by Casey J. Bastian

The existence of racial disparity in federal sentencing practices is a common, well-researched issue. The greatest proportion of studies focus on the aggregate disparity between the imposed sentence length of Black versus white defendants. Research has also been conducted on the total disparate sentences Hispanic defendants receive.

The results are well documented and unfortunate. Concerning federal criminal sentencing, the mean disparities among varying races are substantial. Black and Hispanic defendants typically receive more severe sentences than that of white defendants—a fact particularly prevalent in young male offender sentencings. But the questions concerning disparate sentencing do not end with proving disparity among race, gender, and age.

The next question is: beyond aggregate disparities, is there a significant disparity difference among federal judges? This is something much less researched. Disparity by race is frequently demonstrated, but does disparity by race increase or decrease—is there further disparity—when you compare each individual judge’s sentencing statistics?

The Institute for the Quantitative Study of Inclusion, Diversity, and Equity addressed these inquiries in their 2021 report, “Racial Disparities in Criminal Sentencing Vary Considerably across Federal Judges.” The report examined data to answer four primary questions: (1) What is the average disparity between the sentences of observationally equivalent white and Black defendants? (2) What is the average disparity between the sentences of observationally equivalent white and Hispanic defendants? (3) How much do conditional Black-white sentencing disparities vary across judges?, and (4) How much to conditional Hispanic-white sentencing disparities vary across judges.

The source of the data analyzed was the Judicial System Transparency Through Federal Archived Inferred Records (“JUSTFAIR”). This database is compiled from five public sources representing federal court criminal sentencing decisions. JUSTFAIR contains over 600,000 records for the fiscal years 2001-2018, linking information about the defendant, their federal crimes, the imposed sentence(s), and the judge who imposed the sentence.

The report expounds on and summarizes the JUSTFAIR data pipeline originally presented in a previous study by Ciocanel, M., in 2020. Ciocanel and colleagues had obtained similar data from the U.S. Sentencing Commission (“USSC”) and integrated that data with Federal Judicial Center (“FJC”) Integrated Database to acquire each case’s court docket number. The docket numbers were used to access PACER where the initials of the sentencing judge were collected. Using each judge’s initials, Ciocanel linked to both Wikipedia and the FJC Biographical Directory of Article III federal judges to obtain names, demographics, and educational information of each judge being examined.

As a result of compiling the data from these five sources, JUSTFAIR enables research of the sentenced individuals’ demographic characteristics, name of the sentencing judge, appointment, and education information. A crucial aspect of this current study is JUSTFAIR’s additional data sections pertaining to conviction-relevant law and the factors influencing the recommended sentence.

The study also extended the JUSTFAIR data pipeline by including the 2018-2019 fiscal year sentencing data. This was possible because the USSC individual offender data file is updated yearly, and the FJC is updated quarterly, meaning additional information for consideration became available during the study period.

As a result of several USSC database variables changing, including total prison sentence, the offender-type coding, and the post-Booker reporting, the Ciocanel data processing approach had to be adjusted in this study to remain consistent with JUSTFAIR variables. Therefore, an additional 30,000 cases were analyzed by this newest study. As this study was primarily concerned with racial disparities observed after the Booker decision (which made the USSG’s resultant sentencing ranges advisory rather than mandatory—see United States v. Booker, 543 U.S. 220 (2005)), cases prior to 2006 were situated outside the target data and excluded as well.

Immigration cases were excluded due to the extreme concentration of such cases disposed of in several southwest district courts. The study also did not consider cases for which the sentence length could not be inferred.

The sentencing length for approximately 13,000 cases in the JUSTFAIR database possessed two contradictory data outputs. First, according to the continuous sentence total variable, the resultant sentence length was zero. Yet the categorical variable of imprisonment type did indicate a prison sentence length. As both of these measures cannot simultaneously be correct, the study considered the 13,000 cases to be missing an outcome. Hence, no consideration could be given, and the cases were removed from analysis.

After determinations were made of included and excluded data in the pre-processing stage, the analytic data sample size was representative of 380,000 cases across 1,116 federal judges. The result is a median analyzed case study of approximately 263 imposed sentences per judge. The focus was on this data. The outcome of interest was length of sentence, so the result was interpreted in terms of percentage change as opposed to linear change in months of prison time.

The case-level characteristics of principal interest concerned the defendant’s race and ethnicity. The three categories of race and ethnicity considered were “Hispanic,” “Non-Hispanic Black” (Black), “Non-Hispanic White” (white), and all other cases from the overall 380,000 case data sample received an “Other Race/Ethnicity” designation.

The study indicates that the data groups must be collapsed this way because the sample groups indicative of just race and/or ethnicity in a final data category would not be large enough for them to perform cross-judge analyses in any statistically informative way. As such, the as-defined race and ethnicity data samples indicate each median judge-imposed sentence in 41 Hispanic and 69 Black defendant cases. For these cases, approximately 32 percent of all federal sentences imposed, an extensive set of control variables were instituted to determine across-judge disparity, including: (1) the guideline minimum sentence, with any statutory minimum sentences taken into account; (2) the defendant’s criminal history points; (3) crime type, namely “violent crime,” “drug-related crime,” “embezzlement, fraud, theft,” or “other”; (4) whether the case was settled by plea agreement or trial; (5) sentencing year; and, (6) defendant’s demographics, namely, sex, age, U.S. citizenship status, and educational attainment.

The study’s authors assert that, of the control variables listed above, the Guidelines minimum sentence variable is the most important; a variable that is theoretically standardized based on such factors as severity of offense conduct. Where in a world without racial disparity, the data would reveal that two defendants of different racial or ethnic groups, but having the same minimum sentence Guidelines range, would see roughly equal sentences imposed.

The additional variable of the defendants’ demographics also requires control because it would seem to be an extralegal factor outside the judge’s consideration but can capture relevant legal factor variations.

The authors of the study provide the example of a judge presiding over cases in which a Hispanic defendant might have unique situations not contained in the observable legal variables. If the unique situation were to be an educational variable, then controlling such a variable would account for otherwise unobserved variations in the data. The same would be true for any variable observed but considered “extra-legal.”

The study eventually provides statistical observations on aggregate racial disparities in sentencing before turning to the inter-judge variability in observed disparities. Premised on observed defendant and case characteristics, judges impose sentences that are approximately 13 percent longer for Black defendants and 19 percent longer for Hispanic defendants than the sentences imposed for white defendants.

The study did find that the standard deviation in imposed sentences for all other racial groups was actually 10 percent lower than that of white defendants. The study reveals that its aggregate results concerning racial disparities is in line with previous studies. It can also be inferred from the data that slightly higher aggregate disparities may exist that were not found in data analyses results compiled from earlier time spans using different sampling mechanisms.

A 2011 study found that, even after controlling for considerable legal and extralegal variables, Blacks were found to have received sentences six percent longer and Hispanics one percent longer than whites. The 2011 study used 2000-2002 pre-Booker USSC datasets.

Analyzing 2006-2008 multiagency-based data, a 2014 study found that Blacks receive sentences nine percent longer that whites. A third 2015 study used federal sentencing datasets very similar to this study, but from 2000-2010, finding that Blacks and Hispanics received sentences 1.9 months longer than whites.

Unfortunately, when the study turns to questions three and four, the reported findings are not elaborated upon or discussed in detail. What the study does indicate is that, based on its analytical model, a judge who is one “standard deviation” above average in terms of any Black-white disparity would in-turn impose sentences that are conditionally 39 percent longer than whites; two standard deviations result in an across-judge disparity of up to 71 percent longer for Blacks.

Hispanic defendants receive sentences 49 percent longer than whites from judges one standard deviation above average, while two standard deviations result in 87 percent longer sentences.

The term “standard deviation” is not explicitly defined. It appears to reference that all judges analyzed were assigned a median sentence disparity in their district court. Beyond this median baseline, each individual judge’s sentencing statistics were analyzed, and the specific deviations from the median were assigned one or two deviations to reflect the significant disparate sentencing patterns for that individual judge. These measures account for an individual judge with a greater reflected disparity than the otherwise observable median percent increase in Black or Hispanic sentence lengths.

The study concludes that the findings indicate sizable average sentencing disparities and that nontrivial variabilities persist in individual judge’s racially disparate sentencing patterns. It is noted that these findings of disparity may not be a reflection of direct, unfair treatment based on race. One reason this may be so is that a judge may not sit on cases where Black and white defendants are balanced regarding unobserved legal characteristics despite observed control variables.

The study’s authors also acknowledge limitations in the analyzed JUSTFAIR data, as well as caution being needed in interpreting the results of this study. As cases are not randomly assigned to the judges, systematically differing unobserved factors might create the variability differences noted.

Another limitation in these statistical analyses is that not every case in every district can be observed. The JUSTFAIR database is a sample, not a population, which leaves estimates of measured variability susceptible to random sampling errors. JUSTFAIR contains no sentencing data from sealed cases, or from the following: Eastern District of North Carolina; Southern District of West Virginia; Southern District of Texas; Middle District of Tennessee; Northern District of Illinois; District of Guam; or the Northern Mariana Islands. JUSTFAIR contains less than 33 percent of the USSC cases’ sentencing data of: Northern District of Texas, Southern District of California, District of Oregon, District of New Mexico, Western District of Oklahoma, and the Northern District of Florida as well.

The study’s authors believe that future research should also extend to state courts, something this study does not examine. The majority of sentences imposed in America occur in state courts, and to truly understand inter-judge variability in racial disparate sentencing, it is necessary to gather this information as well. It will require extensive work to gather the data, but helping to move the justice system towards greater equity is worth the effort.  

Source: SocArXiv.org, qsideinstitute.org

Subscribe today

Already a subscriber? Login