Americans due to these weekly accounting
procedures (however, we suspect that these are unevenly applied and are not systematically
administered).

Aggregate Units of Analysis
To evaluate the behavior of the NCSHP at the aggregate level, it is necessary to discuss
levels of aggregation (county, district, or alternative areal measures) that warrant study. There
are only five aggregate units of analysis that seem relevant to our study. Starting with the
largest, there are the eight regional Patrol Troop Areas (each covering about one eighth of the
state). The second largest is the Patrol District Area (there are 53 NCSHP districts). A patrol
district is usually made up of one to two counties. The third largest unit is the county (100
counties state wide). Fourth is the county area level (roughly one third of a county), an area that
the NCSHP uses to locate citations in a subarea of a county. Last is what we call a highway
area, defined as a stretch of highway between the borders of a county area. Although there is
virtually no limit to how areas could be defined, (for example, a researcher might define an area
58
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

as one side of a highway between intersections), there is little practical reason for looking at
citations or citizen behavioral proxy measures for areas that are extremely small because there
are too few observations of trooper intervention to study statistically over a short time span of a
year or even three or four years. Figure 2.1 shows how the units of aggregation differ. For
analysis purposes we will largely stick to two or three levels of analysis, using the highway area
for some of the analysis, the 53 patrol district areas (PDA) for other analysis, and – for
presentation purposes – the eight regional patrol troop areas (PTA).
There are tradeoffs with the choice of a unit of aggregation. At the most micro level
studied here (highway area) we argue that there is less measurement error introduced due to what
we call “spatial heterogeneity” (essentially the mismatch between where troopers patrol and
where drivers drive within an area). At the same time there is more measurement error of other
types at the highway area level. Because highway areas are quite small, there will be
measurement error attributable to relatively small numbers of observations (for example, few
accidents may occur and few citations may be awarded in that particular stretch of highway).
There are many reasons for measurement error that we can only imagine, and they probably vary
with the type of data collected. For accident data, which we propose as a measure of a baseline
against which to compare the citation rates of African Americans,6 troopers may have the option
of letting another law enforcement agency (county sheriff or local municipal police) write up and
record the accident report. The data we have on the number of accidents reported per county
indicate that it is primarily the NCSHP who is responsible for handling the write up of accidents
in counties (and not, for example, the county sheriff’s office), except in those counties with large

6

That is, we propose comparing the proportion of accident involved drivers who are
African American with the proportion of those cited who are African American.
59
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

cities. In the more urban counties, a large proportion of the accidents will be written up by the
local (city) police. It is not obvious what this means in these urban counties for the rates of
African Americans cited when compared to the rates of accidents involving African Americans.

Figure 2.1

Four Possible Aggregate Units of Analysis
NCSHP District (2 counties)

County Areas

Segment of Highway
Within County Area
‘Highway Areas’

60
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

County

It might mean, for example, that if a high proportion of African Americans in urban counties
have their accidents written up by the local police (perhaps because African Americans drive
proportionately more within the city than around the city),7 then the proportion of accidents
recorded by NCSHP involving African American drivers outside of the city limits may be
relatively low. Accident rates are generally lower on large, four-lane highways, relative to the
number of miles driven on those roads. Yet, speeding may be more prevalent on the large, fourlane highways – leading the NCSHP to cite the speeders there, even if the number of accidents
per million miles driven is relatively low.

Stops, Citations and Written Warnings By Race
Bearing in mind these methodological issues, we begin to address the question of the
extent of racial disparity by presenting the breakdown of stops, citations, and written warnings of
drivers by reason for the citation. We show a breakdown by type of behavior in part to make
clear what are the primary responsibilities of the NCSHP. The NCSHP works the highways
primarily to stop, cite and warn drivers for driving misbehavior so as to make the highways
safer. Unlike the local or city police, the primary mission of the NCSHP is to ensure safe
highways— most directly through the enforcement of North Carolina traffic law. It is not to
control, limit, and “fight crime” as crime is traditionally defined (such as robbery, burglary,
larceny, or assault). Only a small fraction of the NCSHP—in recent years approximately twelve
troopers— has as its primary purpose the task of screening drivers for suspicion of transporting
illegal contraband. As can be seen in Table 2.1, the most common form of behavior that triggers

7

This is plausible, assuming greater African American participation in the service
industry, for example.
61
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

a NCSHP trooper response is vehicular speeding. Data are presented for the year 2000, the year
the NCSHP began to collect data on all of their vehicular stops (relatedly, see Appendix C on the
charge likely to be the reason for the stop).
There are several interesting aspects of Table 2.1. First, the NCSHP stops and cites many
drivers over the course of time, as was pointed out in Chapter 1.8 More than .5 million citations
were issued in 2000 (the licensed driver population of North Carolina is approximately 6.5
million), resulting in roughly one citation issued for every fourteen North Carolina drivers (some
more than once, so the prevalence of citations is somewhat less than one in fourteen). As for the
racial breakdown of the citations, stops, and written warnings, we can compare initially the
results to the overall percentage of drivers with licenses who are African American, which is
21.2 percent of all of the white and African American drivers.9 As can be seen in the table, 24.7
percent of all those stopped are African American, so there is some evidence of racial disparity
in this simple comparison (3.5 percent absolute difference between 24.7 and 21.2 percent). The
difference is slightly greater for citations (3.7 percent), and lower for written warnings (2.5
percent).
As such, the results in Table 2.1 tell us a lot about the extent of possible NCSHP racial
disparity state-wide. We can rule out the possibility of large scale racial bias against African

8

We present data on stops here even though we know that a high percentage of stops
seem to be “missing” or at least not in the data base that we have available to us (roughly a third
of the records seem fall into that category -- partly due to lack of stop data on checkpoint stops).
Later in the analysis we will focus on citations and written warnings, as they seem to constitute a
more complete data base.
9

Note here that the base for the percentage calculation is all white and African American
drivers, excluding all others. Also note that 21.1 is within one percentage point of other baseline
denominators such as percentage of drivers in accidents who are African American or percent of
our estimate of “drivers driving” who are African American – described elsewhere in the report.
62
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Americans of the magnitude claimed against police by the American Civil Liberties Union in
other jurisdictions, in which the chances of an African American being stopped were seventyfive times that of a white (various ACLU documents on racial profiling can be found at the Web
site, see Reference under ACLU). Nevertheless, the 3.5 percent absolute difference in stops
represents an approximately 16.5 percent relatively higher chance of an African American being
stopped (3.5/21.2) than there are African Americans with licenses (across all charge types). Of
course, what some may define as a small difference may still be a matter of great moral and legal
concern.10
There are several reasons why the racial disparity shown in this table, however, should
not be interpreted necessarily as evidence of racial bias— although bias of the various forms we
have discussed earlier in the report could account for the racial disparity. Something as basic as
routine variations in patrolling by district could explain the disparity (for example, more
patrolling in urban areas, where African Americans are disproportionately found on the
highway). Still, and although preliminary, the data analysis does indicate a first glimpse into the
extent of racial disparity, and it seems to be in the direction of more action against African

10

The relatively higher odds of an African American being stopped (16.5 percent)
should also be interpreted with caution because the proportion of African American among
licensed drivers is a small percentage – only 21.2 percent. As the denominator in any ratio
becomes smaller, very large relative odds can be generated. For example, take the findings for
whites in the same table: 75.3 percent of the whites and African Americans stopped are whites,
compared to 78.8 percent of the licensed drivers. The -3.5 percent difference (75.3-78.8)
represents a relatively lower rate of being stopped of -4.4 percent. To say that whites have a
relatively lower rate of being stopped (compared to their representation among licensed drivers)
of 4.4 percent appears to be a small number, but it is the “same finding” as what we reported for
African Americans – a 3.5 percent absolute difference (recall that only African Americans and
whites are in this analysis, so the percentage of all licensed drivers and the percent of drivers
stopped must each sum to 100 percent)
63
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Americans, at least relative to their number in the licensed resident driver population for the state
as a whole.
Initially reviewing this indication of racial disparity statewide, we observe that the
percentage of those cited or warned who are African American varies across types of behavior
(for example, speeding, lack of driver’s license, faulty equipment, et cetera). These variations
are interesting, in part because we did not expect there to be large differences in the percentages
of cases involving African Americans across types of charge (why would bias be exercised for
some types of charges and not others?). However, when presented with differences, we find it
useful to try to understand whether the differences might be due to variations in behaviors across
offense types. For example, should we interpret the high prevalence of African Americans
among those cited for speeding as evidence of bias against African American speeders or as
evidence that African Americans speed more than whites? Could the difference be due to, in
part, relatively more African Americans on the highways at times of the day when more
patrolling occurs (we address this later in the report)? Should we interpret differences in
regulatory violations (such as failure to produce a valid drivers license or vehicle registration) as
evidence of racial bias for that type of offense (note that troopers rarely know before stopping a
vehicle that the driver does not have a license)? Or might it be due, perhaps, to differences in the
promptness with which African American and white drivers renew their licenses and
registrations?
Thinking about these issues in the abstract, it seems plausible to us that the educational
and social class background of the driver may enter into the explanation of disparities in some
citation rates. It is plausible that those with less education or who are of a lower social class may
not have the same financial resources as higher educated and higher social class drivers, and
64
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

therefore may be less likely to renew registration or licenses in a timely way or fix faulty
mufflers at the first sound of trouble. It is generally the case that African Americans (on
average) have lower income levels and are more likely to have less formal education and to be
disproportionately from lower social classes. Unfortunately, we have no measures of social class
in our data so as to verify some of these observations. Nevertheless, some readers may prefer to
interpret the patterns of data as evidence of class-related behavioral manifestations whereas
others may prefer to see the patterns as indicative of bias.
The results shown in Table 2.1 indicate that by far the largest racial disparity in citations
is due to license-registration-insurance violations. Although we do not know the social class
background of the specific individuals involved in citations, it seems less plausible to assume
that troopers’ racial bias manifests itself more for license-registration-insurance violations than
for all other types of violations than to assume that there is a behavioral basis for the disparity in
these types of violations. Our reasoning is that citations and written warnings associated with
license-registration-insurance infractions are clearly the responsibility of the driver (whereas
there may be some doubt for a somewhat more subjective determination such as a charge for
“vehicular weaving” if a driver is seen crossing the center-line).
If there is trooper racial bias operating in this class of violation, it could be that the
troopers “look the other way” when they find that a white driver has a revoked license.
However, given the seriousness of this offense (license revocation and/or no insurance, for
example, require a court appearance), it seems unlikely that giving whites a “pass” would occur
frequently, and even less likely that ignoring such an offense would be more common than
“looking the other way” for less serious or more common violations such as speeding below a
triggering threshold or “driving too close.” That is, these data provide some prima facie
65
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

evidence that some cause other than racial bias (perhaps social class), may be responsible for
racial variations in the behaviors that drivers are exhibiting on the highways, resulting in their
receipt of citations and warnings. A racial bias explanation of all the patterns in the table would
involve a specification that there are some behaviors where bias is exhibited and others where it
is not.
We assumed, at the beginning of this research project, that the more subjectively assessed
behavior data, such as “driving too close” or “vehicular weaving,” would indicate more disparity
against African Americans because the subjectivity lent itself to cognitive bias manifestations.
That is, all else being equal, we assume that it is easier to be biased in reacting to the driving of
an African American if an objective indicator, such as a radar gun, is absent. The results in
Table 2.1, however, suggest that this is not the case. Behaviors that we thought would involve
greater subjectivity, such as “unsafe movement” or “failure to stop/yield,” show quite low racial
differences in percentages (for citations— however, the percentages are somewhat higher for
stops).11
One can see in Table 2.1 that there are variations in the percentages of African
Americans who are stopped, cited, or given a written warning. In general, we argue that the
three data sources are not equal, however. Recall that we estimate that the stop data base is
missing about a third of its records (up to, but probably less than, a third of its record since we do
not have data on checkpoint stops – see Appendix G). Thus, these data may be a less reliable

11

However, since we are missing about a third of the stop records, in part due to lack of
data on checkpoints, we do not think the stop records are as reliable a basis for evaluating
disparity as the citation data are. Written warning records are more complete than stop data, but
receiving a written warning is a less severe sanction than receiving a citation, so arguably a
written warning sanction rate that is relatively high is inherently ambiguous as to possible racial
bias.
66
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

source on which to assess the extent of racial disparity. The citation and written warning data
bases are reasonably complete. However, written warning data, we argue, are inherently
ambiguous. Once stopped by the NCSHP, we suspect that all drivers would likely prefer a
written warning to a citation. Consequently, is a high rate of written warnings issued to African
Americans possible evidence of racial bias, or rather, is it evidence that African Americans are
less likely to receive a citation? If one interprets stops that result in written warning as events
that are really a “pre-text” for further investigation and a search, then there must be a substantial
number of searches of African Americans resulting from these written warning events. We do
not find this to be the case (see Chapter 4 for a discussion of the number of searches conducted
by the rank and file NCSHP troopers).
In evaluating the likelihood of racial disparity, we argue that the most complete and least
ambiguous official record source of information that may reflect bias is the citation data base
(See Appendix G for the details of this argument). In Table 2.1 we can see that African
Americans are more likely, relative to their representation among licensed
drivers—approximately 21.2 percent-- to be cited for license-registration-insurance violations,
speeding, equipment violations, other motor vehicle violations, and Driving Under the Influence
(DUI) than are whites. At the same time, African Americans are less likely to be cited for unsafe
movement and failure to stop/yield. Their chances of citation here are approximately equal to
the 21.2 percent baseline comparison on seatbelt-type violations.
As such, the results for citations have a possible “income” or social class interpretation.
Given that African Americans in general have lower incomes and are more likely to hold poorer
paying jobs, it would not be surprising that offenses related to income would be more prevalent
among African Americans. License or registration expirations, as well as equipment violations,
67
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

may occur more often among those having less money to make such repairs as replacing
mufflers or fixing tail lights, and thus may disproportionately involve higher numbers of African
Americans. Put another way, these results represent prima facie evidence that NCSHP stops and
citations may reflect what they see to a greater extent than who they see.
The findings on citations of speeders are also interesting in light of recent data from New
Jersey (Lange et al., 2001) suggesting that, for at least some highway areas (but not others),
African Americans speed more than whites. Similarly, we have found that in a nonrepresentative sample of fourteen highway segments in North Carolina, African Americans
speed more than whites in twelve out of fourteen highway segments (our observational baseline
study is described in Appendix A). However, neither of these studies provide a solid scientific
basis for generalizing the results to other areas. Nor does either study control for obvious sources
of speeding behavior that may be associated with race, such as age, gender, or inter-state travel.
Put simply, we do not know for sure whether African Americans generally speed more than
whites, or if they only do so on some highways under some specific conditions (for example,
where there is a high speed limit, during certain times of the day, or differences in origination
and destination points). Indeed, it may also simply be the case that, among African Americans,
those driving on the highway segments observed in New Jersey and North Carolina are—on
average—younger, more likely to be male, or more likely to be driving long distances. All are
common correlates of speeding behavior.

68
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 2.1 Frequency of Type of Citation, Stop and Written Warning by Race, African
Americans and Whites Only (Row Percentages)
Type of Citation

White
Citations
(most
serious
charge)

African
American
Citations
(most
serious
charge)

Speeding

216,142
(75.2
percent)

Unsafe Movement

White

Stops*

African
American
Stops*

White
Written
Warnings

African
American
Written
Warnings

71,432
(24.8
percent)

133,155
(75.3
percent)

43,761
(24.7
percent)

66,807
(77.2
percent)

19,698
(22.8
percent)

17,495
(79.3
percent)

4,576
(20.7
percent)

5,275
(73.6
percent)

1,893
(26.4
percent)

20,438
(74.3
percent)

7,079
(25.7
percent)

Failure to Stop or
Yield

11,870
(80.2
percent)

2,937
(19.8
percent)

4,618
(76.1
percent)

1,454
(23.9
percent)

–

--

DUI

15,390
(73.5
percent)

5,540
(26.5
percent)

3,431
(70.7
percent)

1,421
(29.3
percent)

–

--

Vehicle Equipment

1,557
(76.0
percent)

492
(24.0
percent)

13,365
(74.9
percent)

4,482
(25.1
percent)

Seatbelt, Helmet,
Child Restraint

84,916
(78.5
percent)

23,311
(21.5
percent)

37,000
(79.3
percent)

9,660
(20.7
percent)

Registration,
License, Insurance

50,640
(68.5
percent)

23,319
(31.5
percent)

13,337
(79.4
percent)

3,458
(20.6
percent)

Other

3,913
(71.9
percent)

1,527
(28.1
percent)

21,376
(68.8
percent)

9,692
(31.2
percent)

Total

401,923
(75.1
percent)

133,134
(24.9
percent)

231,557
(75.3
percent)

75,821
(24.7
percent)

24,359
(73.4
percent)
–

63,733
(77.0
percent)
--

175,337
(76.3
percent)

8,812
(26.6
percent)
--

18,994
(23.0
percent)
--

54,583
(23.7
percent)

*We estimate that approximately one third of the stop records are missing (partly due to the lack of stop
data on checkpoint stops) and thus caution should be exercised in interpreting these two columns of data.
Percentages for African Americans can be compared to the 21.2 percent of African Americans with
licenses in North Carolina.

69
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Although written warnings are somewhat ambiguous to interpret, we will discuss the
pattern of findings in the table.12 Written warnings results indicate that African Americans are
over-represented in the unsafe movement violations resulting in written warnings. Thus, it
seems that African Americans are stopped more often for unsafe movement, but that they receive
a written warning instead of a citation for the offense. One interpretation could be that a written
warning is consistent with the level of seriousness of the infraction. An alternative interpretation
is that the troopers use “unsafe movement” written warnings as a pretext for evaluating the
driver and the vehicle prior to conducting a search. The pattern is similar for vehicle equipment
violations, but less pronounced. Given the infrequency of searches by the regular NCSHP
troopers, however, we do not think that there are very many “pre-textual” written warning stops.
It might be that these higher warnings for unsafe movement and equipment violations are a
pretext to check licences, registration, or for the presence (or smell) of alcohol. Still, we do not
suspect this to be the case. In our conversations with troopers, they sometimes described
discretionary written warnings more as a “gift” to drivers (or as fulfilling a directive such as in
the case of equipment violations) rather than an investigative tool.13 To the extent that the
differences reflect or are interpreted as racial bias, it seems more likely that they would arise out

12

Written warnings, which the troopers are expected to write for most equipment
violations, speeding that is not too excessive, and some other violations, are also issued in
accordance with the NCSHP’s policy of promoting good community relations. In general, the
average citizen would rather get a written warning than a citation. However, written warnings
also have been claimed to be “pre-textual” in nature: troopers stop African Americans ostensibly
to give a warning, but in reality the trooper is “checking them out” for more serious offenses.
13

By perceived “gift” we do not mean to imply that warnings are issued to individuals
based on subjective factors, as opposed to objective ones, such as the seriousness or nature of the
offense that the officer is confronting. Presumably a written warning is more likely if the drivers
is only driving a few miles over the limit, as opposed to 10 or 15. Thus, in general, warnings are
for less serious charges.
70
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

of a more subtle cognitive bias process in which low level driving violations appeared more
serious to the troopers when the driver was African American.
In summary, Table 2.1 indicates that when compared to the number of licensed drivers,
African Americans are over-represented among the ranks of those stopped, cited, or issued
written warnings. No clear pattern emerges to differentiate stops from citations and from written
warnings. Generally speaking, the percentages of individuals stopped, cited or issued a written
warning indicate that there is no large-scale racial disparity (of the magnitude often presented
and discussed in the media). Some racial disparity is found in the citation data, but we have no
basis to rule out the possibility that there are differences in driving location and driving related
behaviors associated with race. Written warnings are clearly more commonly issued to African
Americans. However, given the fact that written warnings are preferable to citations and the fact
that there are very few searches conducted of African Americans as a result of these types of
stops, we see an ambiguous pattern of disparity.
In addition to our interest in written warnings, we are especially interested in verbal
warning stops because they are a new source of information on trooper-citizen contact. Verbal
warnings began to be recorded by the NCSHP in January, 2000. They represent a form of
trooper intervention that was not systematically recorded previously by the NC SHP. Note that
verbal warnings represent a very small proportion of reported stops (less than 3 percent). We do
not conduct as extensive an analysis of verbal warning stops as we do of citations primarily
because we have less information about such stops (due to the questionable quality of the stop
data – See Appendix G) and because they very rarely result in an automobile search.14 Table 2.2
14

See discussions in Appendix G of the limitations of the stop data, in which we suggest
that about a third of the stop forms seem to be “missing” from that data base – in part because of
lack of data on checkpoint stops, and the likelihood that the citation and written warning data
71
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

shows the percentage of written warnings and verbal warnings issued to African Americans (we
repeat the findings on written warnings from Table 2.5 to provide some comparative reference
for the verbal warning findings). The results indicate that 22.8 percent of the drivers issued a
written warning for speeding are African American, whereas 30.7 percent of those issued a
verbal warning are African American. In general, the results show that African Americans
represent a higher proportion of verbal warnings than they do of written warnings across all
violation categories.
The over-representation of African Americans among those given a verbal warning can
be interpreted in two different ways. The first interpretation would be consistent with the
hypothesis that the NCSHP shows greater leniency to African Americans, in part due, perhaps,
to increased scrutiny of the NCSHP by the media, state legislature, and others. Alternatively,
verbal warnings have been discussed as a result of a “pre-textual stop” (to reiterate, a pre-textual
stop is a stop for further police scrutiny, commonly expected to be for the purpose of conducting
a search of the vehicle.) While neither interpretation can be ruled out, it should be noted that
“only” a couple of thousand (2,023) verbal warnings were recorded in the stop data as issued to

base have records not accounted for with stop records. In addition the identity of the trooper is
recorded differently on the stop forms (a special assigned identity number is used rather than the
trooper’s registry number). This identity number exists for the purpose of protecting the trooper,
but it also limits accountability in data processing. For example, if stop records are not always
turned it by a trooper, there would be no systematic way to determine that fact. We believe,
based on comments by troopers in the focus groups that we conducted, that troopers occasionally
do not fill out, file or enter electronically (or have entered for them by a secretary) the stop form
information. Thus, for example, in the year 2000 there are more “citation events” (occasions
where at least one citation is issued) than stops -- even when we exclude citation events where
accidents are involved. We expected there to be more stops than citations since the vast majority
of citations result from stops (as opposed to resulting from an accident investigation). That is not
the case, however. Because stops, citations, and written warnings are different data bases with
different processes associated with them, one should not expect the number of written warnings,
verbal warnings, and citations to sum to the number of officially recorded stops.
72
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

African Americans in 2000, out of nearly one million stops or events resulting in citation or
written warning. Thus, the observed verbal warnings of African Americans represent a very
small proportion of all the interventions of the NCSHP with citizens.
The idea that troopers stop African Americans as a “pretext” is a plausible interpretation
if in fact African Americans are: a) stopped and given written warnings more often than whites;
and b) have their vehicles searched more often than whites. The first condition is not sufficient
as evidence of pre-textual stops, as written warnings are less serious forms of interventions than
citations, and if such interventions are occurring more often for African Americans than whites,
this could be interpreted as evidence of leniency toward African Americans. It could mean, also,
that troopers are making legitimate stops for violations of traffic law that warrant no greater
sanction than a written warning. The second condition does not hold as far as the rank and file
trooper is concerned. The vast majority of troopers do no proactive searches at all (see the
discussion in Chapter 4 regarding searches). Thus, we have no empirical basis upon which to
suspect that the average trooper in the NCSHP is stopping drivers to issue a written warning in
order to check out the driver or the vehicle for signs of more serious violations (drugs or guns),
else such searches would be more prevalent.
From the analysis above, it could be said that there is the possibility of some low level of
bias, for some driving violations, against African Americans, as African Americans are generally

73
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 2.2 Frequency of Type of Written Warning and Verbal Warning to African
American Drivers (Counts and Percentages*)
Type of Violation

Written Warning to
African Americans

Verbal Warnings to
African Americans

Speeding

19,698
(22.8 percent)

615
(30.7 percent)

Unsafe Vehicular Movement

7,079
(25.7 percent)

225
(35.4 percent)

Failure to Stop or Yield

--

36
(32.7 percent)

DUI

--

125
(33.3 percent)

8,812
(26.6 percent)

251
(37.4 percent)

--

93
(23.5 percent)

18,994
(23.0 percent)

116
(28.0 percent)

--

562
(29.7 percent)

Substandard Vehicle Equipment
Failure to use Safety Devices
Vehicle Registration, License, Insurance, or Inspection
Other

Total
54,583
2,023
* Percent = percent of all African Americans and whites issued a written warning or a verbal warning. Example:
“22.8 percent of all the whites and African Americans issued written warnings for speeding in 2000 are African
American.”

over-represented relative to the 21.2 percent of all African American and white drivers. Yet,
there are a number of possible alternative explanations, such as the possibility that African
Americans drive more frequently on highways where the NCSHP happens to work, or that they
drive more at times of the day when the NCSHP is working (for example, night versus day).
The plausibility of these explanations was enhanced when we conducted our baseline
study at fourteen sites (each consisting of 10-15 mile stretches of busy North Carolina
highways), finding that the proportion of drivers who were African American on different
74
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

highways within the same county could vary considerably. This suggests that a mismatch
between the extent of patrolling by the NCSHP and the prevalence of African American drivers
could easily exist. We also learned from the observational data that the NCSHP’s citation
behavior is segmented: some parts of some highways may account for a rather large proportion
of all citations (more on this below).
As a consequence of the analysis done so far, we began to understand that, in order to
properly account for the likelihood of mismatch, it would be reasonable to explore analysis of
data at relatively small units of analysis, such as what we call “county highway areas” (the
continuous stretch of a single highway across about one third of a county). We chose this unit
because it was the only unit of analysis available that was smaller than a “county area”—all of
the highways within about one third of a county. We also surmised that there was probably more
similarity in the drivers and driving behavior on a given, single highway area than on an
intersecting highway area in the same county. Thus, the county highway areas would be
reasonable units of analysis for addressing the mismatch issues discussed above (although far
from perfect units of analysis). In addition, we also realized that we only have one type of
baseline measure of driving behavior at the county-highway-area level—accident data (for
example, the proportion of drivers involved in accidents who are African American), so that
comparisons with other proxy measures would only be possible at higher (larger) units of
analysis, such as the NCSHP’s fifty-three districts covering the state (on average, a district
consists of two counties).

75
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Racial Disparity in Districts
The preliminary evidence on racial disparity in stops, citations, written warnings, and
verbal warnings leads us to consider the possibility that citizen behavior could be responsible for
the disparity. Unfortunately, we have no data base with which to test alternative hypotheses
concerning disparity due to other factors, such as the types of racial bias that could be
manifesting itself in the form of cognitive biases against African Americans. This unknown
plagues us throughout the report. It is difficult to account for bias processes when, for the most
part, we are limited to outcome data. Still, the extent of racial disparity in the findings that we
have reviewed so far is far smaller than many would have anticipated.
The data in the tables above were for the entire state of North Carolina, and because the
state is large, the possibility exists that we might find greater (or lesser) degrees of disparity if we
were to look at sub-sections of the state. Researchers refer to this as “dis-aggregating” the data
or, more precisely, aggregating the data to smaller units of analysis than the state as a whole. (We
say “aggregate” because the process involves combining information for individual troopers to
some areal unit of analysis such as a county or a district.)

Proxy Measures of Citizen Driving Behavior
One methodologically weak assumption, relevant to the tables above, is that we compare,
for example, the percentage of drivers cited for a traffic violation who are African American to
the percentage of African American drivers assumed to be on the highways in the state as a
whole. Although the vast majority of those with North Carolina driver’s licenses presumably
drive on North Carolina roads, they do not do so equally. In our attempts to interpret the data on

76
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

citations it would be preferable to have a baseline of data against which to compare citation rates
for smaller units of analysis than the state as a whole.
Except for baseline data collected at fourteen cites (see Appendix A), we do not have data
on citizen driving behaviors on the highway. Yet, several proxy measures of driving behavior are
available, including the U.S. Census data, the N.C. Division of Motor Vehicles licensed drivers
data base, and the NCSHP accident record data base. In some of our earlier analysis, we
examined the U.S. Census data and found that a measure such as “proportion of residents who are
African American living in a district” was somewhat less highly correlated with the proportion of
citations issued to African Americans than was the proportion of African American licensed
drivers, and so we dropped the U.S. Census data from further consideration. We reasoned that the
number of licensed drivers would be expected to be a more accurate reflection of drivers likely to
be on the highway speeding and committing other violations such that the presumed superiority of
the Division of Motor Vehicles’ data seemed reasonable. Still, the reader should be cautious of
baselines associated with census counts as a proxy for who is really on the highway.
Below, we compare the validity of three types of proxy measures of driver behavior:
number of licensed drivers, an estimate of what we call “drivers driving,” and drivers involved in
accidents. The Division of Motor Vehicles made available to us the data on licensed drivers with
information on the county in which the driver resides. The demographic characteristics of the
drivers on the highways in a county could be estimated by making the strong assumption that the
drivers driving on the highways and violating the law (speeding, driving erratically) mirrored the
drivers who were residents in a county. Of course, we know this not to be true with regard to
gender (males drive more than females) and age (the young drive more than the old), but we have

77
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

no data to suggest that the racial representativeness of the drivers is substantially different from
that of the residents.
Our second proxy measure of driving behavior takes into consideration the fact that
drivers drive outside of their county of residency, and for some counties this consideration may
affect the measure of the racial composition of the drivers in a county. For example, counties
adjacent geographically to a highly urban county with a large African American population may
have a higher percentage of African Americans on their highways due to the spillover of drivers
from the more populated county. Alternatively, counties disproportionately African American
may have large numbers of white drivers pass through them, especially on the interstates,
lowering the proportion of drivers who are African American, relative to the population
proportion.
To take into consideration the invasion of drivers from one county to another we used the
following estimation technique to estimate the proportion of “drivers driving” who are African
American: we made the assumption that drivers who drive outside of their county do so in a
manner proportionate to their representation in the population of their county of residency. That
is, if 20 percent of the residents in a county are African American, then we assume 20 percent of
the drivers who drive in another county from the county in question are African American. We
estimate the number of drivers from a county who are in another county by using data on citations
and matching the records with the DMV data base. Thus, for example, if 500 individuals cited in
County Y were residents of County X, then we assume that approximately fourteen times that
many drivers from County X actually were driving in County Y in the year in question (fourteen
is the approximate multiplier actually used, base on the general risk of receiving a citation in a
given year, roughly one in fourteen). The assumption we make is that the demographic makeup
78
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

of these “invader” drivers parallels that of the residents of County X. Through this method we
were able to estimate the number of drivers driving in each county, and compute the proportion of
“drivers driving” who were African American. We refer to the measure as our estimate of
“drivers driving.”
The third proxy measure is the most powerful measure in the sense that we can count the
number of drivers who are African American and who are involved in accidents for relatively
small areas, such as highway areas. A highway area is the area along a highway in what the
NCSHP call an “area” of a county (roughly one third of a county). The NCSHP record most of
the accidents that occur in the rural areas of the counties, with the exception being the large urban
counties where the local police record many of the accidents. Earlier in the report we showed that
there is a reasonably high correlation between the proportion of drivers in accidents who are
African American and the proportion of speeding drivers who are African American. We would
expect that the proportion of accidents involving an African American driver and the proportion
of drivers cited who are African American should approximate one another.
Although we argue that the accident data provide the best proxy measure at small units of
analysis (See Appendix G), our initial data inquiries discovered an interesting pattern. There are
many accidents on the relatively narrow roads in rural areas of the state, where patrolling is
generally light. These roads do not have as high a volume of traffic as do the relatively more
well-traveled interstate, U.S. and N.C. highways, and thus individually these “country roads” to
not warrant heavy patrolling. By contrast, the busier highways are safer (fewer accidents per
vehicular miles driven), but because of the high volume of traffic, they warrant heavier patrolling
than the rural highways (in part the patrolling is justified because more vehicles can be slowed by
the visual presence of troopers on the busy highways). One implication of this for race is that the
79
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

majority of rural paved highways in North Carolina have a higher percentage of white drivers
compared to the busier highways, which are the more heavily patrolled highways. As a
consequence, we find that the more rural counties tend to show excessively low rates of citations
of African Americans, and the more populous areas slightly higher rates of citations of African
Americans when we compare the proportion of cited drivers who are African American to the
proportion of drivers who are African American and in accidents. We will discuss this more
below.
Accidents are much more infrequent than citations, and so we found in necessary to
aggregate three years of accidents to get a sufficient number of accidents to justify statistical
comparisons. We examined the data to see if there was any evidence of a change in the
percentage of drivers involved in driving accidents who were African American. Generally the
correlations across years were very high, indicating stability in the demographic composition of
the drivers over a three-year period (results not shown here). The demographic profile of drivers
involved in accidents was explored further to determine whether the “contributing circumstances”
of the accident, which had been recorded for only about half of the drivers involved in accidents,
were correlated with the types of driving behaviors that elicited citations. (A “contributing
circumstance” is perhaps a benign way to express finding fault with the driver’s behavior.) For
example, a factor such as “excessive speed” may be considered a “contributing circumstance.”
We thought initially that if, for example, 20 percent of the drivers cited in an area for
speeding were African American, then we might expect that 20 percent of the drivers in accidents
in which speeding was a “contributing circumstance” would also be African American. Although
there was generally a strong correlation between speeding citations and speeding “contributing
circumstances,” our analysis indicated that the behavior-specific accident measure was a less
80
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

valid indicator of who was driving on the highway and a less valid measure of who was violating
a law (such as speeding) than was the measure of the proportion of drivers involved in accidents
(regardless of “contributing circumstances”). Part of the measurement error here may lie with
the fact that only about half of the accidents have data on “contributing circumstances.” The
“lost” observations (accidents with no contributing circumstances indicated on the official forms)
are substantial in number, and their omission makes the correlation with citation data worse. In
summary, among the accident measures (specific-behavior versus accidents in general), the best
correlate of the proportion of citations issued to African Americans (or for any specific type of
behavior resulting in a citation) is the proportion of drivers in accidents who are African
American.

Statewide, the differences across the various proxy measures in the percentage of

drivers who are African American is rather small: Census 21.8 percent, resident licensed drivers
21.2 percent, “drivers driving” 21.6 percent, and accidents 19.8 percent.15 Of the four, the
percentage of drivers in accidents who are African American represents the lowest percentage,
and thus the more conservative of the four baselines against which to compare stop, citation, and
written warning rates. We say “conservative” because accidents are the baseline with the lowest
estimate of the percent African American, and thus it is most likely to generate a positive
difference score (or a larger positive difference score) relative to the African American rate of
stops, citations, and written warnings.
Because racial differences in statewide data could easily be accounted for by the presence
of more patrolling in the more urban areas (and urban areas are generally more African
American), it is appropriate to examine the data at smaller units of analysis. Figures 2.2 through
15

The figure 19.8 represents the percent African American of all white and African
Americans involved as drivers in accidents in the NCSHP accident data file for the aggregated
years 1997 through 2000.
81
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

2.4 show the correlation and “scatter” involved in the relationship between three proxy measures
of behavior and the proportion of citations issued to African Americans: resident licensed drivers,
“drivers driving,” and drivers in accidents. Note that the correlations are calculated at the troop
district level (N=53). The correlations are essentially the same across these three proxy measures
(r’s range from .93 through .96), such that most researchers would say that the three measures are
indistinguishable in terms of predictiveness. That is, the proportion of citations issued to African
Americans at the district level can be equally predicted by using residency data, “drivers driving”
data, or accident data.
Figure 2.2 provides an example. Here the proportion of resident licensed drivers who are
African American is graphed (the horizontal axis) along with the proportion of cited drivers who
are African American (the vertical axis). The middle line in the figure is the linear regression
line, or our statistical model’s best estimate of what proportion of cited drivers should be African
American, given the general pattern. The small, hollow boxes that are scattered around the
regression line are actual observations of districts (fifty-three of them). As one can see, the
scatter of boxes is both above and below the regression line, with most of the scatter occurring
within a 95 percent confidence interval around the line. That is, most of the scatter is within a

82
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.2 Proportion of Citations Issued to African Americans by Proportion of Resident
Licensed Drivers Who Are African American, by Troop District (N=53 Districts)

83
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

range such that one could not claim that any one observation is different statistically from the
estimated point on the middle line.
At first glance, the evidence of Figure 2.2 seems to indicate a remarkable similarity
between the proportion of residents and citations involving African Americans. The correlation
here is .941. However, there are four points to consider in analysis of these data. First, seemingly
small general differences are not assumed to exist in a specific district. For example, there is a
district with only .18 African American residents but that has .27 of drivers cited as African
American. Such a district should not be ignored. Nine percent more of those cited are African
American than who are African American residents. Second, the intercept value (where the
regression line crosses the vertical axis) is approximately .04, indicating that more African
Americans are cited than are resident by about 4 percent. Thus, the overall disparity evident in
the figures is in the direction of excess citations of African Americans. Third, it is important to
bear in mind that there are only fifty-three observations in this figure, such that the slope of the
regression line could change with only a few observations “tipping the balance” one way or
another. More racial disparity, as defined as distance from the regression line shown here, could
be the consequence in some districts, less in others. Fourth, the data on the fifty-three districts are
aggregated from all the smaller areas in the districts. There could be areas within a district where
African Americans are treated disparately, but we cannot determine that using the aggregated
numbers presented in the figure because they represent an average for the whole district. We will
take up this theme in greater detail below. Thus, in summary, the information in the figure is
hardly sufficient grounds to terminate further evaluation, but rather it gives us reason to look
further.

84
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figures 2.3 and 2.4 show the results for our estimate of “drivers driving” and drivers in
accidents (proportion who are African American), respectively. Figure 2.3 seems at first glance
to look a lot like Figure 2.2. There are several differences, however. For one, the slope is
steeper, indicating that as the proportion of estimated “drivers driving” who are African American
increases, the proportion of cited drivers who are African American becomes even higher than is
the case for changes in the resident proportion African American. For example, compare the
intercept value, which is near zero, with the value of the cross of the two reference lines at the .35
proportions. We can see that the regression line estimate lies approximately .03 above the point
where the two reference lines cross. Again, as was the intercept value in Figure 2.2, the direction
of bias is one slightly detrimental to African Americans. The higher the proportion of drivers
driving who are African American, the greater the disparity in the citations to African Americans.
In Figure 2.4, we have the scatter of observations in a regression with proportion of
drivers who are African American as recorded in accident data compared to proportion of drivers
cited who are African American. Here we see yet a different pattern. This time the comparison
of the reference line cross point with the predicted value from the regression line shows an even
larger discrepancy, roughly .06, and the disparity grows with the higher proportion of African
Americans among the drivers in accidents. That is, where there are more African Americans on

85
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.3 Proportion of Citations Issued to African Americans by Proportion of “Drivers
Driving” Who Are African American, by Troop District (N=53 Districts)

the highway (as measured by this baseline), there is a larger disparity in the proportion of African
Americans cited, relative to the proportion in accidents. Thus, it would appear that, despite the
high correlations indicating that approximately 90 percent of the variance in the proportion of
drivers cited who are African American can be accounted for with one or the other of the three
proxy measures of behavior in the figures above, there are nevertheless reasons to examine the
data further because they are suggestive of racial disparity and possible racial bias. We need to

86
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

see if there is more conclusive evidence of bias than has been presented thus far, since we have
been just scratching on the surface of the data. It should also be noted that the evidence lies in the
direction that some might hypothesize using a “racial threat” hypothesis: the more African
Americans present in an area, the greater the tendency for them to be issued citations. However,
we must also point out that we have been looking at the data in a rather superficial way, and that
further analysis is necessary at smaller areal levels to justify such a claim.
When we use the term “superficial,” we mean that there are several basic reasons to
question the relationships between the variables in the figures above. For example, some of the
districts having the highest proportion African Americans are in urban areas where many of the
accidents are handled by the local police, thus possibly driving down the number of accidents
handled by the NCSHP. That alone would not affect the proportion African American involved
in accidents, except for the fact that it is possible that the NCSHP “work” accidents and patrols
somewhat differently. They may patrol some highways where the local police typically write up
the accidents, yet the NCSHP still issues citations in those areas. They may handle all of the
accidents on all highways in more suburban areas, where the more dangerous local roads are
responsible for a higher proportion of accidents relative to that of the busier highways (where
more tickets are written). That is, the bulk of the patrolling may occur on the busier highways,
resulting in a discrepancy between the proportion in accidents who are African American and the
proportion cited who are African American.

87
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.4 Proportion of Citations Issued to African Americans by Proportion of Drivers
in Accidents Who Are African American, by Troop District (N=53 Districts)

More important than the relative locations of accidents and citations in urban areas,
however, is the fact that there is variation by time of day in the proportion of drivers who are
African American and in the patrolling of highways. In the more urban areas or where an
interstate passes through, there is relatively more patrolling at night, when African Americans are
more likely to be on the highway. This driving pattern is not a well known fact, so we need to
establish it.

88
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Nationwide Personal Transportation Survey
The data presented in Figure 2.5 are taken from the 1995 Nationwide Personal
Transportation Survey (U.S. Department of Transportation, 1997). The purpose of that study was
to assess the amount and nature of personal travel in the United States. The figure represents the
time of day that respondents who live in the South Atlantic states leave for work, by race. White
Americans are more likely to be leaving for work during the hours of 5 a.m. and 9 a.m. than are
African Americans. Between the hours of 10:00 a.m. and 1 p.m., there is a steady decline in
whites leaving for work, while at the same time there is an increase in African Americans leaving
for work. From 2:00 p.m. until approximately 6:00 p.m., African Americans are slightly more
likely than whites to be leaving for work. After 6:00 p.m. until about 9:00 p.m., there is not much
difference in the time that white and African American respondents are driving to work. Finally,
between the hours of 10:00 p.m. and 11:00 p.m., African Americans are again more likely to be
driving to work. Overall, the pattern indicates that African Americans are slightly more likely
than whites to be driving to work during late-night hours and mid-afternoon hours.
Figure 2.6 shows the time of day by race that respondents are traveling for any purpose
(not just work). There appears to be even larger differences in the times of day that white and
African American respondents are on the road than were observed above for leaving for work.
African Americans appear to be leaving home in the early morning hours and late night hours
more frequently. Between the hours of 12:00 a.m. (midnight) and 5:00 a.m., they are more likely
to be on the road. This trend significantly decreases between the hours of 7:00 a.m. and 10:00
a.m., when whites are more likely to be leaving home for some type of travel. Further, between

89
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.5 Hour of Departure of Trips to Work by Time of Day, South Atlantic States, 1995
Survey

the hours of 12:00 p.m. (noon) and 5:00 p.m., whites also more likely to be on the road. After
5:00 p.m., the trend swings back to the greater representation of African Americans. Overall,
much like for work travel, the pattern indicates that whites and African Americans are on the road
at significantly higher rates during different times of the day, with greater African American
representation at night and in the early morning hours.

90
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.6 Hour of Departure for All Trips, By Race and Hour of Day,
South Atlantic States, 1995 Survey

91
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

One more point needs to be made regarding the citation rate of African Americans at night
versus during the day. One might argue that the NCSHP actually makes decisions to patrol more
at night in some areas as a result of racial bias or cognitive racial bias. However, in an analysis
not presented here, we found that the accident rates generally determine the citation rates across
districts and across time of day, so it would not seem likely that race is involved in the decisions
to patrol more or less at night, although we cannot rule out the possibility that some small
differences that are observed may be accounted for by such bias.
As can be seen in Figure 2.7, the ratio of night to day citations (number of) substantially
varies across districts. In Districts 3 and 46 (not their actual district numbers), for example, there
are more citations issued at night than in the day, whereas for most districts the majority of
citations are issued in daytime. Since we know that the proportion of African Americans on the
highway varies with the time of day and the amount of citations written in a district varies across
districts by time of day, it is necessary to control for time of day in the analysis. It should also be
noted that although the involvement in accidents of African Americans (and whites) increases at
night, they increase more for African Americans. Thus there is some explanation for an increase
in the proportion of citations issued to African Americans at night. However, it should be noted
that further information could be useful in evaluating the question of whether or not the amount of
patrolling at night is warranted by the volume of accidents at night.
In the analysis below, we will present results separately for day and night time
stops/citations to control (at least crudely) for the variation in the time of day differences across
races. Before doing so, however, we need to discuss another possible explanation of disparity in

92
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.7 Ratio of Night Citations to Day Citations, Fifty-Three Troop Districts

the proportion cited and the proportion driving who are African American, what we call “spatial
heterogeneity” (Smith et al., 2000).

Spatial Heterogeneity
Troopers patrol areas where the highways have many accidents and where there are
relatively many violators of traffic laws. They also patrol where it is more convenient to patrol
(there are turn-around capabilities— for example, overheads, or no guard rail or fence between
93
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

opposing-direction-bound lanes, and other factors). The NCSHP’s choice of specific locations
may also be made based on the prevalence of certain types of citizen behaviors— such as
speeding. If drivers tend to speed on some highways more than others, and those highways have
more accidents or are simply more convenient to patrol, there will be more patrol activity, and
more citations will be issued there.
In Figures 2.2 through 2.4 above, we saw that whether one used the proportion of licensed
drivers who are African American, or the proportion of “drivers driving” who are African
American, or the proportion of drivers in accidents who are African American, the correlation
with proportion of cited who are African American is very high (.93 through .96). Figure 2.8
shows the same variables in a slightly different format (although we omit the information here on
residents as it is very similar to that of “drivers driving”). Here the proportion of “drivers
driving” who are African American is subtracted from the proportion cited who are African
American. Similarly, the proportion of drivers in accidents who are African American is
subtracted from the proportion of cited drivers who are African American. These difference
scores show essentially the same information as in Figures 2.3 and Figures 2.4 above, only here
we examine difference scores, whereas above we examined a regression of one measure of
“proportion African American” on the proportion of cited who are African American. In Figure
2.8, however, it seems easier to “see” the differences in terms of disparity unfavorable to African
Americans and disparity in the opposite direction.
What we see in Figure 2.8 is that several districts have differences in the proportions
unfavorable to African Americans larger than .05 (for example, Districts 15, 18, and most of the
districts between District 44 and 52). Because there will almost always be some differences
between the measures in question and unknown sources— or what is called “random error”— we
94
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.8 Differences Between Proportion Cited who Are African American and
Proportion of ‘Drivers Driving’ Who Are African American and Proportion Cited Who Are
African American and Proportion Drivers in Accidents who Are African American

need to adopt some standard of difference below which we will call “likely measurement error”
(and which we will ignore here for the sake of discussion), versus a difference that we define as
large enough to require further investigation.16 Here we use a difference of +/- .05 somewhat
arbitrarily, but too small a difference would lead to a large number of false positives and false

16

We say “likely measurement error” but we do not formally derive the estimate of what
is likely measurement error, but rather assume that the relatively high variations observable in
the figures are more likely to be areas where bias might be occurring in decision making.
95
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

negatives. The +/-.05 criterion gives us a basis for making comparisons in some further analysis
below.
In the present graph, we are interested to see if the pattern of “positives” and
“negatives”— districts with difference scores above .05 and districts below -.05, respectively—
are similar across the two measures. They are somewhat distinct, in that the difference scores
with the accident measure tends to identify more positives and more negatives than the difference
score based on “drivers driving”(ten accident difference scores are above .05 compared to four
“drivers driving” difference scores; while six negative difference scores are below -.05 compared
to only two “drivers driving” scores). But the accident difference score (again comparing the
proportion cited who are African American to the proportion of drivers in accidents who are
African American) also has more “negatives”—districts with less than -.05 difference scores (in
other words, fewer African American cited than in accidents—six to two). What the reader
cannot learn from the graph is that the districts (which have an arbitrary identifier code attached
to them to identify them only to the researchers) are generally geographically adjacent. So, for
example, most of the districts to the far right of the graph are near one another geographically, as
are many of the districts with difference scores in the middle half of the chart. The implication of
this is that the patterns of positives and negatives are likely to be due to some shared unmeasured
characteristic (perhaps some organizational or geographic characteristic associated with the
difference score variables or their component variables). Ideally as researchers, we would like to
identify those characteristics.
The comparison in Figure 2.8 of “drivers driving” and accidents as a baseline against
which to evaluate the proportion issued citations who are African American, indicates that more
districts qualify as “positives” using accidents as a baseline. For that reason (to be conservative
96
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

in testing the racial disparity hypothesis),17 as well as the fact that accident data are available for
measurement at relatively small units of analysis, we will proceed to analyze further the accident
data, and return to a discussion of drivers driving and resident licensed drivers at the end of the
chapter.
The possible advantage of small units of analysis centers around the question of whether
patrolling occurs evenly across geographic areas within districts compared to the distribution of
African American drivers (and their driving behavior) within districts. The issue we raise is that
there may be a mismatch of where patrolling occurs (too little or too much) relative to where
African American drivers drive. There is a possible mismatch if “spatial heterogeneity” occurs in
both the geographic distribution of African American drivers and in the distribution of patrolling.
For illustration purposes, Figure 2.9 below shows an area of Nash county in North
Carolina, where two major highways intersect, I-95 and U.S. 64. (the other roads on the diagram
are fictitious). The proportion of drivers who are African American among those we define as
“threshold speeders” is presented in the figure. We traveled each of these highways for a week as
part of what we call our observational baseline study (see Appendix A) to determine which
proportion of drivers there drove at speeds higher than the local “speeding threshold.” The
speeding threshold is defined as the speed at which a citation for speeding was likely to be issued
(this we define more precisely as the median speed at which drivers were issued citations). The
proportions of African American drivers at or above this threshold are shown for each highway,
and they differ substantially (12 percent absolute difference). The aggregated data for Nash
County of the number of citations of African American threshold speeders who were issued

17

By conservative here we mean that we would prefer to err on the side of rejecting the
null hypothesis of no racial difference.
97
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

citations would be greatly affected if it happened to be the case that there was more patrolling of
I-95 than U.S. 64 (the proportion of citations issued to African Americans would be lower), or if
there was more patrolling of U.S. 64 (the proportion of citations issued to African Americans
would be higher). If the patrolling was evenly distributed across highways, there would be no
“spatial heterogeneity” issue, as the overall rate comparison would be based on a composite
average of the citations and the speeding drivers.
The highways in Figure 2.9 are two of fourteen that we have observed (see Appendix A).
In the four counties where we observed two different highways, we found that the proportion of
threshold speeders who are African American varies substantially between the two highways,
with an average difference of about 5 percent. In other words, there is considerable spatial
heterogeneity in the distribution of African American speeders between highways. Since our
measure of their speeding is reasonably accurate and objective (in other words, based on
observation, not on official records), we must conclude that there is little doubt that there is
variation within a district in the racial composition of speeding drivers. (We will show below that
patrolling varies considerably from highway to highway.) Thus, comparison of rates at the level
of county or district (such as the fifty three districts in the figures above) will be prone to biased
estimates in the face of a mismatch of where drivers drive and where the NCSHP monitors
vehicles. This mismatch is a manifestation of the more general problem in areal analysis of
“spatial heterogeneity.”
To address the questions of the mismatch of where drivers drive and where the NCSHP
looks for violations (spatial heterogeneity), we present difference scores based on small units of
analysis, the highway area (as defined above). It consists of a stretch of highway covering a
portion of a county (usually about a third of a county). The advantage of the highway area as a
98
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

unit of analysis is that we have at least two reasons to believe that the mismatch of drivers and
patrolling discussed above is lessened by studying highway areas as a unit of analysis rather than
larger units of analysis such as counties or districts. First, we have examined the distribution of
the proportion African American cited along stretches of highway within areas where mile
markers exist, and where they are recorded in the citation data base, as compared to the
proportion on intersecting highways. We find greater consistency (less heterogeneity) in the
proportions along stretches within the county areas. Secondly, we know that patrolling often
takes place in the form of driving up and down rather long segments of highway (we rode with
troopers and observed troopers in their routines of patrolling), thus providing some evidence that
areas of highways are patrolled in segments. Although far from ideal in terms of testing a
“mismatch” hypothesis, we think that there should be less of a mismatch between trooper
patrolling and vehicular driving behavior within smaller geographic units of analysis (the smaller
the unit of analysis the less mismatch is likely).

99
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.9 Two Major Highways in Nash County, North
Carolina

Example of Spatial Heterogeneity in Threshold Speeding Behavior
In Relation to Heterogeneity in Patrolling Behavior (Nash County)
I95
26% Th. Speeders
Are Black

NC_3

US64
38% Th. Speeders
Are Black

NC_5

If we had more observations, we could define an extremely small unit of analysis, such as
the area of highway between intersections. For these extremely small units of analysis,
presumably the demographic make up of that segment of highway would be homogeneous across
its span (since by definition no turnoffs exist).18 Within such a unit there would be a decrease in
the number of issues associated with spatial heterogeneity. Essentially, the amount of patrolling
(frequency of patrolling) of a highway segment would be assumed to be equal between the two
intersections, allowing us to compare the proportion cited who are African American to the

18

Thus once a vehicle enters a stretch of highway at an intersection, the vehicle cannot
exit the highway until the next intersection (except in the rare instance of a U-turn on the
highway).
100
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

proportion involved in accidents on those same segments of highway. We do not have such data,
however.19 We only have data statewide for the accidents and citations in a county highway area.
One problem with studying highway areas, however, is that we must be concerned with a
small number of observations within a highway area. There can be either too few citations or too
few accidents in an highway area to justify comparing the two proportions (African Americans
cited to those involved in accidents). For purposes here, we find that requiring five or more
citations and five or more accidents, and then constructing weighted averages of all those
highway areas meeting those criteria, gives us the same results as requiring twenty or thirty
citations and twenty or thirty accidents in a highway area.
To demonstrate that the “five or more” restriction does not lead us to radically different
conclusions, we compare the difference scores counting all accidents and citations versus the
aggregated difference scores based on a minimum of five. Figure 2.10 shows the relationship
19

It is not obvious that a geographic unit of analysis defined as a stretch of highway
between two intersections (they have been called “face blocks” in other research) would not
suffer from possible spatial heterogeneity concerns. Assuming that a patrol car that enters such a
geographic area must exit it, however, it can be argued that all parts of the area are equally
patrolled, and that over time the proportion African American cited for violations should be
constant or homogeneous within each unit (face block). There could be variation across areas
(face blocks) in the extent of patrolling, as some areas are patrolled more often than others.
However, when all the areas of a county or district are aggregated to the county or district level,
and weighted by each area’s contribution to the count of citations and accidents, the difference
between the proportion cited who are African American and the proportion in accidents who are
African American should be unbiased. Take an extreme case where NCSHP rides up and down
one highway in a county, ignoring all others in the county (except when they are called to
accidents). One might think that in such a situation the estimate of proportion cited who are
African American would be biased relative to the proportion in accidents who are African
American, but if the data are weighted by the contribution of citations and accidents, there could
only be an estimate based on the highway patrolled, since there could be no calculation of the
proportion of those cited who are African American in the other districts (no citations occurred
there, and division by zero is impossible). Take another example, where 90 percent of the
citations occur on one highway in a county, and 10 percent are scattered across other highways.
Again, because the data are weighted when aggregated to the county level, and because the unit
of analysis ensures homogeneity within units, there will be no bias in the estimates.
101
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

between the aggregated difference score measures using weighted averages of the proportion
cited who are African American minus the proportion in accidents with African American drivers
(aggregating all highway areas with five or more citations and five or more accidents), as
compared with the difference scores between citations and accidents for the districts as a whole
(for example, what we presented in Figure 2.9 above). The weight used for the “five or more”
difference scores is the sum of the number of incidents (citations plus accidents) in a district
divided by the total number of incidents (citations plus accidents) for the district. Our intent with
the figure is to illustrate that the dropped highway areas do not have a large impact on the two
types of difference scores compared here (all accidents versus five or more accidents/citations in a
year). Yet, even though there is little overall difference in the proportion African American using
all accidents or five or more per year, results vary for specific districts, depending on which
difference score is used. For example, in District 18, the aggregated difference score results in an
8 percent higher estimate for citations to accidents using the “five or more” restriction than do the
aggregated counts, which are then calculated as proportions and then subtracted from the
proportion of citations issued to African Americans.
It should also be noted that the requirement of “five or more” citations and accidents is
designed to limit the variation in the error when calculating “proportion African American cited
or proportion African American driving vehicles in accidents,” but is in no way intended to be a
sufficient empirical basis for comparing proportions for any one highway area. Rather,
comparisons are made using the aggregated highway area data (summing across highway areas
within a district to the district level), as opposed to differences for any one highway area. By
aggregating the highway area differences and weighting them, the district comparisons are
predominantly affected by the highway areas with the larger numbers of incidents (citations and
102
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

accidents). When the highway areas are aggregated to the district level, hundreds of highway
areas are typically aggregated and weighted by the proportion of all the citations and accidents
each highway area represents. The effect of these weights is to limit the amount of error
generated for each district with small numbers of observations.20
In summary, thus far we have seen that time of day and the mismatch of patrolling and
driving in local areas can lead to differing results in the comparison of driving behavior to the
citation receiving experiences. What remains is to combine the aggregated difference scores at
the highway area level separately for daytime and nighttime, so that we control for both time and
space mismatch of patrolling and driving behavior (as measured by accident involvement).

20

We tested for whether or not 1 or 5 or 10 or 15 cases or up to 200 made any difference
to the analysis by comparing correlations between the proportion cited who are African
American and the proportion drivers in accidents who are African American, and found that the
correlations deteriorated below .500 when less than 5 citations/accidents were used. The number
of highway areas drops from 10,474 to 3,233 when we require 5 or more citations and accidents,
so we lose information on a substantial number of highway areas where there is little action
(accidents or citations). We know of no a priori reason to believe that bias should be exercised
in less trafficked areas than in more trafficked areas. Also, when we examine the data are large
units of analysis, the racial disparity is not large in absolute terms, so it is very doubtful that
racial bias is occurring in less busy places.
103
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.10 Comparison of a Baseline with All Accidents and a Baseline with Highway
Areas of 5 or More Accidents and 5 or More Citations

Day and Night Rates
In the analysis below, we distinguish between citations and accidents occurring between 6
p.m. and 6 a.m, (night) from those occurring between 6 a.m. and 6 p.m. (day). Figures 2.11 and
2.12 illustrate the proportion of citations issued to African Americans and the proportion of
accidents involving African Americans for the fifty-three districts for day and night times,
respectively. In Figure 2.11, the differences in the positive direction that we observe are
substantially fewer in number than we observed previously in Figure 2.10. That is, examining

104
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.11 Difference in the Proportion Cited African American and the Proportion in
Accidents Who are African American, Daytime Only (Min of 5 Accidents and 5 Citations
Per Highway Area)

daytime rates and controlling partially for spatial heterogeneity mismatches result in only 3
districts having greater than a .05 difference score (proportion of those cited who are African
American minus proportion in accidents who are African American). Examining negative
difference scores, there are now nine districts with less than -.05 difference score values
(indicating that African Americans are under-represented among the ranks of those issued
citations in those areas).

105
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 2.12 Difference in the Proportion Cited African American and the Proportion in
Accidents Who are African American, Nighttime Only (Min of 5 Accidents and 5 Citations
Per Highway Area)

In Figure 2.12, the difference scores are compared across districts for citations and
accidents at night. Here, there are six districts with difference scores of .05 or higher, while there
are fifteen with negative difference scores of -.05 or lower. Again, we are controlling for time
because we are limiting the observations to nighttime (again 6 p.m. to 6 a.m.). Looking at
difference scores calculated at the highway area level and aggregating to the district level, we see
yet a different evaluation of the districts than that which we observed by simply comparing the
differences in the proportions at the district level.

106
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

What are the Mechanisms that Generate Positives (and Negatives)?
Several of the districts in Figures 2.11 and 2.12 show racial disparity in the proportion
cited who are African American relative to the proportion in accidents who are African American,
and several of these districts are geographically adjacent to one another. Not visible in the figures
is the fact that several are part of the same troop. This raises several questions about these
districts. Are these districts home to troopers who engage in some form of racial bias? If so, how
does the bias manifest itself? Can we determine if the bias is likely to be at the individual trooper
level, or can it be accounted for at the aggregate level? Furthermore, is it inadvertent, or possibly
even rationally based (for example, the result of deployment to highways with high accident
rates)? Still, one need to address the districts in the figures with a negative differential (less than
-.05) in proportion cited who are African American to the proportion in accidents who are African
American (those areas with too few African Americans issued citations)? What could be
accounting for the “reverse discrimination” pattern? Could it be some form of reverse racial bias
or aversive bias (avoidance of African Americans) or even attempts on the part of some troopers
to avoid the appearance of racial bias by giving breaks to African Americans by issuing them
fewer citations?
We will address possible racial disparity on the part of individual troopers more directly in
the next chapter. Here we consider further subtleties in the processes being studied here. We
begin by examining the six positives in Figure 2.12 above (nighttime positives), Districts 4, 15,
46, 47, 48, and 49. One possible explanation for these positives has to do with where patrolling
occurs relative to where accidents occur. We discussed earlier that there is a tendency for the
NCSHP to “go where the action is” in the sense of traffic, rather than patrol where the accidents
are (so as to prevent accidents through their greater visibility in the area). Table 2.3 shows the
107
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

breakdown of accidents and citations for the six districts with more than a .05 difference in
proportion cited who are African American to proportion in accidents who are African American.
The percentage of all accidents and of all citations occurring on the road type is presented. Thus,
for example, in District 15, 8.2 percent of the accidents occur on the interstate, whereas 13.5
percent of the citations occur on the interstate. In general, there is a mismatch between where
accidents tend to occur and where troopers patrol (almost one-half of the accidents occur on rural
paved roads across districts [District 49 excepted], whereas about one-quarter of the citations
[varying up to 33.8 percent] occur on rural paved roads, suggestive of under-patrolling of rural
paved roads). Of course, this does not mean that the NCSHP troopers should change what they
are doing and focus on rural paved roads. Rural paved roads are numerous and constitute the
most frequent type of road and the most cumulative road miles (data from the N.C. Department of
Transportation, not reviewed here, indicate this to be the case). That is, it may be inefficient to
patrol such highways because they are too dispersed, whereas the N.C. highways, U.S. highways,
and interstates are not as dispersed (indeed, they are often the central traffic arteries of an area.).
In three of the six districts shown, the type of road most heavily patrolled is the type of
road with the highest or second highest proportion of drivers in accidents who are African
American. Thus in essence, the NCSHP “over-patrols” the type of highways in these districts
where African Americans are more prevalent drivers. This over-patrolling could account for
much of the positive difference between proportion cited and proportion in accidents who are
African American. In two of the remaining three districts, there are small differences in the
proportion of citations issued to African Americans across highway types, and in both districts
there is a relatively high percentage of rural paved roads patrolled (about a third of the citations
occur on these roads and the proportion cited who are African American is substantially higher
108
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

than the proportion in accidents). While this could be the result of racial bias, another possibility
is that the daytime rush hour traffic, which is predominantly “white” (assuming the National
Transportation Survey results hold here), accounts disproportionately for many of the accidents,
driving down the proportion of African Americans involved in daytime accidents. By contrast
African Americans are somewhat more likely to be on the road in off-rush hour times – such as at
night, when accident rates are generally lower. See Appendix G where we discuss the fact that
the ratio of accidents to injuries varies considerably from high-volume traffic roads to lessvolume roads: the “fender-bender” type accident may be more numerous in rush-hour traffic, but
also more likely to involve whites. This is a possibility that warrants further exploration.
One exception to the general pattern in Table 2.3 is District 49, in which virtually all of
the accidents are occurring on the interstate, and citations there are relatively less than called for
by the accident rate. Still, 81.6 percent of the nighttime citations in the district are written on the
interstate. In this district, citations are also issued on U.S. highways, and both the interstate and
U.S. highway citations tend to drive up the proportion African American number cited. We are
not sure what could be the cause (racial bias or something else), but, again the predominantly
white rush hour may be accounting for a high percentage of the accidents, driving down the
proportion African American rate.

109
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 2.3 Percent of All Accidents and Percent of All Citations by Road Type, Six “Positive”
Districts, Nighttime
District

Interstate

U.S. Highway

N.C. Highway

Rural Paved Roads

Accid’s

Citation

Accid’s

Citation

Accid’s

Citation

Accid’s

Citation

4

-

-

37.2

46.5

28.4

31.8

34.4

21.7

15

8.2

13.5

20.4

35.9

29.8

31.3

41.6

19.2

46

-

-

38.8

58.1

6.4

8.4

54.8

33.5

47

-

-

25.5

59.5

18.0

14.1

56.5

26.4

48

4.4

7.8

12.1

35.8

37.4

22.6

46.2

33.8

49
99.0
81.6
.6
8.3
.2
8.6
.2
1.5
Read: “37.2 percent of all accidents in District 4 occur on U.S. highways, whereas 46.5 percent of
all citations occur on U.S. highways in that district.”

For nighttime differences in the proportion of those cites who are African American
relative to the proportion in accidents who are African American, the results for the districts with
positive difference scores greater than or equal to .05 (not shown here) also reveal a relatively
high percentage of the drivers in accidents who are African Americans on the highway types
where the highest proportion of citations are issued. The patrolling in these districts at night time
is most often (modal category) on interstates or on U.S. highways, and it is on these same
highways that African Americans have a relatively high representation in the ranks of those
involved in accidents (indicating presence on the highway, and to some extent culpability or
driver error).

110
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Citation Zones as a Causal Mechanism
In addition to variations in rates on road types within districts and within what we define
as “daytime” and “nighttime” hours, we think that it is useful to consider the importance of
“citation zones” in the citation process within a stretch of highway. That is, there is a very local
geographical distribution of citations, and examining them may shed light on how deployment
might account for the positives and the negatives in the figures above.
In Figure 2.13, we review a stretch of I-85 in North Carolina that has several major
intersections. The hash marks indicate mile markers, and the longer lines indicate intersections.
What is striking about the figure is the variation in the number of traffic stops across miles of
highway. They vary considerably from 814 to nineteen (with only 2 miles separating these two 1
mile stretches of I-85). The mechanism by which stops occur in the district where this segment of
I-85 is located, is probably a function of deployment (defined simply here as where the NCSHP
troopers stop vehicles) and also somewhat on the behavior of those vehicles. Notice how the
number of citations are substantially higher near the major intersections. This is not because the
density of traffic is necessarily higher at those points—there are no other intersections along this
stretch of I-85.21 Drivers’ behavior may be involved to the extent that those entering the highway
quickly attempt to approximate the traveling speed of the other vehicles on the highway.
Obviously this involves accelerating the vehicle, and may involve surpassing not only the speed
limit, but the “threshold” speed limit in doing so (the speed limit as enforced on that stretch of I85, roughly about 15 mph above the posted speed limit). Other factors may include geographic
inclines or declines, but we did not collect data on such attributes.

21

Thus once a vehicle enters a stretch of highway between an intersection and another
intersection, the vehicle cannot get off that highway until the next intersection.
111
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Essentially, the areas in the figure with hundreds of citations issued could be classified as
“citation zones” (including what some may call “speed traps”— although more than speeding
could have triggered a stop in that location). As discussed earlier, by the term “citation zone,” we
mean to include more than what is usually meant by the term “speed trap,” typically troopers with

Figure 2.13 NCSHP Stops on Interstate 85 by Milepost Number of African American and
White Stops

radar guns sitting stationary, hiding in bushes or behind an overpass. A citation zone is located
where there is a high volume of stops for any reason. Citation zones account for a high
proportion of all stops (the top 4 miles in total stops account for 66 percent of all the stops on this
13-mile stretch (in other words, 31 percent of the miles account for 66 percent of the stops).
We do not know if accidents co-occur proportionately to the stops because we do not have
milepost (locational) data for accidents. However, it would be surprising to find a close
correspondence of accidents by milepost markers. For example, we find it hard to believe that

112
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

there would be forty-three times as many accidents in a 1-mile stretch of I-85 as compared to a
stretch 2 miles away (we have examined the distribution of accidents across county areas—one
third of a county— for I-85, and find that the length of the highway in an area and its proximity to
large urban centers are strongly associated with the number of accidents). Even if, say, the mile
with 814 stops had disproportionately many of the total number of accidents along this 13-mile
stretch of highway, it is unlikely that the stops would be proportionate to the accidents. Rather,
other factors, such as the volume of violating behavior and availability of turn-around spots are
associated with making a “stop place” attractive. More research needs to be done on regarding
this consideration.
In summary, one of the key “mechanisms” by which citations are generated along a busy
interstate is the “citation zone” (anywhere a disproportionate number of tickets are written).
Accidents in a district, by contrast, tend to occur in areas where the citation zones are probably
seldom employed: small, relatively narrow, rural roads. Recall Table 2.3 above, which shows the
distribution of accidents by type of road across the districts with “positive” scores in Figure 2.12.
Also shown are the proportion of citations by type of road. Clearly, there is a mismatch. It would
seem that troopers do not necessarily patrol roads where accidents are more likely to occur, but
rather patrol highways “where the action is”—the larger interstates, U.S. highways, and N.C.
highways. The type of accident and the prevalence of types of highway could account for the
mismatch of patrolling and accidents. Accidents due to road design (such factors as narrow
roads, poorly maintained road shoulders, steep inclines, or the presence of stop signs or lights) are
associated with less dense traffic, such as accidents on what are called “the rural paved” roads.
Also, there are many such roads across the rural areas of North Carolina, such that the few
NCSHP troopers on duty on any one shift would not be wisely using their time to patrol these low
113
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

density highways. Rather, more violators (and possibly more deserving violators) are available
on the busier highways where the troopers are best deployed. It is ironic that troopers may be
best deployed on the relatively safe (but highly used) roads.
How then might the positives of Figure 2.11 and 2.12 be generated? Deployment
decisions to patrol the busier highways (making NCSHP cars “highly visible”) could account for
the disparity in the racial component relative to accident rates. If African Americans happen to be
on the larger, busier highways more so than whites, the rates of citation will be higher for African
Americans, independent of racial bias on the part of troopers. If deployment decisions are
irrationally based, as opposed to rationally based (“we patrol there because the fishing is good”
rather than “we patrol there because preventable accidents occur there”), the argument could be
made that the racial disparity is inadvertent, but preventable by a change to a more statistically
grounded deployment strategy.

Summary
In this chapter we have seen how the data, when aggregated to the entire state level, show
some disparity in the citation rates, indicating over-stopping and over-citing of African
Americans. However, the extent of racial disparity varies considerably by type of charge. Only
some types of charges show disparity against African Americans. Moreover, there are many
possible reasons for such statistics at the statewide level, potentially having little to do with bias.
As we disaggregate22 the data to smaller units of analysis, we begin to get a clearer picture of the
extent and whereabouts of disparity, and of the possible bias forms it could take (conscious bias,

22

By “disaggregate” here we mean more literally that we aggregate the data to relatively
smaller units of analysis, such as the highway area, as opposed to the trooper district.
114
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

cognitive bias, inadvertent-irrational bias, or rational bias). Looking at the a somewhat
disaggregated analysis for districts (N=53), we see some differences in terms of racial disparity,
with some districts having more disparity evident than others.
The results speak to the possible value of using different baselines, against which to
compare the percent of African Americans cited or warned. We can compare citation and written
warning data to the resident licensed population proportion who are African American, or to the
“drivers driving” who are African American, or to the proportion of drivers in accidents who are
African American.23 We argue that the latter may be a convenient, if not superior, type of data for
looking for bias because of its widespread availability, and the fact that it can be measured at
rather small units of analysis— here the highway area (a stretch of highway that covers an
NCSHP officially designated area, or about one third of a county).
At the highway area level, when we calculate the difference scores for each highway area
with five or more citations and five or more accidents, and aggregate that information to the
district level (N=53), we find several districts that have high rates of citations of African
Americans (also several with low rates, relative to the accident rates). We focused on these
districts for further analysis, and found that the racial disparity is probably in part due to
deployment. Specifically, it appears that the NCSHP deploys troopers to busy highways, which
generally are not the highways where most accidents occur, but are the highways where African
Americans may more often be found as drivers.

23

We propose that the percent of reported accidents in which the drivers are African
American is a measure of the prevalence of African American drivers on the highway. Some
may prefer to attribute driving ability to be measured by involvement in accidents, but we do not
make that claim. Note that we are referring here to accident reports and not to citations nor to
any attribution of responsibility for the accidents.
115
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Other researchers working in the area of racial bias and profiling have struggled with the
problem of identifying baselines that are appropriate for making comparisons. The most
widespread baseline used is still census counts of residents in an area. We find here that the
accident baseline, while far from perfect, provide information that at relatively small units of
analysis. Because of the availability of these data at small units of analysis, accident data
probably should be used as a baseline more generally in research on policing and possible bias or
disparity.
Looking at the results in terms of the bigger picture, the differences in racial disparity
observed in the early tables of this chapter are probably largely accounted for by deployment
decisions and by variation in vehicular behavior across the races. Further analysis would be
necessary, and further discussions with the NCSHP leadership, to determine the likelihood that
the deployment decisions have been arbitrarily made -- and could be changed – if inconsistent
with a more rational plan to allocate troopers to highways. Such a plan would presumably be
based on preventable accidents (such as is part of the Total Quality Management approach of the
NCSHP in the past few years). Such an analysis and discussion is beyond the scope of this report,
but would nevertheless be worthwhile.
As researchers who are confronted with the data and analysis available to us, we believe
that we have demonstrated that some racial disparity exits in the findings, whether large or small
units of analysis are used, but that what seems to account for the disparity is probably mostly
behavior of the drivers of the vehicles, as well as deployment decisions, whether made by the
commanders, sergeants, or the troopers on the highway. The possibility also remains that some of
these decisions are guided by conscious bias or alternatively by cognitive bias. In the next
chapter we will explore further the possibility that cognitive bias plays a role in the day-to-day
116
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

decisions of troopers. That is, we assess individual trooper propensities to depart from the
citation standard set by other troopers in the trooper’s same district.

117
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

References for Chapter Two
American Civil Liberties Union (ACLU). 2002. Web Citation: www.aclu.org/Profiling/index.html
Kociewiewski, David. 2002. “Study Suggests Racial Gap in Speeding in New Jersey” March 21,
2002. The New York Times.
Lange, James E., Kenneth O. Blackman, Mark B. Johnson. (2001). Speed Violation Survey on
the New Jersey Turnpike: Final Report . Report Submitted to the Office of Attorney
General, Trenton, New Jersey.
Smith, William R., S.G. Frazee, E.L. Davison. (2000). Furthering the Integration to Routine
Activity and Social Disorganization Theories: Small Units of Analysis and the Study of
Street Robbery as a Diffusion Process. Criminology, 38(2), 489-523.
U. S. Department of Transportation, Federal Highway Administration. 1997. Web Citation:
www.fhwa.dot.gov/ohim/sept97.htm

118
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Chapter 3 Racial Disparity at the Individual Trooper Level
Goals of Individual Trooper Analysis
The question of whether or not racial disparity is present, as well as the extent and nature
of it, can only be understood fully if the behavior of individual troopers can be evaluated. In the
previous chapter, we argued that the context in which a trooper works will be important in
accounting for the racial make-up of the citizens he or she stops, cites, and warns. In the present
chapter, we will make use of the measures of context from the previous chapter and assess the
extent to which individual troopers deviate from an expected number of citations written to
African Americans. That expected number (based on a statistical model) may be largely based on
a trooper’s colleagues issuing citations in a similar, or even the same, context.
Our purpose in this chapter is, in part, to consider in some detail whether or not a
methodology can be devised to serve as a source of information to identify troopers who deviate
from the norm (baseline) in their citations of African Americans. We would like to know if we
can explain the citations of African Americans by the contexts in which the troopers work, as well
as by the characteristics of the trooper, such as his or her race, gender, or age. As such, this
chapter’s topic is a matter of much concern to both the NCSHP and the public. Troopers are
genuinely concerned that their behavior might be labeled as racially biased, and the public wants
to know if particular troopers behave in biased manners. It should be made clear from the
beginning that our goal is not to identify or name individual troopers, but rather to develop and
illustrate statistical methodologies that could be used by police organizations to generate
information on racial disparity (generate one type of information) at the individual trooper level.

119
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

This information would require further validation by an organization before any conclusions
could be drawn about whether or not racial disparity is a problem.
Presumably, the methodologies described in this chapter would be supplemented with
additional information to validate that a racial disparity problem exists. Such additional
information could include a more detailed analysis of the patrolling patterns of a trooper who has
a seemingly high rate of African American citations. Or it might include reviews of citizen
complaints, or even discussions with troopers in a district about the possibility of the existence of
a hostile racial climate in that district. We want to be clear that the proposed methodologies here
represent “pieces of the puzzle” rather than a definitive indication of the presence of a racially
biased trooper.

Individual Level Characteristics
As mentioned above, and as discussed in the previous chapter, there are some individual
characteristics of troopers that may be pertinent to any disparity that may be found in their
citation behavior. For example, one might hypothesize that white troopers would be more likely
to display bias toward African American drivers than African American troopers. Tables 3.1, 3.2,
and 3.3 show the demographic breakdown of all NCSHP troopers working in North Carolina in
calender year 2000. As can be seen, only 15.4 percent of the troopers are African American (a
small number of other minorities have been excluded from the table), compared to approximately
22 percent of the adult population of North Carolina. Women are much more under-represented
relative to their representation in the population (only 1.7 percent). The majority of troopers are
between the ages of 31 and 44 (54 percent), with 25.5 percent younger than 31 and 20.5 percent
older than 44.
120
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 3.1 Racial Make-up of North Carolina State Highway Patrol
Race of Trooper

Frequency

Valid Percent

African American

212

15.4

White

1168

84.6

Table 3.2 Gender Make-up of North Carolina State Highway Patrol
Gender of Trooper

Frequency

Valid Percent

Female

24

1.7

Male

1383

98.3

Troopers not only vary in age, but also vary in years on the force and the extent to which
they have participated in training programs. The more training programs that a trooper
participates in may be understood as a measure of professionalism. Troopers who take the time to
participate in various training programs might be considered to be among those who take their
work more seriously than others, who want to improve their job performance, or, alternatively,
who have acquired a more professional attitude toward their work.24
Table 3.4 shows the number of troopers participating in specific types of training
programs, and Table 3.5 shows the number of training programs completed by varying lengths of
time of employment. In Table 3.4, the most popular training program is CPR, with about 82
percent having completed this form of training. The second most popular courses are training
classes for security involving “executives”—usually government leaders or dignitaries for whom
24

The number of training programs completed obviously can vary across officers not
only for the reasons of professionalism, but also because of age, tenure on the force, and other
factors.
121
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

the NCSHP may provide security. The third most popular training program is Division of
Criminal Information (DCI), which involves some training in evidence gathering and basic crime
scene forensics. We will not discuss the specific nature of all of these programs, but present them
to make the reader aware of the nature of possible training programs available to the trooper.
Also, as we mentioned above, the participation of a trooper in multiple forms of training
programs may be an indication of “professionalism” on his or her part, but this is—as of
yet—speculative on our part. Our hypothesis is that the more professional the orientation of the
trooper, the less likely he or she would apply the law in a racially biased manner (at least in an
overt way). Later in the chapter, we will assess whether or not our hypothesis is supported.
In Table 3.5, we take up the question of whether participation in training programs is a
function of time on the job (tenure). As expected, those troopers only on the force for a year or
two have had the least opportunity to accumulate training experience (45.6 percent of troopers in
their first two years have had no training beyond the academy). Approximately 18 percent
troopers have managed to complete three or more of the training programs. In general, the longer
the years of employment, the more training programs are completed. The majority of those
troopers with six years of tenure have completed six or more training programs. At the same
time, it is noteworthy that among troopers on the job for twenty-one or more years, about 47
percent have less than six completed training programs. Troopers on the force for less time have
generally completed more training programs. (Possibly the older troopers already had years of
experience before some of the training programs became available, and many of them may have
thought the training was unnecessary since they were performing ably at the job already.)

122
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 3.3 Age Distribution of North Carolina State Highway Patrol
Age of Trooper

Frequency

Valid Percent

30 years old and younger

359

25.5

31 thru 36

376

26.7

37 thru 44

384

27.3

45 and older

288

20.5

One might argue that the race of the trooper would be important to examine relative to the
race of the citizen, under the assumption that if white troopers are biased, a higher proportion of
the citizens they stop or cite would be African American than the proportion cited by their fellow
African American troopers. Table 3.6 shows that the reverse is true, a result that is puzzling only
because we have failed to take into consideration the fact that African American NCSHP troopers
tend to work in areas where there are high percentages of African American citizens and African
American drivers. In Table 3.6, it can be seen that 31.7 percent of all citations issued by African
American troopers are issued to African American citizens, whereas only 24.2 percent of citations
issued by white troopers are to African American citizens. A similar pattern is observed for
specific types of reasons for citations (speeding, unsafe movement, and failure to stop/yield).
In an analysis not presented here, we found that troopers on the force for more than twelve
years seem to be less likely to issue a citation to an African American than are troopers with less
time on the force. However, like the analysis above for “race of trooper,” this is probably due in
part to the fact that older troopers are more likely to work in areas where there are more whites
(less urban counties). Thus, the context in which the trooper works may help explain the pattern
of data at the individual trooper level. The smaller volume of traffic— and thus greater safety
—can make more rural areas an attractive area for older troopers to work, whereas younger
123
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

troopers must “pay their dues” by working more densely populated areas. Most of the rural areas
of North Carolina are predominantly white, while large African American populations
disproportionately inhabit the larger cities as well as several rural counties.25 The general pattern
of lower percentages of African American citizens cited by white troopers holds true across the
reasons for the citation (speeding, unsafe movement, and failure to stop/yield).

Measuring the Context of the Trooper’s Behavior
In the last chapter, we learned that there are various types of measures of context that
could be used in the analysis, such as the racial composition of the drivers with licenses, or the
racial composition of what we call “drivers driving.” We also saw that the racial composition of
accidents could be a valuable measure because it allows us to measure the context of the trooper’s
work in yet another way (and arguably more precise way)—the racial composition of smaller
units of aggregation than the fifty-three NCSHP districts. We assume that the smaller unit of
aggregation would be the more ideal, as it would lessen the dangers associated with spatial
heterogeneity, that is, the mismatch of where drivers drive and troopers patrol, as discussed in the
previous chapter. Two such units of aggregation seemed attractive to us: the county area (roughly
one third of a county in size), and the county highway area (the segment of a particular highway
in a county area). The analytic question before us is how to link information about the contexts in
which troopers work with the individual level data. That is, what is a reasonable mathematical
model of the processes that we have been discussing at the individual level of analysis?

25

The claim that the work context accounts for the pattern is purely speculative on our
part, but we will take it up again in the analysis below.
124
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 3.4 Type of Training Received by North Carolina State Highway Patrol
Type of Training

Number of Troopers
who Received Training

Percent of Troopers
who Received Training

Advanced Accident Investigation

115

8.2

Accident Reconstruction

103

7.3

CPR Training

1151

81.8

Division of Criminal Information

241

17.1

Drug Interdiction Program

231

16.4

Emergency Medical Tech

129

9.2

Executive Security

322

22.9

Field Training Officer

337

24

Field Training Supervisor

218

15.5

Cert Instru Specialized Defensive Driving

105

7.5

Certified Instru General

296

21

Mobile Data Training

707

50.2

Certified Radar Operator

392

27.9

Riot Control

503

35.7

Mini 14 Rifle

365

25.9

Intoxilyzer 5000

1205

85.6

Certified Vascar and Radar Operator

812

57.7

125
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 3.5 Years Employed with North Carolina State Highway Patrol by Number of
Training Programs
Number of
Training
Programs

1 to 2
Years of
Employ
ment

3 to 6
Years of
Employ
ment

7 to 12 Years
of
Employment

No Training

57
(45.6%)

8
(3.3%)

2
(.6%)

1 to 2
Trainings

46
(36.8%)

14
(5.7%)

2
(.6%)

1
(.3%)

12
(4.6%)

3 to 5
Trainings

19
(15.2%)

169
(69%)

150
(42.5%)

98
(25.4%)

108
(41.5%)

6 to 7
Trainings

3
(2.4%)

48
(19.6%)

130
(36.8%)

138
(35.8%)

79
(30.4%)

6
(2.4%)

69
(19.5%)

149
(38.6%)

60
(23.1%)

245
(100%)

353
(100%)

386
(100%)

260
(100%)

8 or more
Trainings
Total

125
(100%)

13 to 20 Years of
Employ
ment

21 or more
Years of
Employment
1
(5%)

Table 3.6 African Americans Cited by Race of Trooper and Type of Citation
White Troopers

African American Troopers

All Citations Issued to AA

79102
(24.2%)

19638
(31.7%)

Speeding Citations Issued to
AA

56119
(23.6%)

13705
(30.1%)

Unsafe Movement Citations
Issued to AA

3614
(19.4%)

831
(27.1%)

Failure to Stop or Yield
Citations Issued to AA

2247
(18.4%)

589
(25.5%)

Obviously, this exercise could be done at any level of aggregation that matched the
available data associated with a stop/citation or written warning with where troopers actually
patrol. We use the spatial units that follow (primarily county area and county highway area, as
126
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

described below) because they represent the smallest units of analysis that we have with sufficient
observations to warrant statistical analysis. For many applications in other venues, only data at
larger units of analysis may be available. At such levels the general method employed here could
still be used, although, as the spatial unit becomes increasingly larger than the actual patrol area,
we would expect more measurement error.
Before describing the analysis further, it occurred to us that the individual trooper data
allows for yet another measure of the context in which the trooper works—the citations of the
other troopers working in the same contexts as an individual trooper. That is, in addition to the
measures of context that we could examine as per the discussion in the last chapter, the individual
level data allow us another measure: the percentage of those cited by other troopers working in
the same areas as the trooper whose citations of African Americans we are evaluating. To arrive
at this contextual measure, we must first establish all of the areas in which a trooper issued
citations. We do that for two levels of analysis: the county area (roughly a third of a county) and
for what we call the county highway area (all citations on a highway within a county area, minus
the trooper-in-question’s citations). For example, if the other troopers issuing citations on U.S.
64 in a county area issue 25 percent of their citations to African Americans, then we would expect
that the trooper in question should issue citations to African Americans at about 25 percent. That
is, we assume that the 25 percent is a reasonable value measuring the racial composition of the
context in which the trooper works. It represents a baseline against which to compare the
individual trooper’s behavior. Note that below we will suggest a reservation that we have about
the use of the citations of the other troopers for the purposes of identifying possible racial
disparity—a concern that will lead us to suggest that the accident measures of context may be
superior. In a nutshell, our concern is that, if there is racial disparity in a particular trooper’s
127
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

citations, then troopers working in the same district may share in the same bias. This will be
discussed more below.
There are at least two options available to us in defining the metric of the dependent
variable in the mathematical model. We could define a dependent variable as the proportion of a
trooper’s citations issued to African Americans, and compare that to the proportion in the baseline
measures (for example, the proportion cited who are African American, or the proportion of
drivers in accidents who are African American). However, we have found that the analysis of
those dependent variables that are defined as proportions to be rather complex (Cohen and Cohen
1983:73-76). We opt instead to simply use the count of the number of African Americans issued
a citation by a trooper as the metric of the variable.
We will primarily rely upon ordinary least squares regression to model the number of
African Americans cited in the year 2000 by each trooper in the NCSHP (below, we present the
results for average troopers on the highway, minus any troopers holding higher ranks such as
sergeant or lieutenant). The choice of estimation technique reflects the distribution of the
dependent variables. Analyses in other contexts or at other levels of aggregation might have less
normally distributed outcomes and so require other estimation techniques. We will present two
models for each of four measures of context. The first model, which we call the deployment
model, will consist of the percentage of those cited or in accidents in the same county area or
county highway area, plus the proportion of citations issued by a trooper by hour of the day.
These measures constitute one baseline against which to compare the number of citations issued
by a trooper to African Americans.
Another measure expected to be related to the number of African American cited is the
number of whites cited by a trooper. Essentially, we are controlling here for the volume of work
128
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

done by a trooper (some issue many more citations than others). The more whites cited, the more
African Americans should be cited. Recall that we found that time of day was an important
determinant of the proportion African American on the highways and roads of North Carolina.
Troopers work different shifts and those who draw more night shifts are likely to issue a higher
proportion of their citations during nighttime hours, and thus, for example, would be expected to
have a higher number of citations issued to African Americans. We control for this by calculating
the percentage of all of a trooper’s citations occurring in each 2-hour period (that is, twelve 2hour periods).26 In the tables below, only two of the two-hour periods are represented (the others
were not found to be statistically significant and were dropped from the equations). We also
attempted to control for weekday versus weekend (specifically, number of weekday citations), but
that variable had to be dropped because it was too highly correlated with the number of whitescited variable. Finally, we added the percentage of the trooper’s citations issued on the interstates
and on rural highways, as preliminary analysis showed a tendency for African Americans to be
cited disproportionately on interstates and under-represented on rural highways. An individual
trooper patrolling disproportionately more on interstates than U.S. or N.C. highways would be
expected to have more citations of African Americans. A trooper working more on rural
highways would be expected to have fewer citations of African Americans.
In summary, our first model represents what might be called the deployment model. The
deployment model includes statistical controls for the area and highway types in which the

26

We arbitrarily divided the day into twelve 2-hour periods: 12:00 a.m. (midnight) to 2
a.m. was the first period, 2 a.m. to 4:00 a.m. was the second, and so on. The number of citations
of some of the adjacent 2-hour periods are correlated, as would be expected, since a trooper
issuing citations between 2 a.m. and 4 a.m., is probably also issuing citations between 4 a.m. and
6 a.m. We found the correlations to be modest, however.
129
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

trooper issues citations and the time of day he or she works. Note that we think it is reasonable to
call this model a deployment model—in that the variables are measuring spatial and temporal
contexts in which the trooper works. While it is possible that the choice of where one works and
at what time of day is not a racially neutral decision, we think it generally unlikely that racial
disparity would serve as a basis for deciding what district one works in, or shift of work.
Our second model includes all of the variables from the deployment model, plus several
individual level characteristics. These include such demographic characteristics of the trooper as
race (African American or white), age, and gender. It also includes the number of training
programs in which the trooper has participated (as a crude measure of professionalism), and how
long the trooper has been employed with the NCSHP. Finally, we add two measures of the
gender and age composition of a trooper’s citations. Specifically, we include the proportion of a
trooper’s citations issued to those under the age of 23 and the proportion issued to women. We
included these variables because, in some earlier analysis, we found some evidence that the rates
of citations for those under the age of 23 seemed to be rather high, suggesting to us that some
vehicular misbehaviors may be more prevalent among the young. If so, it is possible that some of
the citations of African Americans could be accounted for by the age composition of the trooper’s
citations. That is, the more young cited, the more African Americans cited. We also include the
gender composition variable to explore whether or not most driving misbehaviors occur among
men, and whether or not, by extension, there may be some targeting of African American men in
particular by the NCSHP. If so, we would expect the proportion of cites of women to be related
to the number of African Americans cited (negatively related). However, the percentage female
variable was found to be not statistically significant in any of the models, and so we dropped it
from the equations and tables to follow.
130
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Before discussing the results, it should be noted that the models presented below do not
include the variable “trooper’s age.” It was dropped from preliminary models when it was found
to be too highly correlated with the variable years of experience with the NCSHP. The decision
to drop one versus the other was arbitrary, and one could reasonably include age instead of years
on the force. There are no other multi-collinearity problems in the variables remaining.27 Also, it
should be noted that we have dropped all hours of the day from the equations except 2:00 a.m. to
3:00 a.m. and 10:00 p.m. to 11:00 p.m.. The other hours were dropped because they were not
found to be predictive of any dependent variable, and we wanted to fit the results on a single
page.
Results for the first two context measures are presented in Table 3.7. The first column
lists all of the independent variables in our regression analysis. The deployment model involves
the percent cited who are African American (cited by troopers other than the trooper in question)
within the same county areas as those worked in by the trooper in question. In the deployment
model, we see that for every white cited, .253 African Americans are cited. For every percentage
increase in those cited who are African American (cited by other troopers) 5.5 additional African
Americans are cited by troopers. In the parentheses are standardized coefficients, which are often
used to compare the relative magnitude of variables’ effects. Thus, for example, the contextual
measure of percentage of citations issued to African Americans by fellow troopers in the same
county area is the strongest determinant of the number of African Americans cited by a trooper
(.790). Stated another way, for each one unit change in a standard score of the percentage of
those citations issued to African Americans by fellow troopers, we would expect a .790 change in
27

Multicollinearity in the equation is minimized in part because we have centered all the
variables (subtracted the mean from each variable), so as to reduce the chances of multicollinearity with the constant in the equation.
131
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

the standard score of number of African Americans cited. If there is no asterisk next to the
standardized coefficient, neither the unstandardized nor standardized coefficients are statistically
significant at the .05 level. Thus, for those variables, we cannot safely rule out the possibility that
there is no effect of the variable. If a dash or hash mark is presented in the table, the variable was
not included in that model.
As for other results, we see that the proportion of citations issued in the late evening hours
(10 p.m.–11 p.m.) is associated with .936 more of an African American citation, while 2 a.m.–3
a.m. is associated with 1.138 more citations of African Americans. Thus, a trooper issuing a
higher proportion of his or her citations late at night would be expected to have more African
American citations than a trooper with a lower proportion. Other hours of the day were tested and
found to be statistically insignificant, and were excluded from the table (and from the equation
represented in the tables here), as they were not found to be significant in any of the models. We
also find that the more citations that a trooper issues on rural highways, the fewer citations he or
she issues to African Americans (the unstandardized coefficient is -.251, indicating that for each
percentage increase in the citations in rural highways, there is a reduction of .251 in the number
of citations of African Americans).
The deployment model with the first contextual measure accounts for about 70 percent of
the variance in the number of African Americans cited. On the one hand, that would be
considered by many to be a high proportion of the variance. On the other hand, it indicates that
30 percent of the variation cannot be accounted for by the deployment measures used here.
We turn next to the full model, which includes the deployment variables as well as the
individual level variables discussed above. The results of the full model are similar to that of the
deployment model. Note that we are examining this model for the purpose of better
132
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

understanding what might account for higher levels of African American citation behavior on the
part of the troopers. We are not trying to “explain away” any possible bias in the citation
behavior of troopers by controlling for these individual characteristics. For example, if we found
that white troopers issued more citations to African Americans than African American troopers,
then we could not use the predicted value from this full model as a basis for comparison.
Essentially, the predicted value would itself be biased. We have no intention of using the full
model for such purposes. Rather, our intention here is simply to see if there are any correlates or
determinants of African American citing behavior from these individual characteristics.
As it turns out, there is only one statistically significant determinant of African American
citations: the number of training programs in which a trooper participated. The more training (of
a variety of types) a trooper has, the fewer African Americans cited. This indicates some support
for our “professionalization” hypothesis: the more professionalized the trooper, the fewer the
African American citations. However, one should be cautious in making this interpretation,
since, as discussed earlier, age of trooper is related to the number of training programs in which a
trooper has participated. (Recall that age has been dropped from the equation because of high
multi-collinearity, primarily with the years of experience variable). Also, it may be that selection
could be an interpretation of the finding: those with more training are patrolling highways with
less traffic, and this is correlated with the racial composition of the highway not otherwise
measured in our model.28

28

The model assume linear relationships among the variables and may not adequately
capture the subtleties associated with where and when a trooper patrols. For example, a trooper
may patrol more in one county area than the other troopers, or more during specific hours of day
or even days of the week. Our model may be too crude to rule out such “selection”
interpretations.
133
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

In summary, the county area analysis variable—citations of African Americans issued by
others in a trooper’s work area—shows a strong relationship with the number of citations issued
to African Americans by a trooper. Controlling for workload (number of whites cited), we find
some hour of day effects, type of road patrolled effects, as well as effects of participation in
training programs.
Because there is a possibility that the trooper working in the same county areas as other
troopers may be working different highways, we also evaluate the effect of the percentage of
citations issued to African Americans at the county highway area level. This differs from the
previous analysis, because only those highways on which a trooper has cited someone are
evaluated comparatively with the citations of other troopers. That is, if a trooper has issued a
citation on U.S. 64 in a county area, then only those citations issued by other troopers on U.S. 64
in the same county area are used as a baseline for comparison. The results of the county highway
area analysis are presented in the fourth and fifth columns of Table 3.7.
The results are generally similar to those found for the county area analysis. Both the
contextual effect and the road-patrolled effects are similar, although in the county highway area,
the interstate patrolling is associated with fewer African Americans cited (a surprising finding
given our preliminary hypothesis that there would be more African Americans cited as
proportionately more citations were issued on the interstate). In the county highway area
analysis, the variance explained is somewhat higher (.715 and .717 across the two models) than in
the county area analysis. The number of training programs is again found to be negatively
associated with the number of African Americans cited by individual troopers. The magnitudes of
the effects that are statistically significant are similar to those of the earlier models. The only

134
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

substantive difference in the county highway model from the county area model is that only one
hour-of day effect was found (2 a.m.–3 a.m.).
One concern that we have about the use of the racial composition of citations of other
troopers to measure the context in which a trooper works, is that it may mask some racial
disparity due to the selection processes associated with the assignment and movement of troopers
from one patrol district to another. For example, it is plausible that if racial bias was to manifest
itself in attitudinal or behavioral displays, that the “principle of homogeneity” would operate:
“like attracts like.” Those troopers with bias might, over time, find themselves working with
others who are, like themselves, biased. If so, it would not be useful to compare the racial
composition of a trooper’s citations to his co-workers.
A more independent measure of context, yet one still available at the relatively small units
of analysis, is the racial composition of drivers in accidents (here we use accidents over three
years: 1998, 1999, 2000). Table 3.8 shows the results of the analysis using the percentage of
drivers in accidents who are African American as a baseline, and otherwise using the same
substantive models as in the earlier table. In the first column of coefficients, we show the results
of the deployment model and find that the results are similar to that of the earlier deployment
model. We find that early morning citations (2 a.m.–3 a.m.), and late evening citations (10
p.m.–11 p.m.) are related to African American citations. Here, however, we find that the higher
the percentage of a trooper’s citations on the interstate, the higher the number of African
Americans cites. Note, too, that the explained variance is less than we found using citation racial
composition as a context measure. Here it is approximately .60, compared to .70 using the
citation contextual measures.

135
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

In the full model, we find that white troopers are less frequently citing African American
drivers, as are those troopers who have been employed longer (net of the effects of other variables
in the analysis). Also, the higher the percentage of those cited who are younger than 23 years of
age, the fewer African Americans cited. The latter finding is contrary to our initial hypothesis
that there might be some targeting of younger African Americans. Here we find that the younger
the composition of those cited by a trooper, the fewer African Americans cited. As for the white
trooper effect, it could be construed as evidence of avoiding citations of African Americans
(perhaps due to concern over possible scrutiny by the media, legislature, research evaluators, or
the public in general). It could also be explained by possible selection effects that are not
controlled for statistically in the model. African American troopers, for example, could be
deployed to highways that disproportionately have African American drivers. The results of the
full county highway area model support the latter interpretation, because the white trooper
variable is no longer statistically significant when we conduct analysis at the county highway area
level.
Turning then to the county highway area accident measure of context (the percentage of
drivers driving in accidents who are African American), we see that the explained variance rises
to .65 (approximately). The deployment model is about the same as before, but the full model’s
effects are somewhat different because the individual level measures that previously were
statistically significant (race of trooper and percent younger that 23 years of age) are no longer
significant. Thus, there may be some value in using the county highway area as a unit of
measurement as it lends plausibility to the interpretation that the previously seen white-trooper
effect could very well be a deployment effect. Note that the hour of day effects observed for the
analysis (using the county area accident measure) remain intact for the county highway area
136
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

analysis: early morning (2 a.m.) and late evening (10 p.m.–11 p.m.), in both the deployment and
full models, and similarly for the percentage of a trooper’s citations on the interstates. The only
individual level measure that remains statistically significant between the full model of the county
area and county highway area is the percentage of those under the age of 23 years old, which has
a negative effect. Thus, somewhat surprisingly, the higher the proportion of younger people
cited, the lower the proportion of African Americans cited.
In summary, the results of the analysis of Tables 3.7 and 3.8 indicate that virtually the
entire variance that can be explained with the available measures in the number of African
Americans cited by individual troopers, can be explained by deployment factors.29 The
deployment factors we find to be statistically significant include the racial composition of
accidents in the same context, other troopers’ citations in the same county area or same county
highway area, the percentage of citations issued on interstates or on rural highways, the number
of whites cited, and the hour of day in which citations are written. Of the two generic types of
context measures (citations and accidents), accidents probably represent a more independent
measure of context because the citation measures may mask the extent of bias due to possible
clustering of biased troopers in the same districts. Of the county area and county highway area
models, the latter model is probably preferable, since spatial heterogeneity issues are likely less of
an issue within the smaller unit of aggregation (aggregating or adding together all the accidents in
the county highway areas where a trooper issued citations in 2002.30 In the county highway area
29

We do not have, of course, direct measures of the troopers’ level of cognitive bias or
racial animus, so some would argue that our test for disparity might lead to different conclusions
with such measures.
30

However, a more sophisticated analysis could be done with controls for autocorrelation problems due to the fact that the contexts are not measured independently since the
same accidents recur across the areas where a trooper issues citations. That is, it is likely that the
137
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

analysis, only deployment factors are found to be statistically significant. As such, the results are
very similar to the county highway area analysis using other troopers’ citations to establish a
baseline of percent African American against which to compare an individual trooper’s African
American citations, except in the latter we found that the number of training programs was also
statistically significant.
The equations that we have been evaluating are somewhat crude, in that some of the
assumptions of ordinary least squares regression analysis have probably been violated
(specifically, lack of independence in the error terms of the equation, causing what is known as
auto-correlation). Yet, the analysis does propose a general method that can be used and improved
upon for the purpose of determining whether or not individual characteristics or contexts are
important in determining the racial composition of a trooper’s citations. However, we have one
more methodological contribution to explore that more specifically addresses the question of
whether individual troopers stand out from his or her peers in his or her citations of African
Americans: residual scores.

Identification of Troop Districts with Troopers as Statistical Outliers
We now demonstrate how the models, summarized in Tables 3.7 and 3.8, could be used to
begin to flag the presence of troopers who might be out of line with their fellow troopers in their
troop district (fifty-three troop districts). Because the models used here are presented in order to
demonstrate a method rather than to identify specific individuals, we caution the reader against
coming to any conclusions about the actual presence of racially biased troopers in any troop.

error terms in the equations suffer from auto-correlation because troopers in the same districts
will tend to have similar values in the error term of the equations.
138
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Rather, the figures below are presented to show a method with which to identify troopers who are
“outliers”—those who have more citations of African Americans than would be expected using
the models. It should be noted that the determination of who is an outlier will vary somewhat
depending on the statistical models employed. At best, these models should be seen as providin
information that may be weighed by decision makers and compared to other information to help
diagnose the presence of racial disparity in a troop.
In demonstrating the method, we will utilize the deployment model that uses the accident
measure of context at the county highway area level. By doing so, we only use deploymentrelated variables (percentage of drivers in accidents who are African American, number of whites,
type of highway, and hour of day measures), and not such individual-level variables as race or
gender. Therefore, an outlier will be defined as such relative to deployment factors alone.

139
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 3.7 Number of African Americans Cited by Citation and Other Contextual- and
Individual-Level Variables, Models Control for Racial Composition of Context by Using
Citations of Other Officers in Same Context (Regular Troopers Only; N=925; Standardized
Betas in Parentheses)
1.Model Includes Measure at County
Area Level
Deployment

2. Model Includes Measure at County
Highway Area

Full

Deployment

Full

N Whites
Cited by Trooper

.253
(.482)**

.245
(.467)**

.255
(.486)**

.249
(.474)**

Percent Cited Who Are African American
(Cited by Other Troopers) at County
Area Level

5.522
(.790)**

5.460
(.782)**

-

-

Percent Cited Who Are African American
(Cited by Other Troopers) at County
Highway Area Level

-

-

5.569
(.806)**

5.533
(.801)**

Percent Trooper’s Cites 2 a.m.–3 a.m.

1.138
(.042)*

1.103
(.041)*

1.104
(.041)*

1.088
(.041)*

PercentTrooper’s Cites 10 p.m.–11 p.m.

.936
(.042)*

1.101
(.049)*

.674
(.030)

.799
(.036)

Percent Trooper’s Cites on Interstate

-.085
(-.023)

-.075
(-.020)

-.176
(-.047)*

-.159
(-.042)*

-.251
(-.040)*

-.267
(-.043)*

-.252
(-.040)*

-.271
(-.043)*

Trooper White

-

-3.477
(-.015)

-

-2.229
(-.009)

Trooper’s Years On Job

-

.186
(.014)

-

.206
(.015)

N of Trooper’s Training Programs

-

-2.348
(-.064)*

-

-2.229
(-.061)*

Percent Trooper’s Cites of Those Less
Than 23 Yrs. Old

-

-.286
(-.023)

-

-.122
(-.010)

111.057

114.433

110.850

113.154

.697

.700

.715

.717

PercentTrooper’s Cites on Rural
Highway

Constant
Adj. R2

140
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 3.8 Number of African Americans Cited by Accident and Other Contextual and
Individual-Level Variables, Models Control for Racial Composition of Context Using
Accident Data (Regular Troopers Only; N=925; Standardized Betas in Parentheses)
3. Model Includes Measure at County
Area Level
Deployment

Full

4. Model Includes Measure at County
Highway Area
Deployment

Full

N Whites
Cited by Trooper

.246
(.470)**

.228
(.435)**

.257
(.490)**

.243
(.462)**

Percent Drivers African American in
Trooper’s County Area Level

6.434
(.715)**

6.139
(.682)**

-

-

Percent African American Of Those
Cited by Other Troopers at County
Highway Area Level

-

-

6.805
(.753)**

6.590
(.730)**

3.221
(.120)**

2.941
(.109)**

2.872
(.107)**

2.700
(.100)**

1.270
(.057)*

1.404
(.063)*

1.317
(.059)*

1.411
(.063)*

.327
(.088)**

.286
(.076)**

.285
(.076)**

.266
(.071)**

.071
(.011)

.035
(.006)

.076
(.012)

.045
(.007)

Trooper White

-

-10.266
(-.043)*

-

-5.507
(-.023)

Trooper’s Years On Job

-

-.671
(-.050)*

-

-.439
(-.033)

N of Trooper’s Training Programs

-

-.914
(-.025)

-

-1.104
(-.030)

Percent Trooper’s Cites of Those Less
than 23 Yrs. Old

-

-.880
(-.071)**

-

-.596
(-.048)*

111.388

120.179

111.256

116.135

.601

.610

.652

.656

Percent Trooper’s Cites 2a.m.– 3 a.m.
PercentTrooper’s Cites 10 p.m.--11 p.m.
Percent Trooper’s Cites On Interstate
PercentTrooper’s Cites on Rural
Highway

Constant
Adj. R2

Figure 3.1 shows the box-plots for the first nine districts (of fifty-three) in North Carolina,
in which the difference score has been calculated between the number of African Americans cited
by a trooper on a highway type in a district, and the expected number of African Americans he or

141
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

she “should have” cited as predicted by the deployment regression model. We define outliers as a
proportion of all citations written by a trooper. Thus, most troops show a range, within which
nearly all troopers do not have greater than 20 percent more citations of African Americans than
they “should,” relative to the number of citations of African Americans the model indicates they
would be expected to have (based on the deployment factors). (These nine districts in Fig. 3.1
have been chosen and assigned arbitrary identification numbers, 1 through 9, to protect the
anonymity of the district and the troopers.)
The box plot has four parts. For example, District 1’s box plot has a dark line
representing the median value, which is close to zero. That is, in District 1, the median value is
close to the expected value of the number of African Americans cited. The dark red area31
represents the inter-quartile range of values (all the observations between the twenty-fifth and the
seventy-fifth percentiles). The small “T,” or so called “whiskers,” off of the top and bottom of
the inter-quartile range, represents the range of the observed values (minus what are defined as
“outliers”). Outliers are defined as observations more than 1.5 box lengths from the top or bottom
edge of the box. We limit the analysis here to regular troopers who have more than seventy-five
citations issued in the year 2000, thereby to avoid classifying a trooper as an outlier based on a
small number of observations (roughly 97 percent of the regular troopers remain in the analysis).
In District 4, there is one positive outlier (with relatively many citations of African Americans),
and in District 7, there are two troopers who are negative outliers (having about 40 percent fewer
citations of African Americans he or she “should” have relative to the predicted value from the
model).

31

Of course, if you are looking at a copy that was not printed on a color printer, the color
will be a dark gray.
142
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

As can be seen in a review of the nine districts presented in Fig. 3.1, the ranges of the
number of African Americans cited varies considerably from one district to another, with District
2 having a wide range and relatively high range (from approximately 5 percent above, to 20
percent below the median number of African Americans cited). In Figures 3.2 through 3.7, each
of the fifty-three districts’ citation patterns is shown. In general, the results indicate that there are
many outliers, both positive and negative, across the districts. Specifically, Districts 4, 11, 12, 22,
23, 28, 30, 35, 38, 39, 41, 42, 43, 44, 45, and 49 have at least one positive outlier, and Districts 7,
10, 11, 13, 16, 17, 18, 19, 20, 23, 25, 26, 33, 34, 45, 48, 49 and 51 have at least one negative
outlier. In all, eighteen districts are found to have positive outlier values, while twenty-five
districts are found to have negative outlier values.
An examination of these outliers reveals that the positive outliers are somewhat more
likely to involve citations on interstate and U.S. highways, while negative outliers are
disproportionately from rural highways. Positive outliers were found (analysis not presented here
in table form) to be somewhat less experienced troopers and troopers who work weekends. In
examining the citations of specific troopers, we realized that further fine-tuning of the regression
models could be possible. Although we control for such factors as type of highway and time of
day, the variables are crudely measured. For example, some of the rural highways pass near
mostly African American communities within larger, mostly white areas. While in general, rural
highways in North Carolina would be expected to be associated negatively with the percentage of
drivers who are African American and involved in accidents, there are many rural African
American communities in North Carolina. Thus, some troopers who happen to issue a high
number of citations to African Americans do so on rural roads proximate to communities with
relatively high percentages of African American residents. If such roads happen to be relatively
143
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

safely-constructed rural highways, the accident rate could be low, possibly making the estimate of
the percent African American among those in accidents unreliable. Obviously, this points to the
need to conduct further analysis and to refine the regression model with more information.
It is also possible that the mechanism by which large numbers of African Americans are
cited may have a lot to do with the “citation zones” discussion in the previous chapter. Many of
the positive outliers were noticed to patrol more often on interstate and busy U.S. highways than
the negative outliers did. This is suggestive of the important role “citation zones” may play in
creating a positive outlier. That is, an individual trooper may “work a zone” frequently over the
course of a year and develop a large number of African American citations because he or she
works a selected area where there are a relatively high percentage of African American drivers on
the highway (and not often involved in accidents). It is also possible, however, that the trooper is
biased—perhaps both in the selection of vehicles to pull over, or in the selection of an area where
he or she will “work the zone.”

Conclusions
In conclusion, we have shown in this chapter that it is possible to model the citations of
African Americans at the individual officer level, and to do so with considerable success by
conventional standards of explained variance. Although more complex statistical methods can be
applied,32 the models presented show that statistical outliers can be found in some troops but not
in others. (Our experience with some of the more complex modeling techniques is that although
they can improve on ordinary least squares regression models, the latter will be a reasonable

32

Other techniques include hierarchical linear modeling, Poisson or extra-Poisson
regression, or models that control statistically for the auto-correlation problem.
144
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

approximation of the results found using the more sophisticated models). Our goal here is to
show the methodological principles of baseline and individual covariates, as determinants of the
number of African Americans cited by individual troopers. The techniques show the extent to
which a trooper’s individual citation behavior varies relative to that of other troopers. Outliers
can be neatly displayed in boxplots, and further analysis can be done to help verify whether other
considerations, such as specific highway patrolled, frequency of patrolling, and excessive use of
speed zones, are mechanisms that generate some of these high positive outliers.
Finally, it should be reiterated that the determination of disparity cannot be equated with
the determination of bias. Other pieces of information should be gathered to supplement any
statistical analysis similar to what has been done here. Nevertheless, the analysis here, although it
could be improved upon with further considerations of relevant factors, does demonstrate that a
method can be used to provide yet another indication of what behaviors are occurring in the day
to day actions of individual troopers.

145
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 3.1 Box-plots of Districts 1 to 9, (Arbitrary ID Number) With Positive and Negative
Outliers (Defined as Percent of Trooper’s Citations)

146
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 3.2 Box-plots of Districts 10 to 18, (Arbitrary ID Number) With Positive and
Negative Outliers (Defined as Percent of Trooper’s Citations)

147
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 3.3 Box-plots of Districts 19 to 27, (Arbitrary ID Number) With Positive and
Negative Outliers (Defined as Percent of Trooper’s Citations)

148
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 3.4 Box-plots of Districts 28 to 36, (Arbitrary ID Number) With Positive and
Negative Outliers (Defined as Percent of Trooper’s Citations)

149
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 3.5 Box-plots of Districts 37 to 45, (Arbitrary ID Number) With Positive and
Negative Outliers (Defined as Percent of Trooper’s Citations)

150
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 3.6 Box-plots of Districts 46 to 53, (Arbitrary ID Number) With Positive and
Negative Outliers (Defined as Percent of Trooper’s Citations)

151
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

References for Chapter Three
Cohen, Jacob and Patricia Cohen. 1983. Applied Multiple Regression/Correlation
Analysis for the Behavioral Sciences. Second Edition. Lawrence Erlbaum Associates,
Pub.: Hillside, New Jersey.

152
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Chapter 4 Searches by the North Carolina State Highway Patrol

One of the central issues in many of the national accounts of racially biased policing has
been the consideration of the race of a driver as part of a profile used in vehicle searches for drugs
and other contraband. It is clear that the U.S. Drug Enforcement Agency (DEA) and Federal
Highway Administration have, in the past, offered drug interdiction training to local police forces
suggesting, implicitly or explicitly, that race or ethnicity—along with other vehicle or driver
characteristics—might be used to decide whom to stop and search. There have been a number of
cases around the country documenting that state law enforcement agencies search many more
minority drivers than would occur in the absence of such profiles. To the extent that race is used
as a profiling tool to identify potential carriers of drugs or other contraband, we would expect to
find large disparities in the odds of being searched between African American and white drivers.
Racial profiling is a practice in which a police organization generates or uses a profile
meant to describe a typical offender where that profile includes race as one of the criteria. Racial
profiling—although at first glance, it may seem a useful tool for police work—is a form of
institutional discrimination or institutional racism. Institutional racism refers to organizational
practices which produce racial inequality. In policing, there seems to be at least one area in
which explicitly race-sensitive institutional rules are still used. This is the practice of developing
offender profiles—typical characteristics of drug couriers being a prominent example. There is
little question that the DEA generated such profiles in the past as part of the war on drugs, and
that these profiles include race and ethnicity, among other characteristics, which have been
promoted as useful factors in deciding which cars to stop and search for drugs.

153
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

The exposure of racial profiling by drug-seeking state police in both New Jersey and
Maryland played a prominent role in bringing the issue of “driving while black” to national
attention. Racial profiling, when it exists, is a fairly specialized police practice. Many troopers
never search cars for drugs. Where racial profiling is part of an organizational routine, we would
expect there to be evidence of high levels of racial disparity in searches. We would also expect
the practice to be embraced by most, if not all, troopers doing proactive searches. By proactive
searches, we refer to actual use of the profile, resulting in a traffic stop to look for contraband.
Sometimes troopers observe contraband (or indicators of contraband—smells, drug
paraphernalia). In this case, the trooper is confronted with direct evidence prior to the search and
is free and justified to initiate a probable cause search. Troopers who are looking for contraband,
but do not have probable cause, must ask drivers to consent to a search.

Race and Searches by the North Carolina State Highway Patrol
As early as 1996, journalists investigated racial disparity in searches by the NCHSP.
They reported that the group of troopers specifically assigned to drug interdiction, the Criminal
Interdiction Team (CIT), stopped African American men at twice the rate of other NCHSP
troopers patrolling the same areas. They also reported that of the 3,501 vehicles searched by the
CIT in 1995, contraband was found in only 210 vehicles (Neff and Smith 1996). In this chapter,
we look at search activity by the NCSHP from 1997 to 2000, and pay particular attention to
searches by the CIT.

154
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Troopers’ Accounts of Search Behavior
We spoke with both CIT and regular troopers as part of this project in early 2001. The
CIT troopers were adamant that they did not consciously use racial profiling in deciding whom to
stop or search. Instead, it was their practice to aggressively enforce the traffic laws of North
Carolina. They did describe the use of other characteristics of drivers that made one car more
suspicious than another, and therefore deserving of more attention, and they did note that when
they turned on the blue lights, there was some selectivity involved . This might include legallyreasonable selectiveness of drivers’ actions, including more than a single violation, a serious
violation, or an unsafe movement among a pack of vehicles. Other factors might include more
benign indicators that come to troopers’ attention, such as loud music, stickers referring to music
groups connected to the drug culture (such as the Grateful Dead or Phish), or cars about which
they felt “something” was out of place, or about which they had a “gut feeling.”
When troopers gave examples of the use of indicators (we would refer to them as
stereotypes) at work, they tended to be drawn from both white and African American youth or
sub-cultures. The allusions to African Americans were far more general—rapper music or “the
way young African American males dress”—than the examples given for whites—Phish follower,
tie die shirts, biker with a ponytail. From the troopers’ accounts, we have little reason to expect
racial profiling to be going on in a self-conscious way as late as 2001. Still, there is little reason
to ignore the potential for stereotyping and cognitive bias to influence face-to-face highway
encounters. Further, the stereotypes used to describe suspicious African American drivers tend to
be much broader and more diffuse than the stereotypes more likely to apply to whites.
CIT troopers reported that they typically use a fairly standard method of questioning after
a stop that is designed to see if the driver had a plausible and consistent story, and also to
155
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

determine if the driver became less or more nervous across the course of the conversation. It is
assumed that the non-combative and friendly general conversation should relax the driver.
Troopers said a search would likely follow in three general scenarios: if there was probable
cause—for example, visible contraband or the smell of marijuana; if the conversation resulted in
too many inconsistencies; or the driver was still visibly nervous after the initial interview. In the
latter case, the interview method is a tool that produces reasonable suspicion that something
might be wrong. An excessively nervous or confused driver is viewed as more likely to be doing
something wrong, therefore providing troopers with reasonable suspicions, and hence, a good
choice for troopers to request a consent to search.
The CIT members talked about writing a lot of warning tickets. Someone who was still
nervous after having been written a warning ticket often heightens the suspicions of a trooper,
because warning tickets are viewed as an outcome of the encounter that should relieve most
citizens’ anxiety. Although the CIT’s primary task (and pride) centers on drug interdiction, the
troopers made it clear that their interview protocol was cued to nervousness and body
language—therefore making it a useful tool for identifying felons, people with outstanding
warrants, or other violators. Some troopers in the focus groups recognized that the use of
nervousness might produce unreliable results, but the general tone was one of great faith in their
methods of interdiction by conversation. In fact, the use of the interview method was seen as
counteracting any personal stereotypes (biases) one might have about drivers. The troopers also
viewed their leadership to be intolerant of any racial bias.
There was general agreement among the CIT members that certain cues are used to
determine whether or not to gather additional information, and that these cues, in their totality,
made some people seem more suspect than others. (We have referred earlier to decision making
156
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

processes based on incomplete information as a stereotyping process.) These troopers believed
that their high standards for determining when to search (they noted that “articulable and
reasonable suspicion” is required before a CIT member can request consent to search), their
meticulous attention to recording information in detail, and their aggressively “by the book
sergeant” have precluded racism or stereotyping among the CIT.
We also discussed searches with regular road troopers. We did this, in part, because we
had been struck by the extremely small number of searches attributed to them. We found
searches among this group to be so rare as to simply not be part of a trooper’s normal routine.
The troopers we talked with were only a small fraction of the 1,200 or so troopers patrolling the
state highways, but to an individual they either claimed to do no searches, or claimed to search
only reluctantly. At least part of that reluctance was attributed to safety concerns, because
troopers patrol alone. A search would occur, they said, when contraband was visible or the driver
had to be searched incident to arrest. Searches extend contact with a driver, and are thus seen as
dangerous, because they could lead to a confrontation with the driver or occupants. Additionally,
if contraband, especially of the type and volume they might be likely to encounter, was to be
discovered, it would lead to a great deal of undesirable paper work and potential time in court.
The regular road troopers we talked to were not enthusiastic about searches in the least, and it is
clear that they view unnecessary searches, in general, as a nonproductive use of their time. It also
seems clear that the threshold at which a “cue” or “indicator” would rise to the level of reasonable
suspicion is high for regular road troopers.
Given their accounts, we would expect that searches would be rare among regular
troopers. This expectation will be confirmed shortly. In addition, the accounts given by the CIT
members seem consistent with practices that could produce some racial disparity in searches, but
157
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

not wholesale racial profiling. The CIT accounts look more like cognitive bias processes, perhaps
exaggerated by the routine use of what we refer to as stereotypes, to decide whom to stop but not
whom to search. Since the interview method is an interactional and a conversational
accomplishment, it seems reasonable to suspect that the cues and indicators (stereotypes) built
into the process, and cognitive bias—both individual and organizational—might influence not
only the interaction, but also the interpretation of the interview’s outcome in ways that
disadvantage some, but not all, of the persons stopped. Of course, the conversational method is
biased inherently against anyone who is nervous, confused, or belligerent. We know from prior
research and Chapter 6 of this report that African Americans have lower levels of trust in the
police. This may lead in some instances to nervousness in its own right.
Throughout the discussion below, a distinction will be made between consent searches
and probable cause searches.33 For a consent search, as the name suggests, a search cannot be
conducted if the person in question refuses to grant permission for the search. Yet there must be
grounds for suspicion of the person. For a probable cause search, no such permission is needed as
the trooper has seen sufficient evidence to warrant the search, such as drug paraphernalia or a
weapon. However, in our conversations with CIT troopers, it was stated that often it was the
practice in a probable cause situation for a trooper to ask for permission for the search since a
consent search would lessen the subsequent chance that the trooper would be accused of
33

We ignore searches incidental to an arrest because we assume that if there is sufficient
cause to arrest someone, the search of the person is perfunctory from the point of view of
whether or not there is disparity. While there may be some bias in arrest procedures, it is simply
beyond the scope of this report to assess such processes (as we have no data to assess). We
think, however, that there are some checks and balances in the criminal justice system regarding
cases brought to court with insufficient evidence. Presumably a trooper who repeatedly arrests
without sufficient cause would come to the attention of prosecutors and of his or her superiors.
While such checks and balances are undoubtedly imperfect, it is simply beyond the scope of this
research to address such cases.
158
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

fabricating the probable cause evidence (also, by asking permission to search, the trooper might
be perceived as being more respectful and therefore less likely to antagonize the suspect).
Whether or not the trooper recorded a specific event as a probable cause or a consent search is not
clear. Still, we suspect that during the years of this study, some searches recorded as consent
searches were in fact probable cause searches. Unfortunately, any shifts of this sort make it
difficult to evaluate trends in probable cause and consent searches.
To the extent that the troopers we interviewed capture general NCSHP activity, the above
discussion suggests that regular road troopers should record a higher proportion of probable cause
searches—because of their general reluctance to search and their higher threshold for articulable
suspicion. That would lead us to expect them to also record higher proportions of searches
resulting in the seizure of contraband. The CIT troopers, on the other hand, note that searches are
based on the results of the interaction with the driver (articulable and reasonable suspicion
including nervousness and other signs related to the car or driver), and of course, probable cause,
such as visible contraband. Thus, we might expect a higher proportion of consent searches and a
lower rate for identifying contraband among the CIT.

Recorded Search Activity
In this section we analyze the incidence and distribution of searches recorded by the
NCSHP. We contrast the activity of the CIT and other NCSHP troopers during the 1997–2000
period. We think it is important to recognize that in 1996, the CIT came under public scrutiny for
racial profiling in searches. It is reasonable to suspect that the fairly dramatic change in search
behavior in both the CIT and among other troopers soon after was, in part, a response to that
scrutiny.
159
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 4.1 displays separate time trends for recorded searches by the CIT and all other
NCSHP troopers. As shown, the CIT conducts substantially more searches than other troopers.
In 1997, for example, the CIT recorded 769 searches, while the largest other troop recorded sixtytwo searches. In 2000, the comparable numbers were 387 and 110. There was a rapid decline in
searches by the CIT across the period. The number of troopers assigned to the CIT declined
across this period as well, dropping from twenty-five active troopers in 1997, to thirteen in the
year 2000. These troopers were averaging thirty-one searches per year in 1997, and thirty in
2000. Outside the CIT, many troopers record doing no searches in a typical year. Many of the
troopers who do report searches average only one or two per year. Regular road troopers slightly
increased their volume of searches in 1998. There was a small decline in searches by regular
troopers after 1998.

160
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 4.1 Searches by CIT and Non-CIT NCSHP Troopers

161
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 4.2 shows the time trends in the proportion of all recorded searches of African
Americans. There was a sharp decline in the proportion African Americans searched in 1998 for
both CIT and non-CIT troopers. The percent African American increases somewhat for both nonCIT troopers in 1999, and for CIT troopers in 2000. As we have seen in earlier chapters, about 20
percent of drivers are estimated to be African American. In 1997, 46 percent of searches by the
CIT were of African Americans and slightly more than 30 percent of searches by non-CIT
troopers were of African Americans.
In order to ascertain if this rate of searches of African Americans is excessive, we need to
establish a baseline of drivers at risk to be searched. Because drivers are searched after they are
stopped, a reasonable baseline is the racial composition of drivers stopped. The baseline used in
Table 4.1 and Figure 4.3 is measured at the troop level. Except for the CIT, other troops are
identified by a fictitious identification number. This baseline is potentially misleading to the
extent that the few troopers who actually conduct many searches may patrol in areas within their
troop’s territory with higher or lower concentrations of African American drivers. This potential
problem is not present for the CIT, since all troopers actively search cars. Later, we repeat these
analyses at the individual trooper level and use the trooper’s own distribution of citations and
warnings as the baseline to compare with the racial distribution of searches.

162
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 4.2 African Americans as Percent of Searches, CIT and Non-CIT
NCSHP Troopers
.5

Percent of Searches

.4

.3

.2
Other NCSHP
.1
1997

CIT TROOP
1998

1999

163
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

2000

We use combined data on the racial composition of warnings and citations for 1998–2000
to establish a baseline of drivers at risk to be searched. For 1997, only citations are available, and
so we use the racial composition of citations as our baseline estimate of who is at risk to be
searched. As we saw in Chapters 1 and 2, the racial composition of warnings and citations tends
to be very similar. In previous chapters, we have examined variation in racial disparity for the
fifty-three troop districts in North Carolina. Search events are much too rare to be analyzed at
this level. We use instead the eight regular NCSHP troops plus the CIT as the organizational
units. Table 4.1 displays the racial composition of drivers stopped and searched for these nine
troops for 1997. Many troops recorded very few searches in 1997. In all troops except Troop 7,
African Americans are a higher proportion of searches than they are of drivers stopped and cited.
In all troops, including 7, whites were searched at lower rates than they were stopped in 1997.
The next to last column displays the relative odds of being searched after being stopped
and compares African Americans and whites ([African American searches/stops] / [white
searches/stops]). For all troops in 1997, African Americans have higher odds of being searched
than do whites, although the magnitude of this increased risk varies tremendously across troops.
At the low end, in Troop 8, African Americans were 1.27 times more likely than whites to be
searched after a stop. Troop 7 is almost as low, at 1.33. These two troops record very few
searches. At the other end of the distribution for 1997, CIT members searched African Americans
who had been stopped at more than four times the rate of whites. Although only 48 searches were
recorded by Troop 1 during 1997, African American drivers were searched at four times the rate
of white drivers. For 1997 across the NCSHP, there is substantial racial disparity in searches.

164
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 4.1. Comparisons of Baseline Estimates of Drivers Stopped (Cited) and Drivers Searched for African Americans and
Whites in North Carolina by Troop, 1997
Troop

Percent African

Percent African

Percent

Percent

OddsRatio

Number of

American

American

White

White

(African American

Searches

Stopped

Searches

Stopped

Searched

Searches/Stops)
/(White
Searches/Stops)

CIT

22.4

46.2

72.1

35.8

4.15

769

1

5.3

18.8

90.5

70.8

4.01

48

2

15.3

40.0

78.3

60.0

3.41

5

3

28.3

46.2

65.4

42.3

2.52

26

4

8.0

17.7

87.4

77.4

2.50

62

5

24.8

38.1

67.8

42.9

2.43

21

6

31.5

48.1

63.8

51.9

1.88

52

7

33.8

33.3

56.5

41.7

1.33

12

8

25.9

28.0

61.0

52.0

1.27

25

One of the most striking things about Table 4.1 is that, with the exception of the CIT, the
NCSHP does not routinely search vehicles. At the organizational level, if racial profiling
accounted for the observed disparity in 1997, regular road troopers were not involved. While
individual troopers in these units may have been using a racially influenced drug offender
profiling method, troops as a whole were simply not in the search business.
Because there have been dramatic changes in the search behavior of NCSHP troopers
since 1997, it is appropriate to see if these racial disparities in search rates are stable over time.
Figure 4.3 displays the time trends in the relative odds of African American to white searches for

165
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

the four NCSHP troops recording thirty or more stops in two or more years. Across the whole
period, the vast majority of troops record very few searches.

Figure 4.3 Trends in Relative Odds of Black/White
Searches, Troopers with 30 or More Searches for
Two or More Years
7.0

Odds Ratio

6.0
5.0

6

4.0

1

3.0

3
CIT

2.0
1.0
0.0
1997

1998

1999

166
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

2000

In addition to the CIT troop (I), only a single troop was heavily involved in searches
across the years. The time trends are quite dramatic for all four troops. In 1998, there is a sharp
drop in the relative odds of African Americans being searched for all four troops with thirty or
more searches. By 1999, Troop 6 recorded only twenty-four searches and the time series ends.
Clearly the NCSHP not only drastically reduced the number of searches it conducted in 1998, it
also searched African Americans at lower rates. It is still the case across the entire period,
however, that for the few troops with substantial search activity (as defined here), the NCSHP
searches African Americans at higher rates than it stops them.
The patterns are not the same for each troop. The CIT’s racial disparity in searches drops
dramatically from 1997 to 1998, and plateaus with an African American/white odds ratio of about
2:1. Prior to 1998, the CIT had a very high racial disparity in search rates that, on its face, is
consistent with a practice of racial profiling. By 1998, its level of racial disparity was much
smaller and was also lower than the other two troops examined. This pattern looks very similar to
the time trends for the CIT reported in Figures 4.1 and 4.2. After 1997, the CIT searched fewer
drivers and especially fewer African American drivers. The disproportionate searches of African
Americans by the CIT reported in the press in 1996 and still present in 1997, quickly eroded in
1998. This suggests a dramatic shift in selection criteria for whom to search, with race playing a
much smaller role after 1997.
In 1997, African Americans stopped by troopers in Troop 1 were four times more likely to
be searched than were whites who had been stopped. Although these odds dropped dramatically
in 1998, they rose to three times the white odds in 1999 and 2000.

167
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Troop 3 shows an even more dramatic pattern. In 1997, Troop 3 recorded less than thirty
searches. Therefore, its time series begins in 1998, with an odds ratio just below 2.This then
climbs slightly to 2.3 in 1999, and jumps to 6.5 in 2000. In the year 2000, African Americans
account for 44.1 percent of the cars searched by Troop 3, but only 28.3 percent of those stopped.
While this is a substantial disparity, more striking is the very low rate of white searches. While
whites made up 69.4 percent of stops by the troopers of Troop 3, they account for only 14.7
percent of those searched. People of other races were 10 percent of stops and 32 percent of those
searched by troopers in Troop 1.
There remains evidence that African Americans are searched at twice the rate of whites by
troopers of the CIT in the year 2000, but this represents a dramatic decline since 1997. The
much lower racial disparity in searches is consistent with the abandoning of racial profiling
practices, or increased reliance on nonracial driver indicators by the CIT, given the accounts of
how the CIT make their search decisions. The remaining disparity could easily be generated by
stereotyping and the use of the conversational method for generating suspicion prior to asking for
consent to search. It is also possible that the remaining disparity is produced by some non-bias
process that we have not accounted for.
The other two troops showing substantial search activity in 2000, search African
Americans at higher rates relative to whites who have been stopped. Because very few of the
troopers in these troops ever search a car, it does not suggest racial profiling at the troop level.
Individual troopers could be racial profiling, however.

168
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 4.4 Probable Cause as a Proportion of CIT and Non-CIT
Searches, 1997-2000
.4

Proportion of Searches

.3

.2

.1
Non-CIT
0.0
1997

CIT
1998

1999

169
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

2000

Probable Cause versus Consent Searches
Consent searches have been identified in other jurisdictions as the most likely search type
to exhibit racial disparity. Certainly the visible presence of contraband leading to probable cause
leaves little room for bias, although, of course, some situations attributable to probable cause may
be more subjective (for example, a strange odor). When a trooper’s “sense” about drivers,
including nervousness, influences a request for consent to search, much more room for bias is
provided.
Figure 4.4 compares the proportion of probable cause searches for both the CIT and other
NCHSP troopers. In 1997, less than 10 percent of searches were attributed to probable cause.
Probable cause searches rose rapidly across the period. By 2000, 37 percent of CIT searches and
20 percent of other troopers’ searches were based on reported probable cause. Based upon our
interviews, we had expected to find that searches by regular troopers would be more likely than
those by the CIT to have probable cause attributed to a search. This was not the case. In 1977,
both the CIT and regular troopers who recorded searches rarely reported probable cause (10
percent). After 1998, the CIT was more likely than the few regular troopers conducting searches
to record probable cause for a search. Because regular troopers rarely search, it may be the case
that probable cause searches are under-recorded. It may also be the case that the CIT have made
a greater effort to selectively search and, in so doing, are less inclined to request consent without
substantial reason. Figures 4.5 and 4.6 compare the racial/ethnic proportion of probable cause
searches across the same period.

170
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 4.5 presents searches initiated by CIT members. In 1997, African American and
white drivers are equally unlikely to have probable cause attributed to the search. After 1997, for
both white and African American drivers probable cause becomes more common, although the
line is much steeper for African Americans. This chart also includes searches of people of other
races and Hispanics combined. These represent only forty of the 387 CIT searches in 2000, and
show the same basic pattern of increased probable cause, although consent searches are even
more prevalent among CIT member searches of drivers who are neither white nor African
American. The decline in the use of consent searches is consistent with a decline in the influence
of federal guidelines for drug interdiction by the CIT.

Figure 4.5

Race and Probable Cause in CIT Searches, 1997-2000

.5

.4

Proportion of Searches

.3

.2
Black
.1
Other Race
0.0
1997

White
1998

1999

2000

171
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure 4.6 Race and Probable Cause in Non-CIT Searches

.4

Proportion of Searches

.3

.2

.1

0.0
1997

1998

172
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

1999

20

Figure 4.6 shows the results for non-CIT trooper searches. Searches attributed to probable
cause increased dramatically in the searches of African Americans by regular NCSHP troopers.
Surprisingly, the trends for white and “other race” stops are not so clear. If the decreased
searches of African Americans and the decreased use of consent searches is a reaction to the 1996
charges of racial disparity, it seems to have changed the behavior of the CIT troopers for searches
regardless of the race of the driver. For other troopers, consent searches seem to have declined
only for African Americans. Although we do not report separate graphs here, the racial
composition of probable cause searches has not declined, while the proportion African American
among consent searches has declined since 1997 for both CIT and regular troopers’ searches.
Thus, the decline in searches of African Americans documented in Figure 2 occurred entirely
because of the decline of consent searches.
Our analysis of the incidence of probable cause reveals that between 1997 and 2000,
probable cause searches became a higher proportion of all searches, as the use of consent searches
declined. This pattern is particularly strong among the CIT members, and this occurred across all
racial/ethnic groups compared in the CIT analysis. Regular road troopers do few searches and
have dramatically reduced the use of consent searches of African American drivers—but not of
white or “other race” drivers. The prediction based on troopers accounts that the CIT would
make more use of consent searches is not supported by these analyses. Of course, the vast
majority of regular troopers conduct few or no searches.

173
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Hit Rates
Hit rates, or the proportion of searches which result in the successful identification and
seizure of contraband, are potentially useful indicators of racial disparity in search decisions. If
troopers are searching minorities at higher rates because of race rather than because they have
good professional reasons to suspect contraband, then we would expect that a lower proportion of
minority vehicles will be found to contain contraband. Essentially, in the presence of racially
biased search decisions, we expect more innocent minorities to be searched than innocent
whites.34 Examining hit rates by race will provide some insight as to the likelihood that the search
decision may be influenced by race over and above the magnitude of reasonable suspicion.
Figure 4.7 shows the hit rates for the CIT searches. In 1997, the proportion of searches
that generated contraband were quite a bit lower for searches of African American and “other
race” drives than the hit rate for white drivers. The hit rate for “other race” drivers is
considerably lower than for either African American or white searches. To the extent that race
played a part in the determination of who to search, these results would suggest that CIT troopers
were quite unsuccessful in their predictions and choices related to search decisions. The white
and African American rates converge across the period. While the hit rate for “other race” drivers
does rise, it still remains lower than that of the other two groups. These results overall show
substantial racial disparity in the success rate for searches for 1997. This effect seems to have
disappeared after 1998, as fewer African American drivers have been searched and fewer consent
searches have been initiated by CIT members. Given that the CIT hit rates for African American
34

There are issues of social efficiency around absolute hit rates that should be recognized
as well. Even if there are no racial disparities, if hit rates are low, then many innocent drivers are
being searched. The question then becomes at what level of hit rate is the trade-off between the
invasion of innocent citizens’ privacy and potential drug interdiction acceptable?
174
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

and white drivers converged after 1998, the hit rates for searches of African Americans after 1998
surpass those of whites after 1998, and that the racial disparity in searches has dropped
dramatically, the evidence points to earlier disadvantage for African Americans, possibly as a
result of racially influenced practices by the CIT in 1997, but little or no disparity in current
search practices.

Figure 4.7 Proportion of CIT Searches Yielding Contraband By Race,
1997-2000
.4

Proportion of Searches

.3

.2

Black

.1

Other Race
0.0
1997

White
1998

1999

175
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

2000

Figure 4.8 Proportion of Non-CIT Searches Yielding Contraband By
Race, 1997-2000
.6

.5

Proportion of Searches

.4

.3
Black
.2
Other Race
.1
1997

White
1998

1999

176
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

2000

Figure 4.8 reports the same basic data for searches by the rest of the NCSHP. For all
years, hit rates are higher for whites who were searched than for African Americans or drivers of
other races. After 1998, the hit rates rise for searches of both African American and white
drivers, reflecting the lower use of consent searches documented above. If there is any evidence
of convergence in African American and white hit rates, it does not occur until the year 2000.
There is no evidence of convergence in hit rates for drivers of other races. Although regular
NCSHP troopers do very few searches, and these searches are concentrated by 2000 in only two
troops and among five troopers, there is evidence not only of continued racial disparity in
searches, but also of lower rates of contraband found in minority searches.

Individual Trooper Analyses
The analyses of search disparity reported above suggest that, outside of the CIT, a few
troopers may be responsible for the disproportionate searching of African American vehicles
relative to the proportion African Americans stopped. Inside the CIT, there is evidence of some
race-linked disparity in searches, although it is much reduced compared to earlier years. In both
cases, however, we used the racial composition of stops at the troop level as the baseline. Since
there could easily be spatial and temporal heterogeneity within troops as discussed in previous
chapters, these higher odds of African American searches may simply reflect where and when
troopers patrol, rather than some racial disparity in search decisions.
In the year 2000, there were sixteen troopers with fifteen or more searches. These
troopers include members of the CIT and the handful of regular NCSHP troopers who are actively
(by our measure) searching vehicles. Three of the sixteen showed no excess searches of African

177
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Americans compared to those of whites, when using the trooper’s own racial distribution of
citations and warnings as the baseline from which to compare the racial distribution of searches.
Although the number of searches per trooper in 2000 is low, eight of the troopers had odds ratios
of higher than 2, suggesting that they search African Americans at or above twice the rate of
which they search white drivers. Given the racial distributions of citations and warnings, these
troopers display a slightly elevated proportion of searches of African Americans relative to white
searches. This suggests some racial disparity in the decision to search African Americans’
vehicles.35

Conclusion
There is considerable evidence in this chapter that there was a decline in search activity by
the NCSHP after 1997, and that much of this decline arose from searching fewer African
Americans. In 1997, the vast majority of searches were recorded as consent rather than probable
cause searches, leaving more room for trooper discretion in whom to search. In 1997, all troops
searched a higher proportion of African Americans than they did whites, although this varied
35

There are an additional five troopers who had odds ratios above 1 but below 2. As
stated in the main text, there are eight troopers who have odds above 2, suggesting they search
African Americans at or above twice the rate of white drivers. Two of these troopers had odds
ratios above 4, suggesting that they search African Americans at least four times as often as
whites, given the number of African Americans and whites they stop for routine driving
violations. The absolute number of searches is not high, ranging between 1 and 5 per month.
Given these low absolute number of searches it is even possible that for any one of these
individuals the disproportionate search of African Americans happened by chance. Also, the
low numbers prohibit addressing “selection effect” interpretations of the data, such as time of
day, day of week, and spatial heterogeneity concerns. Moreover, it should be noted that we have
no information on the circumstances under which these searches are conducted. For example,
these select few troopers may be called to the scene of an stop made by another trooper, and are
called in because of their expertise in conducting vehicular searches. As such, the selection
mechanisms for their searches may be quite different from that of other troopers.
178
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

from slightly higher odds of an African American search, to four times the odds in the CIT and in
Troop 1. Across the period, there is considerable evidence of changes in the search-decision
process of the CIT. The African American-to-white odds of a search fell from four-to-one in
1997, to two-to-one thereafter. The use of consent searches declined for all racial groups, and
African American and white hit rates converged, and, significantly, the hit rate for African
American searches surpassed that for whites after 1998. Although the CIT search data suggest
that the drug interdiction profiles used prior to 1998 were possibly not unrelated to race, from
1998 on, the disproportionate searches of African Americans, to the extent they are related to
race, is more likely to arise from the use of stereotypes or the residual nervousness of African
American drivers during the “interview method” used to develop reasonable suspicion. The
convergence of African American and white hit rates suggests that the CIT is not searching a
higher proportion of innocent African Americans than innocent whites. Although there is clear
improvement in hit rates, a higher proportion of “other race” drivers who are searched are not
found to be holding contraband.
The pattern and extent of searches is different for regular NCSHP troopers. First, the vast
majority of troopers do few or no searches. In fact, the vast majority of troops do less than thirty
searches in a year. Two troops consistently generate more than thirty searches in a year and also
search African Americans at relatively high rates. In the year 2000, five individual troopers
accounted for almost all of the search activity in these two troops. In addition, hit rates for
searches are consistently lower for African American and “other race” drivers who are searched
by regular NCSHP troopers, suggesting that higher minority search rates are not, in fact, founded
on the same level of reasonable suspicion.

179
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

The basic conclusion to be drawn from this analysis is that the vast majority of NCHSP
troopers do no searches and so cannot be racially profiling in the search decision. To the extent
that earlier charges of CIT profiling may have described search practices at that time, we find no
substantiation for that claim at present. The analysis of individual troopers suggests that there are
eight troopers with high levels of African American-white disparity in search behavior. It should
be recalled, that even for these eight troopers there may be unmeasured and non-biased
explanations for the observed higher searches of African Americans. While these numbers should
not be ignored by NCSHP administration, they likewise should not be viewed as proof of
profiling given the limited information on the trooper-citizen contact available for analysis.

180
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Chapter 5 Citizen Survey Results: Racial Disparity in Self-Reported Stops

Whereas the earlier chapters focused on official records, this chapter reports the results of
an analysis of police stop experiences using citizen self-reports from the North Carolina Driver
Survey. We find evidence of racial disparity in self-reported stops, after controls for driver
characteristics and reported driver behavior. Local police were found to exhibit a higher level of
racial disparity than does the NCSHP. African Americans also report being stopped for
somewhat more discretionary reasons and, to a small but significant extent, being treated with less
respect during stops. There are no racial differences in the relative incidence of citations, written
warnings, and verbal warnings. The evidence in this chapter, based on self-reported data, points
toward greater racial disparity in the stop decision than in the interaction after the stop.

Introduction
The North Carolina Driver Survey was designed to complement the official statistics we
have discussed previously. Official statistics are essentially the officers’ recorded accounts of
citizen-officer encounters. When all encounters are recorded, they can provide exceptional data
on the racial distribution of stops. Official data do not contain very much information on driver
characteristics, especially the ability to determine if drivers who are stopped report different
driving practices or behavior than those who are not stopped.
If most drivers break the driving laws at some time, then there may be substantial officer
discretion in deciding whom to stop. Where discretion is larger, so is the opportunity for racial
bias in the stop decision to enter into the process. The official data we have been examining

181
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

focuses on the NCSHP. The various local police forces operating in North Carolina actually stop
more drivers than does the NCSHP. The survey data allow us to capture drivers’ self-reports of
all stops by all police and to make distinctions between stops by the NCSHP and those by officers
attached to other police forces.
The survey data also allows us to collect information on typical driving behaviors that
may influence the probability of being stopped. In the survey we not only asked North Carolina
drivers whether or not they were stopped, but also why they were stopped, the outcome of the
stop, and whether or not they felt that they were treated with respect during the stop. Finally, the
survey allows us to explore the linkages among race, stop experiences, and self-reports of driving
practice and behavior.

Sample Characteristics
The North Carolina Driver Survey is a telephone survey of a stratified random sample of
current North Carolina licensed drivers. The sample was stratified by race in order to have
sufficient sample sizes to compare the experiences of white and African American drivers. The
sampling frame included white and African American drivers who had applied for or renewed
their licenses in the pervious six months. Using this method we had expected to get phone
numbers and addresses that were relatively current. Unfortunately, it turns out that the N. C.
Department of Motor Vehicles rarely asks for the telephone number, nor does it require proof of
home address for license renewals. Thus, we had to use a telephone match based on surname and
address to develop useful contact information for our sampling frame. The return on the
telephone match was 48.6 percent, lowest for African American females at 39.0 percent and

182
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

highest for white males at 62.8 percent. Cooperation rates on the survey were much better at 59.1
percent, with a high for African American females of 61.8 percent and a low for white males of
56.5 percent. Data were collected between June 22, 2000, and March 20, 2001. Half of the cases
were collected by September 11, 2000.
A comparison of our final sample to the actual race-gender-age distribution of licensed
drivers in North Carolina shows that our final sample is quite a good match to the state
distributions (see Table 5.1). In all four gender-race groups, young adults age 30–39 are under
represented. In most statistical analyses we weight the data to correspond to the known gender
and age distributions of licensed drivers within the two racial strata. We call this the “DMV
weight” because it refers to the distribution of drivers in the N.C. Department of Motor Vehicles’
license registry.

Racial Differences in Stops
One of the primary goals of the survey is to see if there are racial differences in the
probability of police stops. Because police are expected to respond to driver behavior in making
the decision to stop, we also collected information from respondents about their typical driving
behavior. Information on other attributes of the driver that may motivate police stops or influence
driving behavior such as age, gender, education, home ownership, and car model and age was
collected.

183
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 5.1 Age, Gender, and Racial Distribution of Survey Respondents and North Carolina Registered Drivers, 2000.
African American
White
Survey Percent
DMV Percent
Survey Percent
DMV Percent
Males 18–29
12.4
13.3
9.3
10.2
Males 30–39
9.4
11.8
8.7
10.7
Males 40–49
11.5
10.3
10.9
10.2
Males 50–59
7.2
6.2
9.1
8.1
Males 60+
6.6
6.0
10.4
10.3
Females 18–29
14.3
13.5
10.2
9.8
Females 30–39
8.6
12.6
9.1
10.4
Females 40–49
13.7
11.5
11.8
10.2
Females 50–59
8.2
7.1
8.9
8.3
Females 60+
9.3
7.6
11.9
11.6

We were concerned about potential reporting errors in response to our questions about
police stops and driving behavior. Being stopped by the police for speeding or other reasons is
potentially embarrassing. It is well known in survey research that respondents tend to underreport embarrassing behaviors. As described in Appendix E, we conducted a record check survey
of almost 600 drivers with known speeding stops in the last year in order to ascertain the degree
of under-reporting of stops we could expect in the driver survey. The record check survey
showed that 74.8 percent of whites admitted being stopped in the last year. For African
Americans the number was 66.8 percent. This racial difference in self-reports is statistically
significant at a .019 probability. These results suggest that in the larger driver survey we would
expect to find respondents who claimed no stops in the past year, but who were in fact stopped.
And, while both groups under-report speeding stops, African Americans do so at a higher rate.
This finding is consistent with many past studies that report stronger social desirability effects on
survey responses among African Americans. Thus, the North Carolina Driver Survey data will
tend to underestimate the number of stops for both African Americans and whites and it will also

184
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

underestimate the magnitude of the racial disparity in stops.36 In most analyses we present results
that have been weighted to reflect the expected racial difference in non-response to the questions
about stops.37 We refer to these analyses as using the Record Check weights. In general, the
record check weights increase our estimates of the number of stops and increase the racial
disparity in stops, but should more closely mirror the actual racial gap in stop experiences.

Comparison of Driver Survey Estimates of Stops to Official Records
There are reasons to believe a driver survey might underestimate the actual number of
stops and other reasons to suspect they might overestimate the actual number of stops (see
Bradburn 1983). From the record check survey, we developed race-specific estimates of the
degree of under-reporting, and we can use those estimates to weight the data to correct for this
possible source of under-reporting. In the survey we asked respondents to recall the stops they
had experienced over the last year. Survey questions that ask the respondent to recall events are
also subject to telescoping errors in which respondents report on events that actually took place
outside time period parameters specified in the survey question. In our case we ask for stops in
the last year. If telescoping is occurring, weighting the data to correct for under-reporting may
produce larger estimates of the number of the stops in the last year than actually occurred. To the

36

The record check survey also shows that people who do not report a stop also report fewer
risky driving behaviors and less speeding. For non-threatening questions such as miles driven
there are no differences between those who report stops and those who do not.
37
Given these record check results, we weight white respondents who admit to stops at 1.34228
(1/.78) and African American respondents who admit to stops are weighted at 1.4947 (1/.68).
This requires weighting whites and African Americans who do not admit to stops at less than one
in order to preserve the original sample size. These weights are .9245 for whites who do not
report stops and .8228 for African Americans who do not report stops.
185
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

extent that telescoping is associated with race, our estimates of racial differences in stops may be
effected.
In the Table 5.2 we produce population estimates of stops by the NCSHP in the last year
from the North Carolina Driver Survey with DMV weights and with Record Check weights. We
compare these to the estimates of stops from the official 2000 citation records of NCSHP
troopers.
The citation estimates from the NCSHP trooper’s citation reports are very similar to those
based on the survey responses using DMV weights. For both African American and white
citation estimates the actual NCSHP citation count is within a single standard deviation of the
survey estimates using the DMV weights. When the record check weights are used for both
African American and white drivers, the survey estimates of citations are much higher than the
recorded citations in the NCSHP citation reports. These analyses suggest that the point estimates
based on the DMV weights are surprisingly accurate.
Given the under-reporting we found in the record check survey, we would have expected
that the record check weights would have been closer to the official records. It would appear that
reports that telescope the time period are prevalent, and that telescoping over-reporting and
refusal-to-admit-a-stop underreporting roughly cancel each other out. In general, this suggests
that the yearly incidence of stops is best captured by the DMV weighted data.

186
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 5.2 Comparison of Survey and Official Record Estimates of African American and White Total Citations in Past Year
Survey Population
Confidence
Confidence
Estimate from
Estimate of NCSHP
Interval at 66
Interval at 95
NCSHP 2000
Citations in Last Year
Percent
Percent
Citation Reports
Confidence
Confidence Level
Level
African American Drivers
99,986
87,744
DMV Weights
112,228
to
to
124,470
136,715
101,909
152,947
138,208
167,685
to
to
Record Check Weights
182,423
197,162
White Drivers
248,220
222,374
DMV Weights
274,066
to
to
299,912
325,758
294,241
338,976
309,355
Record Check Weights
368,597
to
to
398,218
427,839

The Record Check weights may do a better job of capturing the racial gap in self-reports,
since they incorporate information from the Record Check survey on racial differences in the
probability to report stops in a survey. These estimates probably refer now to a period closer to
eighteen months. We also use the Record Check weights in the multivariate analyses that follow
because risky driving behaviors, are associated with the failure to report a stop in the record check
analyses, and are some of the primary presumed legitimate causes of stops in the multivariate
models below.
Racial Differences in the Reported Stop Experience
Table 5.3 reports the race-gender-age distribution of stops from the North Carolina Driver
Survey. These distributions are for drivers who self-report being African American or white and
are weighted to correspond to the North Carolina Department of Motor Vehicles distributions of

187
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

age and gender within race.38 Young drivers tend to have more stops than older drivers and
African American male drivers tend to be stopped more often than white drivers.
Young African American drivers report an average of 1.24 stops in the past year, almost
twice as many as young white drivers. Almost all of the racial disparity in stops among young
male drivers is produced by stops by local police. NCSHP stops of young African American male
drivers (17.0 percent) and young white male drivers (15.5 percent) are very similar. This same
pattern holds up for older male drivers. In all cases there are racial disparities in stops, but the
racial disparity is larger for local police than it is for the NCSHP. It is also the case that while the
absolute probability of being stopped drops for African American and white males as they age,
the relative racial disparity in stops (the odds ratio of African American to white stops) tends to
rise with age, especially for stops by the local police.

Table 5.3. Male Age and Race Distribution of Stops from the North Carolina Driver Survey, 2000.
DMV 2000 Licensed Driver Weights
18–22
23–49
50+
African
White
African
African
American
American
White
American
White
Number of Stops
1.24
.68
.51
.33
.26
.17
mean (s.d.)
(2.48)
(1.38)
(.93)
(.77)
(.59)
(.61)
Any Stop
42.1%
32.1%
32.6%
24.0%
18.8%
12.5%
Odds Ratio
1.31
1.36
1.50
Local Officer Stop
35.1%
25.9%
23.4%
14.9%
13.3%
7.0%
Odds Ratio
1.36
1.57
1.90
NCSHP Officer Stop
13.8%
6.0%
5.9%
17.0%
15.5%
10.4%
Odds Ratio
1.10
1.33
1.02
Sample Size

104

90

400

38

385

156

267

Although we sampled only people who were listed as African American or white in the DMV
records a few respondents claimed other ethnic/racial affiliations. These people are dropped from
the analyses. A few people refused to answer the race question, these were assigned the DMV
race and kept in the analyses. The original sample somewhat underestimates people in the 30–39
age group for all four race-gender categories. The DMV2000 weights correct for these
underestimates.
188
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 5.4 displays similar analyses of racial differences in stop experiences among
women. Women tend to be stopped less often than men at all ages and among both African
Americans and whites. Among women, police stop experiences also decline with age. Young
African American women report about a third as many stops as young African American men.
Young white women report about half as many stops as young white men. As among men, there
are racial disparities in stops and the disparity is larger in stops by local police than in stops by
the NCSHP. In fact, among young women, white women report slightly more stops by the
NCSHP than do African American women. While among older women African Americans report
more stops by the NCSHP, the racial disparity is very small. For local police, the relative racial
disparity in stops as measured by the odds ratio increases dramatically with age. Very few older
white women are ever stopped by the local police, while 9.6 percent of older African American
women report stops in the last year by local police officers.
Table 5.4. Female Age and Race Distribution of Stops from the North Carolina Driver Survey, 2000. DMV
2000 Licensed Driver Weights
18–22
23–49
50+
African
African
African
American
White
American
White
American
White
Number of Stops
.38
.33
.36
.23
.26
.01
mean (s.d.)
(.61)
(.63)
(.92)
(.52)
(1.10)
(.31)
Any Stop
32.7%
25.9%
24.6%
19.2%
13.2%
8.2%
Odds Ratio
1.26
1.28
1.61
Local Officer Stop
22.2%
16.0%
16.8%
11.9%
9.6%
3.7%
Odds Ratio
1.39
1.41
2.60
NCSHP Officer Stop
10.2%
12.3%
8.8%
8.4%
4.5%
4.1%
Odds Ratio
.83
1.05
1.10
Sample Size

99

81

411

370

198

294

Table 5.5 reports racial comparisons in stops, driving behaviors, and demographic
background weighted to correspond to the 2000 age-gender-race distributions of North Carolina
189
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

drivers. In order to make good use of the sample sizes available, statistics are compared for white
and African American North Carolina drivers across all age and gender groups. In multivariate
models that follow we statistically adjust for gender, age, and other factors that may explain the
racial gap in police stops.
Slightly more than a quarter (26.4 percent) of African American North Carolina drivers
reported a stop in the last year as compared to 18.1 percent of whites. African Americans are 8.3
percent more likely than whites to report being stopped by the police. African Americans are
stopped more often (total stops average .43 versus .25), are 7.6 percent more likely to be stopped
by the local police, and are 1.3 percent more likely to be stopped by the NCSHP. All of these
racial disparities in stops, except stops by the NCSHP, are statistically significant.
The National Institute of Justice recently released national survey estimates of racial
differences in driver stops by police. Their self-report data on police stops show 10.4 percent of
white drivers and 12.3 percent of African American drivers reporting a stop in the last year. This
suggests that the overall rate of police stops in North Carolina is almost twice the national
average. The racial gap in stops by local police is also quite a bit higher in North Carolina than in
the rest of the country. Stated another way the self-reported racial gap in police stops by the
NCSHP is about half of the national average, but the racial gap in reported stops by local police in
North Carolina is almost four times greater than the national average. This may suggest that local
police more aggressively enforce traffic laws in North Carolina. It is also quite possible that the
national study understated the actual rate of stops.

190
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Racial Differences in Reported Driving Behavior
One of the advantages of a survey approach is that we can collect information on reported
driving behavior. If there are racial differences in typical driving behavior they may explain some
of the observed differences in police stops. The third panel of Table 5.5 displays racial
differences in driving behaviors. The first two entries refer to miles driven. It seems reasonable
to predict that people who drive more are at greater risk of being stopped by the local police.
This risk could be based simply on increased opportunity to encounter police officers. In
addition, if most people occasionally break the driving laws, more driving increases the
probability that a person would encounter a police officer while breaking the law. On average
whites drove significantly more miles in the last week and across the last year than did African
Americans. Whites average more than 3,000 more miles per year than African Americans.39
Thus, miles driven cannot explain the racial gap in stops.
Another way to think about the racial gap in stops is to ask how many miles the average
white or African American drives before being stopped. On average, North Carolina African
American drivers drive 32,681 miles before being stopped. Whites in North Carolina report
driving more than twice as far—68,944 miles per stop. Whites are also significantly more likely
than African Americans to drive on interstate highways.

39

Using the 1999 National Transportation Survey, we compared self-reported miles driven in the
last year by African Americans and whites for the nine major census divisions. In every region of
the country, whites drive more on average than African Americans. The gap was largest in the
West North Central and New England regions—at more than 6,800 miles per year and lowest in
the Middle Atlantic and Mountain states—at less than 1,000 miles per year. The estimate for the
East South Central region, which includes North Carolina, was that African Americans drive
5,400 fewer miles per year than do whites.
191
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

In terms of breaking the traffic laws, whites report slightly higher average speeds in 65,
55, and 35 mph speed zones. We experimented with various measures of speeding at thresholds
of 5, 6, 10, and 15 mph above the speed limit. In general there are no significant racial
differences in self-reports of typically speeding at these thresholds.40 In the multivariate analyses
that follow, we use a three-item speeding scale based on a five mile per hour speeding threshold
for three hypothetical speed limits—35 mph in town, 55 mph on a two-lane highway, and 65 mph
on an interstate highway. We chose this measurement because it was more highly correlated with
the probability of a stop than alternative measures of speeding behavior. Because there are no
average racial differences in self-reported speeding behaviors this variable cannot account for the
racial disparity in police stops.
We also asked drivers if they used methods to avoid getting stopped for speeding. Very
few drivers use any method to reduce the chance of being stopped, although cruise control is most
common. Whites were significantly more likely than African Americans to report using cruise
control, listening to a CB-radio, and watching and monitoring the speed of commercial trucks.
There were no racial differences in the use of radar detectors. If these methods reduce the risk of
being stopped they may account for some of the racial disparity in police stops observed in these
data. In the multivariate analyses we use an additive scale called “Methods to Avoid Speeding
Tickets” based on these four items. High scores mean the respondent rarely uses these methods.
A reliability analysis shows that these items are not highly correlated with each other.

40

The one exception is displayed in the table. Whites are significantly more likely to say they
typically drive 40 mph or above in a 35 mph speed zone than are African Americans.
192
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 5.5. Racial Differences in Stops, Driving Behavior and Demographic Background, Weighted by 2000 DMV
Race-Gender-Age Distributions
African
White
Signific
American
ant Difference?
STOP EVENTS
Any Stop in Last Year
26.4%
18.1%
Yes
Total stops last year
.43
.25
Yes
(1.11)
(.67)
Any Stop by a Local Policeman in Last Year
18.8%
11.2%
Yes
Any Stop by NCSHP in Last Year
10.0%
8.2%
No
Helped by Officer in Last Year
6.5%
5.2%
No
Officer at Accident Scene in Last Year
10.5%
8.1%
Yes
DRIVING BEHAVIORS
Miles driven last week
Miles driven last year
How often drive on interstates?
(1=everyday…7=never)
Average speed in a 65 mph zone
Typically drive 70 in a 65 mph zone
Average speed in a 65 mph zone
Typically drive 60 in a 55 mph zone
Average speed in a 65 mph zone
Typically drive 40 in a 35 zone
Scale of 5+ over limit speeding behavior
Scale of 10+ over limit speeding behavior
To avoid getting a speeding ticket, do you,
(1=All of Time…4=Never)
Use cruise control?
Use a radar detector?
Listen to a CB?
Watch and follow trucks?
Frequency of methods used to avoid speeding
tickets
Do you always use seatbelts?
Do you always use turn signal?
Do you change lanes in order to get somewhere
more quickly?
Do you pass slow cars on a two-lane highway?
Do you ever roll through a stop sign?
Do you ever speed up to get through a yellow light?
Scale of risky driving behaviors

193
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

211
(428)
14,053
(30,515)
3.35
(1.98)
67.5
(4.4)
48.8%
57.3
(4.2)
42.9%
36.1
(3.5)
24.6%
1.17
(1.13)
.34 (.57)

274
(452)
17,236
(34,002)
3.09
(1.75)
68.2
(3.9)
51.9%
57.9
(3.7)
45.0%
36.7
(3.0)
28.0%
1.25
(1.13)
.35
(.58)

Yes

3.28
(.96)
3.91
(.44)
3.84
(.60)
3.60
(.84)
3.66
(.42)
92.3%
85.4%
55.8%

2.95
(1.04)
3.90
(.49)
3.91
(.45)
3.72
(.67)
3.62
(.38)
89.2%
71.9%
55.1%

Yes

74.6%
23.9%
51.6%
2.18

77.1%
34.4%
61.8%
2.57

No
Yes
Yes
Yes

Yes
Yes
Yes
No
Yes
No
Yes
Yes
No
No

No
Yes
Yes
Yes
Yes
Yes
No

DEMOGRAPHIC BACKGROUND
New Driver (four or fewer years driving)
Age
Male
Education Scale (1=No High School…7=Grad
School Degree
Own Home
Urban/Rural (1=City….4=Country)
Model Year of Car Typically Driven

(1.31)

(1.34)

5.2%
41.00
(15.3)
47.7%
3.68
(1.85)
67.8%
2.0
(1.28)
93.02
(5.35)

1.7%
44.8
(15.8)
49.6%
4.14
(1.94)
84.6%
2.5
(1.23)
94.01
(5.01)

Rather they are functional alternatives. None are used frequently, although many people
use one. Only 37.4 percent of drivers report never using any of these methods.
Finally, we asked about a series of risky driving behaviors. African Americans reported
significantly higher seat belt and turn signal use than did whites. They also reported lower
likelihood of rolling through a stop sign or speeding up to get through a red light. There were no
significant racial differences in lane-changing or passing on a two-lane highway.41 For the
multivariate analyses we created a scale of “Risky Driving Behavior” by summing these six
items. Inter-item reliability is only moderate (Alpha=.45).
African Americans report driving fewer miles, are less risky drivers in terms of reported
driving behavior, and report driving slightly slower on average than do white North Carolina
drivers. White drivers, on the other hand, use more methods to avoid being pulled over by the
police. These descriptive statistics suggest that the reported driving behavior is unlikely to
provide powerful explanations for the observed racial disparity in police stops. Because whites

41

In our analysis of the record check survey (reported in Appendix E) we checked to see if either
whites or African Americans who failed to report a stop also reported fewer risky behaviors or
slower driving speeds. They did , so we repeat the multivariate analyses using record check
weights to see if reluctance to report stops or bad driving behavior influences results.
194
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Yes
Yes
No
Yes
Yes
Yes
Yes

report driving slightly worse than African Americans, controlling for driving behavior is likely to
increase the estimate of the size of the racial gap in stops in the multivariate models that follow.

Racial Differences in Demographic Background
The bottom panel of Table 5.5 compares African American and white drivers’
demographic backgrounds. Here there are substantial racial differences. African Americans are
significantly more likely than whites to be inexperienced drivers. Compared to white drivers,
African Americans are also on average younger, slightly less educated, less likely to own their
own home, live in more rural areas, and drive slightly older cars.
These demographic differences are potentially important explanations of the racial
disparity in police stops. The public concern with racial bias in police stops implies that police
discretion coupled with police reaction to driver’s status attributes combine to produce racial
disparity in the pattern of stops. If this is a reasonable model of the causal process, then we
would expect that other status attributes that are associated with police perceptions of driver risk
or dangerousness will also be associated with the decision to stop. In particular, we would expect
males, younger drivers, and economically disadvantaged drivers to have higher probabilities of
stops. African American drivers tend to be slightly less likely to be male than white drivers, but
to be younger, slightly less educated, less likely to own a home, and drive slightly older cars. If
police stop decisions are influenced by status characteristics such as these they may explain some
of the racial disparity in stops.
In addition, in multivariate models we can see the extent to which driving behaviors
explain status linked disparity in stops. To the extent that they do, police may be reacting to
driving behavior rather than status attributes. Strong status attribute associations with stops after
195
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

controlling for driving behavior suggests that police perceptions of status attributes such as race,
gender, or age may be important direct influences on the decision to stop a car.

Modeling Police Stops in a Multivariate Context
The vast majority of drivers who experienced a stop were stopped only once in the
previous year. For this reason we model the likelihood of a stop in a logistic regression
multivariate framework. In this statistical framework a dependent variable, coded 1 to indicate
the presence of the outcome (in this case a stop in the last year) and 0 to indicate its absence (no
stop), is regressed on a series of explanatory variables. The statistical model is estimated using
maximum likelihood methods, and it predicts the log of the odds of the outcome occurring (1)
versus not occurring (0) as the dependent variable. In the tables that follow we display the
exponent of the log-odds coefficient which can be interpreted as the multiplicative change in the
odds associated with a one unit change in the independent variable. So, for example, in Table 5.6,
column 1, the reported odds coefficient associated with race is 1.63. This means that for African
American drivers, their odds of being stopped by the police in the last year are 1.63 times higher
than that of white drivers.42 Odds below one mean that the outcome is less likely. Odds above
one mean that the outcome is more likely as such explanatory variables as race, age, or risky
driving behaviors increase. In addition, these multivariate models are all estimated twice, once
42

These models were estimated using only cases with no missing values. In addition, four cases
were excluded because the respondent claimed not to have driven in the last year. Models
include the key explanatory variables that we have available. Alternative operationalizations of
miles driven and speeding were explored but those selected were the most strongly associated
with the probability of stops. In addition, interactions between miles driven and risky behavior,
methods to avoid tickets, and interstate frequency were explored for all three stop measures. In
no case were these interactions statistically significant. Finally, including a dummy variable for
inexperienced driver in no cases increased the probability of being stopped in the last year
beyond the effect of simple age.
196
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

with data weighted to conform to the 2000 North Carolina DMV records of race-age-gender
distributions of licensed drivers and then with a second weight based on the record check survey
to correct for racial differences in the under-reporting of stops.
The logic of the analysis is to first establish the size of the racial disparity in stops. We
then enter a series of demographic control variables to see if they help explain the observed racial
disparity in the first model. To the extent that these demographic variables are also associated
with the probability of stops they, like race, are associated either with driving behaviors or with
police discretion in stops decisions. The third model introduces the series of self-reported driving
behaviors. To the extent that these behaviors encourage police stop decisions they should be
significantly associated with stop outcomes. In addition, if status characteristics are associated
with different driving behaviors, adding driving behaviors to the models should erode the
coefficients associated with demographic status characteristics. If the coefficients for
demographic background are not eroded then this suggests that police are often reacting to
drivers’ status characteristics rather than their typical driving behavior. The final set of models
separates the analysis by race and allows us to see if police react to both demographic attributes
and driving behavior similarly for white and African American drivers. All analyses are repeated
for any stop in the past year, and for stops by the local police and the NCSHP.
Table 5.6 displays the analyses of any stop in the last year. In the first model we see that
the odds of an average African American driver in North Carolina being stopped at least once in
the last year is 1.63 times higher than it is for an average white North Carolina driver. The
estimate (weighted to account for non-responses from the record check survey) indicates that the
African American odds of being stopped are 2.03 times greater than whites. Because the record

197
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

check weights correct for under-reporting, but not for telescoping of the recall period, the latter
racial gap refers to the odds of being stopped in the last year and half.
Adding the demographic variables to the model, substantially increases the model ChiSquare suggesting that the odds of being stopped are strongly influenced by these non-racial
status characteristics. The coefficient associated with race is reduced by about 10 percent using
both sets of weights. The racial disparity in stops is still statistically significant. We also see that
women are .70 times less likely than men to have been stopped in the last year; that increased
education is associated with increased incidence of stops; and that newer cars are stopped less
often, as are older drivers. Rural areas, urban areas, and home ownership are not associated with
the likelihood of a stop.

198
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 5.6. Logistic Regression of Any Stop Last Year Upon Race, Demographic Background, and Driving Behavior;
multiplicative odds coefficient and significance level reported.
African
Analyses weighted to 2000 DMV population
Model 1
Model 2
Model 3
American
White
count (N=2570)
Race (1=AA)
1.63***
1.46***
1.71***
N=1,209
N=1,361
Gender (1 =Female)
.70***
.84
.77
.96
Education
1.06*
1.02
1.01
1.04
Home Owner
.84
.83
.84
.81
Model Year of Car
.97***
.97***
.97***
.96**
Age
.97***
.97***
.97***
.97***
Rural
1.03
1.03
1.05
1.01
Scale Speed 5+
1.07
1.06
1.07
Miles Driven Year (LN)
1.14*
1.06
1.39***
Risky Driving Scale
1.13**
1.12*
1.13
Fewer Methods to Avoid Ticket Scale
.79
.85
.78
Interstate Frequency (low score=high usage)
.97
.93
1.04
Degrees of Freedom
1
7
12
11
11
Model Chi-Square
26.47
166.85
206.83
84.20
108.29
Model Probability
***
***
***
***
***
Analyses weighted to record check estimate of
non-response bias (N=2588)
Race (1=AA)
Gender (1 =Female)
Education
Home Owner
Model Year of Car
Age
Rural
Scale Speed 5+
Miles Driven Year (LN)
Risky Driving Scale
Fewer Methods to Avoid Ticket Scale
Interstate Frequency (low score=high usage)
Degrees of Freedom
Model Chi-Square
Model Probability
Probability levels of * .05; ** .01; ***.001 or below

Model 1
2.03***

1
70.15
***

Model 2
1.83***
.70***
1.06*
.84
.98**
.97***
1.03

7
243.43
***

Model 3
2.15***
.85
1.02
.82
.97***
.97***
1.03
1.07
1.15***
1.13***
.79*
.98
12
293.50
***

African
American
N=1,215
.78
1.01
.83
.97*
.97***
1.05
1.06
1.07
1.13*
.84
.93
11
104.17
***

White
N=1,373
.96
1.04
.80
.96**
.97***
1.01
1.07
1.39***
1.12*
.79
1.04
11
133.31
***

Model three introduces the measures for reported driving behaviors. Drivers who drive
more miles in a year, those who typically drive using more risky driving behaviors and those who
use more methods to avoid getting speeding tickets all are stopped more often. Self-reported
speeding behavior is not significantly associated with increased stops, although the coefficient is

199
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

in the correct direction. The result for methods to avoid getting tickets is at first surprising.
Drivers who consciously try to avoid getting tickets are stopped more often than other drivers.
Evidently these drivers know they are breaking the law, and their increased scanning behavior to
reduce police contacts is not sufficient to offset their increased speeding driving behavior.
Because African Americans report driving fewer miles per year, less risky driving
behavior, and are less likely to use methods to avoid speeding tickets than whites, the race
coefficient actually gets larger in Model 3. In the DMV-weighted models, after accounting for
demographic and driving behavior differences, the odds that an African American was stopped in
the last year are 1.71 times higher than they are for whites. When the data are weighted to
account for potential non-response to stop questions the level of racial disparity rises to 2.15
times higher odds of a stop if you are African American. Driving behavior coefficients are
substantively identical using both weighting systems, suggesting that social desirability effects on
reports of stops and driving behavior are not large or systematic-enough to influence results.
Introducing driving behavior reduces both the gender coefficient and the education
coefficient to non-significance. This suggests that the higher likelihood of being stopped among
men and among more highly educated drivers is a function of their driving behavior. The effects
of driver age and year of vehicle are not similarly reduced, suggesting that police tend to react to
older drivers and newer cars in a more forgiving manner than they do to younger drivers and
older cars. These results also help highlight that police discretion in stop decisions are not limited
to race. There may also tend to be age and class biases in the stop decision of some police
officers.
In the last two columns the models are split by race to see if any of the factors that
encourage or discourage police stops operate differently for African American and white drivers.
200
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

The only variable that affects the odds of being stopped differently for African American and
white drivers is “miles driven.” Driving more miles substantially increases the odds of a stop for
whites. Miles driven for African Americans seems to be unrelated to the likelihood of a stop.
Since officers cannot observe the number of miles driven, this result suggests that for whites a
key risk factor increasing the likelihood of a stop is miles of potential driving exposure to police
surveillance. For African Americans, their race seems to be a risk factor in its own right, a factor
more powerful than miles driven.
After accounting for other demographic factors which might attract police attention and
driving behavior these models suggest that the degree of unexplained racial disparity, as
measured as the relative odds-ratio of African American to white police stops is somewhere
between 1.71 and 2.15. This is a substantial level of unexplained racial disparity and so potential
racial bias in police stops.
Table 5.7 presents an identical analysis except that it is limited to stops by local police
officers. In Model 1—under both weighting protocols, we find larger racial disparities in stops
than were observed in Table 5.6 for all stops. Again, controlling for demographic background
somewhat reduces the degree of racial disparity in stops, and gender, car age and driver age
influence the likelihood of a stop by a local police officer. Driving behaviors are not particularly
important determinants of stops by local police. The one exception is that drivers who use the
interstate less often are also stopped less frequently by local police. This seems to be particularly
the case for African American drivers.

201
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 5.7. Logistic Regression of Local Officer Stop Last Year Upon Race, Demographic Background, and Driving
Behavior; multiplicative odds coefficient and significance level reported.
African
Analyses weighted to 2000 DMV population
Model 1
Model 2
Model 3
American
White
count (N=2581)
Race (1=African American)
1.84***
1.64***
1.77***
N=1,210
N=1,360
Gender (1=Female)
.66***
.71***
.69*
.77
Education
1.06
1.04
1.00
1.09
Home Owner
1.06
1.05
1.04
1.11
Model Year of Car
.97***
.97**
.97*
.97*
Age
.97***
97***
97***
97***
Rural
.95
.96
.95
.98
Scale Speed 5+
1.01
1.00
1.03
Miles Driven Year (LN)
1.03
1.00
1.15
Risky Driving Scale
1.09
1.04
1.16*
Fewer Methods to Avoid Ticket Scale
1.03
1.09
1.01
Interstate Frequency (low score=high usage)
.94
.91*
.99
Degrees of Freedom
1
7
12
11
11
Model Chi-Square
30.57
110.58
151.08
64.33
64.15
Model Probability
***
***
***
***
***
Analyses weighted to record check estimate of
non-response bias (N=2588)
Race (1=AA)
Gender (1 =Female)
Education
Home Owner
Model Year of Car
Age
Rural
Scale Speed 5+
Miles Driven Year (LN)
Risky Driving Scale
Fewer Methods to Avoid Ticket Scale
Interstate Frequency (low score=high usage)
Degrees of Freedom
Model Chi-Square
Model Probability
Probability levels of * .05; ** .01; ***.001 or below

Model 1
2.22***

Model 2
1.97***
.67***
1.06*
1.08
.97***
.97***
.93

1
68.21
***

7
204.10
***

Model 3
2.10***
.71***
1.04
1.07
.97***
.97***
.95
1.00
1.02
1.08
1.07
.94*
12
215.31
***

African
American
N=1,215
.69***
.99
1.07
.96***
.97***
.93
.99
.99
1.03
1.13
.90**
11
80.36
***

White
N=1,373
.76
1.09*
1.13
.97***
.97***
.98
1.03
1.13
1.15*
1.05
.98
11
77.77
***

In general the determinants of local police stops are broadly similar for white and African
American drivers. Still, there are some exceptions. Controlling for driving behavior, local police
stop African American males, but not white males, at higher rates. The finding from Table 5.5
—that miles driven increases white stops but not African American stops— is repeated here. It
would seem that local police do tend to react to the race of drivers and that African American
202
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

male drivers are at particular risk of being stopped. White drivers increase their risk of a stop
when they drive more, but African American drivers’ risk of a stop by local police is increased by
their race and being male. For both the African American and white populations, being older and
driving a newer car reduces the probability of a stop by the local police. Table 5.8 reports our
analyses of stops by the NCSHP. While there is significant racial disparity in stops by the
NCSHP it is less than half as large as the disparity produced by local police stops. In fact, after
controlling for demographic variables the DMV-weighted racial gap is not even statistically
significant. The record check weighted racial gap in Model 2 is just barely significant and
suggests that African American drivers have 1.38 times the white odds of being stopped. After
introducing self-reported driving behavior in Model 3 the racial gap in NCSHP stops increases for
both weighting schemes. These analyses suggest that there is some unexplained racial disparity
and so there may be some racial bias in driver stops by the NCSHP, but the disparity is
substantially smaller than the potential racial bias present in local police stops.

203
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 5.8. Logistic Regression of NCSHP Stop Last Year Upon Race, Demographic Background, and Driving
Behavior; multiplicative odds coefficient and significance level reported.
African
Analyses weighted to 2000 DMV population
Model 1
Model 2
Model 3
American
White
count (N=2581)
Race (1=AA)
1.30*
1.21
1.53*
N=1,210
N=1,370
Gender (1 =Female)
.66*
.90
.75
1.09
Education
1.05
1.00
1.06
.95
Home Owner
.78
.78
.80
.71
Model Year of Car
.99
.98
.99
.96*
Age
.97***
.98***
.99
.98**
Rural
1.10
1.09
1.14
1.01
Scale Speeding 5+
1.23***
1.25***
1.20
Miles Driven Year (LN)
1.32***
1.19*
1.64***
Risky Driving Scale
1.17**
1.25**
1.08
Fewer Methods to Avoid Ticket Scale
.64**
.74
.57*
Interstate Frequency (low score=high usage)
1.06
1.03
1.12
Degrees of Freedom
1
7
12
11
11
Model Chi-Square
3.85
65.48
131.03
62.48
77.88
Model Probability
*
***
***
***
***
Analyses weighted to record check estimate of
non-response bias (N=2788)
Race (1=AA)
Gender (1 =Female)
Education
Home Owner
Model Year of Car
Age
Rural
Scale Speed 5+
Miles Driven Year (LN)
Risky Driving Scale
Fewer Methods to Avoid Ticket Scale
Interstate Frequency (low score=high usage)
Degrees of Freedom
Model Chi-Square
Model Probability
Probability levels of * .05; ** .01; ***.001 or below

Model 1
1.49***

Model 2
1.38**
.70*
1.05
.78
.99
.97***
1.10*

1
11.91
***

7
84.27
***

Model 3
1.78***
.92
1.00
.77
.98
.98***
1.09
1.24***
1.33***
1.17**
.64**
1.07
12
173.60
***

African
American
N=1,215
.77
1.06
.79
.99
.99
1.14
1.25**
1.21**
1.26***
.74
1.04
11
83.03
***

White
N=1,374
1.10
.94
.70
.96*
.98***
1.01
1.21*
1.62***
1.06
.57**
1.13*
11
96.05
***

Supporting the notion that status bias processes are lower among the NCSHP, neither
gender nor age of vehicle is significantly associated with being stopped in Model 3. Age
continues to be associated with the likelihood of a stop but the coefficient is closer to one than in
previous models. So, while the NCSHP, like local police forces, seems more likely to give older

204
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

drivers a break they do not do so as dramatically as local police do. Also consistent with a more
professional orientation, stops by the NCSHP are very strongly related to all measures of driving
behavior. People who report more miles driven, driving faster, risky driving behavior, and trying
to avoid speeding tickets are all stopped at higher rates by the NCSHP.
When the models are split by race there are some suggestions that NCSHP troopers are
not completely color blind. Being older tends to protect white drivers but not African American
drivers from NCSHP stops. Similarly, owning a newer car seems to protect white drivers from
stops but not African American drivers. On the other hand, risky driving behaviors such as
passing, changing lanes without signaling, and driving without a seatbelt tend to encourage
NCSHP stops of African American drivers but not of white drivers. These findings suggest that
bias processes, to the extent they exist among NCSHP troopers, are subtle. White drivers are
“protected” by their age and class, but African American drivers are not. Conversely, African
American drivers are penalized for risky driving behaviors but white drivers are not. Another
way to say this is that, when the NCSHP troopers react to the race of the driver, it tends to be in
conjunction with other status or driving behavior characteristics. These interactions suggests a
bias process, in which attributions about characteristics other than race are modified by the race
of the driver.
As in the previous analysis miles driven is a risk factor for white drivers but not for
African American drivers. White drivers who use the interstate less often than other white drivers
are stopped more often by the NCSHP. White drivers who use fewer methods to avoid being
stopped are stopped less often.

205
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Discussion of Multivariate Results
These logistic regressions of stop experiences upon race, demographic background
characteristics and driving behavior provide evidence that there may be some racial disparity in
police stops of North Carolina drivers. For all stops, as well as for stops by local police and the
NCSHP, African Americans are more likely than whites to be stopped in North Carolina. The
degree of racial disparity, and so potential bias, in police stops appears to be substantially greater
among local police than that within the NCSHP. Within the NCSHP the pattern of stop decisions
associated with race suggests that while everyone is stopped for speeding, risky driving behavior
can be an aggravating factor for African Americans, while class advantage—in the form of a new
car or age—is a status shield for whites but not for African Americans.

Racial Differences in Stop Experiences
We also asked all drivers who reported a stop what reason they were given by the officer
for the stop. We coded the reasons into “speeding,” “other moving violations,” “non-moving
violations,” “general suspicion”-based stops, “no reason given by officer,” “do not recall,” and
“other.” Speeding probably represents the least discretionary reason, since an officer must
document the speed of the driver. “Other moving violations” (for example, rolling through a stop
sign) and “non-moving violations” (for example, a broken taillight) represent clear violations of
the driving laws but also enhanced police discretion in making a stop decision. “General
suspicion” includes various stops to investigate the license, registration, vehicle, and driver.
These could range from running a plate and discovering that the owner’s license had expired, to
stopping a car on suspicion of carrying drugs. In all cases they require the officer to proactively
investigate the driver or the car in the absence of a driving violation.
206
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 5.9 displays the racial differences in reasons given for the stop. African Americans
are somewhat (3 percent) more likely than whites to be stopped for other moving violations.
They are also 4 percent more likely to be stopped for some form of general suspicion. They are
less likely to be stopped for speeding. The table also suggests that African Americans are only
slightly more likely to be informed by the officer that they were stopped for potentially
discretionary reasons (53% versus 51% for other moving, non-moving, and general suspicion,
combined). In the recent study by the National Institute of Justice on police-citizen contacts
African Americans were 1 percent more likely than whites to be told they were stopped for
generalized suspicion and 2 percent more likely for non-moving violations. Similar to the
previous comparison of national figures to North Carolina ones, racial disparity in discretionary
stops seems to be somewhat higher in North Carolina than in the nation overall.

Table 5.9. Reasons Given by Officer for Stop, North Carolina Driver Survey
African American
Speeding
217
0.42
Other Moving Violations
93
0.18
Non-Moving Violations
48
0.09
General Suspicion
136
0.26
No Reason Given by Officer
8
0.02
No Reason Given or Recalled by Respondent
12
0.02
Other
2
0.00
Total Stops
516
100%
Chi-Square=69.63, 6df, p=.000

White
159
71
26
74
2
4
3
339

0.47
0.21
0.08
0.22
0.01
0.01
0.01
100%

We repeated the analyses of stops reported in Tables 5.7 (local police) and 5.8 (NCSHP)
separately for speeding and other stops. We reasoned that, at least for the NCSHP, there is less
discretion in speeding stops than in other types of stops. Among NCSHP stops, 61 percent were
for speeding. Among local police stops, only 44 percent were for speeding. Although there still
were significant racial disparities in stops, the racial gap in speeding stops by the NCSHP after
207
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

controlling for driver demographics and behavior was lower (odds of 1.44 with DMV weights)
than the racial gap for non-speeding stops (odds of 1.54 with DMV weights). Among local police
the racial coefficients were virtually identical for speeding and non-speeding stops. These
analyses are consistent with the conclusion that racial disparity may be more widespread among
local police than among the NCSHP and that the NCSHP speeding stops display a lower level of
unexplained racial disparity.
Table 5.10 reports the outcome of police stops. The top panel of the table focuses on the
distribution of citations, written warnings, and verbal warnings. There are no statistically
significant racial differences in the distributions of citations, written warnings, and verbal
warnings for all stops, local police stops, or NCSHP stops. Local police are more likely to give
verbal warnings and NCSHP troopers are more likely to issue citations, but neither display
significant racial differences in these outcomes.
We also asked those who had been stopped if they felt they had been treated with respect
by the officer who stopped them. While most people felt they had been treated with respect,
African Americans were 5.6 percent less likely to say they were treated with respect than were
white drivers. The racial gaps in reported respect were nearly identical for local police and for
the NCSHP. These later comparisons were not statistically significant, because of the small
sample size of stops within each category of police. The general pattern is clear, however. Most
drivers report respectful treatment, but African Americans are slightly more likely to report a lack
of respect in the officer’s treatment of them after a stop.

208
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 5.10. Frequency Distribution of Stop Outcomes for First Stops, North Carolina Driver Survey
All First Reported Stops
Local Police Stops
NCSHP Stops
African
African
African
American
White
American
White
American
White
Did you get a traffic ticket, a warning ticket, or just a verbal warning?
Citations
52.0%
50.0%
48.1%
42.2%
59.3%
69.2%
Written Warnings
13.1%
18.1%
12.5%
15.6%
14.2%
21.4%
Verbal Warnings
35.0%
31.9%
39.4%
42.2%
26.5%
18.4%
Significant Racial
No
No
No
Difference?
Do you thing that you were treated by the officer with respect during this stop?
Treat With Respect
79.0%
84.6
79.7%
84.1%
77.6%
85.5%
Significant Racial
Yes
No
No
Difference?
Total

329

238

216

135

113

103

Conclusions
According to the analysis of the survey data, African American drivers are significantly
more likely than white drivers to have been stopped by the police in North Carolina. Even after
controlling for other demographic statuses and driving behavior the odds of a stop by local police
may be twice as high for African American as they are for white drivers. Local police are also
significantly more likely to stop African American males relative to African American females,
while among whites there is no gender disparity in stops after controlling for driving behavior.
The estimated racial disparity in stops by the NCSHP is much smaller, but still statistically
significant after controls for driver characteristics and reported driving behavior. The NCSHP
does not stop African American males at higher rates than African American females net of
driving behavior. Among the NCSHP, race is linked to other attributes in the stop decision.
Older whites and whites driving late-model cars are less likely to be stopped than are other
whites. African Americans who report more risky driving behaviors are more likely to be
209
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

stopped. This suggests that the NCSHP troopers are reacting not simply to the race of the driver,
but to the combination of race and other status attributes for whites and race and driving behavior
for African Americans.
After the stop, differences in white and African American experience are less dramatic.
African Americans are slightly more likely to have been informed that the stop was for a more
discretionary reason. African Americans are also slightly more likely to report that they were
treated disrespectfully after the stop. There are no racial differences in the distribution of
citations, written warnings, and verbal warnings. Although there are some racial differences in
experiences after the stop, they are small.

210
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

References for Chapter Five
Bradburn, Norman. 1983. “Response Effects.” Pages 289-319 in Peter Rossi, James
Wright, and Andy Anderson (eds.) Handbook of Survey Research. New York: Academic Press.

211
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Chapter 6 Racial Differences in Trust in the Police

This chapter examines the processes that produce racial differences in trust in the police.
There are substantial racial differences in trust in the police in general and of both local officers
and the NCSHP more specifically. Race differences in trust in the police result not only from
racial differences in stop experiences, but also from the reported stop experiences of friends and
family. Disrespectful interactions between police and citizens can become affronts against the
local community. In addition, African Americans are more likely to believe that the police target
minorities and are less trustful of government in general. These cultural perceptions, which
reflect both historical and personal knowledge of racial bias, are powerful sources of minority
distrust in the police. To rebuild legitimacy in the minority community police forces would need
not only to control racial bias in policing, but also to change the cultural distrust that arises from
past experiences and group history.

Introduction
While racial disparity in police stops is clearly a problem in its own right, both the
experience and perception of police bias may contribute to minority distrust of law enforcement.
Police forces with no or little racial bias in their ranks might still be perceived as untrustworthy
by minority citizens. Even majority citizens who believe that police discretion is sometimes
linked to status characteristics, rather than to driving behavior, may have reduced trust in the
police. The “driving while black” phenomenon has a powerful perceptual dimension, which can
be expected to threaten police legitimacy independently of actual officer behavior. In this chapter

212
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

we examine citizens’ level of trust in the police. As in the previous chapter, we focus on the
police overall, local police, and the NCSHP (relatedly, Appendix F discusses some of the findings
of our focus groups with citizens about various issues of racial profiling).
Surveys have a long history of use in measuring citizens’ trust in the police and other
government institutions. It is possible to use a survey to establish the degree to which the
experiences of interaction with the police and the more generic belief in racial profiling might
influence the legitimacy of the police force in its citizens’ eyes. Trust in government institutions
in general, and any specific police force, will be influenced by the history of relationships
between groups and those institutions as well as more direct experiences and current events.
Surveys are potentially useful for establishing the degree to which current police-citizen
encounters, beliefs or exposure to current controversies such as “driving while black,” or more
deep seated dispositions are in play.

Background
In the focus groups that we conducted with African American drivers, it was clear that
African Americans’ perceptions of racial bias in policing were influenced not only by their own
stop experiences, but also by the stop experiences of family and friends. In addition, the focus
groups revealed that a generic perception of racial bias in policing – a perception we believe was
produced by a culture in which many African American citizens expect racial bias by the police
because they expect racial bias in many or all institutions in the United States. Focus group
participants would also point toward accounts of racial profiling in New Jersey or the Rodney
King beating as evidence that racism is widespread among police in general. This suggests to us
that trust in the police is a function, not only of personal experiences with the police, but also the
213
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

experiences of family and friends and more general cultural perceptions of bias by the police and
the integrity of institutions.
Previous research has demonstrated repeatedly that trust in the police is lower among
minorities than among white citizens (Decker 1981; Flanagan and Vaughn 1996; Weitzer and
Tuck 1999). This research, however, has rarely had direct measures of police encounters and has
never included measures of police encounters in the respondent’s network of family and friends.43
In addition, previous research shows that white and minority citizens believe that racial targeting
is widespread, but minorities are more likely to hold this belief (Gallup 1999). No previous study
has examined perceptions of other forms of targeting by police. Some previous research has
shown that well publicized controversial police incidents (for example, the Rodney King beating)
tend to reduce trust in the police generally, particularly among minorities (Tuch and Weitzer,
1997; Lasley 1994). We reason that belief in racial profiling may have a similar effect. A recent
study has shown that middle class minorities are more likely to distrust the police than lower class
minorities (Weitzer and Tuck 1999). Finally, many studies have shown that minorities have
lower levels of trust in government institutions than do white citizens (Feagin and Sikes 1994).
We use a general trust scale in the analyses below to control for general cultural influences on
trust in social institutions that may influence specific trust in police officers.
In this chapter we first describe racial differences in trust of the police, perceptions of
police stop discretion tied to status characteristics, and more general trust in government
institutions. We then model trust in the police as a function of race, stop experiences, the stop

Warr and Ellison (2000) have recently pointed out that research on fear of crime has also
neglected personal network influences on fear. They find that a prime source of fear of crime is
fear for the safety of others.
43

214
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

experiences of friends and family, general trust in government institutions, and perceptions of
both racial and other types of bias in policing.

Race Differences in Trust of the Police
The first panel of Table 6.1 reports racial differences in trust in the police. The first
question was “On a scale from 1 to 5, where 1 is always fair and 5 is never fair, do you trust the
police to treat you fairly?” In general, whites report trust levels that average close to 2—
“sometimes fair.” African Americans are on average closer to 3—“neutral.” These differences
are quite similar for all three items, although both African Americans and whites report slightly
more trust in the NCSHP than in their local police forces. These results look quite similar to
results reported by Weitzer and Tuck (1999). In the analyses that follow we examine the
processes that generate distrust of the police in general using a Police Distrust Scale made up of
the three items listed in Table 6.. We also analyze the processes separately for the items
indicating distrust in the respondent’s local police and in the NCSHP.

Race Differences in Personal and Network Stop Experiences
The second panel of Table 6.1 shows that African Americans report significantly more
stops per year of driving experience, more stops in the last year, and being treated with less
respect during stops in the last year than do whites drivers.
We also asked respondents to report stops they had heard about that had been experienced
by members of their household and by friends and acquaintances. We do not think that these
responses represent an accurate count of the stop experiences of friends and family. Rather, we
215
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

see these as indicators of the respondent’s stock of stories about police behavior in his or her
immediate network. We do not know the racial composition of household members or friends
and acquaintances, but assume that the networks of white respondents are predominantly white
and the networks of African American respondents are primarily African American.
In the third panel of the table, we see that African American and white citizens report
comparable numbers of stops of household members and of friends and acquaintances. Although
few people report acts of disrespect by the police, African Americans report significantly higher
levels of disrespect in police behavior during stops of both household members and of friends. In
general the levels of reported disrespect are lower in these reports of network experience than
they are in reports of self-experience. Thus, while African Americans hear more stories of
disrespectful treatment than do whites, these stories are not widespread in either community.
These results suggest to us that the damage to police reputation that is produced by
possible racial disparity in stop behavior is not necessarily limited to the person who is stopped,
but to some extent becomes part of the community context in which citizens live. The results also
suggest that the damage done by possible racial disparity in stops is likely to be magnified in the
retelling by friends and family. Since the reported levels of disrespect in the experiences of
friends and family are lower for African Americans than in their own reported stops, it may be the
case that the consequences of racial disparity in police stops are somewhat attenuated in the
retelling. On the other hand, individuals who have never been treated with disrespect by the
police may hear stories about disrespectful behavior, which are then interpreted as evidence of
racial bias in policing.

216
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Distrust of Government and Belief in Profiling
The items that are used in the general trust in government scale are, again, coded from 1—
“always fair,” to 5—“never fair.” For all of the items, except distrust of county commissioners,
African Americans have significantly higher mean scores than whites, signifying higher levels of
distrust in these institutions. Distrust of county commissioners is particularly high among all
citizens, regardless of race. In general, distrust of the police is not markedly higher in the African
American community than is distrust of other government officials. Among whites, however,
distrust in the police is quite a bit lower than distrust of other government officials.
We also asked these citizens about their perceptions that the police profile drivers. We asked:
Since many drivers speed or otherwise break the traffic laws, it is sometimes hard to tell why any
one person gets pulled over by the police. Do you think that the following kinds of drivers are
more likely to be pulled over by police than other drivers: young drivers? men? African
Americans? Latinos? people driving run-down cars? people driving flashy cars?
Perceptions of racial disparity are widespread among African Americans. Eighty-one
percent of African Americans believe that the police are more likely to pull over African
American drivers. Just under 70 percent of African Americans perceive similar bias against
Latino drivers. In contrast, less than a third of white drivers believe that there is racial bias
against African Americans. Slightly more whites perceive bias against Latinos by the police.
Looking at the racial bias scale (which sums the previous two items and divides by two), we see
that whites are 40 percent less likely than African Americans to perceive racial bias in police stop
decisions.
The last panel of Table 6.1 reports racial differences in other possible types of police
profiling. In all cases, African Americans are more likely than whites to suspect bias in police
217
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

stops, although belief in the profiling of run down cars is not significantly different by race. The
racial gap in beliefs in other forms of police bias averages 20 percent across the scale items, half
as large as the racial gap in beliefs in racial bias in policing. One of the most interesting contrasts
in the table is that a higher proportion (13.2 percent) of white respondents believe that the police
target for non-racial reasons than believe that police are racially biased. Among African
Americans the pattern is reversed, racial profiling is perceived as more extensive than other forms
of profiling by 8.4 percent. The racial gap of the belief that police profile men is almost as large
as the racial gap in the belief in racial profiling.44
The racial comparisons in Table 6.1 suggest that African Americans’ higher levels of
distrust in the police could arise from racial differences in stop experiences and in the stop
experiences discussed in their personal networks. It also seems plausible to expect that racial
differences in perceptions of racial targeting and distrust in government may also contribute to the
higher levels of African American distrust of the police. Whites report higher levels of belief in
non-racial targeting by the police. It is not clear how this might influence racial differences in
trust in police. The differences in white and African American responses to the two profiling
scales suggests that splitting the models by race might be particularly revealing.

In Chapter 5, we saw no gender differences in stops by the NCSHP after controlling for driving
behavior. This was also the case among whites stopped by the local police. Among African
American stops by the local police, males were at significantly higher risk than females, even
after controlling for driving behavior. The higher perception of male profiling among African
Americans than whites may be produced by this pattern of local police stops. It, of course, could
also be produced by the general perception in the African American community that African
American men are particularly vulnerable in contemporary American society.
44

218
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Modeling Distrust of the Police in a Multivariate Context
We model the process that gives rise to distrust in the police in a multivariate statistical
context. We use two types of statistical models—ordinary least squares and ordered logistic
regression. OLS models are appropriate for continuous dependent variables. The general distrust
in police scale clearly meets the requirement of a continuous variable as does the general trust in
government scale. The analyses of trust in the local police and the NCSHP are five category

Table 6.1 Race Differences in Distrust of the Police, Stop Experience, Trust in Institutions, and Perceptions of Police Bias
African
Significant
American
White
Difference?
Distrust of Police in General
2.63
1.95
Yes
Distrust of Local Police
2.69
2.04
Yes
Distrust of NCSHP
2.53
1.89
Yes
Police Distrust Scale (Alpha=.809)
2.63
1.96
Yes
Lifetime Stops/Years of Driving
.28
.24
Yes
Number of Stops in Last Year
.44
.25
Yes
Treated with Disrespect During Stop Last Year
6.7%
3.0%
Yes
Number of Household Stops in Last Year
.25
.21
No
Household Members Report Disrespect in Stops
2.5%
1.6%
Yes
Number of Friends Stops in Last Year
.76
.85
No
Friends Report Disrespect in Stops
5.7%
1.8%
Yes
Distrust in Teachers
2.29
2.03
Yes
Distrust in County Commissioners
2.81
2.77
No
Distrust in Judges
2.62
2.24
Yes
Distrust in Congress
2.68
2.53
Yes
General Distrust of Government Scale (Alpha=.746)
2.61
2.40
Yes
African Americans are Profiled
80.8%
32.7%
Yes
Latinos are profiled
69.3%
36.8%
Yes
Belief in Racial Profiling Scale (Alpha=.815)
75.8%
35.5%
Yes
Run down cars are profiled
43.7%
41.2%
No
Flashy cars are profiled
78.3%
60.6%
Yes
Men are Profiled
66.3%
29.3%
Yes
Young are Profiled
81.9%
66.9%
Yes
Belief in Other Forms of Profiling Scale (Alpha=.458)
67.4%
48.7%
Yes

ordinal scales. We use ordered logistic regression for these two items, as well as for our analyses
of racial and general profiling beliefs. In both types of models a positive coefficient indicates that
increases in the explanatory variable (such as a belief in racial profiling) leads to an increase in
distrust of the police. A negative coefficient indicates that an increase in the explanatory variable
219
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

leads to increased trust in the police (for example, age). For all explanatory variables mean
values are imputed for missing values. We include indicator variables for mean substitution in
regression equations but do not report their values in the tables. These models are weighted to
correspond to the 2000 N.C. Department of Motor Vehicles age-gender distributions within
race.45
The logic of the analyses is to first establish the size of the racial gap in trust of the police,
net of the set of driver demographic and behavior variables employed in the previous chapter
(racial differences in these variables were discussed in the previous chapter and reported in Table
4 of that chapter). The demographic variables (gender, education, home ownership, car age,
individual age, and rural residence) allow us to statistically adjust racial differences in trust of the
police for class, gender, and regional influences on distrust in the police that may be correlated
with race. They also are indicators in their own right of other aspects of a citizen’s identity that
may influence distrust of the police. Driving behaviors (speeding, miles driven, risky driving,
methods to avoid tickets, and interstate frequency) are included in these initial models to adjust
for risk behaviors that may influence trust in the police. We reason that people who break the
driving laws may be less trustful of police because they fear police stops.
The second model in the following analyses introduces the respondent’s reports of his or
her and his or her friends’ and family members’ stop experiences regarding distrust in the police.
We measure the number of stops per driving year and in the most recent year for respondents, but
only stops in the last year for their family and friends. We also include three measures of
disrespectful treatment during stops corresponding to personal, family, and friends’ stops. If

The non-response weights used in the last chapter are not appropriate for these analyses. They
are appropriate only for weighting reports of stops.
45

220
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

coefficients in Model 1 are reduced in magnitude in Model 2 this indicates that the Model 1
estimates partly reflect and are caused by differences in stop experiences.
The third model for our analyses of trust in the police introduces the measure of trust in
general government institutions as well as beliefs in racial and non-racial targeting by the police.
If coefficients in Model 2 are reduced in magnitude in Model 3, this indicates that the Model 2
estimates partly reflect and are caused by differences in general trust and beliefs in police bias.
Finally, Model 3 is estimated separately for African American and white citizens to see if the
processes that lead to distrust in the police are the same for African American and white citizens.
Because general distrust of government institutions and belief in racial and other forms of
police profiling are such strong influences on distrust of the police, we also model them as
outcomes of demographic and driving variables. We begin with an analysis of racial differences
in trust of government officials other than the police and belief in racial and non-racial profiling.

Distrust of Government Officials Other than the Police
The variable indicating distrust in government officials is an additive scale based on level
of distrust in teachers, county commissioners, judges and congressional representatives. In the
analyses of distrust of the police that follows we introduce this variable to account for differences
in generic trust in government that might also spill over into distrust of the police in particular.
We first regress the distrust in government scale upon the characteristics of drivers and stop
experiences to see if these characteristics influence generic distrust in government.
Table 6.2 reports the models of generic government distrust. African Americans have
significantly higher distrust of government officials. The racial gap of .25 is slightly higher than
the racial gap reported in Table 6.1. In addition, female and young African American drivers
221
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

distrust government officials at higher rates (Model 3). African American and white risky drivers
distrust government officials at higher rates.
Being personally treated with disrespect by police officers during a stop in the past year
strongly increases generic distrust of government officials. When the models are split by race, we
find that both African Americans’ and whites’ trust in government is undermined by disrespectful
treatment by the police during a stop. In addition, African Americans who hear of disrespectful
police behavior toward friends and acquaintances trust government officials less.
In general, the table suggests that there are relatively large racial gaps in trust of
government institutions that are not a function of other driver statuses or of driving behavior.
Rather, distrust of government in general is increased when police officers interact with citizens
in ways that are interpreted as disrespectful. Hearing stories of disrespectful police behavior from
friends and acquaintances increases distrust of government for African Americans.

Belief in Racial Profiling
The variable indicating belief in racial profiling is an additive scale based on belief in
racial profiling of African American and Latino drivers. Models are estimated using ordinal
logistic regression because this variable only has three categories. The first model of Table 6.3

222
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 6.2 Regression of Trust in Government Scale on Race, Driver Characteristics, and Stop
Experiences, N=2,284
Model 1
Model 2
Model 3
Model 4
.249***
African
Race (1=African American)
.221***
American
White
Gender (1=Female)
.072*
.093*
.215***
-.003
Education
-.003
.001
.012
-.011
Home Owner
-.007
-.013
.020
-.063
Model Year of Car
-.002
.000
-.000
-.001
Age
-.004**
-.000
-.006**
.000
Rural
-.017
-.015
-.022
-.011
Scale Speeding 5+
.012
.011
.004
.023
Miles Driven Last Year (LN)
.039**
.032*
.023
.051*
Risky Driving Scale
.068***
.065***
.070**
.054**
Fewer Methods to Avoid Ticket Scale
-.027
-.024
-.048
.005
Interstate Frequency (low score=high
-.005
-.005
.008
-.017
usage)
Number of Stops per Driver Year
Number of Stops in Last Year
Treated with Disrespect During Stop Last
Year
Number of Family Stops in Last Year
Family Members Treated with Disrespect
Number of Friend Stops in Last Year
Friends Treated with Disrespect
Adjusted R2
* .05; **.01; ***.001

.052

.021
.015
.455***

.040
.007
.482***

.018
.026
.427**

.014
.142
.006
.176
.075

.028
.155
-.012
.268*
.090

.005
.096
.019
-.003
.051

regresses belief in racial profiling by the police upon the respondent’s race and other driver
characteristics. African Americans are much more likely than whites to believe that the police
profile on the basis of race even after controls for other status characteristics and driving
behaviors. The much higher African American belief in racial profiling is only minimally a
function of class, gender, rurality, or reported driving behavior differences between African
Americans and whites.
Higher education is associated with higher belief that the police profile on the basis of
race among both African American and white citizens, although the effect is much stronger

223
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

among African Americans. Urban residents of both races are significantly more likely than rural
residents to believe in racial profiling. Whites who report typically speeding more than five miles
an hour above the speed limit are more likely to believe in racial profiling than more law abiding
white drivers. African Americans who actively try to avoid tickets more are less likely than other
African Americans to believe in racial profiling by the police.
Introducing the measures of stop experiences provides some interesting results. Both
African American and white drivers who were treated with disrespect during a police stop are
more likely to believe the police profile on the basis of race. In addition, African Americans
whose family members experience more stops and whose friends report being treated with
disrespect are more likely than other African Americans to believe that the police profile on the
basis of race. Thus for African Americans, but not whites, belief that the police profile on the
basis of race is related to the recent experiences of members of their community with the police.
An inspection of the R2 across models makes clear that race is by far the strongest predictor of
belief in racial profiling.

Belief in Other Forms of Profiling
The variable indicating belief in other forms of profiling is an additive scale based on
belief in police profiling of young and male drivers as well as profiling of run-down vehicles and
flashy cars. We use ordinal logistic regression to estimate the models presented in Table 6.4
because belief in other forms of profiling is a five category ordinal scale. Controlling for

224
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 6.3 Ordinal Logistic Regression of Belief in Racial Profiling on Race, Driver Characteristics, and
Stop Experiences (N=2,830)
Model 1
Model 2
Model 3
Model 4
African
Race (1=African American)
1.83***
1.789***
American
White
Gender (1=Female)
-.052
-.022
.023
.086
Education
.112***
.119***
.164***
.062**
Home Owner
.001
-.013
.004
.075
Model Year of Car
-.016
-.012
.009
.013
Age
-.002
.002
.005
.001
Rural
-.121***
-.121***
.009
-.150***
Scale Speeding 5+
.113**
.095*
.024
.128*
Miles Driven Last Year (LN)
-.022
-.035
.006
.097
Risky Driving Scale
.065
.006
.078
.066
Fewer Methods to Avoid Ticket Scale
-.171
-.184
-.447**
.007
Interstate Frequency (low score=high
.004
usage)
-.005
.004
.000
Number of Stops per Driver Year
Number of Stops in Last Year
Treated with Disrespect During Stop Last
Year
Number of Family Stops in Last Year

.097
-.039

.151
.061

.084
.002

.744**

.850*

.722*

.160*

.129

.151

.457

.588

.297

.024
.791**
.257

.052
1.04**
.096

.063
.320
.064

Family Members Treated with Disrespect
Number of Friend Stops in Last Year
Friends Treated with Disrespect
Pseudo R2
* .05; **.01; ***.001

.241

demographic and driving characteristics, African Americans are more likely to believe in these
non-racial forms of profiling as well. African American women are less likely than African
American men to believe that the police profile on these other driver characteristics. African
American homeowners are less likely to believe in these forms of profiling, while older African
Americans are more likely to think that profiling is about characteristics other than race. Whites
who report typically speeding while driving are more likely to report that police profile on nonracial dimensions.

225
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

The only statistically significant stop experience reported in the model for whites indicates
that respondents whose family members were stopped more in the last year are more likely than
other whites to believe that the police profile on these non-racial characteristics. This latter
finding is particularly interesting in light of the previous analysis. African Americans’ belief in

Table 6.4. Ordinal Logistic Regression of Belief in Other Forms of Police Profiling on Race, Driver
Characteristics, and Stop Experiences (N=2,830)
Model 1
Model 2
Model 3
Model 4
African
Race (1=African American)
1.16***
1.126***
American
White
Gender (1=Female)
-.228**
-.202*
-.409***
-.078
Education
-.035
-.032
-.018
-.041
Home Owner
-.163
-.175
-.248*
-.149
Model Year of Car
-.009
-.007
.002
-.011
Age
-.001
.001
.014**
-.008
Rural
-.049
-.005
-.056
-.052
Scale Speeding 5+
.094*
.080*
.000
.134*
Miles Driven Last Year (LN)
-.030
-.044
-.048
-.040
Risky Driving Scale
.027
.016
.028
-.028
Fewer Methods to Avoid Ticket Scale
-.062
-.045
-.004
-.090
Interstate Frequency (low score=high
-.002
.003
-.016
-.056
usage)
Number of Stops per Driver Year
Number of Stops in Last Year
Treated with Disrespect During Stop Last
Year
Number of Family Stops in Last Year

.131
.021

.105
.049

.147
.051

.133

-.094

.624

.116

-.010

.261**

.029

-.001

.107

.037
.285
.134

.101*
.409
.050

-.018
-.294
.053

Family Members Treated with Disrespect
Number of Friend Stops in Last Year
Friends Treated with Disrespect
Adjusted R2
* .05; **.01; ***.001

.126

racial profiling is increased dramatically by the experiences of their friends and family. While not
as dramatic, white belief in other forms of profiling is enhanced by family stop experiences.

226
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

General Distrust of the Police
Table 6.5 reports the multivariate analysis of the general distrust in police scale. This scale
was created by summing the three police trust items described in Table 6.1 and dividing by three
to preserve the original metric. We estimate the models using ordinary least squares regression as
the untransformed scale varies from 3 to 15. Model 1 shows that African Americans are .63 units
more distrustful of police than are whites, even after controlling for other demographic statuses
and driving behaviors. This result is very similar to the bivariate relationship reported in Table
6.1, and it indicates that African Americans are significantly more likely to distrust the police
than are whites. This result holds up across all three models, although the total racial differences
in distrust declines in Models 2 and 3. On balance, this suggests that the racial gap in trust in the
police is partly created by police encounters with the respondent and his or her family and friends,
and partly by the more generalized beliefs in police profiling and distrust of government
examined in the previous two tables.
There are three additional statistically significant effects in Model 1. Older drivers show
lower distrust of the police. It is also the case that faster drivers and those reporting more risky
behaviors are more distrustful of the police. As we expected, drivers who break the law routinely
are less trustful of police than their more law abiding counterparts, suggesting perhaps an
understanding of their increased vulnerability. For speeding, this effect is limited to whites.
Model 2 introduces the indicators of personal and network stop experiences. The racial
gap in stops declines slightly relative to Model 1, suggesting that personal and network stop
experiences produce about 8 percent (.58/.63) of the heightened minority distrust of the police.
Stop experiences also reduce to non-significance the effects of speeding behavior on police
distrust. Speeding increases the probability of a stop, which in turn influences distrust in the
227
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

police. In the split models, however, whites who speed remain particularly suspicious of the
police.
The more lifetime stops per year of driving, the less the trust in the police, but this effect is
particularly present among white drivers. Stops in the past year, either of the respondent or of his
or her family and friends, have no effect on trust in the police—with one exception. Surprisingly,
whites who have been personally stopped in the last year show increased trust in the police.
Being treated with disrespect by an officer during a stop in the last year, however, dramatically
increases distrust of the police for both African Americans and whites. Similarly, and only
somewhat less dramatically, disrespectful treatment during stops to family or friends increases the
level of distrust in the police.
One way to interpret these results is that it takes multiple stops over a lifetime to damage
the legitimacy of the police, but it takes only a single act of disrespect during a citizen encounter
to lower citizen trust. Acts of disrespect are likely to be shared with family and friends and also
to lower trust in the police across the whole acquaintance network of the motorist who has been
treated poorly.
Model 3 introduces the measures of trust in other government institutions and the two
measures of perceptions of police targeting of drivers. The racial gap in trust is still statistically
significant but drops to .35 in this model. Thirty-nine percent of the observed racial gap in trust
in the police is directly tied to these three measures. Perceptions of racial and non-racial targeting
by the police, as well as historically produced distrust of government institutions are powerful
sources of distrust of the police. The effect of general distrust of government officials is three
times larger than the effects of belief in either profiling scale. Ending racial disparity in stops,

228
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

even ending perceptions of racial bias, can only be a first step in building community trust in the
police.
The risky driving scale is no longer significant. As we saw in Table 6.2, risky drivers are
less trustful of government in general. The effects of the three indicators of police disrespect are
all reduced in Model 3. Again, as we saw in Tables 6.2 and 6.3, disrespectful treatment by the
police increases distrust of government in general and the belief that police are targeting minority
drivers.
Models 4 and 5 split the sample by race and estimate Model 3 again. The processes which
generated distrust in the police are not dramatically different by race. As noted previously, even
after controlling for general trust in government, whites who routinely speed are less trustful of
the police. African American drivers who are more active in scanning the road for speed traps are
less trusting in the police than other African Americans.
Disrespectful treatment of drivers, or their friends and family, increases distrust of the
police among both African Americans and whites. Belief systems influence distrust in the police
similarly for African American and white citizens. Among African Americans distrust of the
police is strongly influenced by their general distrust of government officials and belief in racial
profiling, but not by their belief in non-racial police profiling. Whites’ distrust of the police, on
the other hand, is also influenced by belief in non-racial profiling.

229
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 6.5. Ordinary Least Squares Regression of Distrust in Police Scale on Race, Driver Characteristics, Stop
Experiences, Trust in Institutions, and Perceptions of Police Bias
Model 1
Model 2
Model 3
Model 4
Model 5
African
Race (1=African American)
.632***
.580***
.354***
American
White
Gender (1=Female)
-.007
.043
.006
.013
-.012
Education
-.003
.046
-.001
.010
-.011
Home Owner
-.038
-.046
-.034
-.049
-.018
Model Year of Car
-.007
-.003
-.024
.000
-.004
Age
-.009***
-.006***
-.005***
-.005**
-.004**
Rural
-.018
-.013
.001
.020
-.015
Scale Speeding 5+
.048**
.031
.020
-.008
.041*
Miles Driven Last Year (LN)
.002
-.011
-.021*
-.025
-.004
Risky Driving Scale
.063***
.057***
.021
.022
.022
Fewer Methods to Avoid Ticket Scale
-.046
-.061
-.039
-.110*
.031
Interstate Frequency (low score=high
.005
.003
.007
.011
.002
usage)
Number of Stops per Driver per Year
Number of Stops in Last Year
Treated with Disrespect During Stop Last
Year
Number of Family Stops in Last Year

.169***
-.036

.137***
-.040*

.051
.005

.206***
-.112***

.718***

.463***

.355***

.658**

.016

-.007

-.027

.016

.550***

.456***

.415**

.565***

-.007
.572***

-.014
.450***

.000
.272***

-.019
.691***

.617***
.210***

.676***
.231***

.553***
.181***

.174**

.133

.217**

.479

.452

.380

Family Members Treated with Disrespect
Number of Friend Stops in Last Year
Friends Treated with Disrespect
Generalized Distrust of Government
Belief in Racial Profiling Scale
Belief in Other Forms of Profiling Scale
Adjusted R2
* .05; **.01; ***.001

.177

.231

Distrust of the Local Police
Table 6.6 repeats the analyses presented in Table 6.5, but the trust refers specifically to the
respondent’s local police force, rather than police in general. The results in Model 1 are nearly
identical to those in the previous table. African Americans, the young, risky drivers, and self230
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

reported speeders are all more distrustful of the police as in the previous table. Adding stop
experiences in Model 2 also produces nearly identical results to those in the previous table. More
lifetime stops increase distrust of the local police, as do disrespectful interactions with the police,
either with the respondent or with his or her family and friends. Again the effect of speeding
behavior on distrust is mediated by stop experiences. Model 3 introduces the measure of general
distrust and the two measures of belief in police profiling. Again these are powerful predictors of
distrust in the local police. They also mediate a large share of the racial coefficient, suggesting
that these more general cultural perceptions which are strongly linked to race are powerful
sources of minority distrust in their local police. Again, being a risky driver is no longer a
significant predictor, suggesting that they tend to distrust government in general (refer back to
Table 6.2).
Models 4 and 5 compare the processes generating distrust of the local police for African
American and white drivers. The coefficients for demographic and driving behavior variables are
quite similar with one exception. Risky white drivers are particularly distrustful of the police,
even net of controls for general distrust of government. More dramatically, stop experiences
produce much larger coefficients for whites than for African Americans. While both African
American and white drivers’ trust in their local police is diminished by disrespectful treatment,
this is much more so the case for white drivers. Similarly, lifetime stops per year reduces white,
but not African American, trust in the police significantly and dramatically. Finally, both African
American and white distrust of local police is tied to their distrust of government in general, but
belief in racial profiling is a more important source of distrust among African American drivers,
while belief in other forms of profiling by the police undermines white trust.

231
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 6.6 Ordinal Logistic Regression of Distrust of Local Police on Race, Driver Characteristics, Stop Experiences, Trust in
Institutions, and Perceptions of Police Bias (N=2,830)
Model 1
Model 2
Model 3
Model 4
Model 5
African
Race (1=African American)
1.11***
1.036**
.732***
American
White
Gender (1=Female)
.036
.117
.064
.052
.048
Education
-.012
-.024
-.008
.006
-.020
Home Owner
-.132
-.151
-.135
-.197
-.039
Model Year of Car
-.009
-.003
-.001
.002
-.004
Age
-.018***
-.013***
-.012***
-.014***
-.009*
Rural
-.021
-.014
.003
.027
-.025
Scale Speeding 5+
.099**
.073
.058
.024
.077
Miles Driven Last Year (LN)
.053
.029
.002
-.006
.035
Risky Driving Scale
.152***
.143***
.071*
.027
.130**
Fewer Methods to Avoid Ticket Scale
-.067
-.071
-.031
-.226
.149
Interstate Frequency (low score=high usage)
.015
.015
.021
.011
.030
Number of Stops per Driver per Year
.329***
-.041

Number of Stops in Last Year
Treated with Disrespect During Stop Last
Year
Number of Family Stops in Last Year

1.221***

.361***
-.060
.819***

.207
.007

.528***
-.270**

.659**

1.095***

-.016

-.031

-.060

.018

1.006***

1.099***

-.003
1.103***

-.019
.960***

.007
.627*

1.671***
.373***

1.762***
.413*

.359*

.096

.673**

.455

.453

.382

Family Members Treated with Disrespect
1.002**

1.288***

Number of Friend Stops in Last Year
Friends Treated with Disrespect

-.018
1.810***

Generalized Distrust of Government
Belief in Racial Profiling Scale
Belief in Other Forms of Profiling Scale
Psuedo R2
* .05; **.01; ***.001

.149

.202

1.620***

Distrust of the North Carolina State Highway Patrol
Table 6.7 repeats the previous analyses, but now trust refers specifically to distrust of the
NCSHP. Models 1 through 3 reprise the results we have already encountered in the general
distrust and local police analyses. Distrust of the NCSHP seems to be governed by the same

232
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

.302*

generic process that leads to distrust in the police in general. The one interesting new finding is
that African Americans who drive more miles per year tend to have increased trust in the NCSHP.
Models 4 and 5 reveal some interesting differences in the process that generates distrust in the
NCSHP between African Americans and white drivers. African American trust in the NCSHP is
not influenced by lifetime stops or personally being treated with disrespect during a recent stop.
Both of these experiences increase white distrust of the NCSHP. Similarly, belief in racial and
non-racial profiling increase white distrust of the NCSHP but not African Americans’ distrust.
These racial differences in the processes that generate distrust of the NCSHP suggest that African
American’s trust in the NCSHP is less vulnerable to either personal experience or the current
controversy around police profiling. This is certainly consistent with the findings in the last
chapter that racial disparity in stops by the NCSHP is much lower, and a much more subtle
process, than racial disparity in local police stops. These findings are also consistent with our
focus groups with white and African American drivers. In those focus groups white drivers did
not clearly differentiate between different types of police. African American drivers, on the other
hand, were quite clear that they had considerably more trust in NCSHP troopers to act
professionally than they did of their local police.

Conclusions
Distrust in the police is produced by a combination of negative personal experiences with
the police, negative experiences of family and friends, belief in police profiling on both racial and
non-racial grounds, general distrust of government institutions, and being a minority.

233
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table 6.7 Ordinal Logistic Regression of Distrust of the NCSHP on Race, Driver Characteristics, Stop Experiences, Trust in
Institutions, and Perceptions of Police Bias (N=2,830)
Model 1
Model 2
Model 3
Model 4
Model 5
Race (1=African
African
American)
1.117***
1.071***
.764***
American
White
Gender (1=Female)
.082
.150
.088
.138
.017
Education
.026
.040*
.043*
.067*
.013
Home Owner
.025
-.007
.019
-.050
.120
Model Year of Car
-.009
-.005
-.003
-.001
-.002
Age
-.020***
-.015***
-.017***
-.016***
-.018***
Rural
-.012
-.001
.009
.081
-.060
Scale Speeding 5+
.132**
.115**
.094*
.041
.146*
Miles Driven Last Year
-.011
-.031
-.098**
-.081*
-.066
(LN)
Risky Driving Scale
.090**
.081**
.015
-.015
.045
Fewer Methods to
-.041
-.057
-.008
-.130
.076
Avoid Ticket Scale
Interstate Frequency
(low score=high usage) -.001
-.003
.002
.005
-.009
Number of Stops per
Driver Year
Number of Stops in
Last Year
Treated with Disrespect
During Stop Last Year
Number of Family Stops
in Last Year
Family Members
Treated with Disrespect
Number of Friend Stops
in Last Year
Friends Treated with
Disrespect

.284**
-.065

.271**
-.086

.083
.017

.419***
-.284**

.977***

.551**

.200

1.221***

.066

.012

.031

-.012

1.013***

.915***

.803*

1.162**

-.013

-.036

-.030

-.030

.815***

.682***

.636**

.899*

1.568***

1.667***

1.511***

.307**

.225

.357**

.456**

.180

.698**

.421

.404

.346

Generalized Distrust of
Government
Belief in Racial Profiling
Scale
Belief in Other Forms of
Profiling Scale
Psuedo R2
* .05; **.01; ***.001

.152

.187

234
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

The related problems of racial profiling and trust in the police are not simple ones. African
Americans distrust the police, especially local police, because of their personal experiences and
more general cultural orientations. Disrespectful interactions are particularly powerful sources of
both distrust in the police and belief in racial profiling. This is not, however, simply a perception
produced by direct experience. On the contrary, negative encounters with the police by family
and friends generate distrust and increase the belief in racial profiling. In fact, among African
Americans, disrespectful police treatment or stories of disrespectful police treatment can even
undermine trust in government institutions in general. Belief in racial profiling undermines trust
in the police even among whites.
African Americans are more forgiving of the NCSHP than are whites. African Americans
are more likely to translate negative experiences into distrust of local police forces than the
NCSHP. This may reflect their observations of lower bias or more professional carriage by
NCSHP troopers. Whites, on the other hand, are less discriminating. Any perception of disrespect
or profiling undermines white trust in all types of police. Whites are particularly influenced by
perceptions of non-racial profiling, assumedly because these are the types of profiling for which
they would be most at risk. Thus while African Americans are more distrustful of the police in
general than are white citizens, white’s trust in the police seems more vulnerable to recent
experiences and media portrayals.
Citizen trust in police is also influenced by more general dispositions toward trust in
government. This is true for white and African American citizens and for all types of police
examined. This suggests that the legitimacy of the police in general, and of specific police forces,

235
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

is a nested problem. Police legitimacy is undermined by disrespectful treatment (especially
among whites), and belief in racial profiling (especially among African Americans), and belief in
other forms of profiling (especially among whites). Where racial disparity in treatment is lower,
as in the NCSHP versus local police, African Americans do not translate negative experience into
reduced trust. Police legitimacy is more vulnerable among whites. African Americans, however,
have a lower level of trust in the police of all types stemming from their past experiences in, and
cultural understanding of, American society. Some of this can be seen in African Americans’
lower trust in government institutions in general, but most seems to be focused on a specific fear
of the police. Among whites distrust of the police is more strongly tied to distrust of government
institutions in general.

236
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

References for Chapter Six
Feagin, Joe. 1991.“The Continuing Significance of Race: Antiblack Discrimination in
Public Places.” American Sociological Review. 56:101–16.
Feagin, Joe and Melvin Sikes. 1994. Living with Racism: The Black Middle Class
Experience. Boston: Beacon Press.
Gallup Poll. 1999. “Racial Profiling is Seen as Widespread, Particularly Among Young
Black Men.” http:/www.gallup.com/poll/releases/pr991209.asp.
Lasley, J.R. 1994. “The Impact of the Rodney King Incident on Citizen Attitudes Toward
Police.” Policing and Society. 3:245–55.
Tuch, Steven and Ronald Weitzer. 1997. “The Polls—Trends: Racial Differences in
Attitudes Toward the Police.” Public Opinion Quarterly 61:642–63.
Warr, Mark and Christopher Ellison. 2000. “Rethinking Social Reactions to Crime:
Personal and Altruistic Fear in Family Households.” American Journal of Sociology.
106:551–578.
Weitzer, Ronald and Steven Tuch. 1999. “Race, Class and Perceptions of Discriminaiton
by the Police.” Crime & Delinquency 45:494–507.
Weitzer, Ronald. 1999. “Citizen’s Perceptions of Police Misconduct: Race and
Neighborhood Context.” Justice Quarterly. 16:819–846.

237
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Chapter 7 Discussion and Conclusions
When this project began, we had two goals in mind: to determine if NCSHP troopers
engaged in “racial profiling” in making their routine traffic stops and to determine what citizens
thought about racial profiling (perceptions of bias prevalence, attitudes toward the NCSHP,
degree of distrust of police, and so forth). As we began collecting and analyzing the data, we
realized that these general goals could be broken down into several sub-goals: 1) to determine if
there were disparities in the stops and citations of African Americans after controlling statistically
for the deployment of troopers (by place and time); 2) to determine if the troopers specializing in
conducting searches for contraband (the CIT ) stopped and conducted searches based on the race
of the person stopped; 3) to develop methods to provide for information as to districts and
troopers with racially disparate intervention rates (to be used in conjunction with other indicators
of bias, such as citizen and/or trooper complaints); 4) to determine the extent to which surveys of
the general public regarding traffic violations and stops can help shed light on the processes
involved; 5) to assess the attitudes of the general public in their perception of racial bias and
profiling and to see if the authority of the NCSHP was being undermined by the perceptions of
widespread racial disparity in stops, citations and searches; and 6) to assess the points of view of
the NCSHP troopers themselves as they discussed possible racial bias and profiling. We will
discuss each of these goals in turn.

Official Record Evidence of Disparity
In assessing the prevalence of racial disparity, it is necessary to differentiate between the
usual statistical information available to state police organizations (essentially statewide totals of
238
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

the number and percentage of interventions involving African Americans) and information based
on, and required for, a more detailed analysis involving appropriate statistical control variables
that account for the deployment of troopers relative to the racial demographic characteristics of
the drivers on the highway. Data on stops, citations, and written warnings that allow for a simple
statistical summary for the state as a whole show evidence of racial disparity against African
Americans. For example, even though African Americans make up approximately 20 percent of
the drivers in the state, they constitute almost 25 percent of the drivers cited for speeding. In
general, we find overall levels of disparity of the magnitude of 5 to 10 percent in absolute terms,
and up to 50 percent in relative terms (for example, a 20 percent African American driver
baseline and a 30 percent rate of stops/citations of African Americans for license, registration, or
insurance violations represent a 10 percent absolute difference or a 50 percent relative
difference—10% absolute difference/20% African American baseline. It should be noted that we
do not find evidence of larger disparity, such as that suggested by Lamberth (2001), nor do we
find that the disparity increases when we control for the measures of the deployment of troopers.
Throughout the analysis of the official record data, we are limited by the fact that we have
baselines that only approximate the actual volume of drivers on the highways (so called “proxy
measures”) -- and those drivers are assumed to be “available” for a stop -- against which to
measure the racial composition of the drivers stopped and cited by the NCSHP. That is, we do
not have direct measures of the racial composition of highway drivers, and thus cannot make
strong claims as to whether or not the racial disparity that we observe is due to racial bias or to
deployment. (By deployment we are referring to the assignment of patrol cars to areas of
highways by time of day). For example, all else being equal, the more patrolling of areas with
more African American drivers, the greater the disparity in the racial composition of those
239
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

stopped, cited or warned. Deployment by time of day is also relevant since we find strong
evidence that the proportion of drivers who are African American varies by time of day (with
proportionately more African Americans on the highway at night).
Analyses of racial profiling are further complicated by researchers’ inability to directly
and precisely measure what troopers actually observe and react to as they patrol the highways.
That is, while driver/vehicle behavior would likely be a crucial piece of information to use in
accounting for the observed levels of disparity, such information on the actual racial distributions
across different infractions are generally unknown. We were able to obtain a first measure of
driver behavior but it was limited to fourteen highway segments (as discussed in Appendix A).
For those limited geographic areas we found that there were differences in the vehicular speeding
behavior of whites and African Americans, with more speeding above what we call “local
speeding thresholds” on the part of African Americans. We do not generalize our findings from
the fourteen sites to other areas, because we cannot rule out the likelihood that there are racial
differences in driving behavior across locales and conditions. Our opinion, as informed by what
little evidence is available on this topic, is that there are many factors involved in the
determination of the racial composition of drivers involved in traffic violations in an area.
Further data on such behavior are necessary for researchers to be able to make truth claims in this
regard.
Besides not having direct measures of violating behavior, we as researchers do not have
direct measures of the deployment of troopers. Instead, we can only assemble a “paper trail” of
stops, citations, and written warnings issued by troopers. Thus, we do not know how many hours
of patrolling occur on a given highway, nor at what precise times. In controlling statistically for
deployment we can only compare citation rates of African Americans to our best available
240
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

baseline measure at the smallest unit of analysis feasible to study. Here we argue that the data on
accidents provide a baseline measure at relatively small units of analysis (as small as what we call
the “county highway area” or the stretch of a highway within about a quarter of a North Carolina
county). We argue that accident data at the level of the county highway area allow us to take a
look into the problem of the mismatch of where NCSHP troopers look for violators and where
violations occur (the so-called “spatial heterogeneity” problem). Yet, our measures will involve
an imprecise match not only because the racial composition of drivers in accidents is not
necessarily an adequate substitute for the behavior of drivers violating the traffic laws, but
because there still could be a substantial mismatch between where the patrols occur and where the
citizens violate the traffic laws. For example, there could be variation within a given highway
segment (such as what we call a county highway area) in the proportion of violating drivers who
are African American because of the intersecting highways which take drivers from, and supply
drivers to, specific segments of the highway. Thus, we may find that on a highway within a
county, such as U.S. 64 in Nash County, that 25 percent of the drivers violating the law are
African American in one segment, and 20 percent in another because the intervening intersection
takes/provides African Americans to/from a town where relatively many African Americans live.
The limited data we have on such variation within highways indicates that 5 percent changes in
the African American composition are common.
An examination of the stops of drivers by the NCSHP on highways where there are
mileposts indicates even more variation in the patrolling patterns of the NCSHP by mile of
highway. Adjacent miles of a given highway can differ by a substantial degree in the number of
stops that occur, and such variation cannot reasonably be attributed only to variation in driving
behavior. The NCSHP may stop drivers more in one segment of a highway than another due to a
241
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

variety of very local considerations, such as turn-around points to the opposite direction, wide
shoulders to pull over vehicles safely, low density of traffic (stopping vehicles in high density
traffic may cause accidents), and so forth. If the choice of such high volume stop zones happens
to coincide with the disproportionate presence of African American drivers, the disparity rate for
an area may rise accordingly.
With all of these considerations possibly at work—and not measured in the data available
to us—one can neither rule out the possibility that racial bias explains some of the variation in
racial disparity, nor assert that it is unequivocally present in any geographic area or in the
workings of any specific trooper. Put simply, the current level of science is inadequate to the
determination of whether disparity can be explained by bias or by one or the other of the rival
hypotheses just discussed.
At the same time the empirical evidence from the analysis of accident and citation data at
the district level (in which data have been aggregated from the county highway area) suggests that
there are some districts where there are relatively high levels of racial disparity that cannot be
accounted for by the deployment of troopers (within the limits of the analysis as discussed above).
Yet, it must be mentioned that there are even more districts where there is under-representation of
African Americans in citations: there are fewer citations of African Americans relative to
involvement in accidents. It could be that the districts with high rates of disparity in this analysis
are units where there is racial bias, or it could be that there are uncontrolled variables that account
for the disparity. Given the data and analysis limitations discussed above, we cannot say with
certainty.
The analysis in Chapter 3 at the individual trooper level also leads us to a conclusion of
ambiguity regarding whether or not there are troopers who have unduly high rates of citations of
242
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

African Americans. While we have developed mathematical models that explain about twothirds of the variance in the number of African Americans cited, it is not possible to ascertain
what proportion of the remaining variance can be explained by bias or by factors not included in
the model.

Criminal Interdiction Team
As for the results of our analysis of the searches conducted by the Criminal Interdiction
Team (CIT) and by regular NCSHP troopers, we were initially surprised to find that compared to
the very large volume of drivers confronted by the NCSHP every year (upward to a million),
there were about a thousand probable cause and consent searches a year in 1997, dropping to
about 500 in 2000. The majority of the probable cause and consent searches are conducted by the
CIT. In consent searches, the trooper receives the permission of the suspect prior to searching the
vehicle, and in probable cause searches, the trooper has reason to suspect that contraband is
hidden in the vehicle and consent is not needed (however, we were told by some CIT troopers that
sometimes consent was asked for in what were clear probable cause situations). Looking at
trends over time, we see that probable cause searches increased as a proportion of all probable
cause and consent searches. The proportion of all searches involving African Americans declined
over the four years of available data, 1997–2000. At the same time the “hit” rates (finding
contraband) have increased for African Americans and are down somewhat for whites. If racial
disparity in searches is any indication of racial profiling, it would seem that the NCSHP has
lessened—or perhaps even eliminated—such activity over the course of the years we have
examined.

243
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

At the same time, our discussions with CIT troopers indicate that the primary method used
by troopers to initiate searches is the so-called “conversational method” in which the trooper asks
the driver routine questions, and depending on the nature of the answers and the degree of
nervousness of the driver, decides to ask the driver for permission to search the vehicle. The
driver then has the option to either decline or accept the offer (we are told that most accept, even
if they have contraband in their vehicle). As such, the method seems to us to open the door to a
rather subjective process of decision making. When does someone appear to be “excessively
nervous?” What constitutes an “inconsistent story” as to where one is going? In that the process
is a subjective one, the door is left open for “cognitive bias” to influence the decision making of
the trooper, possibly resulting in a disproportionate number of African American searches.

Indicators of Racial Disparity
One of the by-products of our efforts is to provide a road-map for some techniques of data
analysis that others may draw upon in making a decision whether or not a particular troop or even
a particular trooper is racially biased. In Chapter 2, we show how a troop district could be
identified as having a relatively high rate of citations of African Americans relative to any of
several base rates (licensed drivers, “drivers driving,” or drivers in accidents). By statistically
controlling for variations in deployment (by time of day and location), and by collecting sufficient
data at relatively small units of analysis (to limit the effects of spatial heterogeneity or the
mismatch between where troopers patrol and where drivers drive), it is possible to obtain an
estimate of the extent to which a district departs from an expected value. As such, districts which
are found to depart excessively from other districts despite the statistical controls are candidates
for further scrutiny.
244
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

It should be pointed out that our assumption is that an overt kind of racial bias is likely to
manifest itself in multiple ways, and hopefully decision makers would have multiple indicators of
such bias available to them. For example, there is likely to be a trail of citizen complaints about
troopers who behave in racially derogatory ways toward citizens. Other troopers are likely to be
aware of the expression of inappropriate racial attitudes on the part of a given trooper or even a
troop. Records, including personnel files, are alternative sources of information on expressions of
overt racial bias.
Racial bias of a more subtle kind, specifically “cognitive bias,” is not necessarily
associated with citizen complaints, or other indicators of explicit bias. Rather, statistical evidence
of disparity that cannot be accounted for by appropriate statistical controls is probably the best
source of information on the presence of cognitive bias. The kinds of statistical analysis
presented in Chapters 2 and 3 provide a means to ascertain where disparity could well be a sign
of such cognitive bias. At the same time, we admit that the evidence is somewhat ambiguous as
to whether a district has an exceptionally high rate of interventions directed at African Americans,
or if a specific trooper has. The statistical analyses need be supplemented with further
investigations to rule out possible rival hypotheses as to why the statistical disparity is present.

Self-Reported Traffic Violations and Stops
In addition to attempting to assess racial disparity in the official data bases of the NCSHP,
we also conducted surveys of the general public to see if there was any correspondence between
the two different sources of information (surveys and official records), and to determine if we
could learn more about the behaviors of interest by studying the surveys. The results are
245
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

encouraging to those doing research in the area of racial disparity in that there are some broad
similarities between the findings of the survey data with the findings of the official data. For
example, according to the survey findings, African American drivers are significantly more likely
than white drivers to have been stopped by the police in North Carolina. This finding generally
corresponds to the official data findings. However, the survey also found patterns that could not
be found with the limited official data available to us: local police are even more likely to have
stopped an African American relative to their representation in the population of drivers and
controlling statistically for self-reported driving behavior. Other intriguing survey findings
include that African Americans are slightly more likely to have been informed that the stop was
for a more discretionary reason, and they are also slightly more likely to report that they were
treated disrespectfully after the stop. On balance, the use of surveys to study the controversial
topic of racial profiling is promising.

Attitudes of the General Public
Survey data also are useful in that they tell us about people’s opinions as relevant to such
important considerations as whether or not trust in the police is undermined by racial profiling
perceptions. One fact is undeniably true: in the focus groups that we conducted with African
American drivers it was clear that African Americans’ perceptions of racial bias in policing were
influenced not only by their own stop experiences, but also by the stop experiences of family and
friends. Some previous research has shown that well publicized and controversial police
incidents, such as the Rodney King beating, tend to reduce trust in the police generally,
particularly among minorities. Not only do media events affect attitudes toward the police, but so
does the direct experience of citizens in their encounters with police, as well as the experiences of
246
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

friends and family. Although few people report acts of disrespect by the police, African
Americans report significantly higher levels of disrespect in police behavior during stops of both
household members and of friends.
The perception of racial bias on the part of police is widespread: 81 percent of African
Americans believe that the police are more likely to pull over African American drivers. Just less
than 70 percent of African Americans perceive similar bias against Latino drivers. In contrast,
less than one third of the white drivers believe that there is racial bias against African Americans,
and whites also see more bias against Latinos than African Americans. Interestingly, more white
respondents believe that the police target for non-racial reasons than they believe that police
profile on the basis of ethnicity or race.
We find that both African American’s and white’s trust in government, as well as in the
police themselves, are undermined by disrespectful treatment by the police during a stop. In
addition, African Americans who hear of disrespectful police behavior toward friends and
acquaintances are less trustful of government officials—and the police.

Viewpoints of the NCSHP Troopers
Not surprisingly, the NCSHP troopers with whom we spoke in the focus group sessions
say that they do not racially profile, although some admit that there may be “isolated instances” of
it within the NCSHP. “Bad apples” exist in any large organization, the NCSHP not excepted.
Moreover, they say that the nature of the work is one in which they are largely reacting to the
behavior of vehicles, and that it is not easy to see the race of drivers until after the decision has
been made to stop the vehicle (and often not until the trooper approaches the vehicle). At the
same time, troopers talked about targeting highways where there are multiple bars (often lower
247
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

class bars), and not spending too much time around “country club” bars (“fishing those holes”
were not seen to be especially fruitful). As such, and to the extent that such deployment choices
are common, decisions of this sort do suggest a level of class bias. There was also some
acknowledgment of what can be considered to be classic stereotyping, such as regarding Latino
drivers generally being “bad drivers.” There was also discussion about enforcement practices that
might increase or decrease levels of disparity in stop outcomes. While most commonly troopers
suggested they have already made the decision to issue a citation (rather than issue a written or
verbal warning) when the car is pulled over, some troopers said that they can sometimes be
influenced by the resulting interaction with the driver: a display of “bad attitude” (inappropriate
demeanor) may earn some drivers a citation rather than a warning. As such, these statements
point to the discretionary aspects of NCSHP, and all law enforcement, work.
One clear message from all of the regular road troopers with whom we have spoken is that
they are reluctant to become involved in vehicular searches. (The rarity of such searches by
troopers in part validates this claim.) In the past, when the “war on drugs” placed more of an
emphasis on drug interdiction, some troopers suggested that some racial profiling may have
occurred (See Appendix B for a summary of focus groups with the NCSHP).
As for what to do about possible racial bias and profiling, the troopers mentioned record
reviews, use of cameras in patrol cars, court visits to assess the quality of evidence and charges,
and occasional ride alongs as possible avenues toward keeping check on troopers.

248
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Conclusion
On the basis of all of our analysis, we conclude that although there is no conclusive
evidence of widespread racial disparity exhibited in the actions of the NCSHP troopers in their
routine interactions and interventions with drivers, there are some districts and some troopers
whose citation rates of African Americans may warrant further investigation and possibly
ongoing or intermittent monitoring. However, it is not our recommendation that such
investigations necessarily occur. Presumably, the NCSHP would have more direct knowledge of
several of the important factors that we have tried to measure that may help to account for the
observed disparity (for example, more precise measures of deployment, especially as it related to
time of day and location). Thus, the decision on their part to reexamine practice in 2000 would
only be made if there were other information (information not available to us, but available to
NCSHP leadership) of potentially biased behavior of any troop or individual trooper active in the
force in 2000. Our purpose here is only to sensitize the NCSHP leadership to possibilities of
irregular patterns that might be construed as evidence of bias. Given inadequacies in the data and
in the measurement of key concepts, we cannot rule out the presence of low prevalence levels of
bias, perhaps of the “cognitive bias” sort. Thus, we think that more evidence would be required
than currently found with the data available to us, to proceed with further investigation of
potentially biased behavior in NCSHP in 2000. At the same time we think that the methodology
that we have developed here can serve NCSHP and other trooper organizations to monitor
possible racial bias in their ongoing operations.
Our analysis of the actions of the CIT, which represents only a handful of the troopers in
the NCSHP, leads us to conclude that even though there is less disparity than there was a few
249
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

years ago, the CIT troopers in the year 2000 are more likely to search a vehicle driven by an
African American than by a white. The fact that the CIT is more likely to find contraband in an
African American driven vehicle undermines, to some extent, the possible bias interpretation of
their behavior. Nevertheless, the use of the “conversational method” as the primary mechanism
for deciding to conduct so called “consent searches” leaves open the door to processes of possible
bias (“cognitive bias”) in which the “signs” that the troopers are looking for are sought more
often, or are seen to be more compelling, when the driver is African American.
Regardless of the interpretation of the empirical evidence of possible racial disparity, the
perceptions of bias and of inappropriate treatment on the part of police (including the NCSHP)
seems to foster distrust of the police. Aside from eliminating any vestige of possible racial
disparity, the NCSHP would be wise to take actions to ameliorate the perceptions that their
behaviors are unfair.

250
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Appendix A: Baseline Observational Study

This appendix details the methodology we used to study one particular driving
behavior—vehicular speeding—and correlates results from that observational study with official
data on trooper stops for the NCSHP (during three months of the year 2000 – May, June, and
July). It is proposed that the data collection technique pioneered by Lamberth, called the
“carousel method” or “rolling survey,” can be improved upon and serve as a methodologically
sound technique for collecting speeding data on demographic groups (Lamberth 2000). An
analysis of fourteen highway segments in North Carolina indicates that speeding behavior is
strongly correlated to NCSHP stops of citizens’ vehicles for speeding. Some evidence of possible
racial disparity on the part of the NCSHP in the stops for speeding on the highways studied is
found, but in general the results are not statistically significant. In general, NCSHP stop
behavior of speeders seems to be determined by vehicular driving behavior.
Central to our concerns herein is the question of whether or not different races or ethnic
groups have similar behavior rates. If so, then the task of determining whether or not racial
disparity exists in stops, citations, and warnings of African Americans is made considerably
easier. But upon what basis can we as researchers assume equivalent behaviors across
demographic groups?

Studying Driver Behavior
The issue of whether or not assumptions of equivalency in behavior across demographic
groups are warranted has not been adequately addressed in previous research. The omission is

251
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

primarily due to a lack of quality data, which in itself reflects several methodological difficulties
of collecting data. These methodological difficulties center around the difficulties associated with
identifying the race (or age and gender) of drivers traveling at high speeds on highways. A recent
study by Lange and associates (2002) used video cameras and radar speed guns to determine race
and speed, respectively. However, of the 38,747 images collected, only 26,334 were usable
(12,413 were not; as reported in Kociewiewski 2002). That is, in almost one third of the
observations, the race of the driver could not be determined or agreed upon by three researchers
watching the video. This research, therefore bolsters our claim that collecting such data on race
of driver is difficult.
One might think that data collection would be as easy as standing by the side of the road
and observing. Armed with a radar gun and a pad of paper, researchers seemingly could record
the race, as well as perhaps the gender, age, and speed of drivers as they drive by, and then would
be able to compare such data with the rates of stops and citations for a given highway.
Researchers might even be able to observe vehicles weaving unsafely, or vehicles with expired
license plate “stickers,” or perhaps identify drivers they thought were “driving while intoxicated.”
One could conceivably estimate speed of the vehicles on a video tape by measuring the time
during which a known distance is traveled.
These first thoughts on how to collect data on the behavior of drivers, however, have
several problems that make them impractical. One problem is that the glare on windshields and
side windows of vehicles makes it difficult to see clearly the motorist’s race, as well as his or her
gender or age, from a safe distance at the side of the road. The glare is in part due to the tinting
done on most, if not all, windshields and to the angle of light from the sky (even on cloudy days).
Video cameras suffer from the same problem. The skeptic is encouraged to simply try to stand
252
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

(from a safe distance) near a highway with fast-moving traffic, and see how often they can
successfully identify the demographic characteristics of drivers. We experimented with this
method and found that we were frequently unable to do so. Compounding this difficulty is the
high rate of speed of the passing vehicles, as there is little time to assess demographic
characteristics, much less their speed. More importantly, when we tried this technique (roadside
viewing), it did not seem likely that our failure to identify demographic characteristics was
random (else one could argue that misses were “random error” and could be safely ignored).
Rather, some types of vehicles or conditions, such as an open side window, seem to permit
greater visibility than others. In short, we suspect that it is simply too difficult to reliably
ascertain demographic characteristics from the sides of busy roads (also, it is rather unsafe for the
observer, unless he or she is well removed from the highway, but then it is even harder to see into
passing vehicles).46 Despite these concerns, we should note that the general substantive findings
of the Lange et al. study are generally similar to what we report here regarding racial differences
in speeding behavior.
As an alternative methodology, researchers might collect data on drivers by riding with
NCSHP troopers. Researchers could keep track of what troopers encounter and react to on the
highway. One drawback of this approach is that troopers might alter their behavior under such
circumstances. Our concerns, however, were primarily with safety. We did not want to place our
researchers in patrol cars during an extended period of time due to the risk of accidents (which is
much higher than for the average passenger car) and of possible encounters with armed citizens
(unlikely, but with potentially catastrophic consequences).

46

Further analysis of the data collected by Lange and associates, however, may prove our
claims wrong.
253
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Fortunately, there is a viable alternative to avoid the problems discussed above on the task
of collecting data on drivers guilty of at least one type of law breaking: vehicular speeding. The
“carousel method” of identifying speeders and their demographic characteristics is a method in
which a research vehicle is driven at the speed limit, and vehicles that pass the research vehicle
can be examined from the vantage point of the research vehicle, wherein the researchers can
identify the race and other demographic features of the driver. Other researchers found, and we
verified, that at the close range of the proximate traffic lane on a highway, one is almost always
successful at identifying race and gender—and even age (see discussion below). There is little
windshield or car-door window glare at close range. Moreover, in the vast majority of cases, our
research teams were in agreement as to the race of the driver in the adjacent lane. By counting
the race of the drivers who pass (as well as whom the research vehicle passes), we can identify
the prevalence of certain demographic groups on the highway, as well as their prevalence among
the speeding population. As such, the carousel method, as used by Lamberth in Maryland, was a
big improvement over previous research studies which were based on the demographic
characteristics of the area surrounding the highway. With the carousel method, the researcher can
keep track of the racial composition of speeders and have a better denominator than, for example,
Census Bureau population counts of local residents. The number of African Americans on a
particular highway who are speeding could be compared with a numerator based upon, for
example, the number of African Americans stopped for or issued a citation for speeding. Such a
ratio could then be compared to the corresponding ratio of whites in order to determine the
existence of racial disparity—and, by extension, any evidence of racial discrimination.

254
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Modified Carousel Method
The problem with the carousel method as it has been implemented previously, is that not
all speeding is equal. Passing vehicles could be driving 2 mph faster or 20 mph faster than the
researchers’ vehicle—and they would both be counted simply as speeders. Almost all vehicles
travel above the speed limit on major highways, and it is well known that police do not routinely
stop or cite vehicles simply because they are breaking the speed limit. Rather, vehicles must be
exceeding a certain “threshold” of speed to actually trigger a stop and citation (or perhaps only a
warning). The carousel method, as used by Lamberth, did not differentiate speeders based on the
extent of their speed.
We explored the possibility of capturing the speed of other vehicles on the highway by
using two radar guns (called in the business “same lane” radar, which means capturing the speed
of a vehicle moving in the same direction of the patrol car or, here, research vehicle). One radar
gun captures the speeds of vehicles as they approach from the rear, and the other gun captures
their speed as the vehicle pulls away from the moving research vehicle. That is, the radar can
capture the speed of a car in front of the research vehicle or behind. This approach of using radar
has some obvious merits. First, the researcher can ascertain the speed of vehicles accurately, and
at two points (back and front of the research vehicle). The average of the two is plausibly a
reliable single measure of the propensity of the driver to speed. However, there is a serious
drawback to using radar: it sets off radar detectors, triggers an alert on the CB, and slows down
traffic. We suspect that virtually every semi-tractor trailer driver in North Carolina uses a radar
detector, and so do many other drivers. By listening on a Citizen Band radio (CB), we verified
that there is a high prevalence of CB use announcing the presence of radar. When we set off the
255
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

radar, there would almost always be a CB alert broadcast on the local highway, presumably heard
by most truckers on the highway within a few-mile stretch. By constantly triggering the radar
during the course of our study, we would be introducing bias into the study that we determined
would be hard to correct. Another problem with radar equipment is that it is difficult to collect
data systematically on highways while traffic is heavy—vehicles simply approach too quickly to
allow you to get readings of every vehicle if you are driving at or below the speed limit. (It is
generally not possible for safety reasons, of course, to have researchers drive above the speed
limit.) We could have sampled the cars, but that would have meant more time on the highway
and/or introduced more complexity into what was found to be quite a difficult procedure —recording data on passing vehicles and their drivers.
We chose instead a simple, but at first glance seemingly error-prone method: using
stopwatches to record how long it takes vehicles to pass our research van (of known length). The
miles per hour of the passing “subject” vehicle could be measured quite accurately by two
researchers, each using a stopwatch, and averaging the two speeds. The stopwatch is started by
both researchers the moment the subject vehicle’s front bumper crosses an imaginary line
perpendicular from the research vehicle’s (minivan) back bumper, and the watches are stopped
the moment the subject vehicle’s front bumper crosses an imaginary line perpendicular from the
research vehicle’s front bumper. By knowing the time it takes to pass the research vehicle as it
travels at a constant speed, we can easily calculate the speed of the passing vehicle. We discuss
below the fact that we systematically underestimate the speed by this method. However, we
correct for this underestimation statistically to arrive at what seem like accurate measures of the
speed of the passing subject vehicle.

256
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

We estimated that we could observe an adequate number of what we called “threshold
speeders” by spending approximately 24 hours of one week driving a segment of highway
approximately 10–15 miles in length, both ways. We were not sure when we started our research
what the “threshold” of speed was that troopers use to stop motorists, but we estimated that it
would be perhaps 7, or 9, or even 12 miles greater than the posted speed limit. (This was
somewhat in error—see discussion elsewhere in the text.) We chose as many sites as we could
afford to study. Fourteen were studied (although the prevalence of speeding on one of the
segments was greatly reduced by the fact that there was some construction on part of the
segment). These fourteen sites represent highway segments that meet the conditions of being
four-lane highways, and highways that have frequent citation events for speeding such that we
could compare the citation rates with the speeding rates (see discussion below). We omit details
of where the fourteen sites are (that would constitute publishing the whereabouts of “speed
traps”), but nine sites are along interstate highways, four along U.S. highways, and one on a N.C.
highway. All are four-lane highways. Three of the sites are on Interstate 95, another four on
Interstate 85, and two on Interstate 40.
Studying highways with more or less than four lanes was deemed unsafe. Two-lane
highways would not permit vehicles to pass us frequently enough, so that it would take many
weeks to accumulate data on passing vehicles. Six- or eight-lane highways (three or four lanes in
each direction) were found in our preliminary tests to be very busy highways, complicating
further the data collection protocols Moreover, cars that were two lanes away could not as easily
be gaged as to when they crossed the imaginary line extending from the back and front bumpers
of our research van. To have chosen any less busy highways, or any with fewer speeding
citations would have been cost prohibitive, as even more time would have to be spent collecting
257
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

data at each site. We found that the NCSHP concentrates patrols on particular segments of
highway and that we were capturing some of these segments in our research. It is interesting to
note that, although there are stops for speeding during the course of a year between almost any
milepost markers on the busier highways (interstates and U.S. highways), stops are much more
concentrated along certain stretches of highway that run for distances similar in length to those of
our selected highway segments.

Baseline Data Collection
Beginning in January 2000, troopers of the NCSHP were required to collect data on each
vehicular stop they initiated. They record the location of the stop by indicating the closest mile
marker, wherever such information is available. We found that for the highway segments studied
here with milepost information (nine), troopers did so more than 90 percent of the time. We were
able to determine how many vehicles were stopped in the fourteen highway segments, and below
we report a comparison of race-specific rates of stops for speeding to race-specific speeding
behavior in the fourteen highway segments during a time period slightly longer than when we
were collecting data—May through July. (We collected data in May and June, and no highway
segment was studied for longer than one week during that time period).47 We extended the time
period into July to generate more observations so as to be able to compare the proportion of
speeders observed who are African American to those who are stopped and/or cited who are
African American.

47

We added July to have more observations to compare to and under the logic that the
traffic patterns are generally seasonal and that May, June, and July were more likely to be similar
than April, May and June (April being the other month adjacent in time to May and June).
258
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Data were collected using two research vehicles, with four researchers per vehicle, at two
sites per week (one site for each vehicle). Of the four researchers in each van, one drove, one
recorded time of day (to the second), race, gender, and age of driver, as well as other information
(color of passing vehicle, state of the license plate, type of vehicle). The other two researchers
timed the passing vehicle from the moment its front bumper crossed an imaginary line drawn in
extension from the back bumper of the research vehicle to another imaginary line from the front
bumper of the research vehicle.
As mentioned above, fourteen sites were selected using the following criteria. A site had
to be a four-lane highway with sufficient recorded stops (which could result in citations or
warnings) throughout a comparable time period (assumed to be May through July) to be able to
compare rates of speeding behavior to rates of stops for speeding. It was judged unsafe to collect
the data on highways that were too busy (for example, during rush hour) or with more than two
lanes each way (as discussed above). Cars passing the research vehicle on both sides were
distracting to the driver and the other researchers, and created “overload”—too many cars passing
at once. It was all the researchers could do to record the data for each subject vehicle individually
as it passed on the left of the research vehicle. We also found that it was hard for the driver not to
get involved in the data collection process (see discussion below), and if he or she were on a
three-lane highway, safety concerns would arise.
The fourteen segments had to have an appreciable number of African Americans driving
on the segment. Sites to the far west of the state were eliminated due to this criterion. North
Carolina’s African American population largely lives along counties stretching north and
eastward from Charlotte to Interstate 95 where Interstate 95 enters Virginia. Many African
Americans in North Carolina also live in the counties along Interstate 40 between Winston-Salem
259
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

and Raleigh. These were the areas within which we selected our fourteen sites. Each research
site was between 10 and 15 miles in length (one way). The research vehicle traveled back and
forth along these segments for approximately 6 hours a day, four days a week (although for some
segments, data were collected for closer to 8 hours a day for three days). Thus, in general, 24
hours of observation, consisting usually from 9:30 a.m. to 11:30 a.m., and from 1:00 p.m. to 3:00
p.m., and again from 6:00 p.m. to 8:00 p.m., were completed for each of the fourteen sites. Noon
hour and afternoon rush hour times were omitted because troopers frequently are attending
accident scenes at these times, and thus are less likely to be patrolling for speeders then. We also
preferred to avoid the heavier traffic situations as they seem less safe for driving. Also, under
heavy rush hour traffic, speeding is largely limited by traffic density along these heavily-traveled
commuter corridors.
Data were gathered during the second week of May until the end of June 2000. Data were
subsequently entered by several graduate students into a spreadsheet and “cleaned” (corrected)
for all fourteen sites.
For nine of the segments, mile post data were recorded by the NCSHP, such that we can
compare the rate of African American to white stops for speeding with the rate of African
American to white speeding within the same highway segment. When we started the study, we
assumed that if we found that the rate was higher for stops than for behavior (within a margin of
error), this might support claims of racial discrimination in these locales. However, we
subsequently found that the process is much more complicated than that. More observations
would be desirable (and possibly greater accuracy in assessing speed) to conduct a rigorous test (a
test with a narrow margin of error) of the hypothesis at each sight to determine whether or not
there is bias in the workings of the NCSHP.
260
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

As part of the study, but not presented here, we employed NCSHP troopers to drive in the
vicinity of our research vans. We kept in radio contact with them and kept varying distances from
them so as to best simulate an environment in which troopers identify speeders. Our thinking was
that it did little good for us to collect speeding data on drivers outside the context of trooper
location. Thus, it was preferable to collect data on drivers who were driving within a few miles of
an NCSHP car. Troopers stop speeding vehicles that they identify visually—typically through
the use of radar—and arguably, many drivers alter their behaviors depending on the speed of the
vehicles on the highway around them. Thus, we wanted to capture the speeds of vehicles in these
types of contexts. (In a subsequent publication we plan to discuss further the effects that NCSHP
proximity had on driving behavior.)

Validating the Stopwatch Method
In order to test the idea that we could gather reasonably accurate measures of speeding
vehicles using stopwatches, we conducted validation of the stopwatch method by recording the
speed of a passing vehicle using a radar gun. To conduct the validation, we used the NCSHP
highway “track” that is normally used for training purposes. We used three vehicles to validate
the ability of each of six researchers to accurately time how long it took one of our vehicles to
pass another vehicle that was traveling at a faster speed than that of the vehicle being passed. The
passing vehicle will be referred to as such, and the vehicle being passed will be referred to as the
stopwatch vehicle. In the stopwatch vehicle, researchers recorded the speed of the passing
vehicle by timing the amount of time it took the passing vehicle to cross an imaginary line
extending from the back bumper (perpendicular to the side of the vehicle) to another imaginary
line extending similarly from the front bumper. The stopwatch vehicle traveled at a constant 35
261
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

mph, and included a driver and three researchers equipped with a pencil, pad of paper, and a
stopwatch.48 They were instructed to record the amount of time it took the passing vehicle’s front
bumper to pass from the stopwatch vehicle’s back bumper to front bumper (vehicle length). They
were instructed to write down the speed as they observed it and not to communicate to others in
the vehicle what time they recorded.49 The stopwatches were standard “sport” stopwatches, and
the number of seconds it took for the passing vehicle to pass was recorded, not the speed in miles
per hour. The passing vehicle was given a list of “passing speeds” at which the driver was to
drive the passing vehicle past the stopwatch vehicle. These speeds varied from about 2 to 20 mph
faster than 35 mph. Speeds faster than 20 mph above 35 mph (such as 55 mph) were thought to
be too dangerous for the amount of “straight away” available on the training track. It was not
thought possible to accurately record the time for speeds significantly faster than 20 mph above
the speed of the stopwatch vehicle, since such stopwatch times that would be recorded were less
than one second (i.e., less than one second would pass between initializing the stop watch and
stopping it to read the recorded time.) CB radio was used to communicate between the passing
vehicle and a stationary vehicle, which was parked by the side of the road. The researcher in the
stationary unit used a radar gun to estimate the speed of the passing vehicle (as well as to estimate
48

The training track was configured such that a speed higher than 35 mph was not
possible without risk to the vehicle and passengers.
49
The researchers were graduate students, and it is believed that they followed
instructions and did not share time information (in part because the times were different from
one another). We believe the researchers had no motivation to communicate the recorded times,
since they did not perceive themselves as being tested, but rather the method was being tested.
They were told it was acceptable to miss a speed measurement and to record whatever time was
on their stopwatch. The variation in stop times across researchers was somewhat high, relative
to our initial expectations for the accuracy of the stopwatch method, so it is doubtful that the
researchers “cheated” and said their times out loud. Additionally, later tests conducted on actual
highways (so called “same-lane” tests) further validated the stopwatch method, and for those
tests the stopwatch researchers did not know the speed of the passing cars (which were gaged by
the “same-lane” radar gun).
262
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

the speed of the stopwatch vehicle, to verify that it was traveling at the designated speed of 35
mph).50 Also, the passing vehicle’s driver had an assistant to record the speed of the passing
vehicle. CB radio was also used to communicate between the researcher in the stationary vehicle
and the stopwatch vehicle, but using a different channel from that which was used with the
passing vehicle. The primary purpose of communicating with the stopwatch vehicle was to
coordinate the efforts of both the moving vehicles.
The driver of the passing vehicle did his best to maintain a constant speed while passing
the stopwatch vehicle, and the radar gun reading confirmed that this was the case. We maintained
constant speed in order to more accurately measure the speeds of the passing vehicle, and thereby
help us verify that the stopwatch method worked. Differences between the radar reading and the
recorded speedometer reading were small (or at least seemed small to us at the time – see
discussion below), and in only one test run, varied as much as 3 mph. In most cases the readings
were the same, but in some cases the readings from the radar gun differed by 1 or 2 mph from that
of the driver’s speedometer. The training track is a closed circle such that the vehicles (two vans)
could drive continuously, but they had to slow down for the curves. The speeds were recorded
only when the passing vehicle was able to pass on the straightaway section of the track.
Figure A-1 shows the relationship between the actual speed of the passing vehicle
(speedometer speed) and the stopwatch measured speeds for what we call the “first road test,” in
which the first three researchers (out of six total) recorded the times on stopwatches. Because we
only had one stopwatch van, we divided the validation tests into two sessions. In this figure the
measured speed is calculated by miles per hour above the 35 mph of the stopwatch vehicle.
50

Actually, two radar guns were used. One was supplied by the NCSHP and the other
was purchased by the research team. The two were almost identical in the readings made for the
passing vehicle and for the stopwatch vehicle.
263
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure A-2 shows the same information, but for the second road test (the next three researchers).
Separate tests were done for each of the individual researcher’s recorded speeds, and correlations
among the researcher’s speeds were found to be between .88 and .99, indicating quite a high level
of reliability by customary standards. That is, the six researchers tested that day tended to vary
little among themselves in their recorded speeds.
In the two figures, the horizontal axis represents the speed of the passing vehicle as
recorded from the speedometer. Although there is some error in this measurement of the
vehicle’s speed, it is probably more accurate than the radar estimate of the speed, which is also
shown in the graph as the hash-marked line (the uppermost line on the graph). Note that the
speed values are sometimes repeated (for example, 44 mph occurs more than once in the first time
trial. Thus, we should not expect any straight lines across this graph).
Note that as the actual speed of the passing vehicle increased across trials, the researchers
were underestimating to an increasing extent the speeds of the passing vehicle. For example, at
40 mph actual speed the estimated speed is just a little less than 40 mph, but at 52 mph actual
speed, the researchers were estimating about 42 mph —roughly 10 mph under the speedometer
speed. Essentially, as the vehicle passed, it took relatively longer for the researchers to stop their
stopwatches. We noticed that researchers virtually never stopped the watches prematurely,
meaning before the passing vehicle’s front bumper actually passed the front bumper of the turtle
vehicle. But the faster the passing vehicle’s speed, the more distance traveled by the passing
vehicle past the stopwatch vehicle’s front bumper before the researchers stopped their watches.
Nevertheless, the researcher’s were systematic in their error. As discussed in the next section, we
can correct for the systematic underestimation of speed and generate an estimate of the speed of
the passing vehicle within a known error range.
264
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Correcting for Under-Estimation of Speed
Predicted values from the equation in the above paragraph were found to fit the
speedometer speed measure quite well (see Figure A-3 and A-4). Here we took the average of
two estimates, one from each of two research “timers” (we use two stopwatch timers in the
observational baseline study of the fourteen sites). Thus, the speeds of Timer 1 and Timer 2 were
averaged for one estimate for each passing vehicle’s speed, and the estimates of Timer 2 and
Timer 3 were averaged for the other pair’s estimates. Comparisons can be made here with the
radar estimate (the dotted line on the graph) or with the horizontal axis (vehicle speed), as they
are very similar. Here one can see that the “corrected” estimated speeds of 40 and 52 mph, for
example, correspond with varying degrees of accuracy with the observed speeds. Thus, the
systematic nature of the under-reporting of estimated speed can be capitalized on, in order to
correct for the speed underestimation.

265
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure A-1 Stopwatch Validity Test for First Three Researchers

266
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure A-2. Stopwatch Validity Test of Second Set of Three Researchers

267
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

If one looks closely at the differences between the actual and estimated speeds across
values of actual speed, it is arguable that these researchers were somewhat overestimating speeds
in the 40–48 mph range, and somewhat underestimating the speed greater than 48 mph.
However, in Figure A-4, the three researchers there show no such pattern.
Note that on the actual highways of our observational study, our stopwatch van would not
be traveling 35 mph, but between 55 and 70 mph. We are assuming that our ability to record
speeds above the posted speeds on the highway is the same as on the training track. Thus, if a
passing vehicle is traveling 50 mph on the training track or 15 mph above 35 mph, the speed of
the stopwatch van, it is equivalent to a vehicle traveling 80 mph while passing our stopwatch van
traveling 65 mph on an observational study highway segment.
We presumed that the risk of being stopped and cited at 15 mph above posted speed was
substantial for most highways. At the time we collected the data at the training track, we did not
think it likely that speeds of greater than 15 mph above posted speed would be common. As it
turns out, we subsequently learned that 15 mph above posted speed is quite common, and is often
the median speed for which drivers are stopped and cited for speeding on the fourteen highway
segments of our observational study. Thus, it would have been desirable to measure speeds more
accurately at between 20 and 25 mph above the posted speeds. We assume that the risk of being
stopped and cited may go up substantially as a vehicle’s speed rises above the 15 mph speeding
threshold.
The implication of these figures is that the researchers’ stopwatch times (from recording
the time it took for vehicles on the highway to pass them) could be mathematically transformed
into a reasonably accurate estimation of the speed of the passing vehicle by using the regression
equation, discussed in the section above. That is, the estimated speed (raw score speed) could be
268
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

multiplied by 1.73 (subtracting .353), to arrive at an estimated speed of the vehicles. Note that we
corrected the average speed of the two stopwatch times in the collection of the speeds of the
passing vehicles in the road trials across fourteen sites chosen for the baseline observational study
described below (although occasionally only one stopwatch time was available).

Figure A-3 Corrected Average Speeds, Time Trial One

269
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Figure A-4 Corrected Average Speeds, Time Trial Two

270
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Note that, unlike the case with the road testing at the training track described above,
passing vehicles on a real highway might accelerate or even decelerate as they passed the research
vehicle. Thus, there might be greater error in recording the actual speed of the passing vehicle
than we observed in the artificial conditions of the training track. Nevertheless, the stopwatch
timers are measuring an “average speed” of the vehicle as it passes the research vehicle, and as
such, the estimate should be sufficient to test hypotheses as to differences in speeding behavior
across demographic groups, within a margin of error of a few miles per hour. Moreover, we
conducted other tests involving the use of a “same-lane” radar gun, which allowed us to assess the
speed of actual vehicles passing a second research vehicle (from a first research vehicle following
roughly 100 yards or so behind the research vehicle). We found that, in general, the recorded
speed did correspond quite accurately to the speed of the passing vehicle measured with the radar
gun, and that the vast majority of cars passing did so at a reasonably constant speed.
It should also be noted, however, that there is a substantial margin of error associated with
the estimate of the speed of the passing vehicles. One can get an intuitive sense of the range of
estimated stopwatch speeds by looking carefully at the variation in the graphs in A-3 and A-4. In
part, the fact that the lines are not smooth speaks to the variation from the speedometer speed
(also recall that some speeds are repeated on the horizontal axis as some time trial speeds appear
more than once). In the graph, one can more easily see the differences between the corrected
estimate and the radar speed than one can see the differences between the corrected estimate and
the speedometer speed. Therefore, the average of the pairs of timed estimates can be quite
different from the radar speed. Also, at the reference lines, one can see that the stopwatch speed
varies by a few miles per hour from the speedometer speed. We ran some simple regression
equations in which the speedometer speed of the passing vehicle was the dependent variable and
271
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

the average estimated speed (stopwatch speed) for pairs of observers was the independent
variable. The range of standard errors associated with these regression equations varied from
2.61 to 5.24, indicating that, on average, an estimate could be wrong by plus or minus 2.6 to 5.2.
As it turns out, this is a rather large degree of variation relative to the variation associated with
the measure of speed on the highway by the NCSHP troopers (who generally measure speed
within 1–2 mph of the best estimate of actual speed). Also, we were surprised when we later
discovered that there is a big difference between traveling just a few miles per hour faster or
slower than the speeding thresholds on a stretch of highway. When speeding thresholds are
exceeded, they trigger a stop and often a citation. A 1 mph difference in speed can make the
difference between whether or not someone receives a citation. The inaccuracies of the
stopwatch method will attenuate slightly any correlation between the estimated speed and the
actual speed (here we are assuming that is the speedometer speed). As we subsequently learned,
the range of likely speed threshold values (speeds that are likely to result in a stop or citation) is
quite narrow relative to the accuracy with which we can reliably measure speed, thereby limiting
our ability to test hypotheses for each of the fourteen highway segments. (Results of the
observational study can be seen in Appendix G.)

272
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

References for Appendix A
Kociewiewski, David. 2002. “Study Suggests Racial Gap in Speeding in New Jersey” March 21.
New York Times.
Lamberth, John. 2000. “Report of John Lamberth” Web citation:
http://www.aclu.org/court/lamberth

273
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Appendix B: North Carolina State Highway Patrol Focus Groups

Topic: Racial Profiling

Selected Findings

Senior Author of Appendix B:

Dr. Harvey McMurray
North Carolina Central University

274
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Group Facilitators:

Dr. Harvey McMurray
North Carolina Central University
and
Dr. Matthew Zingraff
North Carolina State University

June, 2001

Raleigh, North Carolina

275
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

North Carolina State Highway Patrol Focus Groups
Focus groups were conducted to gauge NCSHP trooper perspectives about the nature and
extent of racial profiling in North Carolina. Six focus groups were conducted in early June 2001.
Due the racially sensitive nature of the topic, four of the six focus groups were race specific (two
were African American and two were white). This was deemed appropriate in order to best
provide a forum where respondents would feel less restricted, although each group noted that they
would feel comfortable speaking in the presence of fellow troopers. The management and
command groups were racially diverse. Our random selection process did not capture any female
troopers (there are few women in the NCSHP relative to men). Focus groups numbered from six
to nine troopers and each lasted approximately 2 hours. The sessions were facilitated by two
members of the research team. Three other members of the research team observed the sessions
shielded by a one-way glass window.
It should be understood that findings from the trooper focus groups cannot be generalized
to all NCSHP troopers. They were not selected, nor were they expected, to be spokespersons for
their peers. We desired insights into the activities of individuals on patrol in order to better
understand what they do and why. Further, the specifics of what we learned from these
individuals cannot be generalized to other law enforcement agencies (such as local city police)
due to the differences in scope of their duties and restricted jurisdiction.
For both the general public and the troopers, there appears to be both confusion and
disjuncture between what is thought to be “good” law enforcement, on the one hand, and “racebased” law enforcement on the other. Law enforcement practice, both organizationally and
individually, is influenced in part by a multitude of experiences brought to the table. These
“experiences” result from an accumulation of identifiable and individual encounters with the
276
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

public. They also include generalized expectations of behavior that may or may not be the result
of specific or accurate information. The latter generalizing process is best considered to be a
process of stereotyping. Certain stereotypes associated with minority members of the community
may subject them to more intensive scrutiny, may result in defining their behavior as more serious
than it is and the alleged party as more culpable, and may lead one to believe that the
transgression is more deserving of official intervention than would occur in the absence of the
stereotype. Whether conscious or unconscious, stereotypes held related to degree of suspicion,
prevalence of crime, or dangerousness can lead to more frequent stops of African Americans and
other minorities, as well as influence decisions to release, warn, cite, or search vehicles and
occupants.
Race as a profile serves as both a predictor and an explanation for citizen’s behavior. It is
all too common for individual’s to answer the questions of “why do you think African Americans
are stopped more often than whites?” or “why do you police make so many stops in a certain
area” with simple answers like “well look at the crime rate, look at the prison population.” It
would seem that in most cases, police and citizens do not recognize the disconnect in their
answers associating the assumed (but often unknown) robbery rate in a particular part of town and
a routine traffic stop in that or another area. Still, the answers are given without blinking an eye
and the practices are considered both good policing and a service to the community.
Leaving the question of efficiency aside for another time, such proactive policing strategies
might be viewed as uninformed, but more benign, if racial stereotypes were absent. However,
given past and current American race relations, it is quite unlikely that we separate race totally
from our thinking about crime. It is precisely this type of discrimination that can be so harmful

277
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

and which has been highlighted in the public controversy over race-based policing. Indeed, the
ACLU (1999) notes that “ ... skin color has now become a proxy for criminality.”
For our purposes here, acting out one’s racial prejudice through law enforcement,
individually or organizationally, is “racial profiling.” Such prejudice can be in the form of
presumed criminality, fear of social integration, or a sense of white privilege. Complicating the
issue of racial profiling is the extent to which law enforcement decisions are influenced by
socioeconomic standing. To the extent that this is true, the demarcation of what is, and what is
not racial profiling becomes blurred.
Below are selected findings from the focus groups. Consistent with the structure of the
focus groups, proper summary findings are presented in the following areas: 1) Decision to Stop,
2) Enforcement Decision, 3) Decision to Search, 4) Minorities and Traffic Stops, and 5) Steps to
Stop Racial Profiling.

Decision to Stop
It appears that troopers, for the most part, engage in enforcement patterns they believe
yield the greatest number of enforcement opportunities. A major determinant in the decision to
make a stop appears to be the “behavior” of the vehicle. The focus on such behavior seems to
vary situationally. The interstate, it is believed, is more likely to yield speeding violations rather
than seatbelt violations. Participants state that it is not possible to know the race of the driver on
the interstate or at night. Rather, they focus on the behavior of the vehicle. For example, in the
case of Latinos, they are believed to be more likely to violate driving laws. Consider the
following participants’ comment:

278
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

And I think it depends on other factors such as where do you . . . put your selective
enforcement. Where are you going to put your people? Just say . . . We have
accidents in . . . I’m sure. But where are you going to have the most fatalities, are
you going to have up here on north of . . . Road or are going to have them down
here at . . . , you know, where you have got eight or ten beer joints down there,
where you know that these people are going to just leave. It has nothing to do with
whether they are black or white, the blacks or whites live in that area, but you
have got all of these beer joints and clubs and stuff down there and these people
are going to leave this area and they are going to get out there and they are going
to speed and we are going to have fatalities, . . . , you are going to concentrate on
the weekend. You are going to put your people in that area where you see that you
have had these accidents, where you have had fatalities. Where you have had
these crashes, that is where you are going to assign your people to, during those
times. That is what this strategic planning and stuff is suppose to be. Putting your
people where you have problems.
I’m the same way. I’m in an area where there’s a heavy influx of Hispanics and to
be downright honest, they’re terrible drivers and they get stopped a lot because,
you know, you may mistake them for an impaired driver. They get stopped and
there’s a big portion of them that are impaired so they account for a lot of our
DWI risks but also they’re stopped and given a lot of warning tickets because
they’re just terrible drivers. They don’t have a lot of driving experience I don’t
think. They come up here and get mingled in this traffic and they cause a lot of
accidents. They’re a victim because of their driving, because of inexperience.
I think stereotyping and violations go together. If you look at a certain group,
stereotype a certain group and in that group you deem it to be, have more
violations, then naturally you will work more at stopping whatever . . .
Participants were asked if they believe traffic enforcement was more likely in a “country
club” area as compared to a “low-income” area. The general response is that the law is enforced
whatever the circumstances without regard to social status. Still, troopers acknowledged that it is
easier, because of reasons out of their control, to do their jobs in some places and not others.
Simply, some citizens are more likely to resist the legality of the trooper’s action, complain to
supervisors, and challenge the citation. Each of these situations is something that most troopers
would like to avoid. Interestingly, but not surprisingly, it was reported that the NCSHP receives
more complaints from whites, as compared to African Americans and Latinos. We had picked up

279
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

the same theme in our citizen focus groups. White citizens tended to see any stop as an
unnecessary intrusion. African American citizens tended to acknowledge and accept
responsibility for stops resulting from clear violations of traffic laws. This raises an auxiliary
issue that compounds the problem of racial profiling: how does the coupling of expected
resistance from a white—and perhaps more affluent segment of society— with possible
stereotypes of the minority community (for example, the perception of a certain expected level of
law violations) impact the numbers and degrees of disparity in stop outcomes? With regard to
resistance, consider the following response (it should be noted that this was not the prevailing
norm):
I mean, it might be true now that the area that would display less resistance. I
mean if you got a, if you’re looking at income, you’ve got a rich area and a poor
neighborhood and if the rich people get nice attorneys and do a lot of things, then
spread the word around as well, then you will be more apt to go to the area where
you get that less resistance, where you don’t get that high-priced attorney putting
you on the stand and then taking you through all these things. I still think you will
work both but you might at some point go to the other one a little more frequent
because there’s less resistance to, you know, what you’re trying to do in the first
place.
Your higher income people generally drink and socialize at places where your
lower income people don’t. Friday and Saturday night bars where your younger
people target, where you have a lot of your calls are searched, a lot of your fights.
Your higher income people don’t socialize at the places that you make a lot of
these arrests and get a lot of these calls from. You know, you’ll be in the toxilizer
room and occasionally somebody will bring one in where they got picked up. They
were working the road and you’ll pick up a high dollar somebody who just, whose
driving is impaired and they pick those up. As a general rule, I don’t think a lot of
your police officers target the type of places that maybe your more influential
people might drink at Holiday Inn lounges and places like this. It normally don’t
call your attention there. You really don’t have a reason to be there like you’re
called to a lot of other places.
The rationale for targeting people of color is founded in one’s perspective about what
crime is, where it occurs, and the perceived risk to social order. This is a grey area where income

280
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

and social status might serve as proxies for race. Still, it should be recognized that even attention
to presumed status is a form of stereotyping. Furthermore, differential treatment as a result of
presumed status may also indirectly but systematically disadvantage African Americans and other
ethnic groups. The following quote reflecting “lifestyle” differences is a case in point.
I would say that is also—I think that one thing that I don’t know if it is factored in
or not, it is a social behavior, economics has a bearing on what the behavior of
some people. For example, if you are in . . . where the upper socially economics
predominately live and there are both minorities and Caucasian up here but in
proportion, they are mostly white. My experience has been, and it is just my
perception, I can’t back it up with anything, but their behavior as to how they
socialize is different. It puts them less at risk than the socioeconomic people on
the lower level, be it black or white. By that I mean if they have a house party and
they all camp out at the house and they never leave whereas the lower
socioeconomic people, both black and white, from my experience in . . . and some
of these other counties I worked, they were traversed to a place like a tavern or
something and our guys are circling those bases be it black or white, they are just
circling them and the time that you depend on when they come out of there and you
observe some kind of erratic driving, they are going to get flagged. So they get
stopped if it is predominantly black home, then most of the people arrested from
there are going to be black.
While there was acknowledgment of the possibility of racial profiling, it was generally
believed that it is an infrequent occurrence today as compared to a more frequent occurrence in
years past. Since the large drug interdiction units were dismantled, and the competitive nature
surrounding the quantity of drugs seized has been lessened, troopers indicated that there have
been fewer complaints. The following comments reflect the general feeling about the past and
present levels of race-based law enforcement, also called racial profiling:
I think when we had our interstate squads Statewide and they were really looking
for drugs and doing more searches, I think since that dissolved I haven’t heard the
complaints about, you know, all these people are being stopped, white or black,
and they’re being searched because I mean it was kind of like they were making
great big drug busts and getting their names in the paper. The Governor was
giving them awards and blah, blah for getting a tractor-trailer load of dope and
$350,000 so I think especially the young guys might have been a little more apt
once they got you stopped to say I may get me a big drug bust here or something
281
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

about this guy ain’t right and I’m going to search him. I think now that’s resolved.
I don’t think the general population Patrol Statewide are doing numerous searches
like they were when we had that interstate squad that operated all the way across
the country. I haven’t heard any complaints in several years now on unreasonable
stops and search since the interstate squad went.
Well I’d be surprised if you don’t have some that do. White and black. I mean,
you know, there’s no way that in 1,400 or 1,500 people you’re going to have 1,400
or 1,500 people that’s going to go right by the rules, you know?
There may be some whites that target blacks and there may be some blacks that
target whites or they both may target Hispanics, I don’t know. As a whole, I don’t
think the Patrol’s got a problem with it. There might be some isolated instances or
there may have been more in the past. I wouldn’t dare stand here and tell you I
don’t think it ain’t never happened cause I’m sure it has.
I think awareness has a lot to do with how people respond. In the last probably
five years, a lot of emphasis has been put on you might say racial profiling or
black/white issues and I think when people see that, naturally I think they sort of
pull back a little bit because they don’t want to be in where the target is. But I’d
like to say that the people you have in the community, when they come into law
enforcement, unless they make a conscious effort to change how they feel and think
about others, then that still is prevalent within them even in law enforcement. Of
course, being smart and wise, a lot of this may not come out to be where you can
really point and pull it out but I still think people must make a conscious effort to
change the way they was brought up or thinking and look at reality versus what is
and what’s not and that kind of thing. But I don’t think the Patrol has a problem
Patrol-wide but I will probably say that there are probably instances that these
things do take place.
Senate Bill 76 mandates that troopers complete the “stop form” as a means to better
understand law enforcement patterns and practices as well as the socio-demographics of citizens
stopped by troopers. Most participants believe such reporting is unproductive and an undue
burden on law enforcement personnel. Among other concerns, troopers reported that there was
no way to match the stop forms with citations. One must at least consider the accuracy of these
forms. Consider the following comments:
Well see you’ve got, you’ve got men writing a certain number of citations and
writing a certain number of warning tickets. Now we don’t know who these people
are writing these stop forms because it’s a secret number so you don’t know if
282
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

you’ve got something wrong in that stop form. They didn’t fill out all the blocks.
What do you do with it? You don’t know who the hell to give it back to and you
don’t know who to talk to about it so what do you do with it?
Some troopers fill those stop forms out daily and put them in the box. You’ll pick
them up. The secretary enters them. Some of them will wait until Sunday night
and fill them all out at one time like that. You know, I don’t recognize the numbers
on the top of them but when I enter some of them, some of them I can recognize the
handwriting. But I couldn’t a bit more tell you what that man’s number is. I don’t
even know what my number is right now.
If you knew who it was then you could check the number of citations they’re
putting down and the number of warnings tickets, but that still wouldn’t take into
account the vehicles they stopped that they didn’t do anything on.

Enforcement Decision
It is not possible for troopers to stop every traffic infraction; therefore, discretion is a
significant consideration. A vast majority of the participants state that they have made up their
mind before the stop as to what action they are going to take, such as a verbal warning, written
warning, or citation. They report doing so is a matter of fairness:
It is the only fair way to do it. You should issue a citation, it should be based on
the violation and not necessarily and not on their personal attitude. If you are
going to write a person a warning ticket for a 65 and the person is nice to you and
that is what you always do you write a warning ticket, I think you should continue,
if you got somebody that ain’t so nice, write a warning ticket. I am not there to
judge a person, I mean you stop and you don’t got to love them because you stop
them. I understand you write some people a citation and they got to hug your neck
because you write them a ticket or whatever. You have got to do your job and just
go out and do it.
There were clear differences among participants concerning written or verbal warnings. It
was expressed by one participant that giving a written warning as opposed to a verbal warning
was necessary to protect oneself in case of a complaint. A warning citation is perceived as a good

283
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

tool to promote vehicle safety as well as an effective public relations medium. But in any case,
the decision is probably made prior to the stop as noted by the following four participants:
And I think the violation of law and to what degree. I mean, most of the time I
think when the guys stop the car their mind is made up as to what they’re going to
do already and if the violation that you write a citation or you write a warning,
you just do that but I don’t think there’s any standard that’s set for, you know,
clear cut substantial violation to issue a citation for. A potential violation, you
give warnings.
You know what you are going to do before you get stopped. If it is a clear cut
substantial violation, they are going to get a ticket. Otherwise, they are going to
get a warning.
It is already decided.
I see the violation and it is a clear cut substantial. Before I hit my blue light, it is
already decided, whoever it is, it is a violation.

Decision to Search
When asked why they decided to search a car, 64 percent (N = 22) of the troopers stated
that they did not search vehicles unless absolutely necessary. Participants cite safety as a major
disincentive for conducting a search.
We have some people that do that but it’s not very wide across the State because
that’s a real good way to get hurt to do a lot of searching a vehicle and you don’t
have any help. Someone to watch what they’re doing while . . .
Thus, there were strong sentiments expressed against conducting a search. The most notable
exception is the “plain view search.” In the state of North Carolina, a police officer is required to
address any contraband that they see in plain view while executing a duty of their office. It was
apparent that the troopers would conduct a vehicle search if the circumstances revealed illegal
items that were in plain view.

284
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

The second most prevalent reason to search a vehicle is a ”search incident to arrest.” The
troopers stated that they conduct these types of searches. North Carolina statutes provide that any
law enforcement officer may conduct a search of an arrestee or the area within the arrestee’s
immediate control. This would obviously include the vehicle where appropriate. The primary
justification for the “search incident to arrest” is officer safety. A similar percentage of troopers
who reported conducting “searches incident to arrest,” also stated that safety was of primary
importance when deciding whether or not to conduct a search.
If I feel like searching that person, it is because I’m doing it for my safety. I do
not want to be riding down the road and have a big old gun come jumping [out] at me.
Troopers understood that they had the ability to ask for a “consent search.” Law
enforcement officers may ask a citizen for consent to search their property as long as a few
conditions are met. First, the law enforcement officer must be sure that the subject does not think
they are in custody at the time the request is made. Second, the officer must not imply menace or
negative consequences for a refusal. Troopers noted that search requests are seldom made.
[You] put yourself in a bad situation because, how are you going to justify that? You go to
an individual and say, you stop the speeding and there ain’t nothing else there; [C]an I
search your car? If you don’t do that on every single stop, there is going to be something
[that] look suspicious.
Minorities and Traffic Stops
Participants report that only a small number of minority stops could be attributed to
“unfair troopers.” They assert that in an organization their size, there would be some “bad
apples.” They concede that American society is racially divided, and that it would be naive to
suggest that no individuals joining the ranks of law enforcement harbor these sentiments. When
those attitudes infiltrate law enforcement agencies via its officers, “then all you do is put on the
285
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

uniform but you are still that same person; therefore, you are prone to have some racial bias.”
However, one third of the participants voiced the belief that, while prejudice exists inside law
enforcement, racial profiling is not a problem. When conducting a routine stop, these troopers
behave in a professional manner.
Indeed, most participants believe that the concern about racial profiling is unfounded.
They believe profiling to exist when things look unusual or out of place, for example, if two
African American males are spotted circling someone’s house in an all-white neighborhood. The
troopers label this as profiling because those individuals are considered to be out of place. They
contend they are trained to profile, that is their profession, but not necessarily racially motivated
profiles. Overall, the troopers maintain that minorities and others are stopped due to violations of
the law and not due to race. They also attribute the stops to the geographic location of the
patrolling trooper. If a trooper is patrolling a predominately Latino or African American area, he
or she is prone to stop the individuals within that ethnic group more because they dominate that
particular location.
With regard to racial differences in driving, a minority of participants maintain that
Latinos commonly operate their motor vehicle without a driver’s license. This was the most
commonly noted response from the troopers, and there were no other differences expressed by the
troopers.
Participants reported different experiences with how fellow troopers treat minorities and
ethnic groups. Some troopers acknowledged that they knew of certain troopers who they believed
were prejudiced against people of color, but acknowledged that they only knew of a few of them.
When asked if people needed to be concerned with racial profiling in North Carolina, one trooper
replied:
286
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Of course. I mean, as much as I like the guys I work with, you hear remarks.
Things that are said. It’s not hardcore derogative towards any group but you can
tell, yeah, I can see if this person’s put in that situation or if well, gee, every black
person, I stop they’re always giving me some mouth you know.
Okay, now, depending on how firm you are with that particular person. Now some
people are just not as understanding. I’m not saying all of them are like that but
some people are not used to being around blacks.
When asked about stereotyping within law enforcement, a few participants agreed that
stereotyping existed in law enforcement. Responses provided by the troopers suggest that
stereotyping serves more as “cues” in law enforcement. Although responses varied in the degree
of stereotyping taking place within agencies, the troopers who answered the question agreed that
it does exist in smaller law enforcement agencies as well as larger ones. Consider the following
participant remarks:
I just want to say that we are all trained and I do not care what
law enforcement agency you are in, you are trained to profile and you act on that.
It is inherent. This business requires that.
I think stereotyping and violations go together. If you look at a certain
group,
stereotype a certain group and in that group you deem it to be, have more
violations, then naturally you will work more at stopping whatever
According to those who believe that stereotyping does not have an impact on general law
enforcement practice, a couple troopers thought that stereotyping did have an impact on who is
searched.
With the variety of responses provided, it would be reasonable to conclude that the
responses are based on the troopers’ individual experiences and evaluations of the behavior of
others. The greater consensus gathered from the troopers suggests that racial profiling, while not
representative of the entire agency, does exist at an individual level—but to a lesser extent today
than it did in the past. They also recognized that many of the prejudices and much of the bigotry
287
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

remains. Generally, they are confident that their professionalism and training in the NCSHP keep
those problems in check. Still, troopers were less certain that the same could be said for law
enforcement, generally. When asked whether or not they were concerned about being victims of
racial profiling outside of North Carolina—each of the white participants said no; however, a
majority of African American participants said yes.

Steps to Stop Racial Profiling
The eradication of racial profiling was discussed by all participants. The most frequent
answer was accountability. The second most frequent response was training and education.
You talk about training, we are taught to recognize, to know or recognize
some of the effects but the majority of us that are not involved on the
interstate team where a lot of those guys are trained into what to look for,
doing extensive interviews.
Only one subject reported that, in his view, mandatory stop forms would reduce racial
profiling in traffic stops. Troopers were asked if they thought there were some who may
misrepresent the race of the person being stopped. One participant stated:
When it first came out I heard people say that they were going to be singled out
and they feel that way because they stopped five blacks and then you stop five
whites.
Having a camera in the patrol vehicle as a deterrent was questioned by oneparticipant.
We just had a trooper transfer from ______, this boy averages six or seven complaints a
week. They put a camera in his car to try to, you know, catch him. Now what good is that
going to do? The boy is that smart. He’s smart, he won’t say anything with the camera.
They can’t get him on anything but let the camera go ahead, he’s the biggest asshole
you’ve ever seen in your life. I’m serious. He talks down to people all the time.

288
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

More participants believed that cameras serve more as a protection ( for example,
protection against complaints), than as a tool to combat racial profiling.
A situation I had . . . when a guy stopped this blue jeep Cherokee. It was coming
out of Alabama, a black female who alleged that the trooper used racial slurs—
because she was stopped by a white trooper. . . Then I asked her did she know that
the incident, from the time the decision was made by the trooper and the car was
still rolling before it ever came to a stop, chasing, and all the way up to the time
when the car pulled off again, after being issued a citation, it had been
documented on video and did she know that. She did not know that and the last
time I heard from her was she hung up the phone. . .
Whether or not Senate Bill 76 will have a substantial effect on law enforcement practices
remains to be seen. The stop form appears to be widely resented and viewed as a burden by
troopers. The increased scrutiny of trooper behavior may result in troopers being concerned
about “balancing the books.” One participant stated:
Yeah, golly, the last four I wrote were whites, so I guess I better write a black
person now, or vice versa. Now, whether they actually go and do that, I don t
believe that happens, but there is a lot of conscious there, more so than there ever
has been in the past. I think most of our people, like me, are dedicated, committed
state troopers who are trying to do a good job if properly guided and led. I
believe that what they will do is when they are faced with it, they will deal with it
and I don’t think it will matter who it is.
Conclusion
Racial profiling is a multidimensional issue, in other words, disadvantage can appear in
numerous ways in a variety of situations. While findings from the focus groups are not
conclusive for the collective organization, they provide a better understanding of law
enforcement’s perspectives on racial profiling, as well as a contextual framework to assist with
the interpretation of stop and citation data. By way of a caveat, findings from these focus groups

289
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

should not be generalized to local and county law enforcement. Their law enforcement functions
and cultures are distinct.
Law enforcement has the burden of history and perceptions, real or distorted. Historically,
racial profiling and racist law enforcement in the United States are unfortunate realities, as noted
in works such as that of the Kerner Commission in the mid 1960s. Such practices have
continued, with and without harmful intent. The issue of prejudice is the engine to such practices,
and stereotypes are the fuel. Examples abound that give credence to the public=s concerns. For
example, Carl Williams, former Chief of the New Jersey State Police was fired for stating that
“mostly minorities” trafficked drugs. As we have argued previously, the result of casting a wide
“race” net is more likely the capture of innocents and not those who are guilty. The wide net
should never be mistaken for “good” police work. Most participants acknowledge that there
might be a few troopers that engage in racial profiling or race-based enforcement; however,
accusations of widespread practices are unfounded. It is generally believed that whatever racial
profiling that might have existed in the recent past (middle 1990s) was abated with the deemphasis of “drug interdiction” patrols. Indeed, a vast majority, if not all, of the participants
note that their decision to initiate a stop is a function of the “behavior of the vehicle.” Thus, they
believe that the issue of race-based decisions to initiate a stop is without merit. However,
deployment and enforcement patterns are another issue. It appears that some deployment
decisions are based on traffic demands (such as a road with a history of accidents or fatalities),
and others are based on areas with a significant “opportunity” factor, for example, opportunities
associated with density of bars. (“where the fishing is good”). Such decisions are more likely to
target low-income people (therefore, disproportionately people of color) than their high-income
counterparts. This is manifested with the presumption that they are less likely to challenge the
290
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

action in court and the higher income areas are involved in less overall criminality and disorder.
Overwhelmingly, participants report that they do not desire to search vehicles and will
only do so as a last resort. This is primarily due to safety concerns. This is contrary to the public
perception of excessive searching vehicles and harassment by troopers. Troopers gave the
impression that they were professional and methodical in their job performance. Indeed, each
noted that they would have decided what action they were going to take before the vehicle was
stopped.
The question now is what can be done to eliminate whatever racial profiling might exist.
Senate Bill 76 require the completion of “stop forms.” Such a requirement undoubtedly creates a
sense of institutional consciousness about racial profiling. Accountability appears to be an active
management concern. Supervisors use multiple strategies to monitor trooper activities, including:
1) periodic riding with a trooper, 2) review of citations, and 3) court visits to assess quality of
charges. The latter management strategies, coupled with a more diverse NCSHP, are possible
avenues to trooper accountability and behavior modification.

291
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Appendix C: Citation Charge versus Reason for Stop

In the analysis of Chapters 2 and 3, we generally use what we call the “citation event
data.” We defined a “citation event”as an event during which one or more citation forms are filled
out. Each form can have one or two charges listed, but multiple forms can be filled out for the
same event. During the year 2000, up to eighteen charges were filed for any given event.

A

driver could have received more than eighteen charges at an event—there is no limit to the
number of changes a trooper may file relating to a single incident. However, no person in 2000
received more than eighteen charges at any single event.
When analyzing the charges, we generally assumed that the most salient offense (and the
most serious charge) of the citation event was the first charge listed for the event. However, the
most serious charge is not always the charge that initially triggers the stop. When we examined
the data in further detail, we determined that, in most cases, the behavior that caused the stop and
a subsequent citation was not necessarily the first charge listed on the citation forms. For
example, we found records of drivers who were stopped for speeding and for having a revoked
license. It is more likely that the driver was stopped for speeding and then the trooper discovered
that the driver had a revoked license. Yet, the revoked license is typically listed as the first
charge. Appendix C examines in detail the assumption that the first charge listed on the form
indicates the behavior that was the likely reason for the stop. We will limit the analysis here to
citations not issued at accidents (under the assumption that our main interest is in trooper-initiated
stops resulting in citations).

292
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table C.1 lists all of the citation events for the year 2000 by type of charge. Type of
charge consists of several hundred codes that have been collapsed into the eleven categories listed
below. Appendix D lists the detailed codes and the collapsing rules used to combine them into
eleven categories. The behaviors which constitute the first charges on the citation forms are listed
down the left side of Table C.2. The behaviors that we, the researchers, believe are likely to have
been the behaviors that triggered the stops are listed across the top of the table. The eleven
categories across the top were arrived at by looking at combinations of charges for citation events
with more than one charge. If there was more than one charge listed in addition to speeding, we
reasoned that speeding would be the behavior most likely to initially trigger a stop —simply
because speeders usually catch the attention of NCSHP troopers from a substantial distance (this
is always the case when using radar, and typically at a distance of many yards—often several
hundred yards). As can be seen, the vast majority of cases fall along the diagonal of the table,
indicating that the first charge is most likely the behavior that triggered the stop.
Other behaviors, such as having an expired or revoked license, would not normally be
known to the trooper until after he or she had pulled over a driver. Equipment violations might be
visible from afar, but if a speeding charge is also listed, it is more likely that an equipment
violation became apparent after the car had been pulled over. Essentially, we have imposed a
hierarchy of charges as follows (from right to left in the table): speeding, unsafe movement,
failure to stop/yield, equipment violations, seatbelt violations, miscellaneous traffic violations,
DUI, license-registration-insurance-type violations, resisting/fleeing, stolen vehicle, and finally a
criminal charge. That is, if we find that a citation event has multiple charges, we rank them
according to this list. For example, if there are three charges, one for speeding, one for equipment
violation and one for revoked license, we would assume speeding was the charge that brought
293
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

about the stop. In the absence of a speeding charge, we would assume it was the equipment
violation.
The table can be read as follows. As we can see from the percentages (the second number
in each cell of the table), in 97.6 percent of instances in which seatbelt violations are the most
likely reason for pull-overs, a seatbelt charge is the first charge listed on the citation form. When
we, the researchers, think that DUI triggered the stop, only 86 percent of citations list DUI as the
first charge. Note that, for some serious charges—namely DUI and license-registration-insurance
violations—where there is only that one serious charge listed on the form, we will never know
what behavior brought about the stop. This is because the trooper decided to simply charge the
driver with the most serious offense and drop the other charges. For example, a driver may have
been pulled over for speeding and then the trooper discovered that the driver was uninsured. The
driver might be cited for the lack of insurance, but not the speeding. We would have no record of
the speeding, but only a record of the insurance violation. As for DUI, it is probable the driver
was observed driving in an irregular way and was stopped and found to be DUI—no charges were
filed for “weaving,” yet it is likely that the driver was stopped for some type of irregular vehicular
movement.
We think that the “license-registration-insurance violations” charge is considered by
troopers to be a serious offense, one that is subject to severe sanctions in the courts. In general, a
clear pattern of these charges emerges on the forms on which other charges are also filed.
Apparently, a driver will often be stopped for unsafe movement, failure to stop/or yield, or an
equipment violation. The trooper then finds that the driver lacks a valid driver’s license,
insurance, or registration, and as a result of that discovery, the trooper lists the latter offense as
the first charge on the citation form. We assume that this occurs because such offenses are
294
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

considered by troopers to be more serious than the violations that likely caused the pull-overs,
and so the license, insurance, or registration violation is written-up as the first charge. Other
common offenses that follow this pattern of being listed as the first change—but not necessarily
being the triggering charge—are failing to use a seatbelt and DUI. In general it can be said that,
even though the first charge is not always the likely reason for the stop, it is usually most serious
of the charges.
The fact that the first charge is the more serious charge—and not necessarily the charge
that initiated the stop—concerns us. We would like to know which specific behavior triggered
the pull-over. Given the inadequacies of the stop data, as discussed elsewhere in this report, the
citation data is the primary source of information about which behaviors seemingly trigger an
NCSHP intervention. To look at this question further, we next examine the same relationship
between the charge that we believe is the most likely behavior to trigger a stop and the first
charge that the trooper listed. In Table C.2, however, we limit the analysis to only multiple
charge incidents. Again, only citation events not involving an accident are included here, and
only citations for African Americans and whites (all others excluded). The top row in each
category is the count of the charge and the second row of the category is the percentage of the
column total. For example, let us examine the charges of DUI. In the fifth column, in 76.0
percent of the multiple-charge citation events with DUI as the likely reason for the stop, DUI is
listed as the first charge. Therefore, in 24.0 percent of the DUI stops resulting in a citation, the
first charge is something else—almost always a license, registration or insurance violation.
Assuming that DUI is a more serious charge than having an expired license, expired registration
or even no insurance, one could say that the NCSHP is not consistent in writing up DUI as the
first charge (although 76 percent of the time they write the first charge up as a DUI ).
295
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

We summarize the findings in Table C.2 as follows: 1) In 92.5 percent of the speeding
pull-overs, the first charge is also listed as speeding. In the 7.5 percent of the cases in which
speeding is not listed as the first charge, the first charge is most often either a license-registrationinsurance violation, or a DUI. Thus, the results follow a pattern of the more serious charge (DUI
or license-registration-insurance violations) as the charge listed first. (We assume DUI and
license-registration-insurance violations are generally more serious than speeding because the
punishments are generally stiffer for these violations.) 2) The pattern of discrepancy between a
first charge and a pull-over charge is similar for other types of violations as it is for speeding
violations: usually DUI or license-registration-insurance violations appear first among the
citations. For example, in 8.3 percent of the seatbelt pull-overs, a license-registration-insurance
violation appears first (the most of any non-seatbelt violation). 3)Where there is a charge of
“unsafe vehicular movement,” there is often a DUI violation as well, as one might reasonably
expect.
To follow up on the analysis of the relationship between the unsafe movement charge
and additional violations listed on the forms, we see that in 35.2 percent of the unsafe movement
pull-overs the driver is cited for DUI as the first violation. In another 30.6 percent of cases in
which unsafe movement appears to be the likely reason for the pull-over, there is also a licenseregistration-insurance violation. In 16.9 percent of these unsafe movement cases a seatbelt charge
is listed first.

296
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table C.1 Likely Reason for Stop by First Charge
Likely Reason for the Pull-over by First Charge: All Non-Accident Citation Events
First
Charge

Crime

Crime

Stolen
Vehicle

Resist
, Flee,
or Escape

License,
Insurance or
Registration

DUI

Misc.
Traffic

Seatbelt

Equipment

Fail to
Stop
or
Yield

Unsafe
Movement

Speeding

#

3589

1

11

155

91

28

47

14

1

33

21

%

100

4.5

7.1

0.2

0.5

1.2

0

0.4

0

0.2

0

Stolen
Vehicle

#

Resist,
Flee,
Escape

#

145

39

10

3

9

1

3

21

24

%

92.9

0.1

0.1

0.1

0

0

0

0.1

0

License,
Insurance
or
Registration

#

71879

2248

395

1933

694

1147

1672

1516

99.7

13.4

17.4

1.7

19.1

10.7

11.5

0.5

DUI

#

1440
3

216

790

103

394

1910

1512

86

9.5

0.7

2.8

3.7

13.2

0.5

%

%

21

2

1

1

95.5

0

0

0

%
Misc.
Traffic

#

1634

4

40

24

10

%

71.8

0

1.1

0.2

0

Seatbelt

#

1126
95

278

839

752

839

%

97.6

7.7

7.8

5.2

0.3

Equipment

#

2502

1

16

9

%

68.9

0

0.1

0

Fail Stop
or Yield

#

8360

54

40

%

77.8

0.4

0

Unsafe
Movement

#

10010

61

69.1

0

Speeding

#

%

291744

%

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

98.6

297

Table C.2 First Charge versus Likely Behavior for Pull-over, Multiple Charge Events Only
Likely Reason for the Pull-over by First Charge: All Non-Accident Citation Events
First
Charge

Crime

Stolen Vehicle

Resist, Flee, Escape

Crime

Stolen
Vehicle

Resist,
Flee, or
Escape

License,
Insur.
Registr
ation

DUI

Misc.
Traffic

Seatbelt

Equipment

Fail to
Stop or
Yield

Unsafe
Move
ment

Speeding

#

538

1

10

126

78

19

41

11

1

26

19

%

100

25

52.6

0.9

1.1

3.2

0.2

1.1

0

0.6

0

#

3

2

1

1

%

75

0

0

0

#

9

27

7

3

8

1

3

21

21

%

47.4

0.2

0.1

0.5

0

0.1

0.1

0.5

0

License, Insurance
or Registration

#

14625

1548

309

1562

526

811

1230

1220

%

99

22.8

52.6

8.3

51

35.6

30.6

2.7

DUI

#

5163

164

631

85

323

1416

1257

%

76

27.9

3.1

8.2

14.2

35.2

2.8

#

93

2

37

22

9

%

15.8

0

3.6

0.5

0

#

17930

226

764

678

754

%

88.5

21.9

33.5

16.9

1.7

#

145

1

16

9

%

14.1

0.1

0.4

0

#

377

52

33

%

16.5

1.3

0.1

#

561

55

%

13.9

0.1

Misc. Traffic

Seatbelt

Equipment

Fail Stop or Yield

Unsafe Movement

Speeding

#

41442

%

92.5

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

298

Therefore, in the majority of multiple charge citation events involving unsafe movement, the
driver is charged first with a violation other than unsafe movement.
This issue of unsafe movement charges is of special interest in this study because it might
be argued that “unsafe vehicular movement” is rather subjectively and haphazardly judged by a
trooper on the highway. For example, how much “side to side” motion of a vehicle is necessary
to have unsafe movement? The present data suggest to us that unsafe vehicular movement (in a
multiple charge situation) is often accompanied by other violations of a more serious nature. Not
only are they more serious, but they are not as subjective as connoted in the term “unsafe
movement.” In the particular cases of license-registration-insurance violations and DUI, the
drivers are almost always guilty of the violation—either the driver has a problem with his or her
license-registration-insurance or he or she does not, and if the driver is charged with DUI he or
she probably failed a breathalizer test. Subsequently, there is little subjectivity involved in
determining whether or not the driver was guilty of these other charges listed first in the multiple
charge citation events. Therefore, if the initial pull-over for “unsafe movement” was pre-textual
by the NCSHP (a pretext for conducting a vehicular search for drugs), the associated charges
would in most instances cast doubt upon that interpretation, primarily because the drivers have
often been found to be unambiguously guilty of something more serious than the subjective
unsafe movement. In Chapter 4, we show that so few vehicular searches are conducted by the
ordinary NCSHP troopers that it would seem unlikely that stops for unsafe movement or any
other behavior would often be a pretext for a drug search (a somewhat different interpretation is
given to the NCSHP Criminal Interdiction Team’s procedures and searches—see also Chapter 4).
Above, we showed that the first charge and the pull-over charge are not often the same
charge. Where there is a difference between the pull-over charge and the first charge, more often
than not, the first charge seems to be a more serious charge than the pull-over charge. A more
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

299

important question for the current research, however, is whether or not there are any racial
differences in the type of charges appearing among the first charge versus the pull-over charge.
Table C.3 shows that there are some racial differences for a variety of pull-over charges as they
differ from three first charges: license-registration-insurance violations, DUI, and seatbelt
violations. (Too few cases of other offense categories occur to warrant including them in the
table and we drop crime charges, stolen vehicle, and resist/flee violations completely as they
involve too few observations to be of concern.)
In Table C.3, we can read the table as follows: in multiple charge citation events, if an
African American and a white are pulled over for a seatbelt violation, the African American is
found to have a license-registration-insurance violation 7.9 percent of the time, compared to 8.3
percent of the time for a white driver. In general, the results of an analysis of Table C.3 suggest
that license-registration-insurance violations are more prevalent among African American drivers
involved in multiple citation events than they are among white drivers. This is not surprising
since African American drivers are more often cited for license-registration-insurance violations
than are whites for one-charge citation events.
Here, however, if the first charge were assumed to be the charge that initiated the vehicle
pull-over, more error is introduced for African American drivers than for white drivers. Use of
first charge as a proxy for pulled charge would result in disproportionately more African
Americans having license-registration-insurance as the violation that caused the pull-over, rather
than the behavior that is most likely to have caused it: DUI, miscellaneous traffic violations,
equipment violations, failure to stop, unsafe movement, and speeding (seatbelts are the one
exception to the pattern where a slightly higher percentage of whites stopped for seatbelt
violations have license-registration-insurance violations (8.3 percent) than have African
300
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Americans (7.9 percent). Note, however, that although the percentage differences appear to be
somewhat large (for example, 28.2 versus 19.9 of the DUI multiple charge events—an 8.3 percent
difference), these represent a much smaller percentage of the all the cases (including the single
charge events). Thus, for example, if one added up all of the African Americans who were issued
a first charge of a license-registration-insurance violation and who were also likely to have been
pulled over for another charge of DUI or other miscellaneous violation (such as a traffic violation,
an equipment violation, or failure to stop/yield violation) they would constitute about 1 percent of
all African American citation events (1,249 of 122,649 African American citations).
Because only a very small bias would be involved in using first charge rather than the
charged behavior that we believe is most likely to have triggered the stop, we use first charge in
the analysis of Chapter 2 and 3. (For much of this analysis there is no distinction by type of
charge and so the issues discussed in this appendix are moot). Although there may be a slight
bias in our analysis by not using the charge we think brought about the stop, it is in a known
direction: more African Americans will appear to have been stopped for license-registrationinsurance violations than actually were. One should bear in mind, however, that even though we
could conceivably use our hierarchically derived charge that resulted in the stop as the charge for
analytic purposes in Chapters 2 and 3, the 1 percent of African American citation events involved
(where there is a difference) is much smaller than the 17.3 percent of all African American
citations in which license-registration-insurance violations are involved. We suspect that, for
most of these 17.3 percent of cases, some behavior triggered the stop, but we do not

301
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

have any record of that behavior. In effect, the 1 percent of cases that we could “correct” by
using our estimate of the charge that initiated the stop would be a very small correction indeed.
Table C.3 Racial Differences in First Charge Relative to Likely Reason for the Vehicular
Pull-over (Non-Accident Citations, African Americans and Whites Only) (Column Percentages)
First
Charge

DUI
AA

Misc. Traffic
White

AA

Seatbelt

White

AA

White

Equipment

Fail to Stop

AA

AA

White

White

Unsafe Move.
AA

White

Speeding
AA

White

License/
Reg

28.2

19.9

56.6

50.1

7.9

8.3

55.9

49.4

48.3

30.3

41.1

26.1

3.4

2.4

DUI

71.2

78.5

27.9

27.9

3.1

3.1

8.8

8.1

14.8

13.9

31.4

36.8

3

2.7

--

-–

--

--

88.9

88.3

20.3

22.5

22.3

38.1

13.6

18.3

1.4

1.8

Seatbelt

Example: “Of multiple charge citation events in which DUI was the initial reason for the pull-over, African
Americans are more likely to have a license-registration-insurance violation listed as the first charge than
are whites(28.2 percent versus 19.9 percent).”

302
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Appendix D: Offense Codes for Citation Analysis
(See recoded variable at end of Appendix D for the hierarchy of offenses used to define the
offense likely to have caused the stop).
920
922
930
935
940
945
950
955
999
1020
1022
1023
1024
1026
1028
1030
1032
1040
1099
1103
1116
1118
1120
1122
1124
1125
1126
1128
1130
1132
1134
1136
1137
1138
1139
1140
1142
1144

14-18 VOLUNTARY MANSLAUGHTER F
A
14-18 INVOLUNTARY MANSLAUGHTER
F
A
14-17 MURDER
F
A
14-17 FIRST DEGREE MURDER F
A
14-17 SECOND DEGREE MURDER
F
A
COMMON LAW
SOLICITATION TO COMMIT MURDER
941003
COMMON LAW
ATTEMPTED MURDER
F
A
14-18.1 SOLICITATION TO COMMIT MURDERF
O
HOMICIDE - FREE TEXT
A
14-39 KIDNAPPING
F
A
14-43 ABDUCTION OF MARRIED WOMAN
F
O
14-41 ABDUCTION OF CHILDREN
F
A
14-42 CONSPIRING TO ABDUCT CHILDREN F
A
14-39 FIRST DEGREE KIDNAPPING
F
A
14-39 SECOND DEGREE KIDNAPPING F
A
14-43.3
FELONIOUS RESTRAINT F
A
COMMON LAW
ATTEMPTED KIDNAPPING
F
COMMON LAW
FALSE IMPRISONMENT M
A
KIDNAPPING - FREE TEXT
A
14-27.2(A)
FIRST DEGREE RAPE
F
A
14-27.4(A)(1) FIRST DEGREE SEX OFFENSE CHILD F
14-202.1
INDECENT LIBERTIES WITH CHILD
F
14-27.2(A)(1) FIRST DEGREE RAPE CHILD
F
A
14-27.3(A)
SECOND DEGREE RAPE F
A
14-27.5(A)
SECOND DEGREE SEXUAL OFFENSE F
14-27.6
ATTEMPT 1ST DEGREE RAPE F
O
14-27.6
ATTEMPT 1ST DEGREE SEX OFFENSE F
14-27.6
ATTEMPT SECOND DEGREE RAPE
F
14-27.6
ATTEMPT 2ND DEGREE SEX OFFENSE O
14-27.4(A)
FIRST DEGREE SEXUAL OFFENSE
F
14-27.7
SEX OFFENSE - PARENTAL ROLE
F
14-27.7
SEX OFFENSE INSTITUTION
F
A
14-27.7A(A) STAT RAPE/SEX OFFN DEF >=6YR
F
14-27.4
STATUTORY SEXUAL OFFENSE F
O
14-27.7A(B) STAT RAPE/SEX OFFN DEF>4-<6YR
F
14-27.7
ATT SEX OFFENSE-PARENTAL ROLE F
14-27.2
ATTEMPT 1ST DEGREE RAPE F
A
14-27.4
ATTEMPT 1ST DEGREE SEX OFFENSE F
303

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

F

O

910408
941003961202
941003

A

930423
940606

A
A
A
950818
O
950818
O
950818
950818
A
A
910408
A
951201
940606
A
951201
A
910408
950818
A
950818

1146
1148
1160
1161
1199
1202

14-27.3
ATTEMPT SECOND DEGREE RAPE
F
14-27.5
ATTEMPT 2ND DEGREE SEX OFFENSEF
14-208.11
FAIL REGISTER SEX OFFENDER(M) M O
14-208.11
FAIL REGISTER SEX OFFENDER(F)
F
SEXUAL ASSAULT - FREE TEXT
A
14-87 ATT ROBBERY-DANGEROUS WEAPON
F

1220
1222
1224
1226
1228
1299
1301
1302
1303
1304
1305
1306

14-87.1
COMMON LAW ROBBERY
F
A
14-87 ROBBERY WITH DANGEROUS WEAPON
F
14-89.1
SAFECRACKING F
A
14-87.1
ATTEMPTED COMMON LAW ROBBERY
COMMON LAW
CONSP ARMED ROBBERY BUS/PERS
ROBBERY - FREE TEXT
A
14-32.1(F)
ASSAULT HANDICAPPED PERSON
M
14-32.1(B)
AWDWIKISI HANDICAPPED PERSON F
14-32.1(C)
AWDWISI ON HANDICAPPED PERSON F
14-32.1(D)
AWDWIK ON HANDICAPPED PERSON F
14-32.1(E)
FELONY ASSAULT ON HANDICAPPED F
14-32.2(A)
PATIENT ABUSE AND NEGLECT
F

1320
1321
1322
1324
1325
1326
1327
1328
1330
1332
1333
1334
1336
1338
1340
1342
1344
1346
1348
1352
1353
1354
1355
1356
1357
1358

14-33(B)(1) ASSAULT ATTEMPT SERIOUS INJURY M
O N99 951201
14-33(B)(1) ASSAULT - SERIOUS INJURY
M
O
910408 940608
14-34 ASSAULT BY POINTING A GUN M
A
14-31 MALICIOUS ASSAULT IN SECRET
F
A
14-32.4
ASSAULT INFLICT SERIOUS INJ(F)
F
A
970101
14-33(B)(1) ASSAULT INFLICT SERIOUS INJURY M
O
951201
14-33(B)(1) ASSAULT - DEADLY WEAPON M
O
910408940606
14-33(B)(3) ASSAULT ON A CHILD UNDER 12
M
O
951201
14-33(B)(2) ASSAULT ON A FEMALE M
O
951201
14-288.9
ASSAULT ON EMERGENCY PERSONNEL
M
A
14-33(B)(7) ASSAULT ON DSS WORKER
M
O
910408 911001
14-288.9
AWDW EMERGENCY PERSON F
A
14-33(A)
ASSAULT AND BATTERY M
A
14-33(B)(4) ASSAULT ON LAW OFFICER
M
O
911001
14-33(B)(1) ASSAULT WITH A DEADLY WEAPON M
O
951201
14-34.2
AWDW ON OFFICER
F
O
951201
14-32(C)
AWDW INTENT TO KILL F
A
14-32(B)
AWDW SERIOUS INJURY F
A
14-32(A)
AWDWIKISI F
A
14-33(B)(5) ASSAULT ON COURT OFFICER M
O
911001
14-33(B)(6) ASSAULT ON SCHOOL TEACHER
M
O
911001
14-33 ASSAULT/AFFRAY M
O
910408
14-33(B)(8) ASSAULT ON A GOVT OFFICIAL M
O911001
951201
14-34.2
AWDW GOVERNMENT OFFICIAL
F
A
911001
14-33(A)
SIMPLE AFFRAY M
A
921027
14-33(B)(9) ASSAULT - SPORTS OFFICIAL M
A
931201
304

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

A
950818
A
950818
951201 971201
A
951201
A
A
F
F
A
O
O
O
A
A

A
A
941003
941003
941003

1360
1362
1364
1366
1368
1370
1371
1372
1374
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
2020
2021

14-28 MALICIOUS CASTRATION
F
A
14-29 CASTRATION WITHOUT MALICE
F
A
14-29 MAIMING WITHOUT MALICE F
A
14-30 MALICIOUS MAIMING
F
A
14-33(A)
SIMPLE ASSAULT M
A
14-30.1
THROWING ACID OR ALKALI F
A
113-290.1
NEGLIGENT HUNTING
M
O
911001
940606
14-34.4(A)
ADULTERATED OR MISBRANDED FOOD
F
A
14-34.4(B)
ADLTRT OR MISBRAND TO EXTORT F
A
14-32.3(A)
ABUSE DISABLE/ELDER SER INJ
F
A
951201
14-32.3(A)
ABUSE DISABLE/ELDER WITH INJ
F
A
951201
14-32.3(B)
NEGLECT DISABLE/ELDER SER INJ
F
A
951201
14-32.3(B)
NEGLECT DISABLE/ELDER WITH INJ F
A
951201
14-32.3(C)
EXPLOIT DISABLE/ELDER(F)
F
A
951201
14-32.3(C)
EXPLOIT DISABLE/ELDER(M) M
A
951201
14-33(C)(1) ASSAULT ATTEMPT SERIOUS INJ(M) M
A
951201
14-33(C)(1) ASSAULT INFLICT SERIOUS INJ(M)
M
A
951201
14-33(C)(1) ASSAULT WITH A DEADLY WEAPON M
A
951201
14-33(C)(2) ASSAULT ON A FEMALE M
A
951201
14-33(C)(3) ASSAULT ON A CHILD UNDER 12
M
A
951201
14-33(C)(4) ASSAULT GOVT OFFICIAL/EMPLY
M
A
951201
14-33(C)(5) ASSAULT SCHOOL BUS PERSONNEL M
A
951201
14-33.2
HABITUAL MISDEMEANOR ASSAULT F
A
951201
14-34.5
ASSAULT LEO/PO/OTHER W FIREARM F
A
971201
14-34.6(A)
ASSAULT EMERGENCY PRSNL(M)
M
A
951201
14-34.6(B)
ASSAULT EMERG PRSNL IBI/WDW
F
A
951201
14-34.6(C)
ASSAULT EMERG PRSNL FIREARM
F
A
951201
14-34.7
ASSAULT LEO/PO/OTHER SER INJRY F
A
961202
ASSAULT - FREE TEXT
A
14-58 ARSON
F
A
14-60 BURNING OF A SCHOOL HOUSE
F
A
910610

2022
2023
2024
2025
2026
2027
2028
2030
2032
2099
2110
2112
2116
2199

14-62.1
BURNING BLDG UNDER CONSTRUCT F
A
14-62 BURNING UNOCCUPIED BLDG F
A
14-67 ATTEMPT TO BURN BLDG OR BOAT F
A
14-62 BURN OCCUPIED BUILDING
F
O
931101
14-65 FRAUDULENTLY BURNING DWELLING
F
A
14-58 FIRST DEGREE ARSON F
A
940606
14-58 SECOND DEGREE ARSON
F
A
940606
14-66 BURNING PERSONAL PROPERTY
F
A
951201
14-62.2
BURN CHURCH/RELIGIOUS BLDG
F
A
961202
ARSON - FREE TEXT
A
14-118.4
EXTORTION F
A
14-118
BLACKMAILING M
A
14-113.13
UNLAWFUL OBTAINING CREDIT CARDF
A
940606
EXTORT - FREE TEXT
A
305

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

2206
2210

14-55 POSSESSION OF BURGLARY TOOLS F
14-54 BREAKING/ENTERING AND LARCENY

2211

14-72(B)

A
F

BREAKING/ENTERING AND LARCENY

2212 14-54(A)
BREAKING AND OR ENTERING (F)
2214 14-54(B)
BREAKING OR ENTERING (M) M
2215 14-54(A)
ATTEMPT TO BREAK OR ENTER (F)
2216 14-56 BREAK OR ENTER A MOTOR VEHICLE

O
F

O

F
A
A
FO940606
F
A

910408

2217
2218
2219
2220
2221
2222
2223
2224
2226
2228
2230
2232
2240
2299
2301
2302
2320
2321
2322
2323
2324
2325

14-56.1
BREAK COIN/CURRENCY MACH (F)
F
A
921120
14-56.1
BREAK COIN/CURRENCY MACH (M) M
A
14-56.2
DAMAGE COIN/CURRENCY MACHINE M
A
14-56 BREAK/ENTER RAILROAD CAR F
A
14-56 ATTEMPT BREAK/ENTER MOTOR VEH
M
A
930308
14-56 BREAK/ENTER TRAILER/AIRCRAFT F
A
14-56 BREAK/ENTER BOAT
F
A
14-57 BURGLARY WITH EXPLOSIVES F
A
14-51 FIRST DEGREE BURGLARY
F
A
14-51 SECOND DEGREE BURGLARY F
A
14-54(B)
WRONGFULLY BREAK/ENTER A BLDGM
O
910408
14-54(A)
ATTEMPT TO BREAK AND ENTER (M) M O 940606 910408
14-51 ATTEMPT FIRST DEGREE BURGLARY F
A
940606
BURGLARY - FREE TEXT
A
14-168.1
FELONY CONVERSION F
A
14-168.1
MISDEMEANOR CONVERSION M
A
14-72(A)
FEL LARCENY - >$400
F
O
911001
14-72(A)
FELONY LARCENY
F
A
911001
14-72(A)
MISDEMEANOR LARCENY
M
A
14-87 AID AND ABET ARMED ROBBERY
F
A
910408
14-72 LARCENY BY TRICK
F
O
910408
14-72(B)
LARCENY OF A FIREARM
F
A
910408

2326
2328
2330
2331
2332
2333
2334
2335
2336
2339
2340
2341

14-72(A)
LARCENY BY TRICK
M
O
14-72.1(D)
LARCENY - REMARKING GOODSM
O
14-72.1(D)
LARCENY BY CHANGING PRICE TAG M
14-75 LARCENY OF CHOSE IN ACTION
F
A
14-72.1(D)
LARCENY - TAG TRANSFER
M
O
14-72.1(D1) LARCENY BY ANTI-INVNTRY DEVICE F
14-74 LARCENY BY EMPLOYEE
F
A
14-74 "LARCENY BY EMPL >=$100,000"
F
A
14-72(B)(1) LARCENY FROM THE PERSON F
A
14-71 RECEIVING STOLEN GOODS (M)
M
A
14-71 RECEIVING STOLEN GOODS (F) F
A
14-71.1
POSSESSION OF STOLEN GOODS (F) F
306

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

940606
940606
A
910408
A
971201

A

940606
971201

2342

14-72.1

SHOPLIFTING CONCEALMENT GOODS

2343 14-71.1
POSSESSION OF STOLEN GOODS (M)
2344 14-81 "LARCENY OF HORSE, SWINE, ETC." M
2345 14-81(A)
LARCENY OF HORSE/SWINE/CATTLE
2346 14-76 LARCENY OF PUBLIC RECORD M
A
2348 14-75.1
LARCENY OF SECRET PROCESS
2350
2351
2352
2353
2354
2356
2358
2360
2363
2390
2391
2392
2399
2404
2406
2408
2499
2503
2505
2508
2510
2520
2522
2524

M
O
F

A

F

A

A

A
910408
910408

14-77 LARCENY OF WILL
M
A
14-82 TAKE HORSE/MULE/DOG TEMP PURP M
A
14-84 LARCENY OF DOG M
O
910408
14-72 ATTEMPTED LARCENY M
A
14-118.5
THEFT OF CABLE TV SERVICE M
A
14-72(B)(2) LARCENY AFTER BREAK/ENTER
F
A
14-72.4(A)
UNAUTH TAKE/SALE DAIRY CASE
M
A
14-81(A1)
LARCENY OF DOG F
A
910408
14-79.1
LARCENY OF PINE STRAW
F
A
971201
14-72.2
UNAUTHORIZED USE OF MOTOR VEH M
A
940606
14-72(A)
LARCENY OF MOTOR VEHICLE (F)
F
A
940606
14-72(A)
LARCENY OF MOTOR VEHICLE (M)
M A 940606
LARCENY - FREE TEXT
A
14-72.2
UNAUTHORIZED USE OF MOTOR VEH M
O
940606
14-72(A)
LARCENY OF MOTOR VEHICLE F
O
940606
14-72(A)
LARCENY OF MOTOR VEHICLE M
O910408
940606
STOLEN VEHICLE - FREE TEXT
O
940606
14-13 COUNTERFEITING COIN F
A
14-13 UTTERING COUNTERFEIT COIN F
A
14-14 POSSESS COUNTERFEITING TOOLS
M
O
941003
COMMON LAW
COMMON LAW FORGERY
M
A
940606
14-119
FORGERY OF INSTRUMENT
F
A
14-119
FORGERY AND UTTERING
F
O
14-120
UTTERING FORGED INSTRUMENT
F
A

2526 14-120
2527 14-120
2528 14-120
2540
2542
2599
2601
2602
2603
2604
2605
2606

M

ATTEMPTED UTTERING F
O
UTTERING FORGED ENDORSEMENT
FORGERY OF ENDORSEMENT F

F
A

910408
A
910408
910408

14-122
FORGERY OF DEEDS OR WILLS F
A
COMMON LAW
FORGERY F
O
930322
940606
FORGERY - FREE TEXT
A
108A-53.1(A) BUY/SELL/DISTRIB FOOD STAMPS
F
A
971201
108A-53.1(B) ILLEG POSS/USE FOOD STAMPS(M)
M
A
971201
58-2-161
INSURANCE FRAUD
F
A
930208
108A-53.1(B) ILLEG POSS/USE FOOD STAMPS(F)
F
A
971201
14-113.13
FINANCIAL CARD FRAUD (M) M
A
14-107
WORTHLESS CHECK
M
O
910415
307

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

2607
2608
2609
2610
2611

96-18(B)
ESC LAW FRAUD VIOL M
O
96-18 GIVING FALSE INFO TO ESC
M
O
96-18(B)
NO REPORT TO ESC
M
O
14-107(3)
WORTHLESS CHECK NO ACCOUNT
M
14-106
OBTAIN PROPERTY WORTHLESS CHK M

920203
920203
920203
A
O

2612
2613
2614
2615
2616
2617
2618
2619

14-113.9
108A-53
14-113.11
108A-53
14-113.13
108A-39(A)
14-110
108A-39(B)

FINANCIAL CARD THEFT
F
FOOD STAMP FRAUD (M) M
A
FINANCIAL CARD FORGERY
F
FOOD STAMP FRAUD (F) F
A
FINANCIAL CARD FRAUD (F) F
PUBLIC ASSISTANCE FRAUD (M)
DEFRAUDING INNKEEPER
M
PUBLIC ASSISTANCE FRAUD (F)

A
M
A
F

A

2620

14-168

HIRING WITH INTENT TO DEFRAUD

M

A

2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2632

108A-63
14-151.1
14-155
14-151
14-113.1
14-104
14-113.1
14-186
14-113.1
14-113.1
14-100

MEDICAL PROVIDER FRAUD
F
INTERFERE WITH UTILITY METER
UNLAWFUL TELEPHONE TAP M
DIVERTING UTILITY USE
M
FALSE TELEPHONE CREDIT
M
FAIL TO WORK AFTER PAID
M
CREDIT VIO W/O AUTH OF ISSUEE
FALSE HOTEL REGISTRATION M
CREDIT VIO W/O AUTH OF ISSUEE
CREDIT VIOL AFTER CREDIT REV
OBTAIN PROPERTY FALSE PRETENSE

A
910408
M
A
A
A
A
A
M
O
940606
A
F
O
940606
MO910408 940606
F
A

A
A
A

2633 14-100
2634 20-30(5)
2636 14-114

"OBT PROP FALSE PRET >=$100,000" F
FRAUDULENT LICENSE PERMIT
T
FRAUD DISPOSAL MORTGAGE PROP M

A
O
A

971201
940606

2638 14-113.13
2639 14-113.13
940606
2640 14-221.1
2641 14-221.2
2643 14-454(B)
2644 14-117
2645 14-455(A)
2646 14-167

UNLAWFUL OBTAINING CREDIT CARD
UNLAWFULLY OBTAIN CREDIT CARD

FO
M

940606
O

ALTER/STEAL/DEST CRIMINAL EVID
ALTER COURT DOCUMENTS
F
ACCESSING COMPUTERS (M) M
FALSE ADVERTISING
M
A
DAMAGING COMPUTERS(F)
F
FAIL TO RETURN RENTAL PROPERTY

F
A
A

A

A
M

A

2647 14-455(B)
2648 14-214
2649 14-168.4

DAMAGING COMPUTERS(M)
M
FRAUDULENT INSURANCE CLAIM
FAIL RETN PROP RENTD PUR OPT

A
F
M

O
A

308
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

930208
910408

2650

14-106

OBTAIN PROPERTY WORTHLESS CHK M

A

ACCESSING COMPUTERS (F)
F
A
SECRETING LIEN PROPERTY
M
A
WORTHLESS CHECK CLOSED ACCOUNT
WORTHLESS CHK 4TH SUB OFFENSE M

M
A

A

2657 90-108(A)
OBTAIN CS BY FRAUD (M)
M
A
2658 90-108(A)
OBTAIN CS BY FRAUD (F)
F
A
2659 18B-302(E)(3)
OBT/ATT OBT ALC OTHERS ID M

A

60992

2660

18B-302(E)(2)

OBT/ATT OBT ALC FALSE ID

M

A

2661
2662

18B-302(E)(3)
18B-302(E)(1)

OBT/ATT OBT ALC OTHER DL
OBT/ATT OBT ALC FALSE DL

M
M

A
A

2651 14-454(A)
2654 14-115
2655 14-107(4)
2656 14-107

920609

2663 96-18(A)
2664 96-18(B)
2665 96-18(C)
2666 14-107

MISREP TO OBTAIN ESC BENEFIT
MISREP TO PREVENT ESC BENEFIT
EMPL SEC LAW VIOLATION
M
SIMPLE WORTHLESS CHECK
M

M
M
A
A

A
920203
A
920203
920203
910415

2670

14-107

FELONY WORTHLESS CHECK

A

911001

2680
2699
2704
2705
2710
2714

53-276

CHECK CASHING WITHOUT LICENSE
FRAUD - FREE TEXT
A
EMBEZZLEMENT OF STATE PROPERTY
"EMBEZ STATE PROP >=$100,000"
F
EMBEZZLEMENT BY FIDUCIARY
F
EMBEZZLEMENT RAILROAD OFFICER

F

A

F
A
O
F

A
971201

2715
2718
2719
2720
2722
2723
2799
2908
2910
2912

14-94
14-90
14-90
14-96
14-92
14-92

A

971201

14-91
14-91
14-90
14-94

F

"EMBEZZLEMENT RR OFF >=$100,000" F
EMBEZZLEMENT F
A
"EMBEZZLEMENT >=$100,000" F
A
EMBEZZLEMENT - INSURANCE AGENT
EMBEZZLEMENT-PUB OFF/TRUSTEES F
"EMBEZ PUB OFF/TRST >=$100,000"
F
EMBEZZLE - FREE TEXT
A
69-33 NEGLIGENT AND CARELESS BURNING
14-66 BURNING PERSONAL PROPERTY
F
14-160
INJURY TO PERSONAL PROPERTY

2914 20-107
2916 14-132(A)
2919 14-140.1
2920 14-127
2922 20-107(B)

TAMPERING WITH VEHICLE PARTS
DEFACING PUBLIC BUILDING M
BURN WITHOUT WATCHMAN M
INJURY TO REAL PROPERTY
M
TAMPERING WITH VEHICLE STEAL
309

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

971201

940606
A

971201
F
O
940606
A
850701
A
971201
M
O
M
T
A
A
A
T

O

940606
940606

A
O

940606

951020
O

940606

2924
2926
2928
2930
2940
2941
2999
3401
3405
3409
3410
3411
3412
3413
3414
3415
3416
3420
3423
3424
3425
3426
3428
3429
3430
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3470

20-107
TAMPERING WITH VEHICLE
T
O
940606
14-160
DAMAGE TO PERSONAL PROPERTY M
O
940606
14-160.1(A) ALTER/REMOVE NMV SERIAL NUMBER
M
A
14-160.1(B) POSS/SELL/BUY ALT NMV SER NO
M
A
14-138
FAIL TO EXTINGUISH FIRE
M
O
941003950606
14-138.1
FAIL TO EXTINGUISH FIRE
M
A
950717
DAMAGE PROPERTY - FREE TEXT
A
90-113.22
POSSESS DRUG PARAPHERNALIA
M
A
90-95(E)(8) POSS CS W/IN 300 FT OF SCHOOL
F
A
910408
90-95(E)(9) POSS CS PRISON/JAIL PREMISES
F
A
971201
90-95(E)(8) S/D CS W/IN 300 FT OF SCHOOL F
A
910408
90-95(H)(3A) TRAFFICKING IN AMPHETAMINE
F
A
960501
90-95(H)(3B) TRAFFICKING IN METHAMPHETAMINEF
A
960501
90-95(H)(4A) TRAFFICKING IN LSD
F
A
960501
90-95(I)
CONSPIRE TRAFFIC AMPHETAMINE F
A
960501
90-95(I)
CONSPIRE TRAFFIC METHAMPHETAMIF
A
960501
90-95(I)
CONSPIRE TRAFFIC LSD F
A
960501
90-112
SEIZURE/FORFEITURE OF VEHICLE
FO910408
950206
90-95.4(A)(1) CS HIRE/USE MINOR>13 DEF 18<21
F
A
990127
90-95.4(A)(2) CS HIRE/USE MINOR<=13 DEF18<21
F
A
990127
90-95.4(B)(1) CS HIRE/USE MINOR >13 DEF >=21
F
A
990127
90-95.4(B)(2) CS HIRE/USE MINOR<=13 DEF>=21
F
A
990127
90-95.6
PROMOTE DRUG SALES BY A MINOR F
A
991027
90-95.7
PARTICIP IN DRUG VIOL BY MINOR F
A
990127
90-108(A)(14)
EMBEZZLE CS BY EMPLOYEE OF REG FA
910408
90-95(A)(1) SELL MARIJUANA F
A
971201
90-95(A)(1) SELL COCAINE
F
A
971201
90-95(A)(1) SELL HEROIN
F
A
971201
90-95(A)(1) SELL LSD F
A
971201
90-95(A)(1) SELL SCH I CS
F
A
971201
90-95(A)(1) SELL SCH II CS
F
A
971201
90-95(A)(1) SELL SCH III CS
F
A
971201
90-95(A)(1) SELL SCH IV CS
F
A
971201
90-95(A)(1) SELL SCH V CS
F
A
971201
90-95(A)(1) SELL SCH VI CS
F
A
971201
90-95(A)(1) DELIVER MARIJUANA
F
A
971201
90-95(A)(1) DELIVER COCAINE
F
A
971201
90-95(A)(1) DELIVER HEROIN F
A
971201
90-95(A)(1) DELIVER LSD
F
A
971201
90-95(A)(1) DELIVER SCH I CS F
A
971201
90-95(A)(1) DELIVER SCH II CS
F
A
971201
90-95(A)(1) DELIVER SCH III CS
F
A
971201
90-95(A)(1) DELIVER SCH IV CS
F
A
971201
90-95(A)(1) DELIVER SCH V CS
F
A
971201
90-95(A)(1) DELIVER SCH VI CS
F
A
971201
90-95(D)(4) POSS MARIJ >1/2 TO 1 1/2 OZ
M
A
930208
310

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

3475
3476
3480
3481
3482
3483
3484
3485
3486
3490
3491
3492
3493
3494
3495
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530

90-95(E)(5) SELL OR DELIV CS MINOR >13-<16
F
A
990127
90-95(E)(5) SELL OR DELIV CS MINOR <=13 F
A
990127
90-98 CONSPIRE SELL MARIJ F
A
971201
90-98 CONSPIRE SELL COCAINE
F
A
971201
90-98 CONSPIRE SELL HEROIN F
A
971201
90-98 CONSPIRE SELL LSD
F
A
971201
90-98 CONSPIRE SELL SCH I CS F
A
971201
90-98 CONSPIRE SELL SCH VI CS
F
A
971201
14-401.15
CONTAMINATE FOOD/DRINK W/CS
F
A
971201
90-98 CONSPIRE DELIVER MARIJ
F
A
971201
90-98 CONSPIRE DELIVER COCAINE F
A
971201
90-98 CONSPIRE DELIVER HEROIN
F
A
971201
90-98 CONSPIRE DELIVER LSD F
A
971201
90-98 CONSPIRE DELIVER SCH I CS F
A
971201
90-98 CONSPIRE DELIVER SCH VI CS F
A
971201
90-113.22
POSS DRUG PARAPHERNALIA M
O
910408
90-95(A)(1) MANUFACTURE SCH I CS
F
A
90-95(A)(1) MANUFACTURE SCH II CS
F
A
90-95(A)(1) MANUFACTURE SCH III CS
F
A
90-95(A)(1) MANUFACTURE SCH IV CS
F
A
90-95(A)(1) MANUFACTURE SCH V CS
F
A
90-95(A)(1) MANUFACTURE SCH VI CS
F
A
90-95(A)(2) CREATE COUNTERFEIT CS
F
A
90-95(A)(1) SELL OR DELIVER SCH I CS
F
O
971201
90-95(A)(1) SELL OR DELIVER SCH II CS
F
O
971201
90-95(A)(1) SELL OR DELIVER SCH III CS
F
O
971201
90-95(A)(1) SELL OR DELIVER SCH IV CS F
O
971201
90-95(A)(1) SELL OR DELIVER SCH V CS
F
O
971201
90-95(A)(1) SELL OR DELIVER SCH VI CS F
O
971201
90-95(A)(2) SELL OR DELIVER COUNTERFEIT CS F
A
90-95(A)(1) PWIMSD SCH I CS F
A
90-95(A)(1) PWIMSD SCH II CS F
A
90-95(A)(1) PWIMSD SCH III CS
F
A
90-95(A)(1) PWIMSD SCH IV CS
F
A
90-95(A)(1) PWIMSD SCH V CS F
A
90-95(A)(1) PWIMSD SCH VI CS
F
A
90-95(A)(2) PWICSD COUNTERFEIT CS
F
A
90-95(A)(3) POSSESS SCH I CS F
A
90-95(A)(3) POSSESS SCH II CS F
A
90-95(A)(3) POSSESS SCH III CS
F
A
90-95(A)(3) POSSESS SCH IV CS
F
A
90-95(A)(3) POSSESS SCH V CS F
A
90-95(D)(4) FELONY POSSESSION SCH VI CS
F
A
90-95(H)(1) TRAFFICKING IN MARIJUANA F
A
90-95(H)(2) TRAFFICKING IN METHAQUALONE F
A
90-95(H)(3) TRAFFICKING IN COCAINE
F
A
311

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

3531 90-95(H)(4)
3532 90-95(I)
3533 90-95(I)
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575

"TRAFFICKING, OPIUM OR HEROIN" F
CONSPIRE TO TRAFFIC IN MARIJ
F
CONSPIRE TRAFFIC METHAQUALONE

90-95(I)
CONSPIRE TO TRAFFIC IN COCAINE
90-95(I)
CONSPIRE TRAFFIC OPIUM/HEROIN
90-95(D)(2) SIMPLE POSSESSION SCH II CS M
90-95(D)(2) SIMPLE POSSESSION SCH III CS M
90-95(D)(2) SIMPLE POSSESSION SCH IV CS M
90-95(D)(3) SIMPLE POSSESSION SCH V CS M
90-95(D)(4) MISD POSSESSION SCH VI CS M
90-95(A)(1) MANUFACTURE MARIJUANA F
90-95(A)(1) SELL OR DELIVER MARIJUANA F
90-95(A)
SELL MARIJUANA F
O
90-95(A)
PWISD MARIJUANA
F
A
90-95(A)(1) PWIMSD MARIJUANA
F
A
90-95(A)
P/W/I/M MARIJUANA
F
O
90-95(A)
P/W/I/S MARIJUANA
F
O
90-95(A)
P/W/I/D MARIJUANA
F
O
90-95(D)(4) FELONY POSSESSION MARIJUANA
90-95(D)(4) POSSESS MARIJUANA UP TO 1/2 OZ
90-98 CONSPIRE SELL OR DELIVER MARIJ F
90-95(A)(1) MANUFACTURE COCAINE
F
90-95(A)(1) SELL OR DELIVER COCAINE
F
90-95(A)
SELL COCAINE
F
O
90-95(A)
PWISD COCAINE F
A
90-95(A)(1) PWIMSD COCAINE F
A
90-95(A)
P/W/I/M COCAINE F
O
90-95(A)
P/W/I/S COCAINE F
O
90-95(A)
P/W/I/D COCAINE F
O
90-95(D)(2) FELONY POSSESSION OF COCAINE
90-95(A)
SIMPLE POSS COCAINE M
O
90-98 CONSP SELL OR DELIVER COCAINE F
90-95(A)(1) SELL OR DELIVER HEROIN
F
90-95(A)
SELL HEROIN
F
O
90-95(A)(1) PWIMSD HEROIN F
A
90-95(A)
P/W/I/S HEROIN
F
O
90-95(A)
P/W/I/D HEROIN
F
O
90-95(D)(1) POSSESS HEROIN F
A
90-98 CONSP SELL OR DELIVER HEROIN
F
90-95(A)
SELL/DELIV HASHISH
F
O
90-95(A)
SELL HASHISH
F
O
90-95(A)
P/W/I/S/D HASHISH
F
O
90-95(A)
P/W/I/S HASHISH F
O
90-95(A)
P/W/I/D HASHISH F
O
90-95(A)
POSSESS HASHISH F
O
312

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

A
A
F

A

F
A
F
A
A
A
A
A
A
A
O
971201
971201
950206
950206
950206
950206
A
A

F
M
O
A
O
950206
950206

971201
971201

950206
950206
950206
F
A
920224
O
971201
O
971201
950206
950206
950206
O
950206
950206
950206
950206
950206
950206

971201

3576
3577
3578
3579
3580
3581
3582
3583

90-95(A)
SIMPLE POS HASHISH
M
O
90-98 CONSP SELL OR DELIV SCH VI CS
F
O
90-95(A)(1) SELL OR DELIVER LSD F
O
90-95(D)(1) POSSESSION OF LSD
F
A
90-98 CONSPIRE SELL OR DELIVER LSD
F
O
90-95(A)
POSS METHAQUALONE/QUAALUDE F
90-95(A)
POSS METHAQUALONE/QUAALUDE M
90-95(A)
P/W/S/D METHAQUALONE/QUAALUDE

3584 90-98 CONSP SELL OR DELIVER SCH I CS
F
3585 90-95(D)
"SIM POS-CS-SCH II, III, IV"
M
3586 90-113.10
INHALE TOXIC VAPORS M
A
3587 90-113.12
SELL TOXIC VAPORS SUBSTANCE

950206
971201
971201
O
O
F

O
O
M

971201
950206
941003
A
971201
950206

A

3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3604
3605
3607
3608
3609
3620
3622
3624

90-113.11
POSS TOXIC VAPORS SUBSTANCE
M
A
90-95 S/D OF CS F
O
941003
90-95(D)
SIM POS - CS - SCH IV
M
O
941003
90-95(A)
MFG/CREATE CS F
O
941003
90-95(D)
CS - SCH VI M
O
941003
90-95(A)
CS - P/W/I/S/M ETC F
O
941003
90-95(E)(5) SELL OR DELIV CS TO CHILD < 16
F
O
990127
90-95(D)
"SIM POS-CS-SCH II, III, IV"
F
O
950206
90-95(D)
SIM POS-CS-SCH VI
F
O
950206
14-258.1(A) PROVIDING DRUGS TO INMATE
F
A
910408
90-95(A)
GROWING MARIJUANA F
O
950206
DANGEROUS DRUGS - FREE TEXT
A
14-179
MISDEMEANOR INCEST M
A
14-190.9
INDECENT EXPOSURE
M
A
14-178
FELONY INCEST F
A
14-190.1
MISD DISSEMINATE OBSCENITY
M O 940606 941003
14-190.1
FELONY DISSEMINATE OBSCENITY F
A
940606
14-177
CRIME AGAINST NATURE
F
A
14-190.9
USE OF PREMISES INDEC EXPOSURE M
A
14-184
FORNICATION AND ADULTERY
M
A

3626
3628

14-186
14-177

OCCUPY ROOM IMMORAL PURPOSES M
ATT CRIME AGAINST NATURE (F)
F

3630
3699
3705
3706
3799
3802
3804
3805

14-177

M

ATT CRIME AGAINST NATURE (M)
SEX OFFENSE - FREE TEXT
A
14-190.1
DISSEMINATION OF OBSCENITY
14-190.1
DISSEMINATION OF OBSCENITY
OBSCENITY - FREE TEXT
O
14-318.4
FELONY CHILD ABUSE F
O
14-183
BIGAMY
F
A
14-316.1
CONTRIBUTING DEL OF JUVENILE
313

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

O
A

941003

A

941003

M
O
940606
F
O
940606
940606
970108
M

A

3807
3808
3809
3810
3812
3818
3820

14-313(C)
PURCHASE CIGARETTES < 18 M
A
971201
14-313
SALE OF CIGARETTES TO MINORS
M
A
940606
14-313(B)
NO SIGN FOR TOBACCO SALES < 18 I
A
971201
49-2 IV-D BASTARDY M
O
910408
910429
49-2 NON IV-D BASTARDY
M
O
910408
910429
14-322.1
ABANDON/NON-SUPPORT OF CHILD M
O
14-325
INADEQ SUPPORT OF FAMILY(R/83) M
O

3822
3824
3826
3827
3828

115C-378
SCHOOL ATTENDANCE LAW VIOL
49-2 ILLEGITIMATE CHILD/NON-SUPPORT M
14-322
ABANDON DEPENDENT SPOUSE
14-322(C)
NON IV-D NONSUPPORT SPOUSE
14-322
NON-SUPPORT DEP SPOUSE/CHILD

3829
3830
3834
3835
3836
3837
3840
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3872
3899
3901
3903
3904

14-322(D)
NON-SUPPORT OF CHILD M
O
14-46 CONCEALING BIRTH OF A CHILD
F
A
14-318.2
MISDEMEANOR CHILD ABUSE M
A
14-318.4(A) FELONY CHILD ABUSE-SERIOUS INJ F
14-318.4(A1) FELONY CHILD ABUSE-PROSTITUTN F
14-318.4(A2) FELONY CHILD ABUSE -SEXUAL ACT F
14-320.1
FELONY CUSTODY ORDER VIOLATIONF
14-322.1
NON IV-D ABANDON FOR 6 MONTHS F
14-322.1
IV-D ABANDONMENT FOR 6 MONTHS F
14-322.1
IV-D ABANDON/NON-SUPP OF CHILD M
49-2 IV-D NONSUPPORT ILLEGIT CHILD
M
A
14-322
IV-D NON-SUPP DEP SPOUSE/CHILD M
14-322(D)
IV-D NONSUPPORT CHILD
M
A
14-322(B)
IV-D NONSUPPORT SPOUSE
M
A
14-322.1
NON-IV-D ABAND/NON-SUPP CHILD M
49-2 NON IV-D NONSUPP ILLEGIT CHILD M
A
14-322
NON-IV-D NON SUPP SPOUSE/CHILD M
14-322(D)
NON IV-D NONSUPPORT OF CHILD
M
50B-4.1
VIOLATE DOM VIOL PROTCT ORDER M
FAMILY - FREE TEXT
A
14-291
SELLING LOTTERY TICKETS
M
A
14-301
OPERATE/POSSESS SLOT MACHINE M
14-302
OPER/POSSESS GAMBLING DEVICES M

3905
3906

14-304
14-305

910408
910408

3915 14-290
3930 14-291.1
3931 14-292
3932 14-293

M
O
M
M
M

MANUFACTURE/SELL SLOT MACHINEMA
SLOT MACHINE AGREEMENT M
A
OPERATING A LOTTERY M
A
POSSESSION OF LOTTERY TICKETS M
GAMBLING M
A
ALLOW GAMBLING IN PUBLIC HOUSE
314

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

A
O
A
O

941003

A
A
A
A
A
A
O

970109
970109
970109
920203
910408
910408
940606

O

940606

940606
O
940606
O
A
A

940606

A
A

910408

971201

A
M

A

3933

14-297

ALLOW GAMING TABLES

M

3999
GAMBLING - FREE TEXT
A
4010 14-204
AID AND ABET PROSTITUTION M
4012 14-204(7)
PROSTITUTION
M
A
4013 14-204(5)
SOLICIT FOR PROSTITUTION
M
4014 14-204.1
LOITERING FOR PROSTITUTION
4015
4016
4018
4099
4101
4102

A

910408

A
A
M

14-204(1)
MAINT PLACE FOR PROSTITUTION
M
14-204(4)
TRANSPORT FOR PROSTITUTION
M
14-3 SOLICIT CRIME AGAINST NATURE
M
A
COMMERCIAL SEX - FREE TEXT
A
18B-307(B) MANUFACTURE LIQUOR NO PERMIT M
18B-304
"LIQUOR, ILLEGAL SALE/POSSESS." M
940606
4103 18B-401(A) UNSEALED WINE/LIQ IN PASS AREA M
4104 LOCAL ORDINANC
POSSESS LIQUOR FOR SALE

A
A
A
A
O
A
M

A

18B-401(A) TRANS ALC/CONTAINER NOT MANU T
O
910408
4106 18B-406
ILLEGAL TRANSPORT ALCOHOL BEV M
A
4107 18B-301(F)(7)
POSS/CONS BEER/WINE UNAUT PREM M
4108 18B-301(F)(4)
POS/CON F-WN/LQ/MXBV UNATH PR M

A
A

4105

4109 18B-305
SELL/GIVE ALC TO INTOX PERSON
M
A
4110 LOCAL ORDINANC
ALLOW ILLEGAL CONSUMPTION ALC
M
A
4111 18B-302(B)(1)
PUR MTBV/U-WN BY 19/20
I
A
4112 18B-111
POSS/TRAN/SELL NON-TAX ALC BEV M
O
920609
4113 18B-302(B)(1)
ATT PUR MTBV/U-WN BY 19/20 I
A
4114 18B-302(B)(1)
POSS/MTBV/U-WN BY 19/20
I
A
4120 18B-302(C) AID UNDERAGE PURCHASE LIQUOR M
O
950206
4121 18B-302(C) AID UNDERAGE PURCHASE BEER
M
O
950206
4122 18B-302(C)(1)
AID UNDERAGE PUR ALC BY < 21M A
950206
4123 18B-302(C)(2)
AID UNDERAGE PUR ALC BY > 21M A
950206
4132 LOCAL ORDINANC
POSS/CONS BEER/WINE PUBLIC ST
M
A
4134 LOCAL ORDINANC
POSS/CON BEER/WINE UNAUTH PREMM
A
4136 18B-301(F)(2)
DISPLAY ALC ATHLETIC CONTEST
M
A
4138

LOCAL ORDINANC
POSSESSION ALCOHOLIC BEVERAGE M
A
4140 18B-302
PUR/POSS BEER/WINE UNDERAGE
M
O
920609
4142 18B-302
PUR/POSS ALCOHOL UNDERAGE
M
O
920609
315
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

4144

14-329

MANUFACTURE POISONOUS LIQUOR F

A

4146
4148
4149
4150
4152
4155

14-313
18B-102
18B-102.1
18B-302(A)
18B-302(A)
18B-304

SALE OF CIGARETTES TO MINORS
ABC LAW VIOLATION
M
A
SHIP ALCOHOL FROM OUT STATE
SELLING BEER/WINE TO MINOR
SELLING LIQUOR TO UNDERAGE
POSS/SELL ALC BEV NO PERMIT

M

O

940606

F
M
M
M

A
O
O
A

971201
920609
920609

O
M

940606
A

4156 18B-301(F) CON/OFFER WINE - PUB RD/BYWAY M
4157 18B-301(F)(1)
CON/OFFER ALC BEV PUBLIC ROAD
4158
4159
4160
4161
4162
4163
4164

18B-300(B) CON MTBV/U-WN PREM NO PERMIT M
A
18B-301(F) PUBLIC CONSUMPTION M
A
18B-1004
SELL/CONS BEER/WINE/ALC AFT HR M
O
18B-1004
SELL/CON ALC BEV AFTER HOURS
M
A
18B-302(A) SELL/GIVE MALT /WINE TO MINOR
M
O
18B-302(A) SELL/GIVE ALCOHOL BEV TO MINOR M
O
LOCAL ORDINANC
CONSUME BEER/WINE UNDERAGE
A
4166 18B-302(B)(1)
PUR/ATT MTBV/U-WN NOT 19/20
MA
4167
4168
4169
4170
4171

18B-302(B)(1)
18B-302(B)(2)
18B-302(B)(2)
18B-302(A)(1)
18B-302(A)(2)

4172
4173
4174
4175
4195
4199
4210
4299
4401
4402
4403
4404

18B-300(B) ALLOW CON MTBV/U-WN NO PERMIT
18B-111
POSS/SELL NONTAXPAID ALC BEV
18B-111
TRANSPORT NONTAXPAID ALC BEV
18B-401(A) DRINK BEER/WINE WHILE DRIVING
14-444
INTOXICATED AND DISRUPTIVE
LIQUOR - FREE TEXT
A
14-444
INTOXICATED AND DISRUPTIVE
PUBLIC INTOXICATION - FREE TEX
20-141(H)
IMPEDE TRAFFIC BY SLOW SPEED
20-149
OVERTAKEN VEH INC SPEED(I) I
20-149(A)
IMPROPER PASSING ON RIGHT I
20-125
HORN AND WARNING DEVICE VIO

4405

20-117

4406 20-157(C)
4407 20-126
4408 20-157(B)

POSS MTBV/U-WN NOT 19/20
M
PUR/ATT F-WN/LQ/MXBV < 21 M
POSS F-WN/LQ/MXBV < 21
M
SELL/GIVE MTBV/U-WN TO < 21M
SELL/GIVE F-WN/LQ/MXBV TO < 21

920609
920609
920609
M
60992

A
A
A
A
MA

60992
920609
920609
60992
60992

M
M
M
M
M

A
A
A
A
A

60992
920609
920609
940606
940606

M
O
I
A
A
I

O

940606
941003

A

FLAG LIGHT END OF LOAD VIOL

I

A

OBSTRUCTING FIRE OPERATIONS
MIRROR VIOLATION
I
A
FOLLOWING A FIRE TRUCK
I

I

A

316
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

A

A

4409
4410
4411

20-175
20-140.3
20-160

SOLICITING FROM HIGHWAY I
INTERSTATE HIGHWAY VIOLATION
SAFETY ZONE/SIDEWALKS VIOL

A
I
I

A
A

4412 20-146(A)
4413 20-123
4414 20-137.1
4415 20-116

DRIVE LEFT OF CENTER I
A
IMPROPER TOWING
I
A
NO CHILD RESTRAINT SYSTEM I
A
OVER LOAD SIZE/LENGTH/ VEHICLE I

A

4416
4417
4418
4419
4420
4421
4422
4423
4424

20-154
20-146
20-123.2
20-141(C)
20-115.1
20-122
20-129(C)
20-123.1
20-129(D)

IMPROPER SIGNAL
I
O
CROSSING MEDIAN
I
O
IMPROPER EQUIP - SPEEDOMETER
I
SPEED LESS THAN POSTED MINIMUM I
TWIN/SEMI TRAILER VIOL(I)
I
A
TIRE RESTRICTIONS EQUIP VIOL
I
MOTORCYCLE FAIL BURN HEADLAMP
IMPROPER STEERING MECHANISM
I
MOTORCYCLE FAIL BURN TAILLIGHT

910408
910408
A
910408
A
920203
A
920203
I
A
A
920203
I
A

4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451

20-116(G)
20-125.1
20-129(D)
20-127(B)
20-129(G)
20-140.4(A)
20-156(A)
20-155(A)
20-146(C)
20-155(B)
20-129.1
20-130.3
20-146(D)
20-158
20-154(A)
20-183.2
20-157(D)
20-183.3
20-183.6
20-140.2
20-129(A)
20-129(A)(4)
20-153
20-158
20-141(B)
20-141(B)(G)
20-153(C)

IMPROPER LOADING/COVERING VEH I
A
DIRECTIONAL SIGNALS EQUIP VIOL I
A
920203
REAR LAMPS VIOLATION
I
A
WINDSHIELD WIPER EQUIP VIOLI
O
920203951020
BRAKE/STOP LIGHT EQUIP VIOL
I
A
920203
MOTORCYCLE/MOPED HELMET VIOL I
A
FAIL TO YIELD FROM PRIVATE DRV I
A
FAILURE TO YIELD
I
A
DRIVE WRONG WAY ON DUAL LANE I
A
FAIL TO YIELD LEFT TURN
I
A
ADDITIONAL LIGHTING EQUIP VIOL I
A920203
WHITE LIGHT REAR-DRIVE FORWARD
IA
920203
DESIGNATED LANE VIOLATION
I
A
920203
TRAFFIC CONTROL DEVICE VIOL
I
A
920203
IMPROPER BACKING
I
A
920203
INSPECTION VIOLATION I
A
DRIVE OVER FIRE HOSE OR EQUIP
I
A
920203
INSP STICKER NO INSPECTION I
A
920203
ALTERED INSPECTION STICKER
I
A
920203
OVERLOADED/OVERCROWDED VEHICLE
I
A
FAIL TO BURN HEADLAMPS
I
A
NO HEADLIGHTS ON WIPERS ON
I
A
910408
IMPROPER TURN I
A
RED LIGHT VIOLATION I
O
910408
EXCEEDING POSTED SPEED
I
A
SPEEDING I
A
IMPROPER USE OF TRAFFIC LANE
I
A
317

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466

20-158(B)(1)(3)
FAIL YLD STOPSIGN/FLSH RED LGT
20-158.1
FAILURE TO YIELD - YIELD SIGN
I
20-158(B)(1)(3)
FAIL STOP STOPSIGN/FLSH RED LT
20-158(B)(2) FAIL TO STOP-STEADY RED LIGHT
I
20-141
TOO FAST FOR CONDITIONS
I
O
20-127
OBSTRUCTED WINDSHIELD/WINDOWSIO
20-154
UNSAFE MOVEMENT
I
A
20-154
FAIL TO SIGNAL WHEN TURNING
I
20-141(M)
FAILURE TO REDUCE SPEED
I
A
20-142.5
STOP WHERE TRAFFIC OBSTRUCTED I
20-122.1
UNSAFE TIRES
I
A
20-150(A)
UNSAFE PASSING ONCOMING TRAF I
20-150(B)
UNSAFE PASSING CREST OR CURVE I
20-150(C)
UNSAFE PASSING RR OR INTERSECT I
20-165.1
DRIVE WRONG WAY-ONE WAY ST/RD

4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480

20-141(A)
20-141.1
20-150(E)
20-135.2A
20-135.2(A)
20-135.2A
20-135.2(A)
20-141(E1)
20-141(J2)
20-142
20-181
20-141(A)
20-142.1
20-173

EXCEEDING SAFE SPEED I
A
SPEEDING IN SCHOOL ZONE
I
UNSAFE PASSING YELLOW LINE
FAIL TO WEAR SEAT BELT-DRIVER
FAIL TO WEAR SEAT BELT
I
FAIL TO SECURE PASSEN UNDER 16
FAIL WEAR SEAT BELT- PASSENGER
SPEED ON SCHOOL PROPERTY ORD
SPEED IN HIGHWAY WORK ZONE
FAIL TO STOP FOR RR WARNING
FAIL TO DIM HEADLAMPS
I
SPEED FASTER THAN REASONABLE
FAIL TO OBEY RR SIGNAL
I
FAIL PEDESTRIAN RIGHT OF WAY

A
I
I
O
I
I
I
I
I
A
I
A
I

4481

20-134

NO LIGHTS ON PARKED VEHICLE

I

4482
4483
4484
4485
4486
4488
4489
4490
4492
4494

20-152(A)
20-162.1
20-150
20-135.2B
20-128
20-124
20-183.8
20-127(A)
20-129(B)
20-163

FOLLOWING TOO CLOSELY
I
A
ILLEGAL PARKING
I
A
IMPROPER PASSING
I
A
TRANSPORT CHILD OPEN CARGO BED
IMPROPER MUFFLER
I
A
IMPROPER BRAKES
I
A
FICT/OTH IMPROPER INSPECTION
I
WINDSHIELD WIPER EQUIP VIOL
I
DRIVE WITHOUT TWO HEADLAMPS I
LEAVE VEH UNATTENDED/UNSECURE

IA

950818

A
A
A
I

920817
951020

4495

20-162

PARK FIRE HYD/STATION/PRIV DR

A

318
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

I

IA
A
I
A

910408
910408
A
910408
910408
951020

O
910408
910408
A
951201
A
A
A
I

910408
910408
910408
A

A
A

910408
910408
910408

A
A
A
A
O

910408
971201
921001
941003

O
910408
941003
A
A

A

4496
4497
4498
4499
4531
4532
4535
4801
4818
4820
4822
4899
4904
4905
4910
4912
4914
4916
4918
4920

20-162.1
OVERTIME PARKING
I
A
20-37.6(E)
HANDICAPPED PARKING VIOLATION I
20-127(C)(D) DARKENED WINDSHIELD/WINDOWS I
20
INFRACTION - FREE TEXT
I
A
20-11(L)
LIC/PERMIT SEAT BELT VIOL <18
I
20-11(L)
LIC/PERMIT SEATING VIOL <18 I
A
20-11(L)
LIC/PRMT VIOL OTH RSTRCTN <18
I
14-223
RESIST/OBSTRUCT PUBLIC OFFICER M
14-288.5
FAILURE TO DISPERSE M
O
20-114.1
FAIL TO OBEY TRAFFIC OFFICER
T
14-277
IMPERSONATION- PEACE OFFICERS M
OBSTRUCT POLICE - FREE TEXT
O
14-267
HARBORING FUGITIVE M
A
14-259
FELONY HARBORING ESCAPEE F
A
14-255
ESCAPE BY HIRED PRISONER M
A
14-256
ESCAPE FROM LOCAL JAIL
M
O
148-45(B)
ESCAPE FROM STATE PRISON (F)
F
148-45(A)
ESCAPE FROM STATE PRISON (M)
M
15A-722
FUGITIVE F
O
920203
14-266
PERSUADING INMATES TO ESCAPE M

4922

14-256

4924
4926
4999
5000
5001
5003
5004
5006
5008
5010
5020
5022
5023
5024
5025
5026
5028
5029
5030
5032
5034
5036
5038

14-256
FELONY ESCAPE LOCAL JAIL F
A
14-256.1
ESCAPE PRIVATE CORRECTION FAC F
A
990127
FLIGHT/ESCAPE - FREE TEXT
A
75D-7 RICO PERJURY
F
A
940606
15A-727
GOVERNOR S WARRANT F
O
920203
14-209
PERJURY
F
A
14-210
SUBORNATION OF PERJURY
F
A
14-218
OFFERING BRIBES F
A
940606
14-217
RECEIVING BRIBES
F
A
940606
COMMON LAW
OBSTRUCTING JUSTICE M
A
930426
15A-543(B) FAILURE TO APPEAR ON FELONY
F
A
15A-543(C) FAILURE TO APPEAR ON MISD M
A
14-225.2
HARASSMENT OF JUROR F
A
910408
14-226
INTIMIDATION OF A WITNESS M
O
950206
14-226
INTIMIDATING WITNESS F
A
950206
5A-11 VIOLATION OF COURT ORDER M
A
5A-11 CRIMINAL CONTEMPT M
A
5A-11(A)(9A)
CONTEMPT BY PROBATIONER M
A
940509
15A-1345
MISDEMEANOR PROBATION VIOL
M
A
15A-1345
FELONY PROBATION VIOLATION
F
A
5A-15 SHOW CAUSE
M
A
5A-11 BILL OF PARTICULARS M
O
940606
15A-1345
MISD PROB VIOL OUT OF COUNTY
M
A

MISDEMEANOR ESCAPE LOCAL JAIL M

319
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

A
O

951020

A
971201
971201
A
971201
O
940606
940606
O
940606
O
940606
940606

940606
A
A
A
A

5040
5042
5044
5046
5048
5099
5102
5103
5199
5201

15A-1345
FEL PROB VIOL OUT OF COUNTY
F
A
15A-1347
PROBATION REVOCATION APPEAL
M
A
15A-1347
PROBATION REVOCATION APPEAL
M
O
940606
15A-951
MOTIONS M
A
15A-1344.1(D)
MOTION TO WITHHOLD WAGESM
O
940606
PUBLIC ORDER-FREE TEXT
A
14-218
BRIBERY - OFFERING
F
O
940606
14-217
BRIBERY - RECEIVING F
O
940606
BRIBERY - FREE TEXT
O
940606
14-316
PERMIT CHILDREN USE FIREARMS
M
A

5202
5203
5204
5205
5206
5208
5209
5210
5213
5220
5221
5222
5223
5224
5225
5226

14-269(A)
CARRYING CONCEALED WEAPON
M
A
14-269.2
POSS WEAPON ON SCHOOL GROUNDSM
O
931201
14-269.7(A) POSSESS HANDGUN BY MINOR M
A
930927
14-315(A)
SELL/GIVE WEAPON TO MINOR (M) M
A
961202
14-315(A1) SELL/GIVE HANDGUN TO MINOR
F
A
970101
14-49 MALICIOUS USE OF EXPLOSIVE
F
A
14-49(B1)
USE EXPLOSIVE DEVICE CHURCH
F
A
961202
14-269.8
PURCH FIREARM VIOL DOM ORDER F
A
951020
LOCAL ORDINANC
SHOOTING WITHIN CITY LIMITSMO 940606
14-34.1
DISCHARGE WEAPON OCCUPIED PROP
F
A
COMMON LAW
GO ARMED TO TERROR OF PEOPLE
M
A
14-409.9
POSSESSION OF MACHINE GUN (M) M
O
951201
14-409
POSSESSION OF MACHINE GUN (F)
F
A
941003
14-415.1
POSSESSION OF FIREARM BY FELON F
A
14-258.2
POSSESS WEAPON BY PRISONER (M) M
A
14-402
SELL/PURCHASE WEAPON NO PERMIT
M
A

5227 14-258.2
5228 14-280
5229
5230
5232
5234
5235
5240
5242
5244
5246
5299
5301
5302
5303
5304
5308

POSSESS WEAPON BY PRISONER (F) F
SHOOTING/THROWING AT TRAIN (M) M

A
A

941003

14-280
SHOOTING/THROWING AT TRAIN (F) F
A
941003
14-288.7
DEADLY WEAPON OFF PREMISES
M
A
14-288.8
POSSESS WEAPON MASS DESTRUCT F
A
14-269.2
WEAPONS ON EDUC PROP/AID (F)
F
A
931201
14-269.2
WEAPONS ON EDUC PROP/AID (M)
M
A
931201
14-269(A1) CARRYING CONCEALED GUN(M)
M
A
951201
14-269(A1) CARRYING CONCEALED GUN(F)
F
A
951201
14-415.21(A) CONCEAL HANDGUN PERMIT VIOL(I) I
A
951201
14-415.21(B) CONCEAL HANDGUN PERMIT VIOL(M)M
A
951201
WEAPON OFFENSE - FREE TEXT
A
14-288.2
FELONY INCITING TO RIOT
F
A
941003
14-288.2
RIOT - INCITING M
O
941003
14-288.6(A) TRESPASS DURING EMERGENCY (M) M
A
14-288.6(B) TRESPASS DURING EMERGENCY (F) F
A
14-286
FALSE FIRE ALARM
M
A
320

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

5309 14-196(A)(3) HARASSING PHONE CALL
5310 14-223
RESISTING PUBLIC OFFICER
5312
5314
5315
5316
5320
5321
5322
5325
5328
5329
5330
5332
5333
5334
5335
5336

A
A

940606

14-288.5
FAIL TO DISPERSE ON COMMAND
M
A
940606
14-277
IMPERSONATE LAW ENFORCEMNT (M)
MA 940606
14-277(A)(4) OPERATE VEH WITH BLUE LIGHT
F
A
961202
14-277(B)(5) BLUE LIGHT CAUSE STOP/YIELD
F
A
961202
LOCAL ORDINANC
BARKING DOG
M
O
941003
67-12 DOG RUN AT LARGE AT NIGHT M
A
941003
14-196
INDECENT LANGUAGE ON TELEPHONEMO
911003
LOCAL ORDINANC
BLOCKING FIRE EXIT
T
O
941003
14-277.1
COMMUNICATING THREATS
M
A
14-277.3
MISDEMEANOR STALKING
M
A
921005
14-288.4
DISORDERLY CONDUCT M
A
14-275.1
DISORDERLY CONDUCT AT TERMINAL
M
A
14-401.14(A) ETHNIC INTIMIDATION M
A
930927
14-401.14(B) TEACHING ETHNIC INTIMIDATION
M
A
930927
14-277.4
OBSTRUCT HEALTH CARE FACILITY M
A
930927
20-157
OBSTRUCTING FIRE OPERATIONS
T
O

5338 14-196(A)(2)
5339 14-277.3
5341 136-26
5342 136-91
5344
5345
5346
5347
5348
5350

M
M

THREATENING PHONE CALL
M
FELONY STALKING
F
A
DRIVE ON CLOSED/UNOPENED HWY
PUT INJURIOUS OBJECT IN ROAD

136-72
EXCEED BRIDGE LOAD LIMIT M
14-69.1
FALSE BOMB REPORT
F
A
14-69.1
FALSE BOMB REPORT
M
O
14-69.2
HOAX BY FALSE BOMB F
A
14-35 HAZING
M
A
14-188
KEEPING A DISORDERLY HOUSE

A
921027
M
A
M
A

920203

A
971201
971201
971201
M

A

5352 14-401.8
5354 14-416

REFUSE RELEASE LINE EMERGENCY M
HANDLING DANGEROUS REPTILES
M

A
A

5356

PROFANE LANGUAGE ON HIGHWAY M

A

5358 62A-12
5360 14-286.1
5362 14-225

MISUSE OF 911 SYSTEM M
A
FALSE AMBULANCE REQUEST M
FALSE REPORT TO POLICE STATION

A
M

A

5364 14-288.2
5370 14-399(E)
5371 14-399(E)
5372 14-399(C)
5373 14-399(D)

PUBLIC DISTURBANCE M
A
COMMERCIAL LITTERING
F
LITTERING HAZARDOUS WASTE
LITTERING NOT > 15 LBS M
A
LITTERING 15 - 500 LBS M
A

A
911003
F
A
941003
941003
941003

14-197

321
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

5374
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410

14-399(E)
LITTERING > 500 LBS
F
A
PUBLIC PEACE - FREE TEXT
A
20-30(1)
POSS/DISP ALT/FICT/REVD DR LIC
20-30(1)
KNOW FICT/CANC/REV/SUSP LIC
20-141(J)
SPEED TO ELUDE ARREST
T
20-139(A)
DUI - DRUGS (REPEALED /83) T
20-138(A)
DUI - ALCHOLIC BEVERAGE (R/83)
20-138.1
DRIVING WHILE IMPAIRED
T
20-141.4(A1) FELONY DEATH BY VEHICLE F
20-57(C)
NO REGISTRATION CARD
T
20-174.1
IMPEDE TRAFFIC SIT/STAND/LIE
20-30(2)
ALLOW USE OF LICENSE OR PERMIT
20-32 ALLOW UNLICENSE MINOR TO DRIVE

941003
T
T
O
O
T
A
A
A
T
T
T

A
O
O

910408
941003
971201
910408
900628

A
A
A

5411 20-30(3)
5412 20-146
5413 20-140(C)
5414 20-137.1
5415 20-116

DISPLAY/USE ANOTHER LICENSE
DRIVE LEFT OF CENTER T
O
RECKLESS DRIVING AFT ALC(R/83)
NO CHILD RESTRAINT SYSTEM T
OVER LOAD SIZE/LENGTH/ VEHICLE

T

O

910408

T
O
T

O

900628

5416

20-119

DOT SPECIAL PERMIT VIOLATION

T

A

5417

20-106

POSSESS STOLEN AUTOMOBILE

F

A

5418 20-28(A)
DWLR
T
A
5419 20-141(C)
SPEED LESS THAN POSTED MINIMUM T
5420 20-59 FAIL TO SURRENDER TITLE
T
A
5421 20-7.1 FAIL TO NOTIFY DMV ADDR CHANGE
T

O

O
A

5422 20-129(C)
5423 20-12.1
5424 20-129(D)

MOTORCYCLE FAIL BURN HEADLAMPS
DUI-DRIVING INSTRUCTOR (R/83)
T
MOTORCYCLE FAIL BURN TAILLIGHT

T
O
T

5425 20-116(G)
5426 20-157(A)
5427 20-129(D)
5428 20-217

IMPROPER LOADING OF VEHICLE
FAIL TO HEED LIGHT OR SIREN T
DRIVE W/O REAR LAMPS T
O
FAIL TO STOP FOR STOPPED BUS

T
A

O

T

A

5429 20-138
5430 20-140.4(2)
5431 20-138(B)
5432 20-155
5433 20-146

DUI - FOURTH OFFENSE (R/83) T
O
MOTORCYCLE FAIL TO WEAR HELMET
DRIVE W/.1 OR MORE BL ALC(R83)
T
FAILURE TO YIELD
T
O
DRIVE WRONG WAY ON DUAL LANE T

O

5434 20-30(5)
5435 20-166(B)

FICTITIOUS DRIVERS LICENSE T
HIT - RUN PROPERTY FAIL INFO

O

322
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

A
T

T
O

O
O

O
900628

910408

5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453

20-166(B)
20-166(B)
20-166(A)
20-166(C)
20-183.2
20-7(A)
20-141.3(A)
20-141.4(A2)
20-140.2
20-129(A)
20-140(B)
20-153
20-158
20-111(2)
20-141(J1)
20-153(C)
20-141.3(B)
20-138

5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466

20-158
STOP SIGN VIOLATION T
O
20-7(E)
FAIL COMPLY LIC RESTRICTIONS
T
20-141
TOO FAST FOR CONDITIONS
T
O
20-127
OBSTRUCTED WINDSHIELD
T
O
20-154
UNSAFE MOVEMENT
T
O
20-138.1
DWI 2ND OFFENSE
T
O
20-141(M)
FAILURE TO REDUCE SPEED
T
O
20-111(2)
EXPIRED REGISTRATION CARD/TAG T
20-122.1
UNSAFE TIRES
T
O
LOCAL ORDINANC
SPINNING TIRES T
O
20-140(A)
RECKLESS DRVG-WANTON DISREGARD
20-111(2)
FICT/CNCL/REV/ALT REG CARD/TAG T
20-165.1
DRIVE WRONG WAY-ONE WAY ST/RD

5467
5468
5469
5470
5471
5472
5473
5474
5475
5476

20-141(A)
EXCEEDING SAFE SPEED T
O
20-29 FICTITIOUS INFO TO OFFICER T
A
20-7(F)
EXPIRED OPERATORS LICENSE T
20-34 ALLOW UNLICENSED TO DRIVE
T
20-138.1
AID AND ABET IMPAIRED DRIVING
20-138
DUI - SECOND OFFENSE (R/83) T
20-138
DUI - THIRD OFFENSE (R/83)
T
20-28(B)
DWLR PERMANENT
T
O
20-16.1(B)
FAIL COMPLY RESTRICTED DRIVING
20-142
FAIL TO STOP FOR RR WARNING

5477 20-181
5478 20-141(A)

HIT AND RUN - FAIL INFO
T
O
HIT-RUN PERSON-FAIL ASSIST (M)
T
HIT-RUN PERSON FAIL STOP (F)
T
HIT-RUN -- UNATTENDED VEHICLE T
INSPECTION VIOLATION T
O
NO OPERATORS LICENSE
T
A
PREARRANGED SPEED COMPETITION T
MISDEMEANOR DEATH BY VEHICLE T
OVERLOADED/OVERCROWDED VEHICLE
FAIL TO BURN HEADLAMPS
T
O
RECKLESS DRIVING TO ENDANGER T
IMPROPER TURN T
O
RED LIGHT VIOLATION T
O
ALLOW FICTITIOUS REG PLATE
T
SPEEDING T
A
IMPROPER USE OF TRAFFIC LANE
T
SPEED COMPETITION
T
A
ALLOW INTOX PERSON DRIVE(R/83) T

FAIL TO DIM HEADLAMPS
T
SPEED FASTER THAN REASONABLE
323

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

A
A
T
O
O

O
O
O

910408
910408
910408
910408

A
A
TO

900628

A
O

910408

O
O
A

910408
A
T
O
T

910408
A
961202
O

A

941003

T
T

950206
A
O

O
T

O

5479 20-166.1
5480 20-173

FAIL TO REPORT ACCIDENT
T
FAIL PEDESTRIAN RIGHT OF WAY

A
T

O

5481

20-134

NO LIGHTS ON PARKED VEHICLE

T

O

5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495

20-152
20-162.1
20-150
20-111(1)
20-128
20-7(A1)
20-124
20-309
20-166(B)
20-111(1)
20-129(B)
20-7(N)
20-313(A)
20-162

FOLLOWING TOO CLOSELY
T
ILLEGAL PARKING
T
O
IMPROPER PASSING
T
O
DR/ALLOW REG PLATE NOT DISPLAY
IMPROPER MUFFLER
T
O
NO MOTORCYCLE ENDORSEMENT
IMPROPER BRAKES
T
O
NO LIABILITY INSURANCE
T
LEAVE SCENE OF ACCIDENT T
DR/ALLOW VEH NOT REG/TITLED
DRIVE W/O 2 HEADLAMPS
T
LICENSE NOT IN POSSESSION T
OPERATE VEH NO INS
T
A
PARK FIRE HYD/STATION/PRIV DR

O

5496
5497
5498
5499
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522

20-162.1
OVERTIME PARKING
T
O
20-37.6(E)
HANDICAPPED PARKING VIOLATION T
O
20-141.3(B) WILLFUL SPEED COMPETITION T
O
20
TRAFFIC OFFENSE - FREE TEXT
T
A
20-218
SPEEDING SCHOOL/ACTIVITY BUS
T
O
20-50 FAIL TO OBTAIN REG OR TITLE T
A
20-67 REG/TITLE ADDRESS CHANGE VIO
T
A
20-8 OPER MOPED LESS THAN 16 YOA
T
O
20-29 FAIL EXHIBIT/SURRENDER LICENSE T
A
20-21 USE FOREIGN LICENSE WHILE DWLR T
A
20-136.1
LOCATION OF TV IN VEHICLE T
A
20-73 FAIL TO APPLY FOR NEW TITLE
T
A
20-106.1
MOTOR VEHICLE RENTAL FRAUD
F
A
20-117
FAIL TO SEC RED FLAG ON LOAD
T
O
20-138.1(A) DWI - LEVEL 1
T
A
910408
20-138.1(A) DWI - LEVEL 2
T
A
910408
20-138.1(A) DWI - LEVEL 3
T
A
910408
20-138.1(A) DWI - LEVEL 4
T
A
910408
20-138.1(A) DWI - LEVEL 5
T
A
910408
20-138.1(A) DWI - LEVEL 5 - AID/ABET
T
O910408
20-138.1(A) DWI (.10) - LEVEL 1
T
O
910408
20-138.1(A) DWI (.10) - LEVEL 2
T
O
910408
20-138.1(A) DWI (.10) - LEVEL 3
T
O
910408
20-138.1(A) DWI (.10) - LEVEL 4
T
O
910408
20-138.1(A) DWI (.10) - LEVEL 5
T
O
910408
20-138.1(A) DWI (.10)-LEVEL 5 AID/ABET
T
O910408
324

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

T

A

T

A

A
O
T
O
A

A

T

O

910408

941003
941003
910429

920203

941003
910429
910429
910429
910429
910429
910429

5523 20-7 AID & ABET OPERATORS LIC VIOL
T
5524 20-7(L)
LEARNERS PERMIT VIOLATION
5525 20-37.8
SPECIAL ID FRAUD VIOLATION T
5526
5527
5528
5529
5530
5531
5532
5533
5535
5536
5537
5538
5539
5540
5541
5542
5544
5545
5546
5548
5549
5550
5551
5552
5555
5556
5558
5560
5561
5562
5563
5564
5565
5566
5567
5568
5570
5571
5572
5573
5574
5575

A
951201
TO951201
971201
A
951201

20-138.3
DWI - PROVISIONAL LICENSE T
O910408
941003
20-138.5
HABITUAL IMPAIRED DRIVING F
A
910408
20-183.6
REPRODUCTION OF INSP STICKER
T
A
920203
20-183.11
FAIL WEIGH/ENTER WEIGH STATION TO920203
951020
130-185
DOG VACCINATION
M
O
940606
20-28 AID AND ABET DWLR
T
A
941003
20-287
BUY/SELL VEHICLE NO LICENSE
T
A
940606
20-75 FAILURE TO DELIVER TITLE
T
A
940606
20-30(3)
DISPLAY ANOTHERS LIC AS OWN
T
A
951201
20-72(B)
FAIL SURR TITLE/REG CARD/TAG
T
A
951201
20-72(B)
DELIVER/ACCEPT OPEN TITLE T
A
951201
20-111(3)
GIVE/LEND/BORROW LIC PLATE
T
A
951201
20-149
OVERTAKEN VEH INC SPEED(M)
T
A
951201
90-109
TREATMENT W/O REQ LICENSE
MO
940606
86A-1 BARBERING W/O CERTIFICATE M
O
940606
86A-1 NO BARBERSHOP/SCHOOL PERMIT
M
O
940606
130A-25
PUBLIC HEALTH VIOLATION M
O
940606
130A-178
FAIL TO REPORT FOR TB TREATMNT M
O
941003
130-335
SEWAGE DISPOSAL VIOLATION
M
O
940606
14-284.2
DUMP TOXIC SUBSTANCES
F
O
940606
20-218(B)
SPEEDING IN SCHOOL BUS
T
A
941003
20-114.1
FAIL TO OBEY TRAFFIC OFFICER
T
A
940606
20-106.2(B) MV SUBLEASE VIOLATION (M) T
A
940606
20-106.2(B) MV SUBLEASE VIOLATION (F) F
A
940606
20-106
POSSESION STOLEN VEHICLE(MISD) TO910408
930208
20-111(2)
FICT/CNCL/REV REG CARD/TAG
T
A
961202
20-111(2)
ALTERED REG CARD/TAG
T
A
961202
20-28(A)
DWLR VIOL LIMITED DRIVE PRIV
T
A
910408
20-7(A)
DRIVE W/O LIC FOR VEH-NON COMM TO910408
941003
20-30(7)
SELL FALSE DRIVERS LIC/PERMIT
TO910408
941003
20-37.6(C3) SELL HANDICAPPED PLACARDS
T
A
930927
20-30(5)
OBTAIN DR LICENSE BY FRAUD
T
A
940606
20-313
PERMIT OPERATION VEH NO INS
T
A
910408
20-107
TAMPERING WITH VEHICLE PARTS T
A
940606
20-107(B)
TAMPERING WITH VEHICLE STEAL T
A
940606
20-107
TAMPERING WITH VEHICLE
TA
940606
20-138.3
DRIVE AFTER DRINKING PROV LIC
TO910408
951020
20-12.1
IMPAIRED SUPERV/INSTRUCTION
T
A
971201
18B-401(A) DRINK BEER/WINE WHILE DRIVING MO910408 940606
20-63(G)
COVERING/DISGUISING REG PLATE T
A
910408
20-79 IMPROPER USE DEALER PERMIT/TAG T
A
910408
20-102.1
FALSE REPORT OF THEFT OF MV
T
A
920203
325

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

5576
5577
5579
5580
5581
5582
5583
5584
5586
5588

20-130.1(E)
20-138.3
20-396
20-115.1
20-166(A)
20-166(B)
20-166(C)
20-166(C)(1)
20-37.7(E)
20-343

5590
5591
5592
5593
5594
5595
5596
5597
5599
5610
5615
5620
5622
5624

20-109(B)
ALTERING SERIAL NUMBERS F
A
910408
20-71(A)
ALTER TITLE
T
O
910408
941003
20-71(A)
ALTER TITLE
F
A
941003
20-30(7)
SELL FALSE DRIVERS LIC/PERMIT
F
A
941003
20-138.7
OPEN CONT AFTER CONS ALC 1ST
T
A
951020
20-138.7
OPEN CONT AFTR CONS ALC SUBOFN T
A
951020
20-127(C)
WINDOW TINTING VIOL T
A
951020
20-118.1
FAIL TO ENTER/WEIGH STATION
T
A
951020
TRAFFIC OFFENSE-FREE TEXT
A
20-138.2
DWI COMMERCIAL VEHICLE T
A
910408
20-138.2(A)(1)
COMMERCIAL DWI UNDER INFLUENCETA 910408
20-138.2(A)(2)
COMMERCIAL DWI >=.04 T
A
910408
20-138.2A
CONSUME ALCOHOL COMM VEH
T
A
990127
20-138.2B
CONSUME ALCH SCH BUS/CHILD VEHTA
990127 5630
20-7(L)
LEARNERS PERMIT VIOLATION >18 T
A
971201
20-11(L)
LIC/PRMIT TIME LIMIT VIOL <18
T
A
971201
20-11(L)
LIC/PERMIT NO SUPV DRIVER <18
T
A
971201
20-141.5(A) FLEE/ELUDE ARREST W/MV (M)
T
A
971201
20-141.5(B) FLEE/ELUDE ARREST W/MV (F) F
A
971201
20-141.5(B) ELUDE ARRST MV 2 AGRVTG FCTRS F
A
971201
20-141.5(B) ELUDE ARRST MV >=3 AGRV FCTRS F
A
971201
20-7(A)
NO DRIVERS LIC COMM VEHICLE
T
A
910408
20-28(D)
DRIVE CVEH CLIC DISQUALIFIED
T
A
950606
20-37.12
DRIVE CVEH W/C LIC SUS/REV/DQD TO910408
950606
20-37.12(A) COMM DL NOT IN POSSESSION T
A
951201
20-37.12(A) DR COM VEH W/O PROPER ENDORSE T
A
921005
20-101
FAIL TO MARK FOR HIRE VEHICLE
T
A
951201
20-290
FAIL DISPLAY/ADVT LIC OR LIST
T
A
951201
20-396(A)
RADAR DETECTOR COMM VEHICLE T
A
951201
20-396(A)
MOTOR CARRIER LOG BOOK VIOL
T
A
951201
20
COMMERCIAL LICENSE - FREE TEXT T
A
910408
14-227.1
SECRET LISTENING PRISONER/ATTY M
A
14-159.6(B) TRSPSS POSTED PROP PINE STRAW
M
A
971201
14-159.6
TRESPASS ON POSTED PROPERTY
M
A
941003
14-134
TRESPASS WITHOUT A LICENSE
M
O
910408

5633
5634
5640
5641
5642
5643
5655
5657
5660
5661
5662
5670
5674
5680
5682
5699
5704
5705
5706
5707

USE OF RED OR BLUE LIGHT
T
DRIVE AFTER CONSUMING < 21
RADAR DETECTOR COMM VEHICLE
TWIN/SEMI TRAILER VIOL(M) T
FEL HIT/RUN FAIL STOP PER INJ
HIT/RUN LEAVE SCENE PER INJURY
HIT/RUN FAIL STOP PROP DAMAGE
HIT/RUN LEAVE SCENE PROP DAM
NC ID CARD FRAUD
T
A
CHANGE OF MILEAGE VIOL
F

326
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

A
920203
T
A
951020
TO941003
951201
A
910408
F
A
T
A
T
A
T
A
930621
A
930621

5708
5709
5710
5711
5712

14-159.12
FIRST DEGREE TRESPASS
M
14-159.13
SECOND DEGREE TRESPASS
M
COMMON LAW
FORCIBLE TRESPASS
M
62-319
RIDE ON TRAIN UNLAWFULLY M
14-160
INJURY TO PERSONAL PROPERTY

A
A
A
A
M

O

5713

14-460

A

990127

RIDE ON TRAIN UNLAWFULLY M

5714
5716
5717
5720
5729
5730
5799
6030
6040
6041
6042
6044
6045
6046
6048
6099
6170

14-202
SECRET PEEPING M
A
14-134.3(A) DOMESTIC CRIM TRESPASS(M) M
A
14-134.3(B) DOM CR TRESPSS SAFE HOUSE WEAP F
A
990127
14-149
DESECRATING GRAVES F
A
940606
14-159.3
TRESPASS W/ALL TERRAIN VEH
M
A
971201
63-26.1
TRESPASS ON AIRPORT PROPERTY
M
A
INVADE PRIVACY - FREE TEXT
A
130A-185
DOG OR CAT VACCINATION
M
A
940606
90-109
TREATMENT W/O REQ LICENSE
M
A
940606
86A-1 BARBERING W/O CERTIFICATE M
A
940606
86A-1 NO BARBER SHOP/SCHOOL PERMIT M
A
940606
130A-25
PUBLIC HEALTH VIOLATION M
A
940606
130A-144(F) FAIL TO REPORT FOR TB TREATMNT M
A
940606
130A-335
SEWAGE DISPOSAL VIOLATION
M
A
940606
14-284.2
DUMP TOXIC SUBSTANCES
F
A
940606
HEALTH LAW - FREE TEXT
A
940606
105-113.110 CONTROLLED SUBSTANCE TAX CRIME
F
O
910408
951020
6175 105-113.111 TAX ON DRUGS
M
O
910408
941003
6180 105-236
FAIL TO FILE/PAY INCOME TAX
M
A
6185
6188
6190
6195
6199
6205
6206
6207
6208
6209
6210
6211
6212
6213
6214
6215
6216

105-236(9A) AID/ASSIST FRAUD TAX RETURN
F
105-236(7) ATTEMPT TO EVADE OR DEFEAT TAXF
105-236
FAIL TO FILE/PAY SALES TAX M
A
105-308
FAIL TO LIST PROPERTY FOR TAX
M
TAX REVENUE - FREE TEXT
A
14-361
INSTIGATE CRUELTY TO ANIMALS
M
14-360(B)
CRUELTY TO ANIMALS(F)
F
A
113-270.1(B) FISHING WITHOUT A LICENSE M
A
113-270.1(B) HUNTING WITHOUT A LICENSE
M
113-270.1(B) TRAP WITHOUT A LICENSE
M
A
14-360(A)
CRUELTY TO ANIMALS(M)
M
A
14-361.1
ABANDONMENT OF AN ANIMAL
M
113-271
FISHING WITHOUT A LICENSE M
O
113-272
NO TROUT LICENSE
M
A
113-270.2
NO HUNTING LICENSE M
A
113-270.3
NO BIG GAME LICENSE M
A
75A-6 MOTORBOAT W/O LIFESAVING DEV M
A
327

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

A
A

941003
941003

A
A
940606
990127
971201
A
971201
971201
A
971201

6217
6218
6220
6221
6222
6223
6224
6225
6230
6231
6240
6242
6244
6246
6248
6250
6252
6254
6256
6258
6259
6260
6261
6262
6264
6266
6267
6268
6269
6271
6279
6281
6282
6283
6284
6285
6286
6287
6288
6289
6299
7101
7102
7103
7110
7112

113-270.3
HUNT/FISH/TRAP-NO GAME LICENSE M
A
113-291.8(A) FAILURE TO WEAR HUNTER ORANGE I
A
113-294
SELLING/BUYING WILDLIFE
M
A
113-291
TAKE GAME DURING CLOSED SEASON
M
A
113-291
EXCEEDING GAME LIMIT
M
A
75A-4 OPER MOTORBOAT INVALID NUMBER
M
A
75A-6 BOATING W/O REQ LIGHTS/EQUIP
M
A
113-291.1(F) USE UNPLUGGED SHOTGUN
M
A
920928
75A-10(B1) DWI - MOTOR BOAT/VESSEL
M
A
930621
113-134
FISH WITH UNLAWFUL BAIT
M
A
951201
113-291.1(B)(2)
SPOTLIGHTING DEER
M
A
941003
113-291.1(E1)
SHINE/SWEEP LIGHT FOR DEERM
A
941003
113-60.25
OPEN BURNING WHEN PROHIBITED M
A
941003
113-270.3(C) FAIL REPORT/TAG BIG GAME M
A
941003
113-264(A) LITTER GAMELAND/ACCESS AREA
M
A
941003
113-285
HUNT/FISH POST PROP NO PERMIT
M
A
941003
113-291.1(B)(1)
HUNT FROM MOTOR VEHICLE M
A
941003
113-135
DRIVE ON GAMELANDS ILLEGALLY M
A
941003
75A-15
EXCEEDING NO WAKE SPEED M
A
941003
113-135(A) UNLAWFUL CAMPING
M
A
941003
14-362.2
DOG FIGHTING
F
A
971201
14-362
COCKFIGHTING
M
A
941003
14-362.1
ANIMAL FIGHTING
M
A
941003
113-187
TAKE SHRIMP IN CLOSED AREA
M
A
941003
113-292
FISH TROUT WATER CLOSED SEASONM
A
941003
113-152
COMM FISHING NO VESSEL LICENSE M
A
911003
113-154(A) COM/MECH FISH NO SHELLFISH LIC M
A
951201
113-152
GILL NET VIOLATION
M
A
941003
113-135(A) CRAB OR CRAB POT VIOLATION
M
A
951201
113-290.1
NEGLIGENT HUNTING
M
A
940606
113-135(A) POUND NET OR STAKES VIOLATION M
A
951201
113-187(D)(1)
TAKE SHELLFISH POLLUTED WATER MA 951201
113-135(A) TAKE/POSS UNDERSIZE BLUEFISH
M
A
951201
113-135(A) TAKE/POSS UNDERSIZE CLAMS
M
A
951201
113-135(A) TAKE/POSS UNDERSIZE FLOUNDER M
A
951201
113-135(A) TAKE/POSS UNDERSIZE OYSTERS
M
A
951201
113-135(A) TAKE/POSS UNDERSIZE RED DRUM M
A
951201
113-135(A) TAKE/POSS UNDERSIZE SPOT TROUT M
A
951201
113-135(A) TAKE/POSS UNDERSIZE STRPD BASS M
A
951201
113-135(A) TAKE/POSS UNDERSIZE WEAKFISH M
A
951201
WILDLIFE - FREE TEXT
A
127A-131(A) CONVERSION OF MILITARY PROP.
M
O
941003
127A-131(A) DESTRUCTION OF MILITARY PROP
M
O
941003
127A-131(B) FAIL TO REGISTER MILITARY PROP M
O
941003
14-401.4
REMOVE/ALTER ID NUMBER M
O
941003
14-434
RETAIL UNLAWFUL RECORDINGS
M
O
941003
328

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

7199
PROPERTY - FREE TEXT
O
941003
7301 75D-7 PERJURY - RICO F
O
940606
7399
PUBLIC ORDER - FREE TEXT
O
940606
8410 LOCAL ORDINANC
CITY/TOWN VIOLATION (I)
IA
940606
8499
LOCAL ORDINANCE(I)-FREE TEXT
I
A
8501 LOCAL ORDINANC
TRAFFIC CONTROL DEVICE VIOLI
920203
8502 LOCAL ORDINANC
8503 LOCAL ORDINANC
8504 LOCAL ORDINANC
A
8505 LOCAL ORDINANC
8506 LOCAL ORDINANC
8507 LOCAL ORDINANC
8508 LOCAL ORDINANC
8509 LOCAL ORDINANC

PARKING VIOLATION
T
O
PARKING VIOLATION
I
A
NOISE ORDINANCE VIOLATION
LOITER FOR DRUG ACTIVITY
URINATE IN PUBLIC
M
BEACH STRAND VIOLATION
CITY/TOWN VIOLATION (M)
SPINNING TIRES I
A

MA 940606
A
940606
MA 940606
MA 940606
941003

8510

LEASH LAW VIOLATION M

A

LOCAL ORDINANC

8511 LOCAL ORDINANC
SCREECHING TIRES
M
A
8512 14-399
LITTERING PUBLIC/PRIV PLACES
M
O
8514 LOCAL ORDINANC
LITTERING BEER/WINE CONTAINER
A
8516 LOCAL ORDINANC
ILLEGAL DUMPING
M
A

910408
910408
M

941003
910408
M

8518 LOCAL ORDINANC
NO CITY DOG TAG M
A
8520 91-2 UNLICENSED PAWNBROKER
M
O
8522 LOCAL ORDINANC
PEDDLE LICENSE VIOLATION

920817
M
A

8526 LOCAL ORDINANC
NO CITY TAGS
T
A
8528 67-12 ALLOW DOG RUN AT LARGE/NIGHT M
O
8530 LOCAL ORDINANC
DEFRAUDING TAXI DRIVER

M

941003
A

8532 LOCAL ORDINANC SELLING BEER/WINE W/O LICENSEMO
941003
8534 69-31 VIOL OF FIRE EXTINGUISHER LAW
M
O
941003
8536 LOCAL ORDINANC
LOITERING M
A
8538 LOCAL ORDINANC
OBSTRUCT PEDESTRIAN SIDEWALK M
A
8540 LOCAL ORDINANC TRESPASS OR SLEEP IN PARK VIOLM A951201
8543 LOCAL ORDINANC
8544 LOCAL ORDINANC
A
8546 LOCAL ORDINANC
8555

14-399(C)

DISCHARGE FIREARM IN CITY MA
POSSESS FIREARM ON CITY PROP

940606
M

CARELESSNESS WITH FIRE

A

LITTER-NOT>15LB/27CUFT
329

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

M

M

O910408

911001

8556
8557
8558
8559
8560
8561
8562
8563
8570

14-399(C)
LITTER BEACH-NOT>15LB/27CU FT
911001
14-399(D)
LITTER - 15-500LB/27-100CU FT M
14-399(E)
COMM LITTER-ANY QUANTITY
14-399(E)
LITTER QUAN>500LB OR>100 CUFT
14-399(E)
LITTERING HAZARDOUS WASTE
14-399(C)
LITTERING-NOT > 15 LBS.
M
14-399(D)
LITTERING 15 - 500 LBS. M
O
14-399(E)
LITTERING > 500 LBS.
F
O
LOCAL ORDINANC
BARKING DOG
M

8572
8599
9901
9902
9905
9906
9910
9911
9912
9914
9916
9918
9919
9920
9922
9923
9924
9926
9928
9930

LOCAL ORDINANC
BLOCKING FIRE EXIT
M
A
941003
LOCAL ORDINANCE-FREE TEXT
A
15A-727;733;734
EXTRADITION/FUGITIVE OTH STATE FA
920203
17-1 HABEAS CORPUS F
A
920203
14-18.2(B)
INJURY TO PREGNANT WOMAN(F)
F
A
990127
14-18.2(C)
INJURY TO PREGNANT WOMAN(M)
M
A
990127
14-7 ACCESSORY AFTER THE FACT (F)
F
A
971201
14-7 ACCESSORY AFTER THE FACT (M)
M
A
971201
14-6 ACCESS. BEFORE THE FACT (R/81)
F
O
COMMON LAW
FELONY AID AND ABET F
A
COMMON LAW
MISDEMEANOR AID AND ABET M
A
14-2.4(A)
FELONY CONSPIRACY F
A
14-2.4(B)
MISDEMEANOR CONSPIRACY M
A
941003
90-95.1
CONTINUING CRIMINAL ENTERPRISE F
A
14-7.1 HABITUAL FELON F
A
14-7.7 VIOLENT HABITUAL FELON
F
A
961202
14-254
CORPORATE MALFEASANCE F
A
14-344
SCALPING TICKETS
M
A
14-118.1
SIMULATION OF COURT PROCESSES M
A
14-230
WILLFUL FAIL DISCHARGE DUTIES M
A

9954
9955
9956
9958
9960
9962
9964
9966
9968
9974

20-16.5
20-16.5
20-138.3
20-138.1
18B-401(A)
20-141.1
20-141(B)
90-108(A)(7)
90-108(A)(7)
14-410

CIVIL REVOCATION DR LIC (10)
CIVIL REVOCATION DR LIC (30)
DRIVE AFTER DRINK-PROV LIC T
AID AND ABET DWI
T
O
DRIVE-CONS MALT BEV PASS AREA
SPEEDING - SCHOOL ZONE
T
EXCEEDING POSTED SPEED
T
MAINT PLACE CONTROLLED SUB (M)
MAINT PLACE CONTROLLED SUB (F)
POSSESSION OF PYROTECHNICS

T
O
O
M
F
M

9975
9980

14-410(B)
20-63(G)

SALE PYROTECHNICS TO < 16 YR
COVERING/DISGUISING REG PLATE

M
TO

330
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

MO

910408

O910408
911001
FO910408
941003
FO910408
911001
FO910408
941003
O911001
941003
911001
941003
911001
941003
A
941003

M
M
O

A
A

971201
971201
910408

910408
O
910408
A
A
A
A

951201
910408

9984
9999

20-79 IMPROPER USE OF DEALER PERMIT
OTHER - FREE TEXT
A

T

O

910408

Defining Charges:
Below, the variable called “chargen” is recoded into a variable with eleven categories, called
“firstc11.” This variable was then recoded into “fircrnk.” Categories are defined as below. The
variable was reordered to correspond to a hierarchy in which speeding is number 1 (most likely to
be the charge that caused the stop), unsafe movement is 2, and so on.
recode chargen (lowest thru 2399,2503 thru 4299,4900 thru 5339,5342 thru 5345,5346,5347 thru
5364,5370 thru 5399,5406,5443,5453,5470,5471,5479,5523,5531,5532,5533,5540 thru
5552,5704,5711,
5509,5540 thru
5548,5562,5563,5564,5566,5568,5574,5575,5576,5588,5590,5591,5592,5593,5686,5704 thru
6040,
6041,6042,6044 thru 6205,6207 thru 6299,7101 thru
7112,8410,8499,8503,8504,8506,8514,8599,9901 thru 9999=1)
(2404 thru 2499,5417,5555,5567=2)
(4401,4402,4403,4412,4419,4431,4432,4433,4434,4416,4437,4439,4447,4451,4458,4459,4463
thru 4466,4469,
4482,4484,5412,5413,5419,5422,5424,5445,5446,5447,5457,5458,5464,5466,5484,5492=3)
(4448,4452 thru 4455,4476,4479,4480,5428, 5432,5454,5476,5477,5480,5482=4)
(4407,4418,4421,4422,4423,4424,4425,4426,4427,4428,4429,4430,4435,4436,4462,4486,4488,4
490,4491,4492,5424,5427,
5430,5462,5486,5488,5596,=5)
(4449,4450,4456,4460,4467,4468,4474,4475,4478,5442,5450,5456,5460,5467,5478,5498,5501,5
539,5549=6)
(4414,4470 thru 4473,4485,4531 thru 4535,5414,5444, = 7)
(4404,4405,4406,4408,4409,4410,4411,4413,4415,4417,4420,4438,4441,4444,4445,4446,4457,4
461,4477,4481,4483,4494,
4495 thru
4499,5341,5408,5415,5425,5463,5479,5481,5483,5487,5495,5496,5497,5499,5507,5510,5530,55
32,5579,5580,
5589,5597,5599,8509=8)
(4801 thru 4926,5402,5426,5435 thru 5439,5448,5468,5490,5550,5581 thru 5584,5640 thru
5643=9)
(4440,4442,4443,4489,
5400,5401,5407,5409,5410,5411,5416,5418,5420,5421,5434,5440,5441,5449,5455,5461,5465,54
69,5470,5471,5474,
5475,5485,5489,5491,5493,5494,5502,5503,5504,5505,5506,5508,5523,5524,5425,5525,5528,55
29,5531,5533,5535 thru 5538,5451,5452,
331
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

5556,5558,5560,5561,5565,5573,5586,5630,5633,5634,5655 thru 5682,5699=10)
(5403,5404,5405,5423,5429,5431,5453,5459,5472,5473,5511 thru
5522,5526,5527,5559,5570,5571,5572,5577,5578,5594,5595,5610,
5615,5620,5622,5624,5625=11) into firstc11.
recode firstc11 (6=11)(3=10)(4=9)(5=8)(7=7)(8=6)(11=5)(10=4)(9=3)(2=2)(1=1) into fircrnk.
execute.
value labels fircrnk 1'speed' 2 'unsafe move' 3 'fail to stop yield' 4 'equip' 5 'seatbelt'
6 'other misc traff' 7 'dui' 8 'license' 9 'resist or esc' 10 'st veh' 11 'crim charge'.

332
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Appendix E: Self-Reported Police Speeding Stops:
Results from a North Carolina Record Check Survey

Donald Tomaskovic-Devey
Cynthia Pfaff Wright
Abstract
Survey reports of police stops are a potential methodology for examining the magnitude and
prevalence of the “driving while black” phenomenon. Respondents can be asked to report on
police stops as well as their own driving behavior. Estimates of the magnitude or correlates of
racial disparity in police stops from self-reported survey data are potentially compromised if there
are racial differences in the accuracy of self-reports of police stops and driving behavior. We
report on the results of a record check survey in which we directly assess the degree and
consequences of racial differences in self-reports of police stops. In our sample of drivers who
have been cited for speeding in the last year we found that 74.8 percent of whites and 66.8 percent
of African Americans admitted to being stopped. Thus, while both groups under-report stops,
African Americans do so at a higher rate. This finding is consistent with many past studies which
report stronger social desirability effects on survey responses among African Americans. Thus,
survey data will tend to under estimate the magnitude of the “driving while black” phenomenon.
We find that the people who fail to report a speeding stop also tend to report lower levels of such
other undesirable driving behaviors as rolling through stops signs and speeding. There were no
race by reporting interactions in reports of other driving behaviors.

An early draft of this paper was presented at the American Society of Criminology Meeting.
November 17, 2000, San Francisco, California. Contact Tomaskovic-Devey don_tomaskovicdevey@ncsu.edu or at above address.
Introduction
Much of the research on the “driving while black” phenomenon relies on what we can
learn about police stops recorded in the police reports filed related to the incident. Ordinarily,
official records capture information for citations and searches, and some police organizations

333
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

officially record written warnings. As useful as this information may be, a large number of stops
are not routinely and officially recorded and are thus missing for purposes of analysis. This
unfortunate omission of stop information makes it difficult to address two key questions related to
racial profiling: 1) Are minority citizens the target for unwarranted stops in order to further
investigate the driver and/or occupants of the vehicle? (such stops could be considered pure
harassment); 2) Are the post-stop experiences of minority citizens different than the experiences
of majority citizens?
A recent report (Langan et al. 2001) suggests that we can assess the magnitude of “driving
while black” and learn a great deal about the quality of those interactions by directly surveying
and asking drivers about their stop experiences. In a survey context, it is also possible to ask
drivers about their driving behavior and then model, from the drivers’ points of view, the
probability of stops having been initiated by the drivers’ driving behavior as well as their personal
demographics. The types of models we have in mind might regress the probability of being
stopped because of race, while controlling for racial differences in age, gender, miles driven,
highway versus local driving, tendency to speed, and tendency to violate routine driving
regulations (rolling through stop signs, failure to use seat belts). To the extent that self-report
driver surveys can inform racial profiling research, such a technique may prove to be valuable to
law enforcement agencies and communities looking for a simple and straightforward means by
which they can assess police-citizen contacts.
Using self-report survey data may also appear to be as flawed as simply reviewing written
records of stop, citation, and warning data. Survey responses may under-report police stops
because of the sensitive nature of reporting violations of traffic and other laws and drivers’
distrust of the police or surveys in general. Responses may also over-report police stops because
334
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

of the high profile and political nature of the issue. If sources of non-reporting are associated
with the race of respondent, then survey-based analyses may be misleading—either by
exaggerating or underestimating the degree of racial disparity in police stops.
For example, of the drivers in our North Carolina Citizen Survey, 18.1 percent of whites
and 26.4 percent of African Americans report being pulled over by police in the last year. Given
these self-reports, we might conclude that African Americans are 1.45 times more likely to be
pulled over than whites. But are they actually pulled over at that rate? Researchers answering
yes to this question must make the strong assumption that white and African American drivers are
equally likely to accurately report being pulled over. As we will see, a review of the literature
indicates that a significant number of African Americans tend to under-report in response to all
types of questions that they interpret as being potentially threatening, sensitive or embarrassing at
a much higher rate than most whites typically under-report in response to the same questions.
In order to test the assumption that there is no racial variance in under-reporting instances of
being pulled over by police, we conducted a Record Check Survey of North Carolina drivers with
known speeding citations in North Carolina. In a standard record check survey, the investigator
knows the answer to the question before administering the survey, and then he or she surveys
respondents to measure the accuracy of their responses compared to the original records.

Background Literature
Item Under- or Non-Reporting for Sensitive Questions
Our chief undertakings in the record check survey are to discover the levels and types of
inaccuracy in survey data and to identify the characteristics of inaccurate responders. Sudman
and Bradburn (1982) identified four factors related to survey response errors: memory,
335
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

motivation, communication, and knowledge. Motivation errors are the major concern of this part
of the report. Motivation errors are inaccurate answers to survey questions which occur solely
because the respondent wants to manage the interaction in order to be viewed in a more positive
light by the interviewer. Motivation-based reporting errors are most likely to occur when the
survey item elicits a social desirability response. This type of reporting bias can manifest itself as
non-reporting, over-reporting, and/or under-reporting.
Over-reporting is a common occurrence for survey items that measure socially desirable
activities such as voting. Under-reporting is more common for survey items that measure
undesirable activities, for example, drug use (see Sudman and Bradburn, 1979 for a review). In
the literature, questions asking about undesirable activities are often referred to as “threatening”
or “sensitive” questions. These types of questions encompass activities which are thought to be
private, embarrassing, or illegal (such as personal income, party affiliation, religion, sexual
habits, or criminal activity). Respondents generally under-report when answering these types of
sensitive questions because they think that by admitting to having engaged in such behaviors,
theinterviewer would not view them as favorably as they might have having not know about these
so-called sensitive behaviors. Social desirability is thought to be at the root of non-response and
under-reporting to sensitive questions (Kormendi, 1988). These two phenomena are directly
connected to the reduced accuracy or validity of answers. Self-reports of police stops clearly are
instances of threatening questions at risk for a social desirability-based under-reporting.
Threatening question item response rates vary by study. The non-response rates tend to range
from fewer than 5 percent for questions considered to be less threatening (such as witnessing a
crime but not reporting it, Clark and Tifft, 1968), to as high as 73 percent for questions considered
to be more threatening (such as bankruptcy, Bradburn et al., 1979). The topic of victimization,
336
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

often studied with the use of questions considered to be potentially threatening, has been
repeatedly studied. The reviews of this research show that victimization is consistently underreported (Czaja et al, 1994; Yost and Dodge, 1970; Dodge, 1970; and Turner, 1972). In fact,
research indicates a direct relationship: the more extreme the crime, the greater the underreporting. For example, the findings in the victimization studies show larger under-reporting on
questions about assault than for questions about burglary.

Race and Item Under- or Non-Reporting for Sensitive Questions
Stocking (1979) used the Marlowe-Crowne Social Desirability Scale (Crowne and
Marlowe, 1960, 1964) to review these studies. He found that nonwhite respondents are more
likely to attempt to please the interviewer by giving socially acceptable answers to sensitive
questions. African Americans are more likely to complete interviews (thereby cooperating with
the interviewer), but the information may be less valid in response to sensitive questions due to
the respondents’ attempts to provide socially acceptable answers. Indeed, overall, African
Americans, as a group, are more likely than whites to respond to surveys (Groves and Couper,
1998, Cohen and Carlson, 1995, Brehm, 1993, Jackson et al, 1982, O’Neil, 1979, Hawkins,
1975).
For our purposes, those drivers who refuse to be interviewed are not as important to the
study as is our ability as researchers to discern whether or not the information provided by the
driver in the interview is accurate—especially the responses to what may be interpreted by the
drivers as threatening questions. Women, nonwhites, and those with lower levels of education are
more likely to under-report unacceptable behaviors or counter-normative attitudes (DeLamter,
1982:168). Sudman and Bradburn (1974), summarizing previous research on responses to
337
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

attitude questions, report that, for those questions arousing concern, African Americans are more
likely to have a response effect than whites. Witt et al (1992), in a study of item non-response to
questions about drug use, report that nonwhites are more likely to be item non-respondents than
whites. Cox et al. (1992) found that, compared to whites, African Americans and Hispanics not
only had a higher non-response rate, but also higher incidences of inconsistent responses for
answers to questions about drug use.
Previous research using the record check survey methods have found further evidence of
African Americans being more likely to under-report sensitive or threatening questions. Czaja et
al (1994), for example, examined respondents’ strategies for the recall of crime victimization
incidents. They found that 71 percent of whites reported victimization—compared to only 44
percent of African Americans. Hence, the odds of whites reporting their victimization was 1.9
times higher than nonwhites (see also Sparks 1981, Biderman and Lynch 1981, and Dodge 1983,
for similar findings on victimization; see Czaja and Blair 1990, and Czaja et al 1992, for studies
on other types of questions). There is also evidence that African Americans are more likely than
whites to conform, or acquiesce to questions with positive social desirability cues (Lenski and
Leggett, 1960 and Hare, 1960).

Record Check Surveys
The validity of self-reported behaviors that violate state or federal laws are often
questionable because of the afore-mentioned problems of under-reporting and incomplete or
inaccurate reports. It seems that participants, in their attempts to be both good respondents (by
answering the question) and to present a positive self-image to the interviewer, often do not
refuse to answer the question. Instead, more often, they report that they did not engage in the
338
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

threatening behavior being asked about (Bradburn et al 1978). For instance, Clark and Tifft
(1966), using a polygraph to check validity, found that 15 percent of the respondents refused to
answer a question on speeding while 38 percent of respondents under-reported speeding.
In this study, the possible consequences of more under-reporting by African Americans
compared to whites would be to under-estimate the extent of the “driving while black”
phenomenon. Conversely, if African Americans are more likely than whites to report stops, this
factor could result in exaggerated indictment of law enforcement behavior. While the previous
literature strongly suggests that African Americans are less likely than whites to report
threatening behavior, it may be that the current politicization of the “driving while black”
phenomenon encourages African Americans to recall and report driving stops. Since media
reports tend to place the blame for stops on police and not African American drivers, the social
desirability effects may be weakened for reports of police stops of African Americans in the
current political climate.
One method used to identify under-reporting and inaccurate respondent reports is to
conduct a record check survey. The Record check survey is a methodological tool used to
evaluate the validity and accuracy of respondents’ answers. Researchers create a survey, asking
questions about information about the individuals who are surveyed that the researchers already
have. Data collected from the respondents can then be compared to the known answers— thereby
effectively assessing the accuracy of the respondents’ answers. The purposes of the Record check
survey in this study are to determine whether or not drivers who have been stopped by police are
willing to report the stops during a telephone interview, and to assess how accurately the drivers
report the incidents. Using the findings from a Record check survey allows researchers—without
access to respondents’ known behaviors—to statistically compensate for under-reporting.
339
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

The design of the National Crime Survey (NCS) included three record check surveys. In
these record checks, police reports were compared to survey answers for a sample of citizens with
known police contacts (Yost and Dodge, 1970; Dodge, 1970; and Turner, 1972). Based on the
findings from these record check surveys, the National Crime Survey was then redesigned to use
survey items that produced less under-reporting. While we have built some question-wording
experiments into our survey to improve future surveys on police stops of motorists, our primary
objective is to determine whether or not racial differences in the probability of under-reporting
police stops effect estimates of the magnitude or existence of the “driving while black”
phenomenon.
Study Methods
Design of North Carolina Record Check Survey
Since our research question focuses directly on the relationship between the race of the
respondent and his or her propensity to be stopped by the police for speeding, we are specifically
interested in race-based variation in reporting of police stops. While studying the validity of
respondent answers is important in its own right, the overarching goal of our record check survey
is to increase the precision of estimates of race-based response variation in citizen surveys.
For the North Carolina Record Check Survey, a sample of known North Carolina speeders
were selected from a database of North Carolina citizens who had been ticketed for speeding
between June 1, 1999 and June 1, 2000. The list of inclusive names was obtained from the N. C.
Administrative Office of the Courts (AOC). We used the list of names of ticketed drivers within
the one-year period as the population from which we drew a stratified weighted sample of names.
Our targeted goal was 600 completed surveys. We weighted our sample in order to have
approximately one-half African American respondents and one-half white respondents. This
340
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

weighted sample design was advantageous because our research question concentrates on
discovering adjustment measures for race-based differentials in response to police stops. The
cooperation rate for the survey was 70 percent for white respondents and 69 percent for African
American respondents. As in previous research, whites refused to participate at slightly higher
rates and African Americans were slightly more difficult to locate. It is more difficult to find
valid phone numbers for African Americans than for whites,, but African Americans cooperate
with the survey at higher rates than do whites.
One week before the initial telephone contact attempt, advance letters were sent to each of
the persons included in the sample. The letters explained that the survey focused on the driving
experiences of people in North Carolina and their observations of other drivers on North Carolina
roads, and the results would be used to aid traffic safety and policy decisions. The survey itself
was administered by telephone and averaged nine minutes to complete. Most of that time is
allocated to general driving questions, designed partly to reduce the threat associated with
questions about police stops. In addition to our goal of finding estimates of response bias, we
also included an experimental manipulation of the wording in one-half of the questionnaires to
estimate the effects of the wording of threatening questions on question response. We will report
the results of that experimental manipulation in a future paper.
One issue of concern in record check surveys of stopped drivers is the possibility that
interviewers will know or guess that everyone had been stopped and so encourage higher selfreports. As a precaution to minimize such potential interviewer bias, previous researchers have
seeded record check survey samples with respondents who had not experienced the event in
question (Sudman, et al, 1977). Because we were fielding a larger survey with many of the same
questions, and also because the record check survey contained an experimental manipulation, we
341
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

simply informed the interviewers that the three surveys were linked and contained experiments in
question wording and questionnaire design. Therefore, the high proportion of respondents
reporting stops in the last year in the record check survey was disguised by the much larger
number of respondents reporting no stops in the larger driver survey.
Analyses
Since all respondents selected for our surveyhad been stopped for speeding, the record
check survey concerns the probability of a respondent admitting to the speeding stop event. We
are particularly interested in any racial differences in reporting speeding stops, since the larger
project uses survey self-reports as one avenue of exploring the “driving while black”
phenomenon.

Response Bias in Reports of Police Stops
Table E.1 shows the distributions of drivers admitting to having been stopped by the
police. The findings from the record check survey are that we can expect police stops of all types
to be under-reported by about 29 percent, and that there is a significant racial difference in selfreports of stops. Whites are 8 percent more likely to report any stops than are African Americans.
Table E.2 reports the core of the reverse record check, self-reports of speeding stops
among drivers whom we know, from official records, had been stopped for speeding in the last
year. Thirty percent of drivers do not report the specific speeding event which we used to select
them for the sample. There is a significant racial difference in self-reports: African Americans

342
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table E.1 Self -Reports of Stops by Police in the Last Year by Race
One or More Stops Reported

No Stops Reported

Total (n=604)

70.8%

(427)

29.2%

(176)

White (n=305)

74.8%

(228)

25.2%

(77)

African American (n=299)

66.8%

(199)

33.2%

(99)

Chi-Square = 4.639
Probability = .019

are 2.9 percent less likely than whites to admit to a speeding stop which happened in the last year.
This suggests that survey-based self-reports of police stops may under-represent actual racial
disparities in police activity.

Table E.2 Self-Reports of Speeding Stops by Police in the Last Year by Race
Percent of One or More
Stops Reported

Percent of NoStops
Reported

Total (n=602)

63.5

36.5

White (n=305)

69.8

30.2

African American (n=297)

56.9

43.1

Chi-Square 10.85
Probability = .001

Since the probability of speeding and being stopped is tied to other demographic
characteristics that may be associated with race, we examine whether or not this basic finding is
sustained after controls are applied for gender, age, education, and home ownership (as a proxy
for social class). Table E.3 reports a logistic regression of self-reports of speeding stops upon
race as well as a series of demographic and behavioral control variables. The findings from the
models indicate that racial differences in the reporting of police stops is sustained after
controlling for gender, age, education, and home ownership. Model 2 also suggests that younger
drivers are more likely to report their speeding stops than men and older drivers. We also ran
343
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

interactions of race with age, gender, education, and home ownership. The racial differences in
self-reported speeding behaviors do not interact with any of these correlates of race.

Table E.3 Logistic Regression of Self-Reported Speeding Stop upon Race and
Demographics Controls: Logodds, Odds Ratio, (Probability).
Model 1
Race (1=African American)

Model 2
-.44, .64, (.015)

-.47, .63, (.015)

Gender (1=female)

.22, 1.25, (.266)

Age

-.03, .97, (.000)

Education

-.01, .99, (.907)

Home Ownership

.25 1.28 (.262)

Degrees of Freedom
Model Chi-Square

1

5

5.961

23.917

Using Reverse Record Check Estimates of Response Bias to Adjust for Race Differences in
Stop Reports
The previous analyses demonstrate that there are racial differences in the likelihood of
reporting stops in general (Table E.1), and of speeding stops in particular (Tables E.2 and E.3).
The North Carolina Record Check Survey was keyed to speeding stops, but more general surveys
of racial differences in stop experiences would be more likely to focus on all police stops.
Speeding stops actually provide somewhat less room for police discretion in drivers who are
stopped than other reasons for stops, as discussed elsewhere in this report. In this section, we use
the racial differences in any stop reports from Table E.1 to estimate racial differences in police
stops using a larger, general survey of North Carolina drivers.

344
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table E.4 Full Survey Preliminary Estimates of Racial Differences in Police Pullovers, Adjusting for Response Bias
Self-Reported Stops
White

Drivers with Stops

Adjusted for Response Bias

African
American

White
(267 Self-Reported
Stops/.748)

African American
(360 Self-Reported
Stops/.669)

267

360

356

538

Total Drivers

1477

1368

1477

1368

Percent Stopped

18.12

26.32

24.1

39.76

Ratio African American/White
Stops

1.47

1.65

From the main survey of drivers we have data on 1,477 white drivers and 1,368 African
American drivers. Of the 1,477 white drivers, 18.1 percent report being pulled over by police in
the last year. African Americans report being pulled over about 45 percent more often—26.3
percent of the African American respondents report being pulled over by police in the last year.
The reverse record check results suggest that both of these are likely to be underestimates. Recall that in Table E.1 we saw that whites reported only 74.8 percent of actual stops
and African Americans reported even less—at 66.9 percent. We can calculate, based on reported
stops, the likely actual incidence of stops within race. For whites, that number is 356 (267 selfreported stops divided by .748) and for African Americans, our estimated number of stops is 538
(360 self-reported stops divided by .669). Thus, the reverse record check suggests that the driver
survey estimate of racial differences in stops (Table E.4) should be adjusted upward based on
racial differences in self-reports of stops. The self-report data suggest that African Americans are
1.45 times more likely than whites to have been pulled over in the last year. Adjusting for

345
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

response bias based on the reverse record check, the data suggest that African Americans are
actually 1.65 times as likely to have been stopped in the last year.

Response Bias and Self-Reports of Other Driving Behaviors
An additional goal of the driver survey is to identify racial differences (if any) in risky
driving behavior. Using the data from the reverse record check survey, Tables E.5 and E.6
produce estimates of racial differences in reported risky driving behavior among those that report
and fail to report speeding stop incidences. In this analysis, we investigate if the response bias
identified in the record check survey is associated with racial differences in self-reports of driving
behavior.
The dependent variable in Table E.5 is an additive scale of that we refer to as “risky
driving behavior.” It is meant to capture some of the driving behaviors that, while seemingly
minor, could bring one to the attention of a police officer. Risky driving behavior sums selfreports of rolling through stop signs, speeding up for yellow lights, failure to signal, and not using
seat belts. African Americans report .44 fewer risky behaviors, significantly less than reported by
white drivers. This number is reduced slightly after controls for accurate self-reports of speeding
stops, but shows no significant interaction with bias in reporting speeding behavior. Drivers who
fail to self-report speeding stops also report significantly fewer (.26) risky driving behaviors.
In Table E.6, we examine self-reports of typical speeds driven in 35 mph and 65 mph
speed limit zones. There are no significant racial differences in self-reported speeding behavior
in a 35 mph zone. African Americans report driving more than 1 mph slower than whites in a 65

346
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

mph zone. For both speed limits, those who are more likely to admit being stopped also admit to
higher typical driving speeds. In neither case is there a significant interaction with race.51

Table E.5 Regressions of Self-reported Risky Driving Behavior on Self-report of Stops,
Race, and Their Interaction; Metric Coefficient (Significance), n=604.
Risky Driving Behavior
Race (1=African American)

-.439 (.000)

Self-Report of Speed Stop

-.406 (.000)

-.302 (.034)

.262 (.000)

.512 (.069)

Self-Report* Race

-.163 (.350)

Adjusted R2

0.042

0.055

0.044

Table E.6 Regressions of Self-Reported Speeding Behavior on Self-Report of Stops, Race,
and Their Interaction; Metric Coefficients (Significance), n=604
Driving Speed When Limit is 35
Race (1=African
American)

-.594
(.098)

Self-report of Speed
Stop

-.433
(.228)

-.504
(.398)

1.282 (.000)

1.112
(.355)

Self-Report* Race
Adjusted R2

Driving Speed When Limit is 65
-1.252
(.001)

-1.135
(.004)

-.568
(.378)

.911
(.024)

2.271
(.081)

. 112
(.881)
0.003

0.021

0.019

-.888
(.272)
0.016

0.022

0.023

These analyses lead to two conclusions. First, respondents who are truthful on the record check
question are also likely to report higher rates of illegal driving behavior. We interpret this to
represent a general tendency toward a more accurate response to threatening survey questions
among this population. To test this conclusion we also ran a secondary analysis of
51

We repeated the analyses in Table E.6 using a dummy variable for reporting driving 10
or more miles per hour above the speed limit, and the results were the same. We also re-ran the
analyses in both Tables E.5 and E.6, deleting a single African American case with very low
reported normal driving speeds, but the substantive results were unchanged.
347
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

nonthreatening questions-–—self-reports of miles driven last week and per year. In neither case
was non-response on the reverse record check associated with self-reports of miles driven. Thus,
when the question was nonthreatening, there was no bias associated with a tendency to accurately
self-report a speeding stop. The second conclusion is that non-response bias in reporting police
stops is unlikely to effect estimates of race differences in driving behavior.

Conclusions
As in past research, we find that African Americans are more likely than whites to give
socially desirable answers to threatening survey questions. This tendency means that surveys of
drivers designed to estimate the magnitude of the “driving while black” phenomenon will tend to
underestimate police stops for both minority and majority drivers. This tendency will, however,
be greater for African Americans. Therefore, survey reports of police stops will tend to
underestimate the actual degree of racial disparity in police stops.
Respondents who fail to report police stops are also likely to provide more socially
appropriate responses to questions on risky driving behaviors or speeding. There is, however, no
evidence that African Americans are particularly likely to under-report either risky driving
behaviors or speeding. Evidently, the degree of threat in these items is not sufficient to produce
the type of race-linked social desirability responses we see for reports of police stops. This
simplifies the use of survey data on race and police stops.
Survey-based estimates of the magnitude of the “driving while black” phenomenon are
likely to underestimate the true degree of racial disparity in police stops. In the North Carolina
Record Check Survey, we found that 69.8 percent of whites and 56.9 percent of African
Americans who had been stopped for speeding in the last year actually reported such stops. This
348
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

suggests that self-reports of speeding stops by whites will be under-reported by about 31 percent.
Similarly, African American self-reports of speeding stops, at least in North Carolina, are likely
to be under-reported by 43 percent. Self-reports of police stops from survey data should probably
be adjusted upward to reflect these biases. Similarly, multivariate statistical analyses of the
causes of police stops (for example, race, gender, age, or driving behavior) should probably be
weighted so that those who report stops represent their expected proportion in a population.

349
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Appendix F: Citizen Focus Groups

Citizen Perception of the “Driving While Black” Phenomenon:
Research Summary From Six Focus Groups

A version of this Appendix was presented at the annual meeting of the American Society of
Criminology. San Francisco, November 17, 2000. C. Robert Fenlon (North Carolina Central
University) was author.

350
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Introduction
Focus group research in this study of citizen perception of “driving while black” is
mainly used to gather qualitative data to explore several issues. These are the reported and
perceived reasons for police stops, the perceived treatment of citizens by North Carolina State
Highway Patrol (NCSHP) troopers as reported by the respondents, their experiences with other
law enforcement encounters (local and county), how the police-citizen encounter began and
developed, and what knowledge citizens can report about police-citizen encounters by other
community members, friends and relatives. As such, our intent is to gather information on
personal experiences with the NCSHP troopers and other law enforcement officers, as well as
“vicarious” information as reported by the respondents regarding knowledge of racial profiling.
Focus groups can provide a source of detailed information about a particular social
phenomenon that may not be otherwise captured with more traditional methodological
approaches. Some would argue that an over-reliance on strictly quantitative methodology hinders
the complete understanding of the issue being studied (Krueger, 1994). The focus group
technique, as a data collection method used in conjunction with the information obtained from the
larger study, is an effective way to extract respondents’ attitudes, opinions, and knowledge
through the use of probing, specific questions in order to obtain information on an issue such as
racial profiling.
Morgan (1988) points out that focus groups allow for uncovering what individuals think
and why they think the way they do about an issue. Focus groups allow an examination of
interactions in a group setting, wherein how individuals in the group react to other opinions and

351
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

how they respond to opinions that might be somewhat different from their own (Morgan, 1996)
can be assessed.
We gathered indepth information from four African American focus groups and two white
focus groups in four geographically distinct locations in North Carolina.

Literature Overview
Absent the existence of a federal reporting system capturing information on policecommunity conflict, generally, and overall police misconduct specifically, systematically
gathered information about racial profiling is sparse. The issue of racial profiling is grounded in
the history of the often controversial relationship between the police and minority groups.
Evidence does suggest that conflict and tension between the police and the community has
steadily increased since the 1960s, when strained police-community relations were indicated as a
contributing cause of turmoil and civil disturbances in more than 150 American communities
(National Advisory Commission on Civil Disorders, 1968). Several studies (Carter 1985; Dugan
& Breda 1991; Gallup 1997, Henderson et al 1997; Murty, Roebuck and Smith 1990; Radelet
1986; Weitzer and Tuch 1999; and Weitzer 2000) have indicated that minorities (African
Americans and Hispanics) have less confidence in law enforcement relative to whites. However,
most studies to date indicate that the overwhelming majority of citizens—white and
minority—report positive perceptions of police (Radelet, 1998).
Even though a majority of citizens in the United States have positive views of the police,
African Americans are more likely to believe that the police treat their community concerns with
indifference, to feel that their neighborhoods receive inferior treatment relative to white

352
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

neighborhoods, and to believe that they are more likely to be targeted unfairly by the police due
to their race (Weitzer, 2000).
In a more recent examination of public perception of the police, Radelet and Carter (1994)
reported that African Americans indicated the following complaints:
>
>
>
>
>
>
>

Substandard or poor police protection
Substandard or poor service to minorities, especially inner-city residents
The expectation that the police will not treat them fairly
Numerous incidents of verbal abuse and harassment
Stereotyping of minorities as criminals, particularly in “stop and frisk incidents”
Police use of excessive force
Discrimination in police personnel administration.

Still, Murty, Roebuck and Smith (1990) conducted a study assessing public perception of the
police in predominately African American communities in Atlanta, Georgia. They sampled 600
African Americans, and their results indicated that 65 percent expressed a positive attitude of the
police while 35 percent expressed a negative attitude.
Lastly, Weitzer’s (2000) study examines citizens’ perceptions of racialized policing in
three Washington, D.C. neighborhoods: a middle-class white community, a middle-class African
American community, and lower-class African American community. His findings indicate that
there is across the board agreement that police treat African Americans differently from whites.
Most of the white respondents in his study took the position that black crime and criminality leads
to discriminatory, yet justifiable, treatment because “blacks have a greater likelihood or a higher
rate of black involvement in crime in Washington, D.C.” This seems to be the modal explanation
for the white community as to why African Americans are treated more harshly by the police. It
was interesting to note that a minority of African American respondents held the same view,
cutting across similar logical deductions, correctly or not, regarding this rational of police
treatment of African Americans.
353
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Overall, given limited research, minority perceptions of police are more negative than
whites. This is understandable given the history of relations between minorities and law
enforcement in the United States. Racial profiling or “driving while black” seems to be indicative
of the historical relations between the police and minorities.

Methodology
This focus group research study on public perception of racial profiling used an
exploratory methodological design approach to help better understand this issue. The results are
intended to stand on their own without any follow-ups. The research team chose to conduct focus
groups in four major North Carolina cites distributed across the state. Four African American and
two white focus groups were conducted. The research team determined that segmentation was
needed to create groups of similar racial makeup due to the nature of the issue, racial profiling. To
achieve the highest quality of information, maintain an environment conducive to forthright
opinions that are not guarded or masked, and overall, make focus group respondents feel more
comfortable, we believed that intra-racial composition was needed. Each focus group, with
consideration for age and gender, consisted of ten persons from the same race, having a range in
ages from 24 to 60, and had at least four females . The two white focus groups were held in the
same city, in succession after two of the African American focus groups. Respondents were
selected by a research firm located in each selected city.
Two trained facilitators external to the research team were used to conduct the focus group
sessions. The research team believed that an intra-race facilitator and respondent were needed.
Both moderators had considerable experience conducting focus groups on sensitive issues. Each
session lasted approximately 90 minutes and on some occasions went beyond this time mark.
354
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Debriefing sessions were held subsequent to each focus group session, in order to discuss
observations made by the moderator and the research team regarding facilitating style and
participant responses. At each focus group session, the research team took notes in an
observation room within each facility from behind a one-way mirror. Each of the respondents
voluntarily agreed to participate and signed permission forms allowing them to be recorded and
video taped at each of the focus group sessions. However, no last names were used in discussions
or recorded to ensure confidentiality. The audio tapes of each session were transcribed and sent
to the research team along with the video tapes.
The moderator of each focus groups imposed two kinds of structures: areas at issue and
the level of control exerted to complete the information gathering process in the allotted time. At
the onset of each focus group, the moderator opened with an introduction that fully explained the
intent and purpose of the research being conducted, who was doing the research and why, and
what was expected of them, so that each of the respondents understood the nature the research
study. The following issues were discussed:

<

Personal feelings in general about the police;

<

Differences in police behavior by type/jurisdiction of officer (such as local, county, and
state);

<

Personal experiences involving police stops;

<

Reasons why police may target African Americans drivers.

Data Analysis
We were looking for direct information, expressed by focus group respondents, that
exemplified the core of the issues. While preliminary, the results reported here do reflect the
research team’s overall assessment of the tone, flavor, and direction of the focus group analysis.
355
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

This afternoon we will address some of the material where it best highlights the commonalities
and differences in the experiences and perceptions of racial profiling within the white and African
American groups.

Issue One: General Feeling Toward the Police
It was perceived by the researchers that the general feeling toward police overall was
positive in the African American focus groups. African Americans generally expressed a positive
notion about police performance, practices, and behavior when addressing the safeguarding of
their community. While African American citizens held favorable views of police in general,
they expressed a tremendous distrust of police as individuals. For example, one respondent
indicated
“policemen are humans, too, but there are good ones and there are bad ones . . . and
it’s the bad ones that give the entire police department a bad name.”
Another respondent explained his mistrust of police, even though he had not had any negative
experiences with them, in this way:
“Well, I personally, I’d never had a bad run in with any police officer. But, I am
afraid of them.”
In another focus group located in the western part of North Carolina, respondents expressed
similar sentiments as well. They ebelieved that police are there to protect and serve the
community, and that they do a good job as a whole. However, a statement made by one of the
participants illustrates the gulf that exists between acceptance of what law enforcement is
intended to accomplish and the level of trust in the expectation that this will be accomplished:
“They are there to protect and serve but you have to be cautious of them. . . . I have
never had a run in with the police. . . . but I would not necessarily put my guard
down around a police officer.”
356
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Within the same focus group, two respondents had different views of the police in their
assessment.
“Growing up, policemen were portrayed as good . . . now they’re portrayed as
bad . . . when I see a police I always think negative. I don’t think positive.”
The other respondent indicated
“when I was younger, I guess I was a little bit negative about the police . . . but as I
got older, got to meet some of them personally, heard their point of view . . . it is a
very challenging job . . . as a whole . . . I think they’re going a good job”
The white respondents had very positive perceptions of the police. It seemed to us that
white citizen participants saw police activity (enforcement) as external to their community.
Enforcement, as it should be, is focused where “law and order” seem to be more of a problem.
The white respondents, as a whole, seemed to have a sense that police service their community
with little or no conflict; therefore, they felt comfortable with police presence. They seemed
willing to give up on some freedom, such as stops at police check points, and did not see such
inconveniences as an intrusion on their rights. They did not expect such stops to have any great
effect on their neighborhoods, since most of the crime committed is outside their community.
One respondent indicated the following:
“They (police) patrol areas where there are potential problems . . . stop things, you
know, drug areas they drive through and try to prevent things from happening.”
Another respondent had this to say about the police:
“You have got the rule of law and the police are the ones that enforce that . . . they
keep the order . . . and if they (police) have a license check set up . . . have drug dogs .
. . something looks suspicious . . . check it out . . . in doing that it has caught a lot of
people . . . I for one think its worth a slight inconvenience to get that (drugs) off the
road. I very much support him (police).”
357
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Issue Two: Compared Types of Law Enforcement
African American citizens reported a sense that NCSHP troopers were more professional
in their dealings with the public than were local law enforcement. They attributed this differences
to the level of training received by each. Citizens also spoke negatively of the stereotypical
sheriff lounging across the food counter at the convenience store with his shirt hanging out. It
was believed that the NCSHP would enforce the law more consistently and would rely on
personal relationships with citizens less often in determining the final outcome of the stop than
local law enforcement would. On balance, African American citizens expected to be treated
fairly by the NCSHP but not so fairly by local police.
“I feel that they are more professional, the state troopers.”
“If you are speeding, they are going to give you a ticket.”
“If you are breaking the law, they are going to make it right, but a lot of town cops .
. . allow certain folks to do this and do that, and then certain other folks can’t do the
same thing.”
An African American citizen speaking of her white neighbor driving through the neighborhood
speed trap:
“We meet the city cop, he throws up the hand, she throws up the hand, she continues
to go, I feel like I need to break my speed.”
This last comment leads us to white citizens’ assessment of differences among law
enforcement agencies. Generally, there is agreement between whites and African Americans
here. Both groups express the opinion that NCSHP is more likely to treat one in a formal manner
(that is, stop and ticket), while local police are more likely to be idiosyncratic in their policy
enforcement. Whites see an advantage with local police in comparison to NCSHP, and this
advantage is the result of familiarity between the police and citizenry.
358
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Issue Three: Police Stops and Fairness
African American citizens reported a remarkable willingness to take responsibility for
their actions. Especially in cases where speeding was the infraction triggering the stop, these
citizens reported that the stop was fair. Still, there was a serious concern about the treatment after
the stop. Here issues of respect and understanding surfaced. Citizens reported confronting such
questions as “Do you live around here?” and “Is this your car?” It should not be surprising that,
even when a citizen is knowingly in the wrong, such questions will tarnish future interactions
between citizens and police. The situation is magnified when the stop itself is illegitimate, as
reflected in the following:
“I see him (the officer) just looking around my car and what not, and he said, ‘You
know the reason why I pulled you over, don’t you?’ He said, ‘Your tag has
September ‘98 on it.’ I said, ‘It’s only May.’ He was like, ‘That’s right. My fault.’
He gave me my driver’s license and registration back and said, ‘You don’t have any
guns in the car do you?’ I said, ‘No, it’s at home with the dope.’”
Another encounter was the result of mistaken identity. A young man and his brother had the
misfortune of being in the general vicinity of a search. The two brothers were stopped, tackled,
searched, and detained without ever being told the reason for the encounter.
Walk back towards the front of my car. He told me to walk backwards. With my
hands in the air. I walked backwards about five steps. They stopped me there and
told me to lift up my shirt and turn around so that they could see that I didn’t have
anything in my waist. I did that. I took three more steps back and that’s when I was
tackled. I don’t know how many of ‘em grabbed me but I felt at least four of ‘em.
One on each hand, you know, somebody grabbed me around my neck and they
handcuffed me and had me down, patted me down. I was asking, what did I do?
What’s going on? He said, just don’t say anything just be quiet. We’ll tell you later.
Just stand up and get in the back of the car. And, okay, all right. They put me in the
back of the car. Did the same thing to my brother except they only let him stand up
out of the car and when faced forward, a whole bunch of them just ran over towards
him and tackled him into the grass. They held both of us for ten minutes and we
were sitting in the back of the car. They searched my car, they opened trunk, went
through everything in the car. Nobody told me why they were stopping me. Nobody
told me what was happening, my license disappeared for three days. I don’t know
359
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

who took it. One of the officers took my wallet and my license was taken out of my
wallet. My wallet was given back to me but whoever took my license got in their car
and left with it. I got it back in the mail three days later. I don’t know where it
went. I don’t know who took it and don’t know why it was taken.”
No whites reported this type of treatment, although whites did note that police do
sometimes single out drivers. We heard of “driving while blond” and being stopped for “driving
a flashy car.” The point here is that the stop experiences that whites shared were specific to
themselves and, except in very general terms, were not shared by others. Interestingly, while
white drivers do not report pretextual stops, they were much more likely than African Americans
to display a sense of entitlement and even resentment when they were stopped.
“Sure I was speeding, but they should have cut me a break.” Or the often standard “Why are they
stopping me when they could be doing something useful?”

Issue Four: Targeting African American Drivers? (African American Citizens)
Who Gets Targeted by the Police (White Citizens)
A majority of African Americans reported that, other than race itself, police are more
likely to stop them due to the presumption of “black criminality” and physical stereotyping. It is
expected that police are more likely to target African Americans whose dress is less conventional,
who drive cars deemed to be linked to criminal activity, and who have hair that is counter to the
“norm” (dread locks, for example). African American participants recognize that criminal
stereotyping is predictive of disparate treatment. This recognition heightens the sense of unfair
treatment of African Americans by the police. Additionally, there is a presumption by the
participants that African Americans are less likely to challenge what they know to be unfair
treatment by the police. A lengthy court challenge to an unfair ticket, in the long run, may be
more costly in time and money then the actual cost of the ticket.
360
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

“Most of the time, we don’t fight it, we don’t complain, we just pay it.”
“Blacks are less likely to complain, have less resources with which to get things
done.”
For whites, police are seen as on the lookout for teens, certain types of vehicles, types of
individuals “likely to commit a crime,” persons who “don’t fit the car.” While such stereotypes
might suggest a clear racial bias on their face, participants were just as likely to suggest that
young and male, regardless of race, would be enough to trigger disparate treatment. For example,
one subject noted:
“I think all young males—I think if you see a young male in a fancy sports car with
all of the gadgets and the radio is loud, I don’t think they would care if he were black
or white, they would pull him.”
After some recognition that young African Americans, in particular, might be more likely than
whites to be stopped, especially if they were somehow “out of place” (neighborhood, type of car)
whites wanted to discuss how racial discrimination could go both ways.
“I think is exists, but I think it exists on the other side too.”
“A black police officer pulling a white guy.”
“I mean, you see it in your workplace. It is the same thing.”
Whites seemed more comfortable with targeting and profiling when it could be expressed
as affecting both whites and African Americans. In sum, whites know about “driving while
black” from media accounts, but they don’t think that the extent of the problem is as great as it is
portrayed. Indeed, whites tended to not be particularly sympathetic to African American drivers
who were being targeted by police on the basis of race. They suggested that police were stopping
African Americans for good reasons such as drugs or violence. Their examples were not
connected to driving at all, suggesting a disquieting generalization from one realm of experience

361
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

to another. We end this section with one of the most extreme positions noted in the white focus
groups:
“They can ride around all times of the day or night, they don’t bother to have job
any where, but they keep gas in the car, so you have got to suspect that they have got
something going on in order to keep the car full of gas and on the road.”
Conclusions
The issues surrounding racial profiling by police are complex. One pattern is revealed
through the examination of official police stop statistics. We suspect that often a very different
pattern will appear when actions are viewed through the eyes of police, and also of drivers.
While generalizations based upon a small number of interviews should be cautiously avoided, the
information we do have suggests that, although African American and white drivers may agree in
principle that profiling—even racial profiling—exists, they do not agree on what it is or on what
are its benefits or consequences. While disagreements on all of the issues were found within each
of the focus groups, the strongest differences were found to be between the two races when the
groups were examined collectively.
Our focus groups with African American drivers revealed a generally positive evaluation
of the job that the police do. Participants were quick to say that the police had an important and
tough job and that they were grateful for the good work they do. At the very same time, however,
many of the African American drivers had very little trust in individual police officers. They felt
themselves to be at risk of harassment and bias based on race, and they made considered analytic
distinctions for each and every time they were stopped. Some stops were judged fair—typically
when they realized or admitted that they had broken a law and they were subsequently treated
with respect. In general, African Americans in the groups described law enforcement as an
institution as legitimate and reasonable, yet conversely, they described individual police as
362
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

suspect. They attributed racial bias to “bad apples.” There was, however, some disagreement
among African Americans as to how common the bad apples are.
Stops that were not tied to serious illegal driving behavior—the most common was the
“rolling stop” pull-over—were treated as likely instances of racial bias. In many of these cases
African Americans assumed race was the cause of the stop, because they did not recognize any
other legitimate reason. In some cases this assumption was confirmed by the police officer, such
as when reporting that the African American citizen was stopped for being in the wrong—or, in
other words, white—neighborhood (and thus out of place). One young man spoke of the time he
was stopped (with his brother), removed from the car, tackled, and had guns drawn on him for
driving in a neighborhood where another African American man on foot was being pursued by the
police. Here, apparently, “young, black male on foot” was interpreted as “black male anywhere.”
Lack of respect by the police during legitimate stops were also evaluated by some African
American drivers as likely instance of racial bias. Lack of respect in the interaction was
interpreted as an indicator of racial bias, and encouraged the suspicion that the pull-over was
racially motivated as well. Troopers of the NCSHP, in contrast to officers attached to various
local police forces, were singled out as treating drivers professionally and with respect.
A clear pattern emerged in the focus groups, revealing that African American drivers were
less suspicious of the NCSHP than they were of other police officers. While this evaluation
mirrors our findings in the citizen survey (that racial disparity in police stops is lower among the
NCSHP than among other law enforcement agencies in North Carolina) the focus group
participants used respectful treatment, rather than the rate of stops, as the basis for arguing that
the NCSHP was better.

363
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

In general, African Americans were more likely to perceive racial bias in a stop if the
officer interacted with them in a disrespectful manner or they were stopped without what they
believed to be a legitimate driving infraction. They seemed to be more than willing to
acknowledge their responsibility for a “real” violation. Minor violations created a different
perception of teh possibility of racial bias. In these cases, African Americans saw race as the
predictor of the stop, not the violation. This is in contrast with white drivers who tended to see
all stops—legitimate or minor—as discretionary and idiosyncratic. White drivers talked about
“driving while blond” or “driving while a musician” or “he should have cut me a break.” In many
ways, white drivers evaluated the police more harshly than African American drivers did and
were also more likely to generalize “unnecessary” enforcement across agencies. African
American drivers saw many stops as legitimate and some as potentially racially biased. White
drivers saw most stops as illegitimate, but idiosyncratic.
We also found stark contrasts between African American and white drivers in evaluations
of the “driving while black” phenomenon. African Americans tended to see it as just another
example of racial bias. Racial bias in the policing of drivers was seen as a form of discrimination
similar to the other forms of discrimination faced each day. Its existence was confirmed by some
combination of their own experiences, stories they had heard from friends and family, media
reports on police bias (Rodney King was often mentioned), and the existence of general levels of
prejudice and discrimination in the society at large.
White descriptions were considerably simpler and more disturbing. The white focus
groups tended to accept that police targeted African American drivers, but described racial
targeting as at least understandable if not fair and justifiable. Since African Americans were
stereotypically assumed to be more dangerous and thus more culpable, white citizens typically
364
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

saw police stops on the basis of race as reasonable. Whites tended to use stereotypes and
statistical discrimination arguments similar to those sometimes used by police to justify racial
targeting. It seemed very easy for the white subjects to collectively justify discrimination in
policing, even though they were quite resistant to taking personal responsibility for their own
police stops. As such, disgruntled white drivers are no natural allies for African American drivers
who fear they are being harassed because of their race.

365
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Appendix G: General Issues in Measurement

Before we began the data analysis of the stop, citation, and written warning records, we
found that it was necessary for us to evaluate various aspects of the quality of the data themselves.
We argue below that the stop data are less complete than the citation and written warning data,
and thus much of our subsequent analysis centers on the latter data sets. The primary problem is
that we cannot distinguish between citations issued at checkpoints from citations issued as a result
of an officer stopping a vehicle in routine patrolling (the data set does not have a variable that
allows us to make this distinction). We will also describe data on accidents, as well as
observational data on speeders (details of which are in Appendix A).

The Stop Data
Beginning in January of 2000 the NCSHP was required to record every vehicular stop that
its troopers made. The stop information is recorded on a form, (the “stop form”) and then entered
into a data base. These data seem particularly valuable because they are unique in providing us
with information about police-citizen contacts that were previously unavailable to researchers:
stops that do not result in citations or written warnings. These types of stops are of interest
because they may be argued to be “pre-textual” stops – stops initiated as a pretext for the trooper
to ask questions, possibly leading to a search of the vehicle. The data allow us to test the
hypothesis that African Americans are stopped and given “no action” or only a “verbal warning”
more than whites. Thus, the stop data provide us with further measures of the troopers’ behaviors
and thus better enable us to grasp the entirety of police interaction with citizens.

366
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Unfortunately, the value of the stop data can only be evaluated in conjunction with
citation and written warning records, so as to determine if the stop records are generally being
completed as often as they should be. There are no stop forms completed at checkpoint stops
(stops where all or random vehicles are pulled over) and thus we cannot evaluate fully whether all
the stop forms that should be completed are being completed. We came to the conclusion that
there are too few stops relative to the number of written warnings and citation records to warrant
study of the stop records as a stand-alone data base (at least relative to the goals of the present
research). Stated another way, the evidence suggests that the troopers are filling out the necessary
forms to provide supervisors and researchers with a track record of their transactions with
citizens, but sometimes the stop forms are missing, either because they are not being completed
(the form is not filled out) or at least they are not being entered (someone fails to enter the
completed form into the data base) or because they are not required (checkpoint stop). We
estimate that in up to a third of the situations in which we think that stop forms should be filled
out and entered into the data base, they are not being completed or entered into the data base
when citation and written record forms for the same incidents are being recorded. There are
several possible reasons for the under-recording of stop records. For one, as stated above, we do
not know how many checkpoint stops there are. Two, there could be confusion as to what
constitutes a stop (see discussion below). In addition, unlike the data collecting/entry process for
citations and for written warnings, the data processing procedure for stop forms does not include,
as far as we are aware, standard data quality checking procedures, such as ascertaining that all the
stop forms are processed. It appears that often troopers fill out written warning forms, capturing
the race of the driver, but fail to reproduce the information on the accompanying stop form.
While written warnings certainly provides a “paper trail” related to the stop, it makes it
367
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

sometimes impossible for the researchers to verify that a stop record indicating that a written
warning was issued, and a written warning record of a similar event are actually records of the
same event. This complicates the research effort, as considerable data processing and verification
is necessary to evaluate the quality of the data that are analyzed.52
As mentioned above, ambiguities of what constitutes a stop may also play a role in
reducing the number of stop forms filed. Stop forms are to be filled out when a trooper initiates
and completes a stop of a motor vehicle. Thus, stop forms are not filled out when a trooper is
called to the scene of an accident (or witnesses an accident) or at a checkpoint stop. Nor is a stop
form filled out when a driver is already stopped for some other reason such as vehicular
dysfunction or rest. In such cases written warnings and citations would not normally result in a
corresponding stop form. Unclear is the situation when a citizen calls in a report of another
driver’s behavior, resulting in a dispatch call to a trooper and a subsequent stop of a driver, since
the stop was actually initiated by a citizen. Also, reports from a citizen-band radio, resulting in a
driver being stopped should be considered a stop, but may not always be defined as such. We do
not know why the stop forms are not always filled out, but we do have evidence that they often
are not.53
Although NCSHP troopers are instructed to fill out a stop form every time they initiate a
stop, it is not entirely surprising that the form is not filled out when one considers that the data are
52

For example, it would be useful to match records of the same incident across stop and citation data
bases, or
across stop and written warning data bases.
53
We initially hypothesized that in the first few months data were missing because troopers were learning
how to fill out the new stop forms. We examined the data by month and find no trend in the “missing”
data. Also, it should me mentioned that when the stop data were sent to us we verified that the number
of records received equaled the number of records sent, so we did not lose the data by data processing
mistakes.
368
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

largely redundant with information – including race -- already included on other forms. Thus, if a
trooper failed to file a stop form at a stop, yet issued a citation or written warning, a record of the
event was placed on file and subject to our review. A “paper trail” exists. The trooper may think
that filing a separate stop form is unnecessary because presumably researchers or supervisors
should be able to “figure out” who was stopped by examining all the data files potentially
relevant to his or her stop (the citation, written warning or search files). This assumption would
hold if all of the information collected on a stop form were collected on each of the other types of
forms (such as written warning and citation forms). Unfortunately for researchers, not all the
information on the stop form is contained on these other forms. Written warning records, for
example, omit age of driver, time of day, and the area of a county in which the written warning
was issued. Also, the quality of the stop form data themselves is deficient in other ways. For
example, information is not recorded as to the sub-area of a county in which the stop occurred.
Thus, the stop data are not useful for defining a unit of aggregation (a unit of analysis) that we
elsewhere argue is valuable – the highway area (stretch of highway within an area of a county
(roughly a fourth of a county in size).
It would be useful to know whether the failure to record a stop record in the data base is
systematic or random across district or across officers. To evaluate that question is surprisingly
difficult because there is no identification number linking a stop form to either a written warning
record or a citation record. Thus, some matching criteria, such as the demographic characteristics
of the driver, time of day, etc., are necessary to match stops with citations or written warnings.
To link stop forms that indicate a written warning was issued with written warning files, we
initially thought that matching criteria such as the following would allow for accurate matching:
the district in which the written warning was issued (fifty-three districts), the trooper
369
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

identification number indicating who the trooper is issuing the written warning or filling out the
stop form, the month and day of the week, the race and gender of the driver, and some general
information about the nature of the charge. However, we found that there were too many
inconsistencies in the definitions of these criteria across data sets. For example, type of behavior
resulting in a stop or written warning is coded differently in the two data bases so that it is
difficult to determine clearly the “reason for the stop” and resulting written warning for any
offense other than speeding, equipment violation, or for a license/registration irregularity.
Moreover, the reason for a stop may not be the same as the reason for the written warning: the
trooper may have stopped the vehicle for speeding but issued a written warning because the
inspection sticker had expired (dropping the speeding charge altogether).
As researchers, we must be able to distinguish a citation record which has had a stop
record filed from a citation record which does not have a stop record filed in order to know
precisely how many stops a trooper made. If the researcher cannot distinguish stop and citation
records precisely then we can only estimate how many stops were made by the NCSHP. In the
present context this means that fairly liberal matching criteria must be used to achieve a match
across data bases. As it turned out, we can only match records by date, district, trooper and
gender of person stopped, warned or cited. (Using more criteria results in very few matches.) For
matching purposes we cannot distinguish between records of people of different ages or races,
because age is not recorded across record types and race is problematic regarding the definition of
Hispanics (and by implication whites – see discussion below). Because the matching criteria are
liberal, there will be “false positive” matches — some of the matched stop and written warning
(or citation) records may not involve the same actual event. As just stated, we used liberal criteria
that defined two records as matched if they have the same codes for district, trooper, day, and sex
370
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

of driver. Thus, if a trooper issues several written warnings in the same district on the same day
to several males, we cannot be sure which specific stop record information on which males
stopped and warned (in written form) by the same trooper on the same day matches correctly with
a written warning record. Despite this limitation (which we think is minor relative to our primary
analysis goals), the data do provide us with an estimate of how many records do not match
between the data bases. For example, some stop records will not have a corresponding written
warning record and some written warning records will not have a stop record indicating a written
warning that we can match.
To get an idea of the completeness of the stop data by themselves, examine Table G.1
below. Here there is a breakdown of the number of stops by the actions taken by the NCSHP
trooper at the stop. Also included is the number of non-accident citations and written warnings as
recorded in the respective data bases (that is all the citations issued at non-accident encounters
with drivers, and all written warnings issued.) As can be seen, there are substantially more
citations and written warnings recorded at non-accident interactions between police and citizens
than are recorded on stop records. Of course some of these citations and written warnings are
issued at checkpoints stops, which do not require that a stop form be filled out.
One possibility that could account for the discord is that troopers do not consistently fill
out the stop data forms. This is possible, perhaps, as suggested earlier, because the forms are
largely redundant with the data that must be recorded on what are perceived by troopers to be
“more important” forms – the citation and written warning forms. Conversations with NCSHP
troopers in our focus groups suggest that not all troopers give high priority to completing the form
immediately following the stop or submitting the stop forms for data entry daily. For example,
one trooper reported that he fills out the forms only at the end of the week or when he has enough
371
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

time, and does so by copying the information from their citation and written warning forms. As a
consequence of this practice, stops not resulting in a citation or written warning may be
overlooked or forgotten. We estimate that there is a substantial percentage of the stop forms that
are not filed.
Despite the missing stop forms, some unique information is collected on these forms,
making them valuable. Some of the stop forms record “no action” and the “verbal warning”
actions taken by the trooper. These two forms of action, however, represent only 2.4 percent of
all the stop records. Approximately 26 percent of the stops resulting in citations and /or arrests in
calender year 2000 do not seem to result in a stop record being filed.54 More prevalent is the
failure to report written warnings on the stop records, where 55.6 percent of the written warning
records have no stop record equivalent. The relatively high rate of missing stop records for
written warnings partially corroborates the hypothesis that the troopers do not take filling out or
filing the stop forms as seriously as they might, since the written warning outcome would
generally be regarded by troopers as a less serious type of intervention than a citation. (that is, we
would hypothesize that the less the intervention is consequential for the driver, the less likely the
stop form will be filled out.) Because of the missing records, we are reluctant to

54

This is actually a conservative percentage because we do not have available
information on the actual number of arrests (only the arrests recorded on the stop forms). To
attain as complete a list of matches as possible we include arrests at a stop with the citations
because a citation has been filed. Troopers would indicate on a stop form that the most serious
intervention that occurred was a arrest, and not a citation.
372
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table G.1 Comparison of Number of Stops to Number of Citation Events and Written
Warnings
Action Taken by
Trooper at Stop

Number of Stops
Resulting in Action
Taken

Percent of
Records
“Missing”
(percent of
known events not
in stop data base)

Events Known (NonAccident) from
Originating Data Base

Citations (& Arrests)

440,098 (including
11,815 arrests)

25.8

593132

No action

3000

--

--

Verbal Warning

10366

--

--

Written Warning

109550

55.9

248296

try to interpret any patterns with the stop data as a “stand alone” data base because it would
appear that stops are occurring that are not being entered into the data base.
Despite the fact that most of the information on stop records is redundant with information
recorded on other records, it is lack of redundancy that has been responsible for most of our
research concerns. One of our research goals is to verify that the official records of written
warnings and citations are reasonably complete in the sense that the forms that are filled out are
entered to the data base. To determine the extent that forms are being filed, we attempt to match
stop records to both the written warning and citation files. We were concerned that not all of the
stop forms were being filled out, but we were also concerned that not all of the written warning
and citation information was being filled out or processed. To assess the prevalence of missing
written warning and citation records, we initially matched stop records to written warning and
citation records. The matching process was made difficult, however, because the forms differ
373
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

somewhat in the categories of information recorded, making the match rate less than perfect. On
the stop form, for example, race and ethnicity are recorded as separate variables, while on the
written warning and citation forms, race and ethnicity are recorded as one variable (called
“race”). This difference is a source of discrepancy between the different record types. For
example, all else equal, it is likely that more Hispanics will be recorded as Hispanic on a form
that has a separate variable designating Hispanics than on a form that does not have a separate
variable for Hispanics. Before continuing our discussion of the possibility of missing written
warning and citation records, we will discuss our concerns about the measurement of who is
Hispanic in the various official record files.55

Measurement and Classification of Hispanic Drivers
On the citation and written warning records, which have no special variable to indicate
whether the driver is Hispanic, 5.3 and 3.3 percent of the records, respectively, (see Table G.2)
indicate an Hispanic driver, compared to the stop records, which indicate that 5.4 percent of those
stopped are Hispanic (5.6 percent of those getting citations and 3.5 percent of those getting

55

By separate variable we mean that there is a separate question as to whether the driver
is Hispanic or not. Where there is, a separate variable is to be distinguished from the situation
where Hispanic is a category among several others such as “African American” or “white”.
Presumably in the former situation more Hispanics would be counted than in the latter situation,
where some of the Hispanics would be classified as white or African American instead of
Hispanic.
374
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table G.2 Type of Form and Measurement of Hispanic Drivers
Stop Form

Racial Catgegories: White, African
American, Indian, Asian, Other

Hispanic is not a race
option

Written Warning Form

Racial Categories: White, African
American, Hispanic, Indian, Other,
Unknown/Uncertain

3.3 percent Hispanic

Racial Categories: White, African
American, Hispanic,
Indian, Other, Unknown/Uncertain

5.3 percent Hispanic

Citation Form

Stop Form with Written
Warning as the Trooper
Action

Driver’s Ethnicity: Hispanic, Not Hispanic

Stop Form With Citation
as the Trooper Action

Driver’s Ethnicity: Hispanic, Not Hispanic

3.5 percent Hispanic
5.6 percent Hispanic

written warnings). While the similarity in the findings of these two data sets might lead one to
believe that the definition of who is an Hispanic is similar in both the stop records and the written
warning and citation records, we do not know which specific individuals are coded as “Hispanic”
on a stop record and on a written warning or citation record because there is no common
identifier to link the records. We only know that in the aggregate the percentages are only
slightly different (3.3 versus 3.5 percent and 5.3 versus 5.6 percent in the table above).
In general, we are doubtful about the accuracy of the coding of Hispanics. In comparing
N.C. Division of Motor Vehicle data on Hispanics to citation data, our analysis revealed that 21.6
percent of the NCSHP-defined Hispanics cited by the NCSHP in 2000 did not have Hispanic
listed with the DMV as their “race.” Thus, it is possible that NCSHP troopers over-identify

375
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

people as Hispanic or the DMV under-identifies them, or that some Hispanics choose to identify
themselves as white or other. At the same time, of those drivers identified as Hispanic by the
DMV and differently identified by the NCSHP, we find that only .1 percent are identified as
“White,” less than .1 percent as “African American,” while .6 percent are classified as “other” by
the NCSHP. Thus, relative to the standard of racial/ethnicity identification used by the DMV,
NCSHP troopers tend to “over-identify” drivers as Hispanic and only rarely “under-identify”
Hispanics as some other race or ethnicity (most often here as white).
We are especially doubtful about the accuracy of the coding of Hispanics on forms that do
not provide for an Hispanic ethnicity classification (the written warning and citation forms). On
the stop forms, which have an Hispanic ethnicity category, the race of 89.5 percent of those
Hispanics is coded as ‘U’ (unknown/uncertain), with 9.3 percent of Hispanics defined as white
(the remaining 1.2 percent are coded as Asian, African American, or Indian). Only 1.1 percent of
the written warning records are of ‘U’ drivers (another 3.3 percent are explicitly coded Hispanic)
and 3.3 percent of citations are to ‘U’ or ‘O’ drivers (unknown/uncertain or other), with another
5.3 percent listed as Hispanic race. In other words, Hispanics are not generally coded as being of
the white race on the stop form, but rather are “unknown/uncertain,” while on the citation and
written warning forms, Hispanics are coded as Hispanic for race, with some unknown number of
them likely coded as white (we could only know for sure if we had an identification number
matching stop and written warning or citation records, such as the stop identification number.
Such a number exists, but it is not linked to the written warning or citation records).
There is variation on the stop form in the classification of Hispanics as ‘unknown or other’
versus white (recall there is no Hispanic code for race on the stop form). Only 84.5 percent of the
Hispanics on the stop form with a written warning action are coded as “unknown” race, compared
376
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

to 90.2 percent of the Hispanics on the stop forms with a citation action. The roughly 5 percent
difference is substantial, indicating that it is likely that there is some doubt in the troopers’ minds
as to how to code Hispanics as far as a racial category (are Hispanics “white” or
“unknown/other?”). The possibility exists that a trooper might define an Hispanic as “white” on
the stop form, but as “Hispanic” on the written warning or citation form (typically filled out
within a few minutes of the stop form). We do not have a ready way to verify how often troopers
vary their classification of the race of Hispanics from one form to the next (for example, stop
form to citation form) at the same stop incident.56 The fact that 9.3 percent of Hispanics (defined
so under the NCSHP ethnicity classification) are coded as whites instead of “unknown/uncertain”
makes the percentage of Hispanics in the citation and written warning data bases dubious. In
short, it is unlikely that we will be able to identify the number of Hispanics “mis-classified” as

56

A substantial proportion of the ‘unknowns’ who have citation records according to the
stop data are not Hispanic (21.6 percent), Similarly, 25.4 percent of the ‘unknown/uncertain’
category for written warnings are not Hispanic, so we cannot assume that all
“unknown/uncertain” classifications on the written warning and citation records are actually
Hispanic. It is possible that none are Hispanic since the trooper has a race category option for
Hispanic, which is presumably the most likely category chosen by the trooper for someone who
appears to be Hispanic.
377
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

whites in the present data. At the same time, it seems likely that a small proportion—but a nontrivial number— of whites in the citation and written warning records are mis-classified and
should be defined as Hispanics.

Missing Stop Records by District
The error associated with the classification of a driver as Hispanic also makes it less likely
that we will be able to match all of the stop records unequivocally with the citation or written
warning records. Our primary motive for doing so is to ascertain that any missing records (from
the citation or written warning data bases) include a disproportionate number of African
Americans. In some initial data analysis we attempted to determine how many stop records were
“missing” by comparing to all the stop records for the months January through July, of calender
year 2000. (Since we only have trooper identification information for those seven months, they
are the only months worth considering for matching purposes.)57 However, we found that we
could not match a fairly large proportion of cases between the stop and the citation and/or written

57

A modification of the original legislation required that an identification number be used
by each trooper on the stop form after July, 2000, but that “outside” persons would not have
access to this number. This was interpreted by the NCSHP representatives to mean that we, as
researchers, should not have access to any identification number, even one without an
accompanying list of names. This is not an unreasonable interpretation by the representatives
since we as researchers have information on approximately 97 percent of the stops made by the
NCSHP from August through December of 2000 that could in principle result in the
identification of a trooper. Specifically, knowing the time of day and location of the citation and
written warning records, we could make a reasonably accurate estimate of who the trooper was
who was involved in a stop. As it stands, however, we do not know whether the same trooper
was involved in any two stops for the August through December data because no identification
number of any kind was provided to us. The troopers’ identities are known to us, however, for
the months of January through July, 2000.
378
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Table G.3 “Missing” Stop Records (Comparing non-Accident Citations and all Written
Warnings to the Number of Stop Records Resulting in Citation or Written Warning)
District

Written
Warnings
and
Citation
Events

Stops for
Written
Warnings
or
Citations

District

Proportion
of Written
Warnings
and Cites
“Missing”

Written
Warnings
and
Citation
Events

Stops for
Written
Warnings
or
Citations

Proportion
of Written
Warnings
and Cites
“Missing”

1

18311

11,504

.372

28

12670

10589

0.164

2

14314

12,076

.156

29

12818

8,512

.336

3

12147

8,517

.299

30

20068

15,467

.229

4

11767

8,634

.266

31

22970

16,710

.273

5

9205

6,921

.248

32

10416

7,813

.250

6

8149

6,691

.179

33

12557

8,675

.309

7

26443

15,809

.402

34

7264

4,060

.441

8

11780

8,999

.236

35

15556

9,941

.361

9

15920

11,387

.285

36

12728

8,659

.320

10

20416

14,691

.280

37

17580

10,625

.396

11

20316

14,553

.284

38

26482

16,711

.369

12

17410

12,163

.301

39

9519

7,869

.173

13

11125

9,063

.185

40

10833

4,803

.557

14

11106

7,448

.329

41

14199

9,379

.339

15

18088

14,210

.209

42

14646

9,022

.384

16

10566

8,868

.161

43

14658

8,696

.407

17

34050

21,467

.370

44

11454

8,320

.274

18

16612

10,928

.342

45

16142

11,673

.277

19

13170

9,578

.273

46

10078

7,746

.231

20

13807

8,713

.369

47

13243

10,848

.181

21

21889

15,801

.278

48

8770

5,757

.344

22

9727

7,473

.232

49

19925

12,540

0.371

23

21870

17,634

.194

50

11070

8,591

0.224

24

11013

9,100

.174

51

13688

9,261

.323

25

7964

6,292

.210

52

9343

7498

.197

26

19832

14,328

0.278

53

13434

11172

0.168

27

11886

5,440

0.542

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

379

warning data bases for the months from April through July. We suspect that the high failure rate
is due to the fact that there was no data verification for the identity numbers that we assigned to
troopers for April through July.58 (we, the researchers sent each trooper an identity number to
use). Some troopers may not have used the identification number that we provided (although
most seemed to), or may have been confused about which number to use (our identification
number, their regular identification number) or may have simply forgotten their identification
number (we resent identification numbers to several troopers the first month). Others may have
been careless in recording the correct number. Without basic data verification procedures in
place, the April through July data (when the troopers were supposed to use the identification
number that we provided them) are suspect for matching purposes.
It is interesting to note that the proportion of “missing” written warnings and citations
omitted from the stop data base vary considerably across patrol districts (see Table G.3). The fact
that the proportion of “missing” stop records varies as much as it does from district to district
supports the idea that some troopers— especially in some districts— are less attentive to filling
out the forms. As best we understand, there are no systematic monitoring procedures in place. Of
course, use of checkpoints can also vary across districts, and account for some of the “missing”
stops.
We summed the number of citations and written warnings in each of the fifty-three
districts (excluding citations issued at accidents since there is usually no stop associated with

58

Troopers were mailed an id number that only we, as researchers, knew who had what
number. However, we did not have a way to verify how consistently that the numbers were
being used by every trooper for every stop. Most of the troopers, if not all of them, appeared to
be using a number within a range of numbers that we sent them, but we do not know if they
always used the correct number.
380
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

accidents). Across the fifty-three NCSHP districts, the average proportion of written warnings
and citations “missing” is about 29 percent. (In the table, the districts have been arbitrarily
assigned an identification number, so our district number is not the NCSHP district number.)
Several districts are missing more than a third of their stop records, and one district is missing an
estimated 56 percent of its stop records. The fact that what we estimate to be the proportion of
“missing” stop records varies considerably across districts suggests to us that the lack of standard
data verification procedures results in haphazard reporting of stops in some districts. We doubt
that checkpoint citations and written warnings could account for all the “missing” stops.
To simplify the analysis, in the next section we will focus on only the first three months of
data in calender year 2000. Here, troopers were using their regular identification number, the
same number they use on the written warning and citation forms. Thus, there is a greater chance
that the stop records will match with the citation and written warning records.

Comparison of Stop Written Warning Actions to Written Warning Records
We continue our analysis with a comparison of the stop records indicating that a written
warning was issued (the trooper took the action of issuing a written warning) and the actual
written-warning records. From January through March, we could identify (locate and match on
several characteristics) roughly 96 percent of the written warning records when a stop record
indicated that a driver was given a written warning. That is, we could successfully find the
matching written warning record when the stop record indicated that there should be a written
warning record. We matched the stop-written warning to the written warning record information
for the following variables: district of the stop, identification number of the trooper, month and

381
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

day of the stop, and sex of the driver. A considerably higher failure rate was found (in other
words, fewer matches) when we added race as a criterion for the match. This supports our
suspicion that the classification of the Hispanics was not as consistent as we—as researchers—
would have liked it to be.
The missing 4 percent of written warning records could be due to any number of factors,
including the following: the trooper made an error in recording the district number (perhaps the
trooper recorded his/her official district assignment number instead of the number of the district
in which the event occurred), his or her identification number, the date, or the sex of the driver.
In reviewing the records that we could not match, we noticed that the dates were often different
by a few days. This raises the possibility that the trooper simply forgot the date, entered the
wrong date, or that the trooper was filling out the stop form at a later date and simply used that
date on the form rather than the date of the actual stop. The plausibility of the latter was
reinforced during the focus groups with the troopers, when some troopers indicated that they do
not always fill out the stop forms at the time of the stop. Instead, the stop forms are completed at
a later date by copying the information from the written warning records onto the stop records.
Perhaps when doing so, they use the date the form was completed rather than the correct date of
the stop (creating a paper trail but one not specific to the actual stop date).
One possible data analysis strategy would have been to use the written warning stop
records when we could not find the written warning record. However, there are three reasons
why we did not do this. First, it seems unlikely that the 4 percent missing written warning
records were really missing (as opposed to not matching). There would be little reason for a
trooper to report on the stop form that he or she is issuing a written warning, and then fail to do

382
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

so. There is a large surplus of written warning records that do not have matching stop records,
and it seems likely that the real matching record is, in fact, somewhere in that pool. Second, the
information on the stop form regarding written warnings is generally less useful to us because
there is less information about the location of the event. Thus, these observations would drop out
of some of our subsequent analysis in which the location of the event was a variable in the data
base (they would constitute “missing” data for the analysis). Third, there is virtually no
difference in the racial categorization in those records that match versus those that do not match
across data bases. Specifically, there is only a .1 percent difference in the percent of African
American stop-written warning records missing and the percent white stop-written warning
records missing. Thus, there does not seem to be any bias by race in non-matching stop written
warnings.
More common than missing written warning records were missing stop records indicating
that a written warning had been issued. Approximately 33 percent of the written warning records
in January through March had no identifiable stop record associated with it (note that this is a
much lower rate of missing stop records than we observed above for the whole year, 56 percent).
The 33 percent missing stop records suggests that NCSHP troopers do not always fill out stop
records when issuing written warnings (or the written warning was issued at a checkpoint stop).
From a data analyst’s perspective, it is unfortunate that the NCSHP trooper does not always fill
out a stop record, but the fact that a written warning record was filed by the trooper indicates that
a “paper trail” exists, and we can evaluate those warnings for possible racial disparity.

383
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Missing Citation Records
In addition to being concerned about written warning records that possibly were not being
filled out or processed, we were even more concerned about the possibility of missing citation
records (since citations represent a more severe sanction than written warnings). We were
doubtful that there could be missing citation records (for example, a stop form suggests a citation
was issued, but we cannot locate the specific citation record) because it would seem unlikely that
a trooper would not file a citation if he or she had issued one to a driver. Nevertheless, we
conducted an analysis similar to that for written warning records, matching records that were
identical in district, trooper identification, month, day and gender. We were able to match 95
percent of the stop records that indicated that a citation was issued. Stated another way, we found
that 5 percent of all citation events could be deemed missing by this standard (compared to 4
percent of the written warnings discussed above). We think, however, that—similar to the
argument above for written warnings—our failure to identify a citation record as a match to a stop
record indicating that a citation was issued could be attributed to recording errors of any of the
data elements used to match records. Nor do we find anything but a trivial difference in the racial
breakdown of this 5 percent of cases. We, therefore, did not pursue the possibility of generating
citation records from the stop records (in other words, creating a citation record because we could
not find one that we think was issued based on a stop record), but instead proceeded to analyze
the citation records as they had been given to us.
In summary, our analysis of stop record data indicate that a large number of stop records
do not appear to have been filed by NCSHP troopers. However, there does not appear to be any
justification for a suspicion that significant numbers of written warning and citation records are

384
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

also not being filled out or filed by the troopers. At most, 4 percent of the written warnings and 5
percent of the citations could be “missing,” but we think it is more likely that we simply cannot
identify the appropriate record. The records exist; we simply cannot identify them from the rather
large pool of unmatched records. One thing is certain: a single identifier used across all records
would allow both better monitoring of stop reports and assist researchers in the future.

Accidents
In order to determine if there is any racial disparity in stops, citations, or written warnings,
it would be useful to compare the proportion of such interventions involving African American
drivers with some baseline measure of African Americans violating traffic laws. In the absence
of measures of such law violating behavior, we look to other sources of information to
approximate the measures we lack. We argue that accident reports filed by the NCSHP represent
a useful baseline for comparison of the percent of African Americans who have been issued
citations and/or written warnings. Here, we initially ask the question whether or not the accident
data represent a meaningful measure of all the reported vehicular accidents, and whether
accidents can provide a useful basis for assessing who is driving on the highways or who is
perhaps driving dangerously on the highways.
As for the first concern, in large metropolitan areas it is most often the local police who do
the “paper work” associated with official records of accidents, and they will not be represented in
the NCSHP accident files. (We do not have access to the local police accident files). Of course,
the NCSHP generally do not patrol heavily in metropolitan areas (except for major highways such
as I-85 or I-40), so the exclusion of local accident data may be less relevant to our concerns. That

385
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

is, if the local police process accidents in their primary jurisdiction and the NCSHP processes
accidents in theirs— and there is no overlap, then the NCSHP accident data could be a reasonably
complete source of information on highway accidents.59 Unfortunately, we do not have a means
to measure jurisdictional patrol overlaps, so there will be some doubt about the completeness of
the NCSHP accident data for our purposes.
Table G.4 shows some correlations between three measures obtained from the Automobile
Association of America’s (AAA) 1999 data on driving miles and accidents in North Carolina, and
four measures of accidents from the NCSHP data files (AAA, 2001). The most recent year for
which such AAA data are available is 1999, and we assume that the 1999 data would be highly
correlated with the 2000 data. The first AAA variable is a measure (in thousands) of vehicular
miles traveled per county (here the 100 counties in North Carolina have been aggregated into the
fifty-three NCSHP districts). The second measure is the number of collisions per 10,000
vehicular-miles-driven. It should be noted that these collisions are for all jurisdictions in North
Carolina, not only those of the NCSHP. Thus we would expect the correlations between the
NCSHP accident data and the data for the whole state to be somewhat attenuated (and they are).
The third measure is the number of collisions (again across all jurisdictions), and the fourth
measure is the ratio of accidents to injuries (included to show that there is variation in the severity
or seriousness of accidents relative to driving prevalence).
As can be seen in the correlation matrix in Table G.4, there is a very high correlation
between the number of collisions and the number of VMT (vehicular miles driven)— .965. Thus,
much of the variation in collisions across districts can be attributed to the sheer volume of traffic.
59

We would also have to rule out accident reports being filled out by each county’s
Sheriff Department.
386
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

The moderately high correlation of collisions per 10,000 VMT with VMT indicates that collisions
are more prevalent in high VMT areas independent of the sheer volume of the traffic, suggestive
that the density of the traffic may be responsible for the high collision rate, although other safety
related factors may be also involved.
The NCSHP data base on accidents provides us with a count of the number of accidents
per district. It would be expected to correlate more weakly with the VMT because the
jurisdictional area of the NCSHP is a subset of all the areas within a district (VMT is for the
entire area, including the metropolitan areas that are not usually patrolled by NCSHP). The
correlation of .585 indicates that there is, nevertheless, a moderately strong correlation between
the NCSHPs’ count of accidents and the vehicular miles driven. The number of people injured in
NCSHP-recorded accidents correlates .632 with VMT and .580 with the number of accidents
(.511 with number of collisions).
The number of people injured in accidents is not strictly a linear function of the number of
accidents. In fact there is much variation in the ratio of accidents to injuries, varying from 1.68 to
12.47 with a mean of 5.907 and a standard deviation of 2.94. Generally the ratio is higher where
there is more traffic and a higher rate of collisions per VMT. One could speculate that the
“fender bender” type of accident is more prevalent where traffic is heavier. Where traffic is
relatively light, an accident has a higher chance of involving an injury than where the traffic is
heavy. As for deaths, there is virtually no correlation between the number of deaths and any of
the other variables except number of injuries. The number of deaths on NCSHP highways is
relatively rare, compared to the traffic volume or number of accidents (the mean number of deaths
per district in 2000 is twenty-four). The relative rarity of highway deaths suggests to us that the

387
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

number of deaths is quite random and largely a product of such other considerations (not
measured here) as promptness of ambulance service, proximity to hospitals, or highway
engineering factors.
Therefore, we conclude from this preliminary analysis of the NCSHP accident data that
there is a plausible foundation for using such data as a basis for further comparisons. For
example, we could use the accident data to compare the percent of accidents with African
Table G.4 Correlations Between Accident Measures*
VMT
(AAA)

VMT

Collisions
Per 10000
VMT
(AAA)

Collisions
(AAA)

NCSHP
Recorded
Accidents

NCSHP
Recorded
Deaths
Due to
Accident
s

NCSHP
Recorded
Injuries

Ratio
of
Accidents
to Injuries
(NCSHP)

1

Collisions
Per 10000
VMT

0.624

1

Collisions

0.965

0.743

NCSHP
Recorded
Accidents

0.585

NCSHP
Recorded
Deaths

.059
n.s.

NCSHP
Recorded
Injuries

0.632

0.383

1
0.527

1

-.112 n.s.

-0.05

.216 n.s.

1

0.354

0.511

0.58

0.618

1

Ratio of
0.289
0.235
0.293
0.829
-.072 n.s. 0.092
1
Accidents
to Injuries
*All correlations above are statistically significant except where noted as non-significant (n.s.).

388
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

American drivers to the percent of those cited who are African American. We discuss further the
validity of using accident data in the section below on the observational baseline study in which
we observed the race and speed of motorists on fourteen select highway segments.

Baseline Observational Study
In an attempt to determine if there is variation in the speeding of motorists on highways in
North Carolina, we observed the race of passing drivers while ourselves driving at the speed limit
and timed the speed of those drivers with stop watches. The details of what we call our baseline
observational study can be found in Appendix A. Of importance here is the fact that we have
gathered some data on fourteen highway segments (both directions of traffic for stretches of
highway that are 10 to 15 miles in length, one-way), selected because they are areas where the
NCSHP stops many vehicles (both white and African American drivers). The observational study
is a modification of a method pioneered by Lamberth, but in many ways is unique. For example,
we actually measure the speed of the passing vehicles. Also, we measure the location so as to
compare data with stop and citation data from the same or proximate area. Since we were the
first to attempt this method, there are improvements that could be made to the study— primarily,
that it would have been helpful to have gathered more data at each of the fourteen locations. At
the time we did the study (May and June, 2000) we did not know as much about speeding and
stops of speeders as we know today. The implication of this understanding is that the results from
our baseline observational study are somewhat ambiguous as to the possible presence of racial
disparity in the stops and citations on the fourteen highway segments studied. The method,

389
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

however, has been proven useful, in our opinion, to the further study of vehicular driving
behavior as well as policing behavior.
One of the facts that we learned from the baseline study—after we had collected the
data— is that many highways have different “speeding thresholds”— that is, speeds that are
likely to result in a pullover and citation. Importantly, these speeding thresholds are quite high
relative to our expectations going into the study. These thresholds were roughly twelve to fifteen
mph above the posted speed limits, whereas we had expected that seven to nine mph above the
posted speed would be likely to trigger a patrol stop and citation. Thus, when we learned that a
driver on any of the fourteen highway segments had nearly a zero probability of being cited for
driving seven to nine mph above the speed limit, we were surprised. As a consequence of these
considerations, we did not find as many speeders (threshold speeders) as we thought we were
finding when we collected the data. One unfortunate consequence of the high threshold speeding
values, is that we cannot conduct a rigorous comparison of the speeding and ticketing of African
Americans in the same geographic areas— there are too few “real speeders” (threshold speeders)
to do so. However, we will discuss the general pattern of findings below, which indicates that,
overall, there are slightly more African Americans stopped and cited for speeding in the fourteen
areas than there are African Americans found speeding. The differences observed, however, are
generally not statistically significant across the sites.
In Table G.5 below, we present some of the findings from the observational baseline study
of each of the fourteen sites. Note that these fourteen sites represent a convenience sample of
locations in North Carolina. To be included in the study, a highway segment had to have four
lanes of traffic (for safety reasons— see discussion in Appendix A) and had to be among the

390
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

highways with a high number of citations in the early months of the year 2000.60 Because the
sample is a convenience sample, one should resist the temptation to generalize from our sample of
fourteen sites to other four-lane highways or highway segments. We do not have a simple
random sample of locations, so generalizing to other geographic areas would be hazardous.61 As
can be seen in Table G.5 the percent of African Americans who speed is examined relative to two
standards. The more conservative standard is the speeding above the median local threshold
(defined as the median mph above the posted speed limit from those citations issued in 2000 on
the same highway [for example, I-95] in the same county area [roughly one fourth of a county]).62
A more liberal threshold value would be the speed at the first decile of the distribution of those
cited (a decile represents 10 percent of the observations, so the first decile represents the speed at

60

Note that the stop data includes milepost markers as one of the variables, thus allowing
us to identify segments of a highway with relatively many stops. It should also be noted that we
checked to see that there were numerous stops of African Americans as well as of whites in these
segments, although we did not formalize this process by requiring that a certain number or
proportion of those stopped be African American.
61
We did not have sufficient data or information about speeding to draw a simple
random sample of areas. A simple random sample of segments of four-lane highways would
yield mostly highway segments where there are too few stops or citations to compare statistically
to the observed speeding. We also did not feel confident that the stop data base had all of the
stops (as per the discussion earlier in Chapter Two), so that a rigorous sampling of areas with
high rates of stops would be inappropriate. The fourteen sites represent fourteen of the top 60
highway areas in the state in terms of number of stops. Most of the areas not selected were two
lane highways, or in highly traveled areas around large cities (where we thought it unsafe to
drive and collect data on speeders) or too far from Raleigh, where our research team was based,
such that we could not afford to drive to the site every day for a week.
62
In should also be noted that for the non-Interstate highway segments we generally do
not have mile post data, so the location of the stop is only approximate (we know the highway,
county and area of the county in which the stop took place, but not exactly the 10 to 15 mile
segment of the US or NC highway). It is only for the Interstate highways that we have mile post
data and thus know within one mile where the stop occurred. It should be acknowledged,
however, that we have no way to check on the validity of the mile post marker information. If a
trooper mis-records the mile post where the stop occurred, we would not have a means of finding
such an error unless the milepost indicated is in a different police district.
391
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

which Table G.5 Various Measures of “Threshold Speeding” by Highway Segment,
Observational Baseline Study
Segment

Posted
Speed
Limit

Median
Threshold
Speed
(MTS)

Percent
Over
MTS
who are
African
Americ
an

First
Decile
Threshold
Speed

Percent
Over
Decile
Threshold
Speed
who are
African
America
n

Percent
Stopped
for
Speeding
who are
African
America
n

Percent
in Accidents
who are
African
American
(Highway
Area)

I-95-N

70

82

22.9

79

14.6

29.4

21.6

I-95-J

65

80

25.4

78

17.6

27.3

27.2

I-95-H

70

80

25

79

21.5

33.8

30.9

I-85-G

65

80

55.9

79

33.9

40.6

41.5

I-85-W

65

80

40

79

22.8

36.8

31.4

I-85-V

65

80

47.6

78

34.3

40.5

32.1

I-85-RD*

55 & 65

70 & 80

12.5*

69&78

17.7*

25.0*

28.1

I-40-J

70

85

15.4

79

14.7

14

15.3

I-40-P

70

84

15.2

79

11.6

14.1

15.7

US-L

55

70

20

69

22.2

22.0**

28.4

US-C

65

80

12.2

77

14.9

19.5**

14.8

US-G

55

70

26.6

69

24.8

26.6**

28.3

US-N

65

80

42.3

77

31.3

44.9**

35.9

NC-G

65

80

21.1

77

10

28.9**

31.6

*Two posted speeds across this highway segment. Construction at site reduced speeding, so N of
MTS is very low (16).
** = estimates based on the highway of the observational study for the entire area of county.
Since there were no mile posts on these highways, it was impossible to identify stops in the
highway segments studied. Estimates of percent African American are generally higher using the
county area data than using only the highway segment data (by a percentage or two).

392
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

10 percent of the vehicles are stopped and cited). Comparing the two threshold speeds indicates
that African Americans are somewhat over-represented among those who are driving at or above
the fifty percentile speeding threshold within the fourteen sites.63 Note that we do not have
available to us the speeds for which drivers are stopped, only the speeds for which drivers are
issued tickets. Thus, it would be reasonable to assume that drivers are stopped at speeds
generally lower than the speeds for which troopers issue citations. Many drivers stopped for
speeding are given written warnings (roughly a third), and for those drivers we do not have a
recorded speed. Also, we have no direct measure of the risk of being stopped at a given speed,
only evidence on the speed information of cited speeders.
The table may be read as follows: in highway segment I-95-N, where the speed limit is 70
mph, there is a conservative (or high) estimate of the speed that will result in a stop (82 mph) and
a liberal (or low) estimate (79 mph). Half or more drivers are issued citations for speeding at the
conservative threshold, and 10 percent or more are cited for speeding at the liberal threshold.
(“Conservative” here means that we are probably underestimating the risk of being stopped and
cited on I-95-N if one is driving 82 mph, and “liberal” means we are over-estimating the risk of
being stopped and cited if one is driving 79 mph.)
The percentage of drivers observed speeding at or above the conservative speeding
threshold varies across segments— from a low of 15.2 in segment I-40-P, to a high of 47.6 in I63

Generally the first decile value and the median value were quite close, so that there is
only small differences in the percent African American using the two definitions. That is
because there seems to be a “magic” number for each highway area, and often only one mph
difference between the first decile cut-off point and the median cut-off point, e.g, 14 mph more
than posted speed is the first decile value and 15 mph the median value. In other words, 10
percent of those cited were cited for traveling 14 mph more than the posted speed, and 50
percent (or more) were cited at 15 mph more than the posted speed. This was a common finding
across highway segments.
393
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

85-V. Using the more liberal speeding threshold measure, we see that the percent of speeders
who are African American varies from a low of 10.0 in NC-G, to a high of 34.3 in I-85-V. These
results are both interesting and disappointing. We probably are measuring the speed of vehicles
accurately within (plus or minus) 2 mph— see discussion in Appendix A. Yet, the difference
between our conservative and liberal estimate of threshold speed is between 1 and 3 mph. The
difference that a couple of miles per hour makes in terms of the risk of being stopped is quite
large (for that matter, when using radar equipment to measure speed, there is a margin of error
within 1 mile per hour, therefore indicating that there is presumably more error in our
measurement of speed than in a trooper’s). On most of the highways we studied, as the term
“threshold” implies, there is a substantial change in risk of being stopped that occurs as one
approaches speeds of 80 mph. Traveling 77 mph on some highways elevates the risk
substantially; for other highways one needs to be driving 79 mph to incur relatively high risk. By
80 mph, the driver’s risk is substantial on all but three of the observed highways. On those three
highways, substantial risk does not occur until traveling 82, 84, or 85 mph (I-95-N, I-40-P, I-40-J,
respectively).
One of the implications of these numbers is that we would have to be measuring speed
exactly (within .5 mph) to know precisely what percentage of African Americans we should
expect would be stopped for speeding on each of the highway segments. Instead, we can only be
approximate in our estimates. The conservative and liberal criteria initially proposed here suggest
that a percentage of African Americans somewhere between 14.6 and 22.9 percent could be
expected to be stopped for speeding on I-95-N. This is a rather large interval, pointing again
toward the fact that there is some error in measurement that prohibits us from making strong

394
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

claims as to racial disparity in our baseline observational study. As we will discuss below,
however, we think it is even more complicated than that.
The table also shows that there is some association between the percentage of drivers in
accidents who are African American and the percentage driving above the speeding thresholds
(both conservative and liberal). Note that the comparison is imperfect regarding the geographic
area covered. NCSHP accident data do not include mile post data, so we can only compare the
observed speeding behavior within the highway segment with the accident behavior in what we
call the highway area (in other words, all the accidents on the same highway within about a
quarter of a county).64 The actual correlations across the fourteen sites are .793 and .731 between
the percentage of drivers in accidents who are African American and the percentage of threshold
speeders who are African American (conservative and liberal thresholds, respectively). Omitting
the I-85-RD segment (because of low reliability) raises the correlations somewhat—to .842 and
.739 (conservative and liberal thresholds, respectively).
These correlations are encouraging in that there are reasonably high correlations between
a potential baseline measure (percent of accidents with African American drivers) and measures
of police behavior (percent of citations issued to African Americans). The correlation between
the percentage of drivers in accidents who are African American and the percentage stopped for
speeding who are African American is .865 (.874 if I-85-RD is omitted). Thus, there is a slightly
higher correlation between the percentage of African American drivers in accidents and the

64

This comparison of a highway segment with what we call a highway area is generally
comparing a 10 to 15 mile stretch of highway with a longer stretch of highway that includes the
10-15 mile segment. The longer highway we estimate to generally be two to three times longer.
Also, it should be noted that we are counting all accidents from 1998 through 2000 in the
highway
395
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

percentage stopped for speeding than there is between the percentage of threshold speeders who
are African American. Note here that the number of highway segments studied is only fourteen
so the correlations are highly susceptible to fluctuations. In any event, there is an empirical basis
for expecting a reasonably strong relationship between accidents and NCSHP intervention in an
area.
In our analysis, we looked at the stop records for May through July, 2000, for daytime
hours—approximating the hours that we observed speeders (weekdays, 9 a.m. to noon and 1 p.m.
to 5 p.m.)— and found that, for example, 29.4 percent of the drivers stopped for speeding on I-95N (see table) were African American, a disparity in the direction of bias against African
Americans. In seven of the fourteen districts, relative to the conservative threshold, disparity in
the direction of hypothesized bias is found, but in five districts there is more speeding observed of
African Americans than there are stops of African Americans (one district has unreliable data,
and one district is a tie). Using the liberal risk of speed threshold, we find that all but two
highway segments (U.S.-L and I-40-J) have fewer African Americans speeding above the
threshold than those who are stopped.
It should be noted, however, that we do not know much about the selection process by
which troopers decide to pull over vehicles for speeding. For example, it is likely that not every
driver observed speeding at the threshold value gets stopped, but those that speed well above the
threshold values are more likely to be stopped. We would need data on what troopers actually
observe to make a better assessment of the real threshold values that trigger stops. Such values
may vary with how busy the trooper has been that day, or how long it has been since a trooper has

396
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

pulled over someone. It seems likely, however, that the more grievous the speeding violation, the
higher the risk of someone being stopped.
Figure G.1 Percent of Drivers Speeding More Than the Speeding Thresholds by Race

397
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

In Figure G.1 we show that, across all fourteen highway segments of the observational
study, African Americans make up a higher proportion of the “threshold speeders” (up to 30
percent) than they make up of all drivers on the highway (20.2 percent).65 Threshold speeders
here is defined three ways: low threshold of speeding (at or above the first decile of speeds at
which drivers are cited), high threshold speeding (at or above the median at which drivers are
cited) and a flat fifteen mph above the posted speed limit. (we added the 15 mph above posted
speed as a threshold because it is a standard used in a New Jersey study— Lange et al., 2001).
For example, we see that African Americans make up approximately 30 percent of those traveling
7-8 mph above all three speeding thresholds, while they make up only 20.2 percent of those
traveling on the highway (the flat reference line at approximately the 20 percent mark on the
vertical-axis). We can see that African Americans are over-represented among those who are
speeding above the point at which speeding citations are likely. We can say that this pattern is
“robust” across definitions of speeding thresholds in the sense that the pattern is the same
regardless of how the thresholds are defined. At the top of the graph are three lines representing
the percent white of the total number of all white and African American drivers. Thus, the
percent white represents a complement to the percent African American (the two percentages
must sum to 100).
Note, however, that the proportion African American among the threshold speeders does
not rise continually across the speeding values on the horizontal-axis. Thus, the pattern is
dissimilar to that shown by researchers in New Jersey. They found that African Americans made
up a progressively higher percentage of the speeders at the higher speeds (Lange et al., 2002).
65

The percent reflects a count of all drivers observed who were African American or
white. Drivers judged to be Hispanic or Other are excluded from the computation of the rate.
398
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

Here, however, we found that the pattern is not a steady incline but rather a curvilinear pattern of
initial incline (up to 7-8 mph above the speeding thresholds), and then a decline to where the
proportion of threshold speeders who are African American approaches roughly 20 percent (the
same percent of drivers observed on the highway who are African American). Note that Lange
and colleagues did not present measured speeds above 89 mph, whereas we are presenting speeds
up to 96 mph (at least where the speed limit is 70 mph and the speed threshold is 85 mph). It
should be noted, however, that we do not measure speeds very accurately above 90 mph.66 Thus,
it is possible that the Lange data, if presented for higher speeds, might show a pattern similar to
ours.

66

Note that Lange and colleagues present their data only for a speeding threshold of 15
mph above the posted speed limit. They report a continual increase in the percent of speeders
who are African American up to 89 mph. We use three different speed threshold measures, and
all three show the same pattern: a decline in the percent African American after 7-8 mph above
all three speeding thresholds. The value of “11 or more” mph above the speeding threshold for a
15 mph threshold is, of course, equivalent to 26 mph above posted speed (which varies from 55
to 70 in our observational study). Thus, the highest speed that could be included in our graph is
96 (70 + 15+11).
399
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

References for Appendix G
AAA Automobile Association of America. 2001. Report on Crashes in North Carolina Counties,
1999. On the web site for AAA of North Carolina in 2001. Site has been removed.
Lange, James E., Kenneth O. Blackman, Mark B. Johnson. 2001. Speed Violation Survey on the
New Jersey Turnpike: Final Report. Report Submitted to the Office of Attorney General,
Trenton, New Jersey.

400
This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.

This document is a research report submitted to the U.S. Department of
Justice. This report has not been published by the Department. Opinions or
points of view expressed are those of the author(s) and do not necessarily
reflect the official position or policies of the U.S. Department of Justice.