Luc Behaghel

Age Biased Technical and Organisational Change, Training and Employment Prospects of Older Workers

We analyse the role of training in mitigating the negative impact of technical and organizational changes on the employment prospects of older workers. Using a panel of French firms in the late 1990s, we first estimate wage bill share equations for different age groups. As a second step, we estimate the impact of ICT, innovative work practices and training on employment flows by age group in the next period. Training appears to have a positive impact on the employability of older workers, but it offers limited prospects to dampen the age bias associated with new technologies and innovative work practices.

Please Call Again: Correcting Non-Response Bias in Treatment Effect Models

We propose a novel selectivity correction procedure to deal with survey attrition in treatment effect models, at the crossroads of the Heckit model and the bounding approach of Lee (2009). As a substitute for the instrument needed in sample selectivity correction models, we use information on the number of prior calls made to each individual before obtaining a response to the survey. We obtain sharp bounds to the average treatment effect on the common support of responding individuals. Because the number of prior calls brings information, we can obtain tighter bounds than in other nonparametric methods.

Retirement, Early Retirement and Disability: Explaining Labor Force Participation after 55 in France

We analyze the influence of health and financial incentives on the retirement behavior of older workers in France, building upon Stock and Wise (1990) option value approach. The model accounts for three main retirement routes: the normal retirement, disability insurance (DI) and unemployment/preretirement pathways, and is estimated with a combination of microeconomic datasets that include the French data of the European SHARE survey. The estimates confirm that a decrease in the generosity of the pension and DI schemes induces people to stay longer in the labor market, and that people with better health tend to retire later. We present extreme situations simulating what individuals retirement behavior would have been if only one retirement route had existed and in the absence of constraints on work capabilities. We show that average years of work between 55 and 64 are nearly 14% greater when regular retirement incentives are applied to the whole population than when it is DI rules that are systematically applied.

Skilled Labor Supply, It-Based Technical Change and Job Instability

We provide empirical evidence on the impact of IT diffusion on the stability of employment relationships. We document the evolution of different components of job instability over a panel of 348 local labor markets in France, from the mid-1970s to the early 2000s. Although workers in more educated local labor markets adopt IT faster, they do not experience any increase in job instability. More specifically, we find no evidence that the diffusion of IT increases job-to-job transitions, and we find that it tends to reduce transitions to nonemployment among high-school dropouts. Overall, the evidence goes against the view that the diffusion of IT has spurred job instability. Combining local labor market variations with firm data, we argue that these findings can be explained by French firms’ strong reliance on training and internal promotion strategies in order to meet the new skills requirement associated with IT diffusion.

Next please! Estimating the effect of treatments allocated by randomized waiting lists

Oversubscribed treatments are often allocated using randomized waiting lists. Applicants are ranked randomly, and treatment offers are made following that ranking until all seats are filled. To estimate causal effects, researchers often compare applicants getting and not getting an offer. We show that those two groups are not statistically comparable. Therefore, the estimators arising from that comparison are biased and inconsistent. We propose new estimators, and we show that they are unbiased and consistent. Finally, we revisit two applications and we show that using our estimators can lead to sizably different results from those obtained using the commonly used estimators.