On the Link Between On-the-Job Training and Earnings' Dispersion

This paper attempts to devise a methodology that allows estimating the exact impact of training on the dispersion of wages. The approach is derived from a decomposition technique recently proposed by Fields (2003). It extends Fields' approach by taking into account population subgroups and selectivity bias. The empirical illustration is based on French data. The results show that when a distinction is made between workers who receive and did not receive training the between groups dispersion explains only 5.5% of the overall variance of earnings. Most of the earnings dispersion is a within groups dispersion and there is a lot of overlapping between the earnings' dispersion of the two groups. Such findings imply that even though unobserved heterogeneity plays a key role in the selection of those who receive training and thus has an important impact on the between groups dispersion it cannot be a variable lying behind labor market segmentation. There is thus a much greater degree of heterogeneity within that between the two groups.