More than half the population considers lack of security a significant national problem in Peru, and one in four citizens has reported being the victim of a crime. However, empirical evidence remains scarce and does not factor in the social cost of crime. To this end, this paper seeks to contribute by measuring the short-term and long-term impacts of crime victimization on trust in public institutions and identifying the vicious circles of distrust as suggested by the literature. We exploit a vast set of information from victimization surveys, police stations, and local government censuses by combining machine learning and matching algorithms. In line with the theory, we find that crime victimization reduces trust in public safety institutions in the short-term while eroding trust in institutions tasked with upholding criminal sanctions in the long-term. We also find that effects are critical for female and repeat victims. In addition to complying with balance and falsification tests, our results are robust to different types of matching and the potential presence of unobservable variables, which suggest that the findings may be causal.