We exploit the Collective Clemency Bill passed by the Italian Parliament in July 2006 to evaluate the indirect effects of a policy that randomly commutes actual sentences to expected sentences for 40 percent of the Italian prison population. We estimate the direct and indirect impact of the residual sentence – corresponding to a month less time served in prison associated with a month of expected sentence – at the date of release on individual recidivism. Using prison, nationality and region of residence to construct reference groups of former inmates, we find large indirect effects of this policy. In particular, we find that the reduction in the individuals' recidivism due to an increase in their peers’ residual sentence is at least as large as their response to an increase in their own residual sentence. From this result we estimate a social multiplier in crime of 2.