Explore the vast range of titles available.

Kat Albrecht and Kaitlyn Filip on data disadvantaging defendants.

Can machine learning improve human decision making? Bail decisions provide a good test case. Millions of times each year, judges make jail-or-release decisions that hinge on a prediction of what a defendant would do if released. The concreteness of the prediction task combined with the volume of data available makes this a promising machine-learning application. Yet comparing the algorithm to judges proves complicated. First, the available data are generated by prior judge decisions. We only observe crime outcomes for released defendants, not for those judges detained. This makes it hard to evaluate counterfactual decision rules based on algorithmic predictions. Second, judges may have a broader set of preferences than the variable the algorithm predicts; for instance, judges may care specifically about violent crimes or about racial inequities. We deal with these problems using different econometric strategies, such as quasi-random assignment of cases to judges. Even accounting for these concerns, our results suggest potentially large welfare gains: one policy simulation shows crime reductions up to 24.7% with no change in jailing rates, or jailing rate reductions up to 41.9% with no increase in crime rates. Moreover, all categories of crime, including violent crimes, show reductions; these gains can be achieved while simultaneously reducing racial disparities. These results suggest that while machine learning can be valuable, realizing this value requires integrating these tools into an economic framework: being clear about the link between predictions and decisions; specifying the scope of payoff functions; and constructing unbiased decision counterfactuals.

This article develops a new test for identifying racial bias in the context of bail decisions—a high-stakes setting with large disparities between white and black defendants. We motivate our analysis using Becker’s model of racial bias, which predicts that rates of pretrial misconduct will be identical for marginal white and marginal black defendants if bail judges are racially unbiased. In contrast, marginal white defendants will have higher rates of misconduct than marginal black defendants if bail judges are racially biased, whether that bias is driven by racial animus, inaccurate racial stereotypes, or any other form of bias. To test the model, we use the release tendencies of quasi-randomly assigned bail judges to identify the relevant race-specific misconduct rates. Estimates from Miami and Philadelphia show that bail judges are racially biased against black defendants, with substantially more racial bias among both inexperienced and part-time judges. We find suggestive evidence that this racial bias is driven by bail judges relying on inaccurate stereotypes that exaggerate the relative danger of releasing black defendants.

Judges in multiple US states, such as New York, Pennsylvania, Wisconsin, California, and Florida, receive a prediction of defendants’ recidivism risk, generated by the COMPAS algorithm. If judges act on these predictions, they implicitly delegate normative decisions to proprietary software, even beyond the previously documented race and age biases. Using the ProPublica dataset, we demonstrate that COMPAS predictions favor jailing over release. COMPAS is biased against defendants. We show that this bias can largely be removed. Our proposed correction increases overall accuracy, and attenuates anti-black and anti-young bias. However, it also slightly increases the risk that defendants are released who commit a new crime before tried. We argue that this normative decision should not be buried in the code. The tradeoff between the interests of innocent defendants and of future victims should not only be made transparent. The algorithm should be changed such that the legislator and the courts do make this choice.

Court-related fines and fees are widely levied on criminal defendants who are frequently poor and have little capacity to pay. Such financial obligations may produce a criminalization of poverty, where later court involvement results not from crime but from an inability to meet the financial burdens of the legal process. We test this hypothesis using a randomized controlled trial of court-related fee relief for misdemeanor defendants in Oklahoma County, Oklahoma. We find that relief from fees does not affect new criminal charges, convictions, or jail bookings after 12 months. However, control respondents were subject to debt collection efforts at significantly higher rates that involved new warrants, additional court debt, tax refund garnishment, and referral to a private debt collector. Despite significant efforts at debt collection among those in the control group, payments to the court totaled less than 5 percent of outstanding debt. The evidence indicates that court debt charged to indigent defendants neither caused nor deterred new crime, and the government obtained little financial benefit. Yet, fines and fees contributed to a criminalization of low-income defendants, placing them at risk of ongoing court involvement through new warrants and debt collection.

When is it possible for one person to persuade another to change her action? We consider a symmetric information model where a sender chooses a signal to reveal to a receiver, who then takes a noncontractible action that affects the welfare of both players. We derive necessary and sufficient conditions for the existence of a signal that strictly benefits the sender. We characterize sender-optimal signals. We examine comparative statics with respect to the alignment of the sender's and the receiver's preferences. Finally, we apply our results to persuasion by litigators, lobbyists, and salespeople.

Race, ethnicity, gender, and age are core foci within sociology and law/criminology. Also prominent is how these statuses intersect to affect behavioral outcomes, but statistical studies of intersectionality are rare. In the area of criminal sentencing, an abundance of studies examine main and joint effects of race and gender but few investigate in detail how these effects are conditioned by defendant's age. Using recent Pennsylvania sentencing data and a novel method for analyzing statistical interactions, we examine the main and combined effects of these statuses on sentencing. We find strong evidence for intersectionality: Harsher sentences concentrate among young black males and Hispanic males of all ages, while the youngest females (regardless of race/ethnicity) and some older defendants receive leniency. The focal concerns model of sentencing that frames our study has strong affinity with intersectionality perspectives and can serve as a template for research examining the ways social statuses shape inequality.

Using rich data linking federal cases from arrest through to sentencing, we find that initial case and defendant characteristics, including arrest offense and criminal history, can explain most of the large raw racial disparity in federal sentences, but significant gaps remain. Across the distribution, blacks receive sentences that are almost 10 percent longer than those of comparable whites arrested for the same crimes. Most of this disparity can be explained by prosecutors’ initial charging decisions, particularly the filing of charges carrying mandatory minimum sentences. Ceteris paribus, the odds of black arrestees facing such a charge are 1.75 times higher than those of white arrestees.

This article uses a natural experiment to analyze whether incarceration during the pretrial period affects case outcomes. In Philadelphia, defendants randomly receive bail magistrates who differ widely in their propensity to set bail at affordable levels. Using magistrate leniency as an instrument, I find that pretrial detention leads to a 13% increase in the likelihood of being convicted, an effect largely explained by an increase in guilty pleas among defendants who otherwise would have been acquitted or had their charges dropped. I find also that pretrial detention leads to a 42% increase in the length of the incarceration sentence and a 41% increase in the amount of nonbail court fees owed. This latter finding contributes to a growing literature on fines-and-fees in criminal justice, and suggests that the use of money bail contributes to a “povertytrap”: those who are unable to pay bail wind up accruing more court debt.

Two related lines of research have gained traction in the social sciences during the past three decades. One examines the association between race and punishment, while a second investigates stratification and colorism, defined as discrimination based on skin tone. Yet rarely do scholars examine these issues together. The current study uses new data to investigate the association between offender’s skin tone, Afrocentric facial features, and criminal punishment. More than 850 booking photos of black and white male offenders in two Minnesota counties were coded and then matched to detailed sentencing records. Results indicate that darker skin tone and Afrocentric facial features are associated with harsher sanctions and that the latter effect is particularly salient for white defendants. The findings add to existing work on skin tone and stratification and suggest that future research should consider other aspects of appearance, such as facial features, in the study of punishment and inequality.