Abstract
Using data from the General Household Survey 1973–95, it is shown that there were substantial differences between the earnings of whites and nonwhites in the self-employment sector of the British labour market over the period. Fitting a model of earnings determination which takes account of sample selectivity allows these differences to be decomposed into characteristics and coefficients effects. It turns out that both are important. Further decomposition reveals a crucial role for differences in the amount of schooling received by whites and nonwhites and in the returns to age, gender, marital status and qualifications.