Australian Wage Trends and Correction for Sample Selection: A Comparison of Blau and Beller and Heckman Procedures

Two recent papers by Blau and Beller (1988, 1992) examine the trend in relative wages by gender and race in the US. The papers question whether the observed trend in earnings ratios are an illusion, the pattern of wage convergence being confined to the chosen sample rather than the population as a whole. The papers illustrate the potential importance of a changing pattern of labour force participation and thus correction for non-random sampling. Blau and Beller (1988) utilise the Heckman (1979) two-step estimator, the conventional approach to address the problem of sample selection bias. Blau and Beller (1992) introduce a novel alternative, described below, which builds on earlier work by Smith and Welch (1986). Despite the close link between the two papers, no attempt is made to compare results or integrate the approaches to the correction of wage trends for changes in sample composition. The current note attempts to provide such a link in the context of an illustrative analysis of relative male/female wage trends in Australia over the 1980s.