The author(s) shown below used Federal funding provided by the U.S.
Department of Justice to prepare the following resource:
Document Title:

Estimating the Prevalence of Wrongful
Convictions

Author(s):

Kelly Walsh, Jeanette Hussemann, Abigail
Flynn, Jennifer Yahner, Laura Golian

Document Number: 251115
Date Received:

September 2017

Award Number:

2013-IJ-CX-0004

This resource has not been published by the U.S. Department of
Justice. This resource is being made publically available through the
Office of Justice Programs’ National Criminal Justice Reference
Service.
Opinions or points of view expressed are those of the author(s) and
do not necessarily reflect the official position or policies of the U.S.
Department of Justice.

Estimating the Prevalence of Wrongful Conviction
Grant Award #: 2013-IJ-CX-0004

Summary Technical Report Authors:

Kelly Walsh, Jeanette Hussemann, 

Abigail Flynn, Jennifer Yahner, Laura Golian 


Inquiries should be directed to: 

KWalsh@urban.org

202-261-5434
Kelly Walsh c/o Urban Institute 2100 M St. NW
Washington, DC 20037

This project was supported by Award No. 2013-IJ-CX-0004 awarded by the National Institute of
Justice, Office of Justice Programs, U.S. Department of Justice. The opinions, findings, and
conclusions or recommendations expressed in this publication/program/exhibition are those of the
authors and do not necessarily reflect those of the Department of Justice.

This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Abstract	
This study extends research on wrongful convictions in the U.S. and the factors associated with
justice system errors that lead to the incarceration of innocent people. Among cases where
physical evidence produced a DNA profile of known origin, 12.6 percent of the cases had DNA
evidence that would support a claim of wrongful conviction. Extrapolating to all cases in our
dataset, we estimate a slightly smaller rate of 11.6 percent. This result was based on forensic,
case processing, and disposition data collected on murder and sexual assault convictions in the
1970s and 1980s across 56 circuit courts in the state of Virginia. To address limitations in the
amount and type of information provided in forensic files that were reviewed in the Urban
Institute’s prior examination of these data, the current research includes data collected through a
review of all publicly available documents on court processes and dispositions across the 714
convictions, which we use to reassess prior estimates of wrongful conviction.

1
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view 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	
Over the past two decades, more than 300 people in the U.S. have been exonerated of crimes that
occurred in the pre-DNA era (Innocence Project, 2016). As a result, researchers and practitioners
have become increasingly interested in identifying the factors and processes that lead to the
incarceration of innocent people. To date, this research body has offered a limited understanding
of factors associated with justice system errors, and widely divergent views on the prevalence of
wrongful convictions.
This research builds upon the 2012 Urban Institute report, “Post-Conviction DNA Testing and
Wrongful Conviction,” (hereafter called “Part I”) which presented an estimated rate of wrongful
conviction based on post-conviction DNA testing of over 700 felony convictions in the state of
Virginia between 1973 and 1987 (Roman et al. 2012). These 2012 findings relied primarily on
original forensic testing documents (e.g. evidence submission forms, serology reports) and postconviction DNA testing reports to identify convictions that were strengthened or weakened by
new DNA evidence. These documents contained data on victim and suspect demographics, case
progression dates, original forensic collection and testing, and post-conviction DNA testing.
However, these data were inconsistently available for all cases, and the documents did not
include more detailed information on the criminal justice processes that led to the conviction
(e.g., legal representation, case disposition, motions to appeal).
In order to address this gap and supplement the 2012 findings, Urban was awarded funds in 2013
to visit Virginia circuit courts and review case processing records for these felony convictions.
The additional data collected as part of the current study (Part II) provide a new foundation by
which to reclassify case outcomes and calculate an estimated rate of wrongful conviction for
similar convictions in Virginia during the 1970s and 1980s. In an effort to examine the external
validity of this estimate, we additionally analyzed 1985 felony conviction data from 43 states to
determine whether Virginia conviction rates were similar to other states during the time period in
which this post-conviction DNA testing occurred.
This report summarizes the methods and findings for estimating the prevalence of wrongful
convictions. Additional findings on the correlates of wrongful conviction will be presented in
future peer reviewed publications.

Key	Terms	
Throughout this technical summary, we rely on the following terms to discuss the outcomes of
DNA testing and the impact of that testing on the strength of the original conviction, as reported
in Part I:

2
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

-

Determinate	(Outcomes	2,	3,	and	4): Allows a conclusion to be drawn as to whether the

person convicted was a possible source of the DNA developed from the original
evidence.
-

Indeterminate	(Outcome	1): No new DNA evidence was developed in the case or no

conclusion can be drawn about the source of DNA.
-

Inculpatory	(Outcome	2): Describes DNA evidence that adds strength to the assertion that

the person convicted committed a criminal act.
-

Exculpatory	but	insufficient	for	exoneration	(Outcome	3): The DNA evidence that

excludes the person convicted as the source of the DNA, but which does not support a
claim of wrongful conviction due to the context of the case and old evidence (i.e., its
probative value).
-

Exculpatory	and	supportive	of	exoneration	(Outcome	4): The DNA testing excludes the

person convicted as the source of DNA developed from old evidence. Given the context
of that old evidence in the case, this result would support a claim of wrongful conviction,
though it may not be sufficient to prove wrongful conviction.

Prior	Research	
The design and findings of studies of wrongful convictions vary widely. Prior research has relied
on data collected through self-reporting by convicted individuals (Poveda 2001) and interviews
with criminal justice professionals (Ramsey and Frank, 2007; Zalman et al. 2008) across diverse
populations and types of convictions. Estimates of the prevalence of wrongful conviction range
from as low as 0.027 percent (cited in Justice Scalia’s concurrence in Kansas v. Marsh) to as
high as 37.7 percent (Poveda, 2001), though the most widely accepted estimates range between 1
and 5 percent (Gross, Hu, Kennedy, and O’Brien 2014; Gross 2008; Radin 1964; Gould and Leo
2010; Ramsey and Frank 2007; Zalman 2012). However, most of these studies contain sample
bias they are unable to correct for, which may affect the validity of these estimates (e.g., bias
due to the convictions examined, type of data collected, or how a potential wrongful conviction
is determined). For example most studies are not able to account for the process by which DNA
testing of evidence occurs or for variation among forensic practitioners.
Notably, the results presented in Urban’s 2012 study (Part I) come from the first effort to apply
post-conviction DNA testing to a large set of convictions, regardless of any existing claims or
evidence of wrongful conviction. The Part I retrospective study focused exclusively on 715
murder and/or sexual assault convictions disposed in Virginia between 1973 and 1987, for which
3
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

biological evidence from the original case was found. While almost all forensic evidence had
been destroyed in accordance with commonwealth policy at the time, one forensic serologist,
Mary Jane Burton, and the individuals whom she trained, physically attached biological evidence
(including swabs and cuttings) to hard copy laboratory files. Because this evidence was not
stored with other case evidence (e.g. weapons or knives), these potential sources of DNA were
not destroyed and have since aided exoneration cases in Virginia. Importantly, because DNA
evidence and testing was assigned at random to the serologists (as reported in Part I), these 715
cases provide a nearly unbiased sample of convictions from the time.
Urban’s Part I data collection and reporting relied on a review of information included in DNA
files provided by the Virginia Department of Forensic Science (DFS), including basic case
attributes, original forensic testing, post-conviction DNA testing, and some case outcomes. After
data were collected for each of the 715 convictions, the research team determined through
consensus whether the evidence was inculpatory of the convicted suspect’s guilt, or either
supportive of or insufficient to support exoneration. Depending on the determination, all
convictions were coded as indeterminate (Outcome 1), inculpatory (Outcome 2), exculpatory but
insufficient for exoneration (Outcome 3), or exculpatory and supportive of exoneration
(Outcome 4).
The Part I study reported that either 5% or 8% of convictions in homicide and/or sexual assault
cases were wrongful. These findings were important for their rigorous method of producing an
estimate of the prevalence of wrongful conviction from a relatively unbiased sample. However,
the Part I conclusions were based almost exclusively on information in the DFS files, which did
not necessarily provide all the relevant case file information needed to make an accurate
determination regarding the original conviction.

Rationale	for	Research		
This Part II study represents an effort to overcome the limitations of Urban’s Part I study by
reviewing all publicly available files from over 50 Virginia Circuit Courts to collect case
processing and disposition data for each conviction included in Part I. Key research questions for
this project include:
1.	 What is the prevalence of the four conviction-level outcomes identified in the first study
(i.e., the DNA evidence was: inculpatory, exculpatory but insufficient for exoneration,
exculpatory and supportive of exoneration, or indeterminate)?
2.	 What case, victim, and convicted person attributes are correlated with these outcomes?
3.	 How do the Virginia conviction rates compare to other states?
4.	 What is the utility of post-conviction DNA testing as a tool to detect wrongful 

convictions? 

4
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view 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 technical summary focuses specific attention to the methods and results relevant to research
questions 1 and 3.1

Methods	
Data	Collection		

This research relies on two data sources: primary data collection of information in original case
files across 56 courthouses, and a review of existing data available from the Bureau of Justice
Statistics Census of State Felony Courts, 1985.
Courthouse	 Data. Conviction-level data was sought from 56 circuit courthouses in Virginia.

Courthouses were included in this study if they housed at least one file from the Part I data set that
had an exculpatory outcome (Outcome 3 or 4). If so, all cases were requested for review, regardless
of outcome. Most case files were coded at the courthouse. In total, the research team visited 31
courthouses in person; 18 courthouses agreed to photocopy and mail existing files; and 7
courthouses confirmed that the files no longer existed in either on- or off-site storage locations.
Urban’s data collection instrument included 67 variables which covered information on the
person convicted, the offense, and court processing and disposition. The instrument was
developed based on case file information observed at three courthouses (Arlington, Alexandria,
and Fairfax), and was then piloted at two courthouses across multiple coders prior to full
implementation to ensure that the instrument was inclusive and reliable. Because court identifiers
were not included in the Part I data set, convictions were matched using the convicted person’s
name, victim name, and the date and type of offense. Primary sources of data available in the
courthouse files included sentencing and presentence reports, trial transcripts, witness summons,
correspondence between involved parties, and other documents.
Trial and hearing transcripts, when available, were a rich source of information. The testimony
provided information about the circumstances of the crime or potential mitigating factors. This
sometimes included if the victim and defendant knew each other, information on prior
convictions, the presence of an alibi, sentencing decisions, and the number of witnesses that
testified for each side. Each one of these elements were captured in project variables.
Case file storage policies, and therefore data availability and completeness, varied across the 56
courthouses included in this study. Data availability ranged from absence of a case file to full
cases that included multiple forms and documents. Detailed hard-copy case files were available
for 432 convictions while partial files (e.g. sentencing reports only) existed for 131 convictions.
                                                            
1

Questions 2 and 4 will be addressed in future articles submitted for peer reviewed publication.

5
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

For 81 convictions in our data set there was no file available.2 In total, we collected data on 47
convictions with exculpatory evidence (i.e. Outcomes 3 or 4) and 517 convictions with either
inculpatory or indeterminate evidence (i.e. Outcomes 1 and 2).
Based on a review of the VA post-conviction DNA testing effort history, the research team
decided to include five observations in the dataset that were not included in Part 1: exonerations
resulting from post-conviction DNA testing of similar Mary Jane Burton files that did not have
DFS files available for review during Part 1. These were exoneration cases that preceded, and
inspired, the broader VA effort to test all cases with remaining evidence. According to the
conviction offense and conviction year we determined that these five cases would have been
eligible for our study if they had not already been tested prior to this research effort. These five
observations were all classified as Outcome 4 (having exculpatory evidence supportive of
exoneration). There is no indications that other cases, which did not result in exoneration, were
tested and should have been included in our data set. The research team conducted internet
searches on these 5 additional convictions to fill in coding variables.
Bureau	of	Justice	Statistics	Census	of	State	Felony	Courts,	1985. Bureau of Justice Statistics data

was acquired to test the generalizability of the findings of this study. If, for example, it turns out
that Virginia courts were much more or less successful than other states in securing felony
convictions, then the results would have limited external validity. To assess generalizability, we
use data from 1985, which is in but near of the end of the time period covered in our research.
The BJS data report the number of cases filed, convicted, and dismissed across jurisdictions in
the 50 states.
Data	Analysis	
Courthouse	Data	and	Case	Reclassification. To assess the impact of the additional data collected,

we reviewed all Outcome 3 and Outcome 4 cases, as coded in Part I, to determine if they should
be reclassified to a new outcome category. This type of reclassification would have a direct
effect on the estimated rate of wrongful conviction. In total, we reviewed 46 exculpatory case
files, including 15 Outcome 3 cases and 31 Outcome 4 cases.3 Information reviewed to
                                                            
2

This total does not include 15 convictions that were originally selected to be part of the sample. The research team
decided to exclude 11 convictions since they were indeterminate and were located at courthouses that were not
included in courthouse data collection. Six additional convictions were excluded from both courthouse data collection
and the final dataset. One of these convictions was found to be a duplicate of another case. Another five were found
to not have resulted in a conviction—the defendant was found not guilty of the charges or the charges were dismissed.
Finally, five additional observations were added to the dataset to ensure inclusion of exonerations that were a part of
the set of convictions this study focuses on but did not have an available DFS file in Part 1. As such, the N for this
study (714) is different than the previous study (715).
3
We were unable to gain access to case files for three Outcome 3 and seven Outcome 4 convictions. Convictions
without new information coded during this phase of coding did not have their Outcome classification reviewed or
revised. We also did not review the five new observations added to the dataset, since each of those cases resulted in
exoneration. 

6
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view 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 case should be reclassified included the offense location, victim/suspect
relationship, and presence or absence of confessions and alibis. A group of five researchers on
the project team reviewed each conviction independently and then reached consensus on whether
to keep or revise the Part I outcome classification (e.g., shift from exculpatory and supportive of
exoneration to exculpatory and insufficient for exoneration) for each conviction. Outcomes were
reclassified if courthouse coding provided new information that created a probable explanation
for the absence of the convicted person’s DNA and/or the presence of another individual’s DNA
other than an actual wrongful conviction.
Following the reclassification process, the research team calculated a new estimated rate of
wrongful conviction. Building off a conclusion reached in the Part I study, the research team
focused only on convictions that had a sexual assault component because very few homicideonly convictions yielded determinate post-conviction DNA outcomes (n=20). Of the 430 cases
involving a sexual assault component, 231 yielded determinate post-conviction DNA outcomes.
The estimated rate of wrongful conviction in this sample was arrived at by dividing the number
of Outcome 4 convictions among these cases by the total number of convictions with a
determinate outcome and sexual assault component.
This observed rate was then statistically adjusted through inverse probability weighting (IPW) to
correct for differences between determinate and indeterminate cases with a sexual assault
component based on the age of the case (cases less than 40 years old were more likely to yield
determinate DNA evidence than those 40 years and older) and on the nature of the conviction
offense (cases involving a rape or sexual assault conviction were more likely to yield determinate
DNA evidence than those for which murder was the most serious conviction). IPW methods
provide an approach to correcting for non-representation by weighting sample members with
determinate DNA outcomes to have the same distribution of the two key covariates as the full
population of cases. (Hirano, Imbens, and Ridder 2003; Wooldridge 2002).

7
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Bureau	of	Justice	Statistics	Census	of	State	Felony	Courts,	1985	We use data from the 1985 Bureau

of Justice Statistics Census of State Felony Courts to test the generalizability of the estimated
rate of wrongful conviction. This dataset reports the number of felony cases filed in 1985 in
jurisdictions across 50 states, and measures the proportion of cases that result in various
dispositions, including convictions. While this dataset is useful as a tool to examine variation in
the proportion of felony cases filed across states, several limitations to the dataset affect our
ability to rigorously analyze the external validity of our Part II findings. First, there is variation
across states by the unit in which the number of cases filed and disposed are reported.
Jurisdictions in 69 percent of states, including Virginia, rely on a defendant-based unit of count.
In these jurisdictions, cases are reported by defendant, such that a defendant with three charges is
counted as one case. Jurisdictions in twenty-one percent of states rely on a charged-based unit of
count, such that a defendant with three charges would be counted as three cases. The remaining
jurisdictions rely on an indictment-based unit of count, in which a defendant with three charges
is counted as one case, but so too are two defendants involved in the same case. The second
limitation to this data set is the occurrence of missing data. In particular, Virginia did not report
the number of cases disposed by trial, guilty plea, or acquittal. Finally, this dataset does not
differentiate between types of crime, which limits our ability to focus specifically on the filing
and disposing of murder and/or sexual assault cases.
Given the limitations to the Bureau of Justice Statistics data, our analysis and results focus
broadly on felony cases. To ensure that the unit of count is consistent across data points, we
focus only on the jurisdictions across 43 states that rely on a defendant-based unit of count.4
Finally, because we are not able to analyze whether a case is disposed through conviction due to
missing data, we focus on the representativeness of the number of cases filed and the difference
between number of cases dismissed.

Results	
Case	Reclassification	

To reassess the estimated rate of wrongful conviction reported in Part I, the research team
examined the courthouse data collected in this study to determine if the new information
influenced whether the DNA evidence supported exoneration in cases with exculpatory DNA
evidence. During the reclassification process, 14 convictions that were classified as Outcome 4
during Part I were reclassified to Outcome 3. No Outcome 3 cases were reclassified as Outcome
4.

                                                            
4

States excluded from analysis include Arkansas, Connecticut, Georgia, New Hampshire, North Carolina, Utah,
Vermont, and the District of Columbia.

8
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view 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 1 shows the breakdown of Outcome 3 and Outcome 4 convictions from Part I and Part II,
and the breakdown of Outcome 3 and Outcome 4 convictions amongst the subset used to
calculate the rate of wrongful conviction. None of the new information obtained from the
courthouse files affected the number of convictions classified as indeterminate or inculpatory—
new data collected did not (and could not) change the number of convictions where the DNA
evidence strengthened the conviction (Inculpatory/Outcome 2) or change the fact that new DNA
testing did not produce a useful profile (Indeterminate/Outcome 1). To better understand this
reclassification, Box 1 gives two examples. In the first, a conviction was reclassified when the
additional context weakened the probative value of the DNA evidence. In the second, the
additional context did not weaken the probative value of the DNA and did not support changing
the original classification from Part I.
The reclassification process was limited to only the Outcome 4 and Outcome 3 convictions in
which we were able to collect new information, and as a result some of the convictions that
remained classified as Outcome 4 following Part II did not have additional information from
courthouse coding that further confirmed their Part I classification. Of the 31 convictions, 11
convictions had new information that informed our choice to preserve the originally coded
Outcome. Meanwhile, eight convictions had information coded at courthouses, but not for any
variable that would change an Outcome. Additionally, as described above, 5 additional
convictions resulting in exoneration were added to the dataset that were not included in Part I.
Finally, 7 convictions originally classified as Outcome 4 were not found during Part II coding,
resulting in no new information to consider.
Table	1:	Number	of	Convictions	Classified	as	Outcome	3	or	4	

Outcome 3
(Exculpatory but
Insufficient)

Outcome 45
(Exculpatory and
Supports Exoneration

Part I Classification (2012)

18

43

Part II Classification (2017)

30

31

Part II Classification (2017) for
convictions with sexual assault
component

16

29

                                                            
5

 Includes the five additional exonerations that were added to the dataset but not included in Part 1. 

9
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

Box	1.	Examples	of	Conviction	Reclassification
Example	1 Outcome 4 case reclassified to Outcome 3
Suspect was convicted of rape along with a codefendant. The DFS case file coded in Part I provided
little contextual information about the crime. In Part II, the research team learned from the courthouse
case file that the codefendant pled guilty and testified against the suspect, stating that both men
participated in the rape. The suspect confessed to attempted rape, and a urologist testified that it was
possible that Suspect A was impotent at the time of the rape. Since new information provided a
possible explanation for the absence of the suspect's DNA, the conviction was reclassified as
Outcome 3.
Example	2 Outcome 4 case kept as Outcome 4
The defendant pled guilty to statutory rape, and the DFS case file coded in Part I provided little details
about the context of the crime. From the courthouse case file coded in Part II, the research team learned
that the suspect originally denied having any contact with the victim, and that the victim was not initially
confident in her identification of the suspect at a line-up. Given the additional contextual information
that could support the suspect’s innocence, and the absence of additional information providing a reason
for the absence of the suspect’s DNA, the conviction remained classified as Outcome 4
	
Estimated	Rate	of	Wrongful	Convictions	

Ultimately, among the 231 convictions in this sample with a sexual assault component and
determinate post-conviction DNA testing results, 29 convictions (12.6 percent) yielded
exculpatory DNA evidence that would be supportive of the convicted suspect’s exoneration.
Applying the inverse probability weights described previously, the rate can be corrected to 11.6
percent; this adjusted rate provides an estimate of wrongful conviction in the larger sample
(N=430) of both determinate and indeterminate cases. These estimates may be considered an
upper bound on the rate of wrongful conviction for these cases, since it is possible that even after
Urban researchers’ careful review of courthouse information on cases with exculpatory DNA
evidence, there could be some rightful convictions included.
	
External	Validity	

To determine whether the rate of wrongful conviction could be generalized to other states, we
examined case filings and dispositions across jurisdictions in 43 states that consistently rely on
defendant-based case reporting. Due to missing data, this analysis focused explicitly on the
proportion of cases that resulted in a disposition other than dismissal. These data indicate that
33% of cases filed in Virginia in 1985 resulted in a dismissal, compared to other state dismissal
rates which range from 1% to 75% (Alaska and Maine, respectively). The average dismissal rate
10
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

across states included in the analysis is 16%. A test of proportions indicates that the number of
cases that are not dismissed in Virginia does not significantly differ from other states. These
findings provide a basis, albeit limited, to argue that estimates of wrongful convictions in
Virginia may apply to other states and, furthermore, be lower than states that report lower
dismissal rates.

Dissemination	and	Closeout	Activities	
The project team presented initial findings that are documented in this technical summary at the
Annual Meeting of the American Society of Criminology on November 17, 2016 as part of a
thematic panel (Wrongful Convictions, CODIS, and Sexual Assault Case Processing: Findings
from the Urban Institute’s Forensics Research Portfolio).
Prior to the end of the grant period, the research team will upload deidentified data collected
during the study along with code and documentation used to produce analyses to the National
Archive of Criminal Justice Data, in accordance with NIJ requirements. Additionally, the project
team will submit at least one journal article for publication prior to the end of 2017. This may
include findings discussed in this technical summary along with any identified correlates of
wrongful convictions.

Conclusion	
This study extends prior research on the prevalence of wrongful convictions. In particular, we
rely on case processing and disposition data collected on 714 murder and sexual assault felony
cases across 56 circuit courts to calculate an estimated rate of wrongful conviction. Based on
forensic, case processing, and disposition data, we estimate, after weighting, that wrongful
convictions in cases with a sexual assault component occurred at a rate of 11.6 percent, which is
different than prior estimates reported by the Urban Institute in 2012, due to both a more refined
scope and additional context from case files. We also examine Bureau of Justice Statistics data
collected from Felony Courts in 1985 to determine whether this new estimated rate of wrongful
conviction is generalizable to other states across the U.S. These analyses indicate that the rate of
dismissal in Virginia is not significantly different from other states, suggesting that the findings
of this research may be extended - with caution - to other jurisdictions.
This research effort has created the most valuable dataset to date to investigate the prevalence of
wrongful convictions, and provides research and practitioner communities with a new prevalence
estimate for a problem that continues to plague jurisdictions across the country. Importantly, this
research represents the only known effort to apply DNA testing to cases regardless of a person’s
individual claim of innocence. The process by which outcomes are revised by considering court
processing and case disposition information highlights the limits of DNA evidence in identifying
11
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice

potential instances of wrongful conviction. Furthermore, while most post-conviction efforts rely
on DNA testing only if the conviction is a probable wrongful conviction, this work inverts that
process and puts the DNA testing at the front end, which simultaneously uses DNA to identify
both wrongful and rightful convictions.
Future analyses will include an examination of whether the data collected in this study is
correlated with instances of potential wrongful conviction and present findings on the utility of
DNA as a wrongful conviction detection tool. Since this dataset will be archived, it is our hope
that it will be useful to other researchers interested in post-conviction DNA testing and wrongful
conviction.

12
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view 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	
Gould, J. B., Carrano, J., Leo, R., & Young, J. (2012). Predicting Erroneous Convictions: A
Social Science Approach to Miscarriages of Justice. Washington, D.C.: American
University.
Gross, Samuel R. (2008). Frequency and Predictors of False Conviction: Why We Know So
Little, and New Data on Capital Cases. B.O'Brien, co-author. Journal of Empirical Legal
Studies, 5(4): 927-62.
Gross, Samuel R., Chen Hu, Edward H. Kennedy, and Barbara O’Brien. (2014). Rate of False
Conviction of Criminal Defendants Who are Sentenced to Death. Proceedings of the
National Academy of Sciences of the United States of America, 111(20): 7230-7235.
Hirano, K., Imbens, G.W., and Ridder, G. (2003). Efficient Estimation of Average Treatment
Effects Using the Estimated Propensity Score. Econometrica, 71(4): 1161–89
Innocence Project. (2016). http://www.innocenceproject.org/exonerate/.
Kansas v. March, 278 Kan. 520, 102 P. 3d 445, 2006. Concurring Opinion, Justice Antonin
Scalia.
Poveda, Tony G. (2001). Estimating Wrongful Convictions. Justice Quarterly, 18(3): 689-708.
Ramsey, Robert J., and James Frank. (2007). Wrongful Conviction Perceptions of Criminal
Justice Professionals Regarding the Frequency of Wrongful Conviction and the Extent of
System Errors. Crime & Delinquency, 53(3): 436-470.
Roman, J., Walsh, K., Lachman, P., & Yahner, J. (2012) Post-Conviction DNA Testing and
Wrongful Conviction. Washington, DC: Urban Institute.
Wooldridge, J.M. (2002). Inverse Probability Weighted m-Estimators for Sample Selection,
Attrition, and Stratification. Institute for Fiscal Studies, Cemmap Working Paper (cwp
11/02).
Zalman, M. (2012). Qualitatively Estimating the Incidence of Wrongful Convictions. Criminal
Law Bulletin, 48(2): 221-279.
Zalman, Marvin, Brad Smith, and Angie Kiger. (2008). Officials’ estimates of the incidence of
“actual innocence” convictions. Justice Quarterly, 25(1): 72-100.

13
 
This resource was prepared by the author(s) using Federal funds provided by the U.S.
Department of Justice. Opinions or points of view expressed are those of the author(s) and do not
necessarily reflect the official position or policies of the U.S. Department of Justice