We study compliance dynamics generated by a large set of behavioral rules describing social interaction in a population of agents facing an enforcement authority. When the authority adjusts the auditing probability every period, cycling in cheating-auditing occurs: Intensive monitoring induces compliance, but with high compliance there is incentive for lax monitoring; with less monitoring, compliance starts decreasing, and then there is an incentive to intensify monitoring. Thus, the real life phenomenon of compliance fluctuations is ex-plained by the nature of social interaction process rather than by exogenous parameter shifts. For the authority committed to a fixed auditing probability, we derive a sufficient condition for fines to be effective means of deterrence. Our analysis can be applied, among others, to crime, tax evasion, safety regulations, employment and environmental protection.
social interaction compliance random matching behavioral rule learning