Differential prediction and disparate impact of pretrial risk assessments in practice: a multi-site evaluation

Lowder, Evan M.; Diaz, Carmen L.; Grommon, Eric; Ray, Bradley

doi:10.1007/s11292-021-09492-9

Cited by 8 publications

(14 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When implemented with a validated instrument as part of a presumptive release regime, the same factors can be part of a relatively efficient and equitable policy that minimizes detention without compromising public safety (Desmarais et al., 2022; Grossman et al., 2022). Given controversies about risk assessment (e.g., Lowder et al., 2022; Minow et al., 2019), any instrument applied as part of a presumptive release regime must be free of predictive bias by race and implemented with high fidelity. Under those conditions, we agree with Reitz's (2020) compelling argument that a risk‐based policy reform would be a step forward from the “horribly flawed” status quo and could be powerful force in changing absolute incarceration rates—that is, the lives of large numbers of people—in both Black and White communities.…”

Section: Discussionmentioning

confidence: 99%

Understanding racial disparities in pretrial detention recommendations to shape policy reform

Skeem

Montoya

Lowenkamp

2023

Criminology & Public Policy

View full text Add to dashboard Cite

Research Summary Federal pretrial services and probation officers assess defendants and make influential recommendations that defendants be either released or detained, based on their threat to community safety and risk of flight. To inform efforts to reduce disparities in pretrial detention, we examined officers’ decision making about 149,815 defendants across 81 districts. Overall, the probability of a detention recommendation was 34% higher for Black than White defendants. Racial disparities were most pronounced in ambiguous cases that invoked substantial officer discretion—including cases where the defendant had little or no criminal record. Nevertheless, mediation analyses revealed that up to 79% of the racial disparity in detention recommendations operates through institutionalized factors (i.e., pretrial policy) rather than personally mediated factors (e.g., implicit racism or classism). The lion's share of the disparity operates through one institutionalized factor alone: criminal history. Policy Implications This study illustrates an empirical strategy for understanding the pathways through which disparities operate, which is crucial for shaping effective solutions. Providing officers with training and decision guides could reduce personally mediated bias—which is crucial for high discretion cases. However, this study shows that disparities mostly flow through institutionalized bias. So, greater gains may be had by making strategic shifts in policies and their implementation. One promising direction is to corral criminal history by adopting a tight definition that demonstrably predicts violence and failure to appear, and limiting the weight assigned to criminal history versus other predictive factors, when making recommendations. Another promising direction is to adopt risk‐based release policies that leverage an existing tool to reduce both detention rates and racial disparities.

show abstract

Section: Discussionmentioning

confidence: 99%

Understanding racial disparities in pretrial detention recommendations to shape policy reform

Skeem

Montoya

Lowenkamp

2023

Criminology & Public Policy

View full text Add to dashboard Cite

show abstract

“…With respect to the latter concern, a recent study found that even when a pretrial risk assessment instrument demonstrated better predictive validity in White than Black people, its use still resulted in a decrease in pretrial detention and bail assignments for both groups. In fact, the bail amounts assigned to Black people decreased more than twice as much as the bail amounts assigned to White people with the use of the instrument (Lowder et al, 2022). Other studies have similarly found decreases in pretrial detention across groups defined by race or ethnicity when decisions are informed by pretrial risk assessment instruments but have not included consideration of bail assignments (Viljoen et al, 2019).…”

Section: Money Bail Pretrial Risk and Racementioning

confidence: 99%

“…With respect to the former concern, studies that have examined racial bias in the predictive validity of pretrial risk assessments have produced mixed findings. Some studies find comparable levels of predictive validity across racial subgroups, whereas others find slightly better predictive validity for White than Black people (though performance still typically exceeds minimum thresholds for acceptable predictive validity in both groups; Brittain et al, 2021;Desmarais et al, 2021;Lowder et al, 2022). With respect to the latter concern, a recent study found that even when a pretrial risk assessment instrument demonstrated better predictive validity in White than Black people, its use still resulted in a decrease in pretrial detention and bail assignments for both groups.…”

Section: Money Bail Pretrial Risk and Racementioning

confidence: 99%

“…While stakeholders debate the merits of maintaining money bail or implementing pretrial risk assessments, there is relatively little empirical evidence to inform this debate. To our knowledge, only two studies have examined how pretrial risk assessment scores influence the assignment of bail (Lowder et al, 2021, 2022). No studies have directly compared the association between money bail and whether a person fails to appear in court or is rearrested to the association between pretrial risk assessment scores and whether a person fails to appear or is rearrested in the same sample.…”

Section: The Present Studymentioning

confidence: 99%

See 1 more Smart Citation

Comparing the relationships between money bail, pretrial risk scores, and pretrial outcomes.

Zottola¹,

Desmarais²

2022

Law and Human Behavior

View full text Add to dashboard Cite

Objectives: There has been much discussion around the use of both money bail and pretrial risk assessment instruments. We examine how bail and risk scores compare in terms of their associations with failure to appear in court and rearrest during the pretrial period. Hypotheses: Our research questions included whether bail and risk scores differed between people who did and did not experience pretrial outcomes and whether pretrial scores were associated with outcomes when controlling for bail and other relevant covariates. Method: To examine these associations, we drew a sample of 492 people (33% women; 60% Black) booked into county jail in a jurisdiction not yet using a pretrial risk assessment instrument to inform release decisions. We completed the Public Safety Assessment (PSA) for this sample and collected data on failure-to-appear and rearrest incidents for 1 year following initial bookings. We examined the associations between bail amount, failure to appear, and rearrest and between PSA subscale scores, failure to appear, and rearrest. Results: Bail amount was not associated with either failure to appear or rearrest. People who failed to appear or were rearrested had higher bail amounts, on average, than people who did not. In contrast, PSA subscale scores were significantly associated with outcomes in the expected direction. Conclusions: Our findings do not support the use of money bail for ensuring that people return to court and avoid rearrest. Instead, our findings suggest that using pretrial risk assessment, instruments could result in more accurate and appropriate release decisions.

show abstract

“…One of the primary functions of RAIs is to estimate the likelihood that an individual will reoffend in the future, and these estimates can be used to assist justice professionals in making key decisions, including treatment and service allocation, supervision, and various types of sanctions (Hoge, 2002). In the context of R/ED in RAIs, therefore, it is fundamentally important to ensure that a given RAI score has the same meaning regardless of race and ethnicity (Lowder et al, 2019(Lowder et al, , 2022Skeem & Lowenkamp, 2016). This is commonly tested by examining the extent to which the predicted recidivism matches the observed recidivism when using a common regression line (Cleary, 1968).…”

Section: Predictive Bias For Raismentioning

confidence: 99%

Racial/ethnic disparities of the pact in predicting recidivism and court dispositions for justice-involved youth.

Li¹,

Hein²,

Quintana³

et al. 2023

Law and Human Behavior

View full text Add to dashboard Cite

Objective: Responding to the concern about racial/ethnic disparities (R/ED) in the use of risk assessment instruments (RAIs) in justice systems, previous research has overwhelmingly tested the extent to which RAI scores consistently predict recidivism across race and ethnicity (predictive bias). However, little is known about R/ED in the association between RAI measures and court dispositions (disparate application) for justice-involved youths. This study investigated predictive bias and disparate application of three risk measures—criminal history, social history, and the overall risk level—produced by the Positive Achievement Change Tool (PACT) for White, Black, and Hispanic justice-involved youths. Hypotheses: Given the mixed evidence in existing research for predictive bias and lack of evidence for disparate application, we did not make any specific hypothesis but conducted exploratory analyses. From a clinical perspective, however, we anticipated little or no evidence to support predictive bias and disparate application of the PACT among White, Black, and Hispanic youths in the jurisdiction we examined. Method: The sample consisted of 5,578 youths (11.4% White, 43.9% Black, and 44.7% Hispanic) who completed the PACT while in the Harris County Juvenile Probation Department, Texas. The outcome variables included recidivism (general and violent reoffending) and court dispositions (deferred adjudication, probation without placement, and probation with placement). We ran a series of moderating binary logistic regression models and moderating ordinal logistic regression models to evaluate predictive bias and disparate application. Results: Race and ethnicity influenced how the criminal history score related to violent recidivism: This compromised the validity of the score as a predictor of recidivism. Moreover, evidence showed that the overall risk of reoffending was associated with harsher sanctioning decisions for Black and Hispanic youths than for White youths. Conclusion: Ensuring that RAI results are consistently interpreted and used in informing decisions is as important as ensuring that RAI scores function equally well in predicting recidivism regardless of race and ethnicity.

show abstract

Differential prediction and disparate impact of pretrial risk assessments in practice: a multi-site evaluation

Cited by 8 publications

References 48 publications

Understanding racial disparities in pretrial detention recommendations to shape policy reform

Understanding racial disparities in pretrial detention recommendations to shape policy reform

Comparing the relationships between money bail, pretrial risk scores, and pretrial outcomes.

Racial/ethnic disparities of the pact in predicting recidivism and court dispositions for justice-involved youth.

Contact Info

Product

Resources

About