Predicting labour force types

A small group of people accounts for a large majority of flows between labor market states and of spells in un- and non-employment. In this paper, we ask whether it is possible to identify those weakly attached to the labor market during their prime working-age years using information available early in their lives. First, we use information on labor force transitions between ages 30 and 50 contained in the long panel provided by the NLSY79 to identify those weakly connected to the labor market during their prime age. To do so, we use k-means clustering on moments describing observed spells in employment, unemployment, and non-employment between 30 and 50. This points to a group of less attached individuals who are disproportionally female, less educated, and in poor health. In a second step we predict, using information collected at various points before age 30 - which we do not use in clustering - whether individuals will turn out to belong to the weakly attached type in their prime age. We find that information from ages 22 to 29 allows predicting membership of the low-attachment group with high precision. Particularly influential is information on early labor market experiences and health. The fact that we can predict weak and strong labor market attachment during prime working age using variables observed in individuals' twenties suggests the presence of persistent heterogeneity that shapes labor market experiences throughout the life cycle.