Audits as Signals

A broad array of law enforcement strategies, from income tax to bank regulation, involve self-reporting by regulated agents and auditing of some fraction of the reports by the regulating bureau. Standard models of self-reporting strategies assume that although bureaus only have estimates of the of an agent's type, agents know the ability of bureaus to detect their misreports. We relax this assumption, and posit that agents only have an estimate of the auditing capabilities of bureaus. Enriching the model to allow two-sided private information changes the behavior of bureaus. A bureau that is weak at auditing, may wish to mimic a bureau that is strong. Strong bureaus may be able to signal their capabilities, but at a cost. We explore the pooling, separating, and semi-separating equilibria that result, and the policy implications. Important possible outcomes are that a cap on penalties increases compliance, audit hit rates are not informative of the quality of bureau behavior, and by mimicking strong bureaus even weak bureaus can induce compliance.