We document the level, type, and severity of crime in rural Bangladesh. Crime affected 36% of the households. The average cost of crime was worth two weeks' household consumption. We then study how microcredit affects crime. Our theoretical model shows how microcredit may increase cost of crime at the village level while reducing it for borrowers. We estimate a negative household level and a positive village level effect of microcredit on cost of crime, thus providing evidence that observable private protection against crime generates a negative diversion externality of crime towards other households.