We argue that the decision to bribe bureaucrats depends on the frequency of corruption within a society. We provide a behavioral model to explain this conduct: engaging in corruption results in a disutility of guilt. This implies that people observe a lower probability to be involved in corruption if on average the guilt level of others within a country is higher. We also explore whether - and to what extent - group dynamics or socialization and past experiences affect corruption. In other words, we explore theoretically and empirically whether corruption is contagious and whether conditional cooperation matters. We use the notion of ?conditional corruption? for these effects. The empirical section presents evidence using two data sets at the micro level and a large macro level international panel data set covering almost 20 years. The results indicate that the willingness to engage in corruption is influenced by the perceived activities of peers and other individuals. Moreover, the panel data set at the macro level indicates that the past level of corruption has a strong impact on the current corruption level.