Abstract

This study examines the impact of risk-based tax audits on future compliance behavior. Using administrative data from the Rwanda Revenue Authority (RRA) on nearly 9,000 domestic tax audit cases between 2018 and 2024, the study applies an adapted difference-in-differences methodology for multiple treatment periods. The study finds that tax audits are associated with significant decreases in reported tax declarations in the years following an audit, even as the tax authority’s assigned risk scores decline. These patterns suggest that taxpayers may reduce their reported declarations after being audited, despite appearing less risky to the administration. Taxpayers who incur higher audit penalties exhibit even sharper drops in declarations post-audit, alongside increases in their assigned risk scores. These results are primarily driven by taxpayers identified as non-compliant during their audit, pointing to a strong loss-repair mechanism in which taxpayers may under-declare in an effort to recover audit-related financial losses. However, the findings are not limited to non-compliant taxpayers: even those who faced no penalties experience a temporary decline in declarations up to two years post-audit before returning to prior levels. Taken together, these dynamics indicate that tax audits may reduce compliance in some contexts — even among initially compliant taxpayers, albeit temporarily — a concerning behavioural response that may not be fully captured by the tax administration’s risk scores.