This paper estimates a dynamic demand model for liquor in the United States using panel data from 43 states. Because of taste changes over time and between states in liquor consumption, purely time series or cross sectional studies do not elicit reliable price elasticity estimates. This study makes the case for pooling and shows how one can control for individual state effects and endogeneity of the regressors using estimators suited for a dynamic demand model. Our results indicate that the long-run price elasticity is in the -0.7 range. The findings also support strong habit persistence, a small positive income elasticity, and very weak evidence of bootlegging from adjoining states. The magnitude of the long-run price effect suggests that sin taxes can serve not only as an important income source but also as a significant deterrent effect.