Racial Bias and In-group Bias in Judicial Decisions: Evidence from Virtual Reality Courtrooms

Samantha Bielen, Wim Marneffe, Naci H. Mocan

NBER Working Paper No. 25355
Issued in December 2018, Revised in May 2019
NBER Program(s):Health Economics, Law and Economics, Labor Studies, Public Economics

We shot videos of criminal trials using 3D Virtual Reality (VR) technology, prosecuted by actual prosecutors and defended by actual defense attorneys in an actual courtroom. This is the first paper that utilizes VR technology in a non-computer animated setting, which allows us to replace white defendants in the courtroom with individuals who have Middle Eastern or North African descent in a real-life environment. We alter only the race of the defendants in these trials, holding all activity in the courtroom constant, creating arguably perfect counterfactuals ( Law students, economics students, practicing lawyers and judges are randomly assigned to watch with VR headsets, the trials that differed only in defendants’ race. Background information obtained from the evaluators allowed us to identify their cultural heritage. Evaluators made decisions on guilt/innocence in these burglary and assault cases, as well as prison sentence length and fine in accordance with the guidelines provided by the relevant law. There is evidence that evaluators are harsher towards defendants of their own race during the guilt-innocence decision. In the sentencing phase, we find in-group bias in the opposite direction: evaluators are more lenient towards defendants of their own race. This is the first paper that identifies such opposing effects of bias within the same decision sequence. Regarding overall racial bias, we document that minorities are more likely to get convicted, and that they are given stiffer punishment, conditional on conviction. We find only scant evidence that the concerns of the evaluators about terrorism, about immigration, or their trust in the judiciary or the police have an impact on their judicial decisions, suggesting that the source of the bias may be deep-rooted. Merging a small sample of judges and prosecutors with the sample of lawyers provides results that are very similar to those obtained from the analysis of lawyers.

download in pdf format
   (1523 K)

email paper

Machine-readable bibliographic record - MARC, RIS, BibTeX

Document Object Identifier (DOI): 10.3386/w25355

NBER Videos

National Bureau of Economic Research, 1050 Massachusetts Ave., Cambridge, MA 02138; 617-868-3900; email:

Contact Us