A distributional analysis of union-wage effects in South Africa: Evidence from panel data.

This paper examines whether individual workers' time-invariant unobservable characteristics influence estimates of the South African union wage premium across the length of the conditional wage distribution. It employs the 2001–7 South African Labour Force survey and a fixed effects quantile regression estimator. Results show a relatively large (small) wage premium at the bottom (top) of the conditional wage distribution when workers' time-invariant unobserved characteristics are ignored. Accounting for this set of factors substantially reduces the wage premium at all points of the distribution. In fact, the wage premium becomes somewhat constant across the conditional wage distribution, suggesting that unions in South Africa have little wage compressionary effects. [ABSTRACT FROM AUTHOR]