Abstract: The U.S. Prohibition experience shows a remarkable policy reversal. In only 14 years, a drastic shift in public opinion required two constitutional amendments. I develop and estimate a model of endogenous law enforcement, determined by beliefs about the Prohibition-crime nexus and alcohol-related moral views. In turn, the policy outcomes shape subsequent learning about Prohibition enforcement costs. I estimate the model through maximum likelihood on Prohibition Era city-level data on police enforcement, crime, and alcohol-related legislation. The model can account for the variation in public opinion changes, and the heterogeneous responses of law enforcement and violence across cities. Results show that a 15% increase in the homicide rate can be attributed to Prohibition enforcement. The subsequent learning-driven adjustment of local law enforcement allowed for the alcohol market to rebound to 60% of its pre-Prohibition size. I conclude with counterfactual exercises exploring the welfare implications of policy learning, prior beliefs, preference polarization, and alternative political environments. Results illustrate the importance of incorporating the endogenous nature of law enforcement into our understanding of policy failure and policy success.