Despite widespread popular accounts linking crack cocaine to inner-city decay systematic research has analyzed the effect of the introduction of crack on urban crime. We study this question using FBI crime rates for 27 metropolitan areas and two sources of information on the date at which crack first appeared in those cities. Using methods designed to control for confounding time trends and unobserved differences among metropolitan areas find that the introduction of crack has substantial effects on violent crime but essentially no effect on property crime. We explain these results by characterizing crack cocaine as a technological innovation in the market for cocaine intoxication and by positing that different types of crimes play different roles in the market for illegal drugs. In a market with incomplete property rights and inelastic demand, a technological innovation increases violence on the part of distributors but decreases property crime on the part of consumers. We also find evidence that the increase in urban crime during the 1980's occurred in two distinct phases: an early phase largely attributable to the spread of crack and a later phase largely unrelated to it.