In a national sample of the Italian population, surveyed four times between October 2002 and January 2007 (N = 2,008), we performed a multilevel longitudinal study aimed at predicting the increase in crime risk perception as a function of three families of independent variables, respectively lying at the within individual level (direct victimization and indirect victimization), at the between-individuals level (being a woman, being an older person, being a poorly educated person and size of area of residence) and at the ecological level (county's crime rate, unemployment rate and immigration rate). Direct and indirect victimization, being a woman, being an older person, living in a large town and in a context characterized by high crime and unemployment rates positively influenced the change in crime risk perception, while the other individual and ecological predictors we used in our predictive model did not. Strengths, limitations, implications and future developments of this research are discussed.