Tax evasion contributes to the low tax-to-GDP ratios observed in many low-income countries, and tax audits are one of the main tools governments use to combat it. To what extent tax audits increase revenue collection is therefore a key policy question.
Henning & Okello Ayo (2026) estimate the causal effect of comprehensive tax audits conducted on firms by the Uganda Revenue Authority (URA) in financial year 2020/21. These intensive audits involve direct interaction with firms and may examine any part of a firm’s tax affairs over the previous five years. In 2020/21, the URA ranked firms using data-driven risk scores and assigned the highest-risk firms to comprehensive audits, creating a sharp assignment cutoff: crossing the cutoff increased the probability of receiving a comprehensive audit by 64 percentage points. Firms just below the cutoff instead received lighter issue or desk audits. The paper’s estimates capture the effect of assigning one additional firm to a comprehensive rather than a lighter audit.
Provides No Benefit on Average
The net revenue raised is calculated as the sum of (1) increased tax revenue due to tax corrections issued as a result of the audit (~ +$29,000), (2) reduced VAT liability reported over the first five months after the audit (~ -$31,000), and (3) reduced CIT liability reported for the year after the audit (~ -$5,400). The reduction in reported VAT and CIT liability results from a combination of firm exit, movement to the informal sector, and reduced sales by remaining firms.
The net revenue raised is then: $29,000 – $31,000 – $5,400 = -$7,400 per marginal comprehensive audit. This is the paper’s conservative case: it assumes the URA eventually collects the full correction amount, that conducting the audit is costless, and that the behavioral effects last only as long as the observation window. If instead the assumptions are that the URA collects 60% of corrections (as its audit reports show) and the VAT effect persists for a full year rather than five months, the net cost is -$62,400 per marginal audit. Neither figure includes the direct cost to the URA of conducting the audit.
The paper assumes the firms’ willingness to pay to avoid the audit is positive.
Because firms’ willingness to pay to avoid a comprehensive audit is positive while the net revenue effect is negative, the MVPF for conducting a more intensive audit on the marginal firm is negative (Boning et al., 2025). In other words, it provides no benefit on average.
Boning, William C., Nathaniel Hendren, Ben Sprung-Keyser, and Ellen Stuart (2025). ‘A Welfare Analysis of Tax Audits Across the Income Distribution’. The Quarterly Journal of Economics 140(1): 63–112. DOI: https://doi.org/10.1093/qje/qjae037.
Henning, David, and Joseph Okello Ayo (2026). “Tax Audits and Their Distortionary Effects.” Working Paper. https://djhenning.github.io/assets/pdf/Henning_JMP.pdf