Firms face increasingly large applicant pools: a larger pool of candidates may improve match quality, but it also raises the cost of screening. This paper studies how expanding the set of candidates made visible to a recruiter affects the firm's hiring decision: whether the firm selects a candidate, whom it selects, and how fast.  In order to overcome the endogeneity issue that appears when considering vacancies and their applicants, I exploit the assisted preselection service of the French Public Employment Service, in which quasi-randomly assigned caseworkers screen applications and send a shortlist to the employer. One additional transmitted candidate raises the probability that the firm makes a contract proposal from the shortlist within three months by 3.1 percentage points (10 percent of the baseline), with strongly diminishing returns: gains are concentrated where shortlists are small, wages are low, and labor markets are tight. Expansion changes the composition of the choice set: marginal candidates are less experienced and fit the vacancy less well, yet the number of candidates above a basic quality threshold also increases. Despite the added noise, firms do not select observably worse candidates and do not take longer to decide. A sequential screening model rationalizes these patterns: as evaluation is optional, an extra credible candidate is an option the firm can decline, so its option value remains always positive. The results imply that public intermediation is most effective when it targets shortlist expansion toward thin candidate pools rather than uniformly increasing applicant volumes.

Using data from the French Public Employment Service linked to social security records, we examine the shifts in the supply of and demand for information technology (IT) skills that followed the recent emergence of generative artificial intelligence (GenAI). We show that the large-scale deployment of GenAI tools between late 2022 and late 2023 was followed by a sharp decline in vacancies for IT-specialist jobs and, more broadly, for jobs requiring IT skills. While relative wages remained largely unchanged, jobs requiring IT skills also experienced a substantial increase in applicants per vacancy and a marked decline in hiring. These findings are consistent with a model in which GenAI and natural-language coding boost the productivity not only of expert programmers but also of ordinary software users, particularly the least skilled.

Age discrimination remains a significant barrier in the French labor market, despite policies aimed at extending working lives. This study provides the first experimental evidence on the role of human capital stereotypes in hiring decisions for senior workers. Using a large-scale correspondence study with over 12,000 applications, we examine how distance to retirement, inter-firm mobility, and training influence callback rates. Our results show a strong age penalty across four occupations: on average, the probability of receiving a callback falls by one third between ages 59 and 62, declining from 12.5% to 8.1%. We then examine whether this penalty can be explained by differences in retirement horizon, recent training, or career mobility. Taken separately, none of these dimensions has a statistically significant effect However, when combined, these signals allow a 62-year-old candidate to catch up with a 59-year-old, and a 59-year-old to catch up with a 49-year-old. We also document heterogeneity by contract type: extending the retirement horizon has a positive and statistically significant effect for permanent contracts (CDI), but a negative effect for fixed-term contracts (CDD). Overall, these findings indicate that increasing seniors’ remaining work horizon alone is insufficient to eliminate age-related hiring penalties. Supporting career trajectories upstream remains a necessary condition for improving older workers’ labor market prospects.

This paper constructs global harmonized historical series on labour hours by gender, employment status in 57 core territories – 48 main countries and 9 residual regions – covering all world regions across the 1800-2025 period. We quantify the global decline in labour hours and relate it to the long run rise in productivity, with sizable variations across regions and periods . At the global level, hourly productivity (net domestic product per work hour) rose from about 0.7€ in 1800 to 16€ in 2025 (PPP 2025 €). In 2025, hourly productivity ranges from 4€ across Subsaharan Africa to 55-60€ in the USA, Sweden, Germany or France. In the long-run, about 35-40% of the rise in productivity was used to reduce labour hours and obtain extra leisure and 60-65% to raise production. We also stress the role of power relations and unpaid labour in the changing structure of labour hours throughout the 1800-2025 period. In particular, we find that the gender gap in hourly pay is currently much larger than usually thought once we include unpaid domestic work. Using this definition, the gender pay gap reaches 40-50% in rich countries, as opposed to 10-20% in conventional estimates. Finally, based on historical trends, we discuss future trajectories for labour hours, productivity, gender inequality and structural transformation over the 2025-2100 period. In our central scenario, we estimate that global hourly productivity could reach about 100€ in all countries by 2100, together with substantial reduction in work hours and gender gaps and large sectoral reallocation of labour time away from the most polluting sectors.

We examine the impact of the death of an adult child on household wealth accumulation at old age. Using a combination of theoretical modelling and matching difference-in-differences empirical design, we provide evidence on how the death of a child alters economic decision-making of parents. We develop a life cycle model of liquidity-specific wealth accumulation incorporating a dynastic component in the bequest motive. Our theoretical framework predicts that wealth accumulation and wealth composition diverge based on the death of a child and dynastic disruption. Using Dutch register data, we empirically disentangle the effect of the death of a child (bereavement effect) and the effect of dynastic discontinuity (dynastic motive effect). Our results suggest that the dynastic motive operates primarily in the composition rather than the level of wealth, through portfolio adjustment liquidating symbolically-loaded illiquid assets for liquid ones. These findings contribute to the broader debate on bequest motives and the role of dynastic preferences in wealth accumulation at old age, highlighting the differential responses across dynastic structures.