Reducing Racial Disparities in Bail Decisions
How to cite
Dobbie, Will; Yang, Crystal. 2020. "Reducing Racial Disparities in Bail Decisions." AEA RCT Registry. AEARCTR-0006871. https://doi.org/10.1257/rct.6871-1.1
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
In this project, we are partnering with state and county courts from around the country to help improve decision-making and reduce racial disparities at the pretrial justice stage. We plan to tackle racial disparities driven by bias and inaccurate stereotypes through a combination of three interventions that are low cost and have shown tremendous efficacy in other settings. The first intervention provides objective information on the pretrial risk of white and non-white defendants to judges to correct inaccurate stereotypes that exaggerate the relative danger of non-white defendants. The second intervention provides a simple benchcard to judges to slow down and systematize their decision-making. The third intervention provides detailed feedback to judges on their own outcomes over time, giving them the motivation, information, and tools necessary to reduce racial disparities in their pretrial decisions. We will estimate both (i) the impact of the full set of interventions on long-run outcomes and (ii) the impact of each individual intervention on medium-run outcomes. Our key outcomes of interest are overall pretrial detention rates as well as the racial gap in detention rates for white and non-white defendants.
Intervention
We have designed a set of three interventions to improve decision-making and reduce racial disparities at the pretrial justice stage.
(1) Objective Information on Misconduct Rates by Race: Our first intervention provides objective information on the pretrial risk of white and non-white defendants to reduce judges’ reliance on inaccurate stereotypes that exaggerate the relative danger of non-white defendants. The goal of this first intervention is to reduce bias in judges’ beliefs about differences in defendant risk by race by providing accurate information on true differences. In particular, we aim to correct mistaken beliefs that can arise if judges erroneously assume that non-white defendants are far riskier compared to white defendants than supported by data. We will create accessible, engaging materials, such as a two- to three-page informational handout, that provide descriptive statistics of evidence-based measures of risk by race using historical data. We will contrast the information on evidence-based risk with “status quo beliefs on risk,” which is defined as the risk that judges must believe in order to justify observed racial differences in pretrial outcomes. For example, we will present a simple bar diagram to show judges what the racial gap in pretrial release rates should be based on evidence-based measures of risk versus status quo beliefs on risk. This information will be based on historical data from the judge’s own local area.
(2) Generalized Benchcard: Our second intervention provides judges with a simple and short “benchcard” to slow down and systematize their decision-making. The goal of this intervention is to prompt the judges to consider different areas important to the bail decision-making process and guide any questions that they might ask. In particular, we aim to highlight the most important features of bail cases in each court system to help judges avoid errors due to forgetfulness and limited attention. For example, we ask the judge to rank the importance of the defendant’s past criminal history, as well as the defendant’s risk of flight and danger to the public. We then ask the judge to rank the importance of employment, housing, and other life circumstances. Finally, we ask the judge to rank the importance of whether there are less restrictive non-cash alternatives in the case. We also ask the judge to list any other factors that are important in the current case.
(3) Individualized Feedback on Release and Misconduct Rates by Race: Our third and final intervention aims to leverage the power of our information and benchcard interventions by providing individualized feedback to judges over time. We will provide judges with individualized feedback on their past performance, including their own release and “success” rates, where the success rate is defined as the percent of released defendants who appear for their scheduled court appearances and who are not rearrested for another crime pretrial. We would start by providing historical information, before providing updated information on a regular basis (e.g., every 3 months) as the experiment progresses.
We will provide judges with a set of performance statistics including what percent of all, white, and non-white defendants they released, as well as the appearance rate, arrest-free rate, and violent arrest-free rate among those released. We will also show where the judge currently stands in terms of their release and appearance rates relative to three points of comparison. The first point of comparison is a “performance benchmark,” or what is theoretically feasible in terms of simultaneously maximizing release and good conduct rates using all the information we can observe about defendants. The second point of comparison is the set of “best judges” in the same area, which we define as the approximately 25 percent of judges who have the highest combined release and good conduct rates. The third point of comparison is the judge’s own historical data, facilitating self-comparisons. This information will be presented in the aggregate and disaggregated for white and nonwhite defendants.
Experimental design
We will partner with bail judges across the United States to implement an RCT to assess the effectiveness of the three interventions. Half of the judges will continue making decisions under the status quo and will serve as the control group for the study. The second half of judges will receive all three interventions and will serve as the main treatment group for the study. We will follow judges for 24 months after randomization.
There are two key components of our data collection plan. First, we will collect survey data at baseline and endline. The survey will ask judges for baseline demographic information, questions about which factors they deem important to pretrial decision-making, questions assessing their current bail practices, questions about their perceived risk of defendants, and questions about how judges trade off release rates and pretrial misconduct rates. We will also administer a baseline Implicit Association Test (IAT) to all judges. This survey will be repeated at the end of the study, where we will also ask additional questions in order to better understand how judges interpreted and utilized the intervention they received.
Second, we will collect administrative court data on pretrial release decisions and misconduct outcomes (as measured by new crime and failure to appear) at the case level. These data will also include judge identifiers, defendant identifiers, defendant demographics, and additional case and defendant details such as the crime type and prior criminal history.
Design details
Cited by
No citation data available for this registration yet.