The Value of Information for Regulatory Enforcement

Abstract:

Regulatory enforcement often depends on information produced by agents who may have incentives to distort it. We study this tension in the context of the Italian Court of Auditors, where local public finance auditors report to national judges. We model this process as involving two frictions: imperfect detection of irregularities and incentives to underreport them. A reform that randomized auditor assignments provides a natural experiment that potentially increasesthe first friction (by reducing local knowledge) while reducing the second one (by increasingindependence). Combining novel administrative data with a machine-learning-based measureof fiscal risk, we find that enforcement increased and became better targeted toward high-risk municipalities. We trace this improvement to the transmission channel: randomly assigned auditors report more irregularities, especially in high-risk municipalities and where pre-reform local ties were strongest. Experienced judges translate these improved signals into enforcement, concentrating deliberations on high-risk cases. The results suggest that in this setting,the benefits of auditor independence dominate the costs of reduced local knowledge.