Inequality in South Africa is the enduring legacy of racial discrimination. We use a dynamic perspective to show the linkages between persistent effects of discrimination in the labour market and the efficacy of redistributive fiscal policy in reducing inequality. We present a machine-learning analysis based on household survey data in the Post-Apartheid Labour Market Series to predict the main drivers of the relationship between workersâ€™ heterogeneous socioeconomic characteristics, the behaviour of variables related to labour market status, and labour income inequality. The empirical investigation covers the period 2000â€“17. Drawing on this preliminary evidence, we build a dynamic stochastic general equilibrium model with a dual labour market and job search frictions that represents the structural features of South Africaâ€™s economy, which can be used to assess the effects of fiscal policy on inequality in the postapartheid period and to simulate the effects of alternative fiscal measures and labour market reforms.