Abstract

Making use of a unique administrative data set consisting of the universe of administrative filings in Rwanda, this paper investigates the impact of tax audits on businesses’ reporting behaviour. The evidence suggests that tax audits have a positive impact on corporate income and corporate tax liabilities reported for three years after the start of the audit process. The results also suggest that the type of audit matters. While ‘comprehensive’ tax audits have a significant positive effect on compliance, ‘narrow-scope’ tax audits exhibit both a positive and a negative effect during a three-year period after the audit, with the net impact being negative. The implication of this, from a tax compliance perspective, is that ‘narrow-scope’ audits are ineffective and that doing more of those and less of comprehensive ones might have a negative impact on tax compliance. Effective tax compliance strategy therefore requires the careful evaluation of all types of audits.

Introduction

Recent estimates have it that achieving the Millennium Development Goals requires increasing domestic revenues in low-income countries by around 15 percentage points of GDP, a target which requires the implementation of key policy reforms ( Gaspar et al., 2019 ). While the effort to boost domestic revenue mobilisation in developing countries continues, it is now even more challenging, and pressing, given the recent and much needed relief measures implemented to ease the impact of the COVID-19 pandemic. 1 Amongst the policy reforms for improving revenue mobilisation is strengthening tax administration capacity, an issue which has come to the fore for many countries around the world during the last two decades.

An integral part of tax compliance is operational audits, and the extent to which they are conducive to future compliance. There is a growing literature on audit assessment providing mixed evidence regarding the impact of tax audits on future compliance. 2 Surprisingly, however, little attention has been paid to the compliance impact of the types of audit, which are broadly categorised as ‘comprehensive’ and ‘narrow-scope’ (with the latter further categorised as desk-based or issue-oriented). Comprehensive audits are in-depth and in-person examinations conducted across different tax bases and fiscal years, whereas narrow-scope audits focus on a limited number of fields in a tax return. This distinction in audit types, and their impact on tax compliance, is at the heart of the contribution of this paper.

Results

This section presents the results, starting with the aggregate ATT, followed by the audit-type-specific aggregate ATT, and concludes with a back-of-the-envelope calculation of the return on investment (ROI) of audits. In Appendix B we present further sensitivity analysis which validates the results presented in the main text and the methodology used. 37

۶٫۱٫ Aggregate ATT

A crucial assumption behind any DID analysis is the existence of a common previous trend in the outcome variable at the time of the treatment ( Meyer, 1995 ) which, in the present context, means that one should observe a similar pattern (trend) in reporting behaviour of audited and unaudited taxpayers before treatment. To test for this, as well as estimate the period-specific audit effects on the outcome variables, we rely on Weighted Fixed Effect regressions based on panel data from 2013 through 2018 and weights obtained from our CEM stratification. In the regressions we also include both individual and year fixed effects. The excluded category is the last year before the treatment is applied (2015). Fig. 5 , Fig. 6 present the results of this analysis for taxable income and tax liability, respectively, largely confirming that there is no statistically significant difference in trends between audited and matched controls before audits take place. After treatment, the estimates indicate a positive and increasing effect of audits on subsequent tax reporting behaviour of audited taxpayers for the three years after the treatment and the two outcome variables.