What Marginal Outcome Tests Can Tell Us About Racially Biased Decision-Making

The US Pretrial System: Balancing Individual Rights and Public Interests

“…The results from Arnold, Dobbie, and Yang (2018), along with related discussion in Hull (2021), also make clear that outcome tests are purposely narrow, isolating racial bias from other important sources of disparities such as illegal statistical discrimination on the basis of race. The fact that outcome tests also combine behavior that is clearly disparate treatment, such as Becker taste-based discrimination, and behavior that a court may not find illegal under this standard, such as unconscious/implicit bias or stereotyping, further complicates the interpretation and use of these tests.…”

“…In the context of pretrial decisions, the idea is that marginal White defendants will have higher rates of pretrial misconduct than marginal Black defendants if bail judges are racially biased against Black defendants. Recent work has clarified that outcomes can vary across groups at the margin due to Becker taste-based discrimination, inaccurate racial stereotypes (such as those modeled by Bordalo et al 2016), and potentially other form of racial bias (Arnold, Dobbie, and Yang 2018;Hull 2021;Gelbach 2021). The outcome test also captures de facto racial bias that might arise though seemingly race-neutral characteristics such as type of crime and neighborhood by allowing non-race characteristics to differ for marginal White and marginal Black defendants.…”

“…The fact that outcome tests also combine behavior that is clearly disparate treatment, such as Becker taste-based discrimination, and behavior that a court may not find illegal under this standard, such as unconscious/implicit bias or stereotyping, further complicates the interpretation and use of these tests. 6 The outcome test cannot be used to test for more restrictive definitions of racial bias that hold fixed all non-race characteristics of White and Black defendants without additional restrictions: for example, see the discussions in Canay, Mogstad, and Mountjoy (2020) and Hull (2021). 7 New work by Grau and Vergara (2020) proposes an alternative implementation of the outcome test that relies on using the predicted status (or propensity scores) of defendants to identify defendants at the margin of release.…”

Systemic Discrimination Among Large U.S. Employers

“…Much of the economics literature has focused on separating the contributions of tastebased and statistical discrimination to observed disparities, an exercise that requires inferring the extent to which employer conduct is motivated by beliefs regarding the productivity of different groups of workers (Becker, 1957(Becker, , 1993Aigner and Cain, 1977;Charles and Guryan, 2008;Bohren et al, 2019). Recent empirical and methodological work looks at group differences in the treatment of equally-productive individuals in bail decisions, motor vehicle searches, probation revocations, and other settings (Arnold, Dobbie and Yang, 2018;Arnold, Dobbie and Hull, 2020;Canay, Mogstad and Mountjoy, 2020;Rose, 2020;Feigenberg and Miller, 2021;Hull, 2021). In the employment context, it is widely understood that both taste-based and statistical discrimination typically involve disparate treatment of individuals according to legally-protected characteristics, which is prohibited by the Civil Rights Act.…”

Unveiling the Cosmic Race: Racial Inequalities in Latin America

“…using historical or natural experiments (Albright et al 2021;Derenoncourt and Montialoux 2020;Derenoncourt 2021); and new developments on methods also provide causal evidence of racial discrimination (Arnold, Dobbie, and Hull 2021;Hull 2021). But for most cases with cross-sectional data, it is not straightforward to provide causal estimates of racial discrimination.…”