Deterrence in networks

Abstract

We propose a deterrence mechanism that utilizes insider information acquired by criminals through customary practices. Under this mechanism, a suspect caught committing a criminal act can nominate a peer who has committed a similar offense, with only the more severe offender facing penalties. Theoretical analyses indicate that, under general conditions, our mechanism drives the best-response dynamic downwards compared to the commonly used regulatory practice of penalizing only the first suspect. Experimental data confirms the mechanism's deterrence effect, but unveils deviations from equilibrium predictions: the deterrence effect is weaker than anticipated and insensitive to network structures summarizing insider knowledge. To understand this, we analyze post-experiment questionnaire responses and find evidence that some participants employ level-k rather than Nash strategies. Structural estimation confirms that the level-k specification better fits the data than Nash. These findings inform policymakers of the potential usefulness and constraints of the peer-informed audit mechanism.