Both the demand for skilled labor and the skill wage premium have become increasingly dispersed across U.S. local labor markets over recent decades. This paper examines how withinoccupation technological change contributes to these uneven regional developments. Using a novel measure based on shifts in task intensities across 430 detailed occupations, the analysis shows that occupations evolve toward cognitive-intensive task content more rapidly in densely populated labor markets. These developments are closely linked to innovation diffusion, with innovation-related occupational task reallocation being substantially stronger in dense regions. The results further show that greater exposure to within-occupation cognitive-biased technological change is associated with faster employment growth among college-educated workers and lower wage growth among workers without a college degree, thereby widening local skill premiums. Decomposing these wage effects reveals that occupational re-sorting explains only a modest share of the observed changes. Instead, approximately 80 percent of the increase in wage inequality is attributable to within-occupation adjustments.

Although more than 20 per cent of the workforce changes their occupation every year, we still do not fully understand the mechanisms behind the observed mobility. This paper focuses on analysing the relationship between work-hour instability and occupational mobility in the US labour market. I use the longitudinal dimension of the Current Population Survey (CPS) to measure individuals’ intra-year work-hour variation and analyse their mobility through a balanced occupation panel. Being in the highest quartile of work-hour variation is associated with a higher mobility rate of 0.33% for men and 0.81% for women compared to an average monthly mobility rate of 1.71%. Analysing the predicted marginal effects across different household compositions suggests that the substantial gender gap can be explained by the intra-household specialisation of men and women. The last part of this study shows that only workers with highly volatile work hours sort themselves into more stable occupations.

Does artificial intelligence displace workers or make them more productive? We argue that it does both, through distinct channels that conventional one-dimensional exposure measures cannot separately identify. We distinguish these channels by linking the text of U.S. patents to occupations’ tasks and tools with natural language processing: innovations resembling an occupation’s tasks are classified as automation, those resembling its tools as augmentation. Exploiting the gradual spatial diffusion of pre-2000 AI breakthrough innovations, we estimate economically large but opposing effects through these two channels. A one-standard-deviation increase in local AI automation lowers occupational employment by 25 percent, with little effect on wages; AI augmentation instead raises wages by 7.5 percent and employment by 10 percent. Moreover, relative to comparable non-AI innovation, AI strengthens both the labor-displacing effects of automation and the productivity-enhancing effects of augmentation. Our findings suggest that the labor-market consequences of AI depend critically on the balance between automation and augmentation rather than on overall AI exposure alone.

Recent research by Autor and Thompson (2025) develops a novel framework for measuring occupational expertise from occupational task descriptions and shows that expertise explains an important share of wage differences across occupations as well as changes in occupational expertise over time. Building on this work, our project develops a multilingual framework for measuring occupational expertise that exploits Wikipedia's hierarchical knowledge structure to recover comparable expertise measures across languages without relying on English-specific linguistic resources. These measures provide a foundation for studying how occupational expertise evolves across European labor markets over time, enabling comparisons across countries and across local labor markets within countries. The project aims to improve our understanding of how the evolution of occupational expertise contributes to divergences in wages, employment, and regional economic development across Europe.

In this project, we leverage unique Polish registry data on ex-prisoners and exploit a policy reform that increased the minimum wage of prisoners to the same level as the statutory minimum wage. After a statewide implementation of this policy, it became more difficult for incarcerated individuals to find a job, which is essential for them to collect work days that qualify for unemployment benefits upon release. This setting creates a compelling quasi-experiment to study the effect of liquidity on re-employment outcomes for ex-prisoners. For causal identification, we use the prison entry date, which determines exposure to the high-cost prison work regime, as an instrument for qualifying for unemployment insurance. Naïve OLS estimates suggest that benefit claimants reintegrate into the labor market more quickly compared to non-claimants. In contrast, using the IV identification strategy reveals that claimants fare worse than non-claimants, remaining unemployed for longer. These findings underscore the importance of accounting for selection into prison work and suggest that a within-prison work policy that eases liquidity after release may unintentionally discourage job search.

Women are significantly underrepresented among innovation leaders, including startup founders, patent inventors, and venture capitalists. This raises the question of whether technological change is gender-biased in ways that systematically favor masculine work values and male-dominated occupations, and how such dynamics shape the evolution of the female labor market. I investigate these questions for the U.S. by constructing a measure of gender-biased technological change based on time-varying occupation data combined with female employment shares across occupations, and by linking this measure to patent data, decennial Census data, and political survey data. The pre-analysis shows that recent technological change is associated with declining demand for skills and work values traditionally emphasized in female-dominated occupations, alongside relative demand shifts toward male-dominated, cognitive-intensive jobs. These patterns suggest polarizing effects within the female labor market, with college-educated women adjusting more flexibly to technology-driven changes than women without a college degree, widening gaps in job content, family orientation, and political ideology between low- and high-skilled women.

This project investigates whether artificial intelligence changes the returns to worker ability within occupations. Combining individual-level panel data with novel occupation-specific measures of AI automation and augmentation, we examine whether AI amplifies or compresses wage differences among workers performing the same occupation and identify the mechanisms through which these effects arise.