The vast majority of the empirical literature on crime has focused on the effects of "supply-side" shocks such as the severity of laws and enforcement. In this paper we analyze the effects of a large and unexpected "demand-side" shock: the drop in daytime population in Washington, DC caused by the government shutdown of October 1-16, 2013. We derive implications from a simple theoretical model where criminals choose effort and allocate it across different criminal activities. We test these implications using the city of Baltimore as the comparison group, and employing difference-in-differences methods. Consistent with the model's predictions (and inconsistent with alternative explanations), we find a 3% decline in crime in DC during the shutdown period, with the net effect resulting from a 9% decline during the day hours, and a 5% increase in crime during the evening and night hours, indicating reallocation of criminals' effort induced by the shutdown.