South Africa exhibits extreme levels of income inequality and is ranked as one of the most unequal countries in the world. In order to measure these severe levels of inequality, it matters how we account for the different parts of the income distribution. Although the approach has gained international attention, there has not been any attempt at combining tax administration data with household survey data in order to account for incomes at all parts of the distribution and especially from the top of the income distribution in South Africa. This paper uses a novel technique to identify the optimal method of combining tax administration with household survey data. Our results show the dramatic effects of accounting for reporting bias in household surveys by using tax administration data. When combining the two datasets, we find a significant decrease in overall inequality of taxable income in South Africa between 2011 and 2014, the two years under observation. Nonetheless, income inequality in South Africa remains high. For our analysis, we use two waves of the National Income Dynamics Study, a national representative household survey, and compare the information to a sample of almost 1.2 million records on personal income tax for the 2011 tax year and about one million records for the 2014 tax year.