Employment of young and older workers: three policy evaluations

The dissertation has the goal of providing a better understanding on the effectiveness of specific active labour market policies that aim at integration of young and older workers. The first Chapter evaluates the effectiveness of the 2003 reform of the Italian apprenticeship regime is assessed. This reform raised the age-eligibility and revised the training component. The different timing of the implementation of the reform in the Italian regions and sectors is exploited. To estimate the ATT of the apprentices in the reformed regime (compared to the old regime), the Covariate Balancing Propensity Score estimator is implemented (CBPS - Imai and Ratkovic, 2014) employing a large set of covariates. The inflow sample of about 18,000 apprentices hired in 2007 is drawn from administrative data of the Social Security Institutions (INPS). Four years after hiring, the reform induced an increase in the transition rate to permanent jobs in the same firm and boosted the average wage of the apprentices. Finally, by a DiD estimator on the LFS, it is found that the higher diffusion of the apprenticeship among the youth becoming eligible is offset by a reduction of other temporary jobs. The research related to the second and the third Chapters assesses the effectiveness of Belgian federal policies to boost the employment rate of the older population. In this research we rely on an endogenous stratified sample of administrative data containing about 244,000 individuals (aged between 52 and 61 years old in 2002) with their employment history since 1957. The second Chapter assesses the impact of a Belgian employers’ Social Security Contributions reduction for workers older than 58. The analysis is performed on multiple repeated cross-sections even if panel data are available, to account for age-varying confounding factors. We use a CDiD estimator (Heckman et al., 1997) and when needed a trend-adjusted version of it (Wolfers, 2006). To facilitate the integration of endogenous sampling weights in this estimator, we implement it as an Inverse Probability Weighting (IPW) estimator, which we extend to allow for multiple cross-sections in the before and after periods. We find small positive short-run impacts on working time and larger ones on the employment rate, but only for employees at high risk of leaving to early retirement. The wage is not affected. In a Cost-Benefit-Analysis, we estimate that during the 1.25 years after its introduction the subsidy imposed a net monthly cost of 330€ per saved job to Society. Had the subsidy been targeted to sectors where early retirement schemes are widely used, Society would instead have gained 2,080€ per saved job. The third Chapter evaluates the impact of the Belgian part-time Time-Credit scheme for older workers. The policy measure allows older workers to reduce their working time by 20% (or 50%) with the goal of postponing their retirement decision and possibly improve their work-life balance. Workers receive a lump-sum in-work benefit of about € 215 (€ 385), granting an average income replacement of 90% (66%) of the full-time wage. We assess the ATT on the survival in employment and we control for selection on observable (IPW) by using their whole employment history. As control units can enter the treatment in later periods, we take into account the dynamic treatment selection (Vikström, 2014). Our estimates indicate a positive employment effect in the short-run followed by a negative impact after four years with insignificant health effects. The policy does not pass the Cost-Benefit-Analysis test.