Poverty Amongst the Self-Employed

Conventional wisdom suggests that the self-employed in the South African labour market are involved in survivalist activities with very low returns. This paper analyses this assumption of widespread poverty amongst the self-employed. Hence, the earnings distribution amongst the self-employed and its determinants are examined. To this end, four models are presented that estimate the impact of these variables on the earnings from self-employment. The four models are a standard log-linear function, the logit and probit likelihood functions, and a log-linear earnings function that separates the sample ex ante into those below and those above the poverty line. Each model, from the standard earnings function, to the log its and probits shows that numerous covariates such as education, age, race, gender and location are important predictors of self-employment earnings. While the results differ between models, the differences in themselves suggest certain common characteristics of the self-employed, and also provide a point of departure for relevant policy intervention.