We examine the social dynamics of crime by means of evolutionary game theory, and we model the choice of boundedly rational potential victims to privately self-protect against prospective offenders. Negative externalities from self-protection, as the socially transmitted fear of victimization, can influence the strategic choices of victims even with constant or declining crime rates, and this circumstance may lead to Pareto inefficient equilibria with excessive expenses for private protection. Providing higher levels of public security (or of appropriate social care) financed through discriminatory taxation of private defensive behaviors can prevent crime and reduce superfluous self-protection, thus driving the social dynamics toward a more efficient equilibrium. Public policy can therefore be effective in implementing the social optimum. This article extends previous work by Cressman, Morrison, and Wen by increasing the range of possible dynamics and the scope for public intervention. Consequently, in our model, public policy can deter crime and improve the welfare of victims by addressing the intangible aspects of crime, that is, the social dynamics of fear.