Anticipated Monitoring, Inhibited Detection, and Diminished Deterrence

Monitoring programs—by creating expected costs to regulatory violations—promote compliance through general deterrence, and are essential for regulating firms with potentially hazardous products and imperfectly observable compliance. Yet, evidence on how monitoring deployment affects perceived detection probabilities and—by extension—compliance, is sparse. Beginning in May 2020, pandemic-related protocols in Maricopa County, Arizona, required routine health inspections to occur by video-conference at food establishments with vulnerable populations (e.g., hospitals and nursing homes). Unlike conventional on-site inspections—which continued at most food establishments—these "virtual" inspections were scheduled in advance, and thus, easily anticipated. The virtual format also likely inhibits observation of some violations, further reducing detection probability. Tracking five violations that are detected by tests in both inspection formats, I find evidence of substantial anticipation-enabled detection avoidance. Comparing against contemporaneous on-site inspections, virtual inspections detect 53% fewer of these specific violations relative to pre-treatment levels, and that decrease reverses entirely when treated establishments are subsequently inspected on-site. Detected counts of all violations decrease 41% in virtual inspections. Consistent with general deterrence, this decrease is more than offset in establishments' first post-treatment on-site inspections, where detected counts exceed the pre-treatment average by 28%. Across establishment types and compliance histories, deterrence-effect heterogeneity suggests a simple dynamic enforcement rule would better allocate existing inspection resources, and might meaningfully reduce social noncompliance costs.