Marginal Deterrence at Work

We test the rational economic model of marginal deterrence of law enforcement --- i.e., the need for graduating the penalty to the severity of the crime. We use a unique data set, which combines individual-level data on sentence length for a representative sample of US inmates with proxies for maximum punishment and monitoring costs across US states over 50 years. We show that the penalty is increasing in the level of the offense. Consistent with the marginal deterrence framework, we also document that a decrease in maximum penalty or an increase in monitoring cost are associated with longer sentences and higher monitoring rates. We also provide evidence that the effects of maximum penalty and monitoring cost are stronger in states where income inequality is higher. Finally, we show that steeper sanctions are associated with less harmful crimes. Overall, these findings favor the marginal deterrence framework over the maximal penalty principle and other competing theories of justice.