Primary Submission Category: Causal Inference and Bias/Discrimination
Can a risk-based release policy reduce unnecessary pretrial detention and racial disparities in detention?
Authors: Lina Montoya, Jennifer Skeem, Christopher Lowenkamp,
Presenting Author: Lina Montoya*
In the US, over 400,000 people are in pretrial detention. Some scholars argue that risk assessment instruments (RAIs) – or checklists that score a person’s likelihood of reoffending or failing to appear in court – can increase the capacity to release defendants safely. Others reject these tools as racially biased, concerned they will perpetuate disparities in the criminal legal system. Using federal data from the Administrative Office of the US Courts, we assess the extent to which defendants’ counterfactual outcomes would have improved had release recommendations been given using a dynamic treatment regime based on the Pretrial Risk Assessment (PTRA; a RAI), compared to the status quo (i.e., detention decisions made with no RAI support). Using Targeted Maximum Likelihood Estimation, we examine these effects overall and by racial categories. For all defendants, when examining the impacts of PTRA-based decisions, we find a substantial reduction in detention (RD: -37.47%; 95%CI: [-37.29, -37.65]) and an increase in positive outcomes (RD: 35.77%; 95%CI: [35.21%, 36.32%]), with little public safety cost (RD: 1.72%; 95%CI: [1.19%, 2.25%]). Further, under this RAI strategy, compared to White defendants, Black defendants would have had fewer detentions, more positive outcomes, and comparable risk events. These findings add to RAI research, including how these tools can offset biases embedded within current decision-making practices in the US criminal legal system.