Assessing Unemployment Traps in Belgium Using Panel Data Sample Selection Models

In this paper we investigate whether unemployment traps exist and are significant in the transition from unemployment into employment in Belgium. In order to assess them, we use panel data sample selection models. Specifically, we estimate a parametric random effects models composed by a wage equation and a selection equation by maximum likelihood techniques. The income ratios have been computed for every individual in the sample by using the predicted wages obtained from the estimation of the wage equation corrected for sample selectivity. The empirical analyses has exploited the data extracted from the waves 1993-1997 of the Panel Study of Belgian Household and has been led separately on (unbalanced) samples of men and women. The estimation results suggest significant differences in the behaviour of these two groups. The experience of long periods of unemployment in the past is particularly important: long-term unemployed people have difficulties in re-integrating the labour market and they obtain low salaries when they succeed in finding employment. Long unemployment spells are likely to have a "scarring" effect on subsequent earnings. Moreover, the computation of income ratios for all individuals highlights the importance of unemployment traps for the women present in the sample. Indeed, their expected wage is often lower than their income while being unemployed. A significant proportion of the available samples (men and women) is shown to enter employment although this transition is accompanied by a substantial loss in their disposable income