The Speaker: Oskar Nordström Skans
Topic: The Heterogeneous Earnings Impact of Job Loss Across Workers, Establishments, and Markets
rganizer: Maria Nareklishvili
Time: 16th May 2024 9:00-10:00 (BST)
On May 16th, 2024, Maria Nareklishvili hosted a webinar featuring Oskar Nordström Skans at the Royal Statistical Society (RSS) to discuss "The Heterogeneous Earnings Impact of Job Loss Across Workers, Establishments, and Markets." Professor Skans, from the Department of Economics at Uppsala University, is known among others for his theoretical and empirical research on labor market dynamics, including network models, labor market transitions, wage determinants, productivity, and worker sorting.
The webinar informed us about the effects of mass layoffs on workers' wage trajectories. Utilizing generalized random forest approach and comprehensive Swedish administrative data, the article shows that the earnings impacts of job displacement due to establishment closures are highly heterogeneous across different worker types, establishments, and markets. The findings indicate that the decile of workers most affected by job displacement experiences a 50% reduction in annual earnings in the first year post-displacement, with cumulative losses reaching 200% over seven years. Conversely, the least affected decile incurs only a 6% loss in the first year. Notably, workers in the most affected decile had lower pre-displacement wages and declining earnings trends. These workers are also more susceptible to adverse market conditions. The research highlights that older workers in routine-task-intensive jobs face the highest predictable displacement effects when employing simple targeting rules.
The webinar was highly relevant for an audience interested in applied statistics, econometrics, and economics. The speaker introduced machine learning methods to estimate the heterogeneous effects of mass layoffs on workers’ earnings dynamics. The intersection of machine learning and empirical labor economics has recently gained prominence due to at least two key factors. Firstly, the availability of high-quality administrative data allows researchers to develop and apply statistical methods specifically designed to investigate workers’ labor market outcomes. Secondly, advancements in causal machine learning literature enables applied economists to estimate policy effects with greater flexibility. This highlights the significance of emerging statistical and modern machine learning techniques in understanding individual behavior and decision-making