South Africa corporate tax gap study: A bottom-up approach based on audit assessments

While anecdotal evidence suggests widespread corporate tax evasion, systematic empirical data remain scarce. In this paper, we estimate the corporate income tax gap, the discrepancy between reported liabilities and true underlying tax obligations, for the entire population of small, medium, and large incorporated businesses in South Africa. Our analysis employs a bottom-up approach, leveraging a rich dataset of tax returns and the universe of taxpayer audits to quantify detectable non-compliance. By applying a random forest machine learning model to the microdata, we analyse the structure of evasion and estimate an underreporting gap of 13% across small, medium, and large firms. Our findings also reveal that the tax gap differs markedly across firms of different sizes and industries. Multinational firms exhibit systematically smaller tax gaps than national firms, whereas businesses in special economic zones exhibit systematically larger tax gaps than other firms