The correlation between hot weather and crime is well documented but not fully understood. We combine millions of administrative records, victimization surveys on unreported crime, and daily weather information to analyze the effect of temperatures on crime in Mexico. We find that sample selection cannot explain the observed positive correlation between temperature and crime. Moreover, we find that shifts in alcohol consumption and time use on weekends are responsible for 28 percent of temperature-induced crimes. We also observe changes in the hour and location of crimes, providing new evidence on the importance of time use as a determinant of crime. (JEL D91, K42, O13, O17, Q54)