Recently, there has been progress in understanding the underlying mechanisms of normative behavior. Altruistic punishment can solve the second order collective good problem of norm enforcement. However, the problem to detect norm violators has been overlooked in most recent advances. We combine ideas from labeling theory and social constructionism in a mathematical model and analyze the interaction between subjective beliefs about the dark field of norm violations on one hand and actual norm violations and control behavior on the other. We derive an inverse self fulfilling prophecy effect: If inspectors or norm targets believe in a large unknown extent of norm violations, actual norm violations typically decrease. Finally, we provide an experimental design to test our analytical results.