Prediction Errors, Incarceration, and Violent Crime: Evidence from Linking Prosecutor Surveys to Court Records

Incarceration is often justified by a defendant’s risk of future crime. To what extent do biased beliefs about predictors of crime distort incarceration decisions? We survey prosecutors about how violent rearrest rates vary by defendant age and criminal history. Surveyed prosecutors make systematic errors: They underestimate the decline in rearrest with age and overestimate the increase with criminal history. By linking prosecutors’ beliefs to their quasi-randomly assigned cases, we show that prosecutors’ beliefs predict incarceration patterns by defendant age and criminal history in their cases. Finally, we find that prosecutors with more accurate beliefs simultaneously reduce violent rearrest and incarceration. (JEL D83, K14, K41, K42)