We estimate the "incapacitation effect" on crime using variation in Italian prison population driven by eight collective pardons passed between 1962 and 1995. The prison releases are sudden – within one day –, very large – up to 35 percent of the entire prison population – and happen nationwide. Exploiting this quasi-natural experiment we break the simultaneity of crime and prisoners as in Levitt (1996) and, in addition, use the national character of the pardons to separately identify incapacitation from changes in deterrence. The elasticity of total crime with respect to incapacitation is between -20 and -35 percent. A cost-benefit analysis suggests that Italy's prison population is below its optimal level.