The Department of Business and Management Science and NoCeT invite you to a faculty seminar with Michael Best, Columbia University.
The threat of audit is at the core of efforts to limit tax evasion. Studying a nation-wide random audit program in Pakistan, we show that audits detect unpaid taxes in one third of cases, and in these cases they detect unpaid taxes of 40% of sales on average. However, only 22% of this amount is recovered. While the detected and recovered amounts are largest for the very largest firms, so are the costs of the audits, such that on net, the benefits of audits are largest for firms around the 90th percentile of the firm size distribution.
Exploiting a switch from random to a public audit targeting rule, we find no evidence of behavioral responses to either the probability of audit or the event of being audited. Counterfactual simulations suggest that alternative targeting rules using the same information as the actually-implemented rule could have achieved 50% higher detection or 45% higher tax recovery. Our results highlight the role of the state's capacity to collect detected unpaid taxes, without which the deterrence effect of audits is limited. We show how this limitation to fiscal capacity can dramatically change the way that audits should be targeted.