This study examines how drug injection and needle sharing propensities respond when a needle exchange program (NEP) is introduced into a city. I analyze 1989-1995 Drug Use Forecasting data on adult male arrestees from 24 large U.S. cities, in nine of which NEPs opened during the sample period. After controlling for cocaine and heroin prices, AIDS prevalence, fixed effects for cities and years and city-specific time trends, the model indicates that the presence of a NEP is associated with declines of 13 percent in drug injection and 20 percent in needle sharing among drug injectors.