Thursday, November 17, 2011
Traditional economic models of criminal behavior have straightforward predictions: raising the cost of expected cost of crime via apprehension probabilities or punishments decreases crime. Estimating the extent to which increased punishments deter crime is often complicated by omitted variable bias, as the severity of a crime is often a key determinant of the punishment. I overcome this obstacle by taking advantage of discrete thresholds for blood alcohol content (BAC) which determine punishments for driving under the influence (DUI). Regression discontinuity derived estimates suggest that having a BAC above the DUI threshold reduces recidivism by up to 2 percentage points (17 percent). As previous DUI violations increase future penalties for drunk driving, this is consistent with Beckerian models of criminal activity. However, enhanced penalties for aggravated DUI also reduce recidivism by an additional percentage point (9 percent), despite the fact that the enhanced punishments only affect the current penalties. This suggests a form adaptive expectations play a role in criminal behavior, whereby expectations of future punishments are based upon previous punishments experienced.