In this paper we estimate the impact of a nationwide public preschool expansion that took place in Spain over the 1990s on criminal behavior later in time. We exploit variation in enrollment rates across Spanish regions and birth-cohorts, and we link education data to a unique administrative crime dataset recording offenses committed in the region of Catalonia over the period 2009-2014. We find that for the average birth cohort, Catalan municipality and year, a 1 percentage point increase in preschool exposure at age 3 yields 1.6% fewer crime actions during youth and young adulthood. We are able to account for region of origin, birth cohort, time and local fixed effects, as well as several region and time-specific controls. Leveraging detailed information on types of crime committed, we propose a categorization of offenses into those likely to have been rationally planned and driven by economic motives, and those in which emotional factors and lack of self-control play a significant role. On average, we find the benefits of preschool to be larger and more robust on crimes belonging to the latter category, suggesting that non-cognitive skills play an important role in explaining the overall effect.