Prof. Elliott Ash: Can Information Quality Explain Judge In-Group Bias? Evidence from Wisconsin Criminal Courts

VIRTUAL BEHAVIOURAL AND EXPERIMENTAL ECONOMICS RESEARCH SEMINAR

Talk by Prof. Elliott Ash, ETHZ

This paper studies in-group bias in Wisconsin criminal courts. Using records from 1.5 million cases from 2000-2017, we provide evidence that reconciles conflicting results on judicial in-group bias. We start by documenting baseline disparities and show that black defendants are more likely to get a jail sentence than comparable defendants of other races, while male defendants are more likely to get a jail sentence than comparable female defendants. In the aggregate, there is no in-group bias -- that is, judges do not tend to favor defendants of the same race or gender on average. However, we do find a racial in-group difference in the response to a recidivism risk score that we produce ourselves using a machine learning model trained to predict reoffense: judges are more lenient for same-race defendants who are low risk but harsher for same-race defendants who are high risk. We rationalize this through a model of judges' decisions with recidivism risk predicting that being in the in-group allows judges to have a more precise signal on the riskiness of defendants. In line with the model, we find that experienced judges are more responsive than inexperienced judges to recidivism risk in their sentencing decisions. Overall, the evidence is suggestive of statistical discrimination with better information about the in-group rather than taste-based discrimination based on out-group animus.

The link for participation in the event is the following: https://fernuni.zoom.us/j/61604483707.