Abstract: We study how artificial intelligence (AI) affects workers’ earnings and employment by combining German online job vacancies with administrative records from 2017–2023. AI exposure is measured at the occupation–region–year level as the share of vacancies requiring AI skills and instrumented with national trends in AI demand. On average, we find no meaningful displacement or productivity effects. Yet, expert workers gain while non-experts often lose, with returns shaped by occupational task structures. We also document AI-induced expansions in analytic and interactive tasks that are linked to earnings gains. Overall, our results highlight distributional concerns in the early stages of AI. (Current Version, Work-In-Progress)
(joint Myrielle Gonschor & Marc-Justin Schmidt)
Abstract: Economic crises can prompt firms to reorganize their workplaces and adjust the skills they demand —but which of these changes persist, and which are merely transitory? We study this question by exploiting the COVID-19 pandemic as an exogenous shock to workplace organization, leveraging data on pre-pandemic working-from-home (WFH) feasibility and detailed skill information from over 11 million German online job vacancies from 2017 – 2024. Using an event-study framework that compares skill demand within firms, but across occupations with varying degrees of pre-pandemic WFH feasibility, we find that shifts in firms’ skill demand have been sizable but transitory, typically lasting between one to three years. Our analysis also sheds light on post-pandemic skill demand. Notably, demand for analytic and interactive skills showed temporary adjustments in WFH-feasible jobs, possibly reflecting crisis-induced workplace reorganization, before reverting to pre-pandemic levels by late 2024. Similarly, changes in demand for manual skills associated with “essential work” proved short-lived. Mechanism analyses further reveal that these pandemic-induced skill adjustments were (i) driven by specific skill clusters, (ii) mediated by firm characteristics, and (iii) shaped by sector-specific exposure to remote work. Combined, our findings suggest that while economic crises can trigger short-run adjustments in how firms organize work and define skill needs, they do not necessarily lead to persistent changes in the underlying skill content of jobs.
(Work-in-Progress, Draft coming soon!)
(joint w/ Niklas Benner, Felix Heuer & Rebecca Kamb)
Description: We study how firms adjust their recruitment behavior when labor markets tighten. Using German online job vacancy data and official statistics, we measure tightness at detailed occupation–region levels and track how it shapes firms’ vacancy posting, wage offers, and skill requirements. Our framework focuses on both extensive adjustments (e.g., whether firms post more vacancies or start advertising wages) and intensive adjustments (e.g., skill and digital content of jobs). These perspectives allow us to shed light on whether firms in tighter markets substitute scarce skills with digital technologies, raise wage offers, or adjust hiring standards.
(Work-in-Progress)
(joint w/ Niklas Benner & Roman Klauser)
Description: We study how emerging digital technologies reshape labor demand. Using German online job vacancy data linked to administrative establishment records, we track the diffusion of general-purpose technologies such as artificial intelligence and cloud computing across regions, industries, and occupations since 2019. We measure their automation and augmentation potential by analyzing shifts in skill requirements within vacancies and across establishments. Combining descriptive evidence with a shift–share design, we examine how firms adjust employment structures and wages in response to exposure to these technologies. This framework allows us to identify whether emerging technologies substitute existing tasks, expand new ones, or alter the overall demand for labor.
(Work-in-Progress)
(joint w/ Niklas Benner, Verena Malfertheiner & Michael Stops)