The Effects of Judge Race and Sex on Pretrial Detention Decisions

Abstract

The pretrial detention decision has critical implications for a defendant’s employment opportunities, family ties, likelihood of conviction, and length of prison sentence. While prior researchers have assessed the disparities that exist in the bail decision based on defendant and case characteristics, little systematic empirical attention has been paid to the effects of the pretrial detention judge on decisions at this stage of criminal case proceedings. Here, we focus specifically on judge race and sex, exploring not only the unconditional effects of judge sex and race but also whether the effects of these judge characteristics at the bail decision are conditioned on the sex and race of the defendant. Using newly collected pretrial detention data from 22 federal district courts from 2003 to 2013, we empirically examine the effects of judge and defendant race and sex on whether defendants are released on their own recognizance before trial or, instead, are given a more punitive pretrial outcome. Our results indicate important judge and defendant-based differences in bail setting leniency provided to defendants including that Black judges are more likely to grant pretrial release without hefty conditions to white defendants than are white judges. We also find that female judges are more likely to detain or require monetary bail for male defendants and less likely to do so for female defendants relative to male judges.

Keywords:

• Pretrial detention
• federal district courts
• race
• gender

Notes

1 The data collection began with the drawing of a stratified random sample of defendants from the Federal Judicial Center’s Criminal Integrated Database by district-sentence year from 2006 to 2013. After the sample was drawn, we collected the case’s docket sheet and other relevant case documents from PACER (“Public Access to Court Electronic Documents”). The case documents were used to collect the dependent variable (detention decision), the detention judge’s name, the defendant’s name and year of birth, and other relevant case and defendant information. We then matched the defendant with his corresponding observation in the publicly-available U.S. Sentencing Commission’s Offender Datafiles so that additional defendant-specific variables would be available to us. This matching was done with a series of case and defendant characteristics including district; sentence date; prison, probation, and supervised release details; fine, special assessment, and restitution details; individual statutes for conviction charges; and other defendant information. We had over a 92% successful match rate.

We exclude judges and defendants from the data who are neither Black nor white. Doing so simplifies the data and better allows us to test conditional theories of judge decision making. We leave it to future research to further flesh out the theoretical and empirical implications for judges and defendants of other races and ethnicities. In alternative modeling, we exclude all defendants who are non-U.S. citizens. Doing so does not affect the key results presented below.

The 22 districts covered in the data include: D. Arizona, Central D. California, District of Columbia, Middle D. Florida, Southern D. Florida, Northern D. Iowa, Northern D. Illinois, Southern D. Illinois, D. Kansas, Eastern D. Kentucky, Eastern D. Louisiana, D. Massachusetts, D. Maine, Eastern D. Michigan, Middle D. North Carolina, D. New Jersey, Southern D. New York, Eastern D. Pennsylvania, Middle D. Pennsylvania, D. South Dakota, Western D. Virginia, and D. Vermont. Because the data collection focuses on sentence date, the pretrial detention hearing dates within the data occur as early as 2003.

2 This project simplifies the measurement of the trial judge detention decision into two groups – lenient (ROR) and more punitive (monetary bail and detention) for both empirical and theoretical reasons. In terms of the underlying theory and literature, we struggle to develop judge race and sex and defendant race and sex expectations that distinguish between the two punitive worlds of monetary bail and detention. Empirically, while the PACER case documents do not always enable us to observe whether a defendant made bond, our review of the available case documents on these matters reveals frequently lofty monetary bail amounts and conditions and a wide range of success among defendants in meeting their bail conditions and avoiding detention. To us, this suggests a level of punitiveness toward pretrial defendants that simply does not exist for those defendants released on personal recognizance. We look forward to future research that further untangles the theory and empirics of detention decisions and opens the door to studies examining judge differences across ROR, monetary bail, and detention separately.

3 Scholars have argued that race is a social construct that remains fluid and that racial group membership is socially contested (Omi and Winant 1994; Hirschman, Alba and Farley 2000; Prewitt 2006). Sen and Wasow (2016) make a compelling argument that race is an aggregate of many components such as societal values, cultural traits, and institutional power relationships, to name a few. We recognize the validity of the argument and do not seek to essentialize Black judges to a single dichotomous variable that treats them as consisting strictly of immutable characteristics. However, data limitations prevent us and many studies like ours focused on judge and defendant race from capturing and including the variety of components of a more wholesale and nuanced accounting of race in this study’s modeling. Nonetheless, a shared group consciousness based on common experiences connected to membership among those of the same racial identity, including within the judicial system (i.e., Dawson 1994; McClain et al. 2009; Bratton and Haynie 1999; Heider 1958; Kulik, Perry, and Pepper (2003) means there is still value in examining the role of race in judicial proceedings like those studied here. We look forward to future data collection and research efforts in the law and courts arena that enable a more wholesale accounting of judge race and sex.

4 Because of the high percentage of magistrate judges in our data, some may be concerned with the proxy judge ideology score used for magistrate judges. We note that excluding the judge ideology measure from our statistical model reported below does not affect other key findings that we report.

5 We utilize base offense level rather than the final offense level because it is the “starting point” for a Guidelines calculation (United States Sentencing Commission 2013) and is, at least in theory, closer to the pure factual circumstances of the crime in question. By contrast, a final offense level is the product of findings by the sentencing judge relating to crime-specific aggravating and mitigating characteristics that adjust the base offense level. Since “criminal sentencing scholarship has taken note that the factual circumstances under the Guidelines that lead to adjustments are often vague and allow for considerable discretion by the judge” (Schanzenbach and Tiller 2006, 28) we opt for the use of the base offense level to avoid using judicial discretion at sentencing to predict judicial discretion at bail. Ultimately, substituting final offense level for the base offense level does not alter our key findings.

6 Many of our defendant-specific control variables, like Defendant Criminal History and the trigger variables, are measured based on data collected by the U.S. Sentencing Commission at the time of sentencing (i.e., after the pretrial detention decision). These sentencing-stage data allow us to capture a more detailed picture of a defendant than any alternative source, and we expect that much of the same underlying information about the defendant and the offense are available to the pretrial detention hearing judge (albeit in a less accessible manner). Where missingness occurs in the Sentencing Commission data, we exclude the observation.

Additional information

Funding

The authors thank the anonymous reviewers, the editor, and the special issue editors for their helpful feedback. Boldt and Boyd acknowledge funding from National Science Foundation grant SES-1729077. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. We are grateful to Peppermint, Ryan Brock, Julia Spicer, Hunter Rahn, Jacob Kepes, Surendra Mahadeo, and Monique Alavi for their research assistance.