Do general amnesty programs lead to reductions in the crime rate among immigrants? We answer this question by exploiting both cross-sectional and time variation in the number of immigrants legalized generated by the enactment of repeated amnesty programs between 1990 and 2005 in Italy. We address the potential endogeneity of the "legalization treatment" by instrumenting the actual number of legalized immigrants with alternative predicted measures based on past amnesty applications patterns and residential choices of documented and undocumented immigrants. We find that, in the year following an amnesty, regions in which a higher share of immigrants obtained legal status experienced a greater decline in non-EU immigrant crime rates, relative to other regions. The effect is statistically significant but relatively small and not persistent. In further results, we fail to find any evidence of substitution in the criminal market from other population groups - namely, EU immigrants and Italian citizens - and we observe a small and not persistent reduction in total offenses.