The top tail of South Africa’s earnings distribution 1993-2014: Evidence from the Pareto distribution

Abstract

We estimate the parameters of a Pareto distribution for South African earnings as measured through the October Household Surveys, Labour Force Surveys and Quarterly Labour Force Surveys, as assembled in the Post-Apartheid Labour Market Series (PALMS). We develop an outlier detection algorithm consistent with this distribution and then adjust the Gini coefficient for inequality in the top tail, using the robust estimation technique of Cowell and Flachaire. That procedure suggests that wage inequality is a bit higher than conventionally estimated. We also show that the top tail of the South African earnings distribution is ‘thick tailed’ and explore what that means. Our analyses show big shifts in the distribution in some of the surveys in ways that suggest measurement changes rather than changes in the underlying distribution.