This paper studies the non-market effects of education on crime using a panel dataset for the 20 Italian regions over the period 1980-1995. Our empirical results suggest that education reduces crime over and above its effect through labour market opportunities (employment rate and wage rate). Because of the absence of a credible instrumental variable for education for Italy, our empirical strategy is to include in our econometric specification region fixed effect, year fixed effects and region-specific time trends together with an extensive set of socioeconomic and deterrence variables. Our results are robust to model specification, changes in the typology of crimes and finally, to alternative definitions of education. (Contains 4 tables.)