This paper addresses two questions: First we identify indirect deterrence effects of a policy increasing expected prison sentences. Second we estimate endogenous social effects in crime. We use a unique dataset of former inmates released pursuant to the Collective Clemency Bill passed by the Italian Parliament on July 2006. We study how an individual’s post-release behavior (i.e. recidivism) varies with a) the average expected sentence at the time of exit from prison of former inmates imprisoned in the same facility; b) the prevalence of recidivism of the peers who served the sentence in his same prison. The identification strategy relies upon some features of the Collective Clemency Bill. The bill provides a three years reduction in detention for most prisoners in Italy. Inmates face different expected sentences for any potential crime because of different residual sentences at the time of the release. Indeed, former inmates will have to serve the residual sentence – varying between 1 and 36 months – suspended by the pardon in addition to the sentence given for the new crime if they recidivate. Given that the longer is the residual sentence the lower is the probability that a former inmate recidivates, in the model estimating the recidivism of an inmate we use the residual sentence of his peers as a source of exogenous variation in the recidivism of his peers. We find a negative impact of others’ expected sentence on individuals recidivism and a positive endogenous social interactions effect.