Research

Who will gain and who will lose as AI automates tasks? While much of the discourse focuses on job displacement, we show that job transformation - a shift in the task content of jobs - creates large and heterogeneous earnings effects. We develop a quantitative, task-based model where occupations bundle multiple tasks and workers possessing heterogeneous portfolios of task-specific skills select into occupations by comparative advantage. Automation shifts the relative importance of tasks within each occupation, inducing wage effects that we characterize analytically. To quantify these effects, we measure the task content of jobs using natural language processing, estimate the distribution of task-specific skills, and exploit mappings to prominent automation exposure measures to identify task-specific automation shocks. We apply the framework to analyze automation by large language models (LLMs). Within highly exposed occupations, like office and administrative roles, workers specialized in information-processing tasks leave and suffer wage losses. By contrast, those specialized in customer-facing and coordination tasks stay and experience wage gains as work rebalances toward their strengths. Our findings challenge the common assumption that automation exposure equates to wage losses.

I show that unobserved sorting patterns of firms and workers across space can account for the tight link between rising aggregate wage inequality and rising spatial inequality in West Germany. Two-sided sorting patterns of workers and firms interact with a change in technology to produce a spatially concentrated increase in inequality, driving up regional disparities. These sorting patterns are determined jointly in equilibrium and depend on theoretical objects that are difficult to measure in the data. This paper develops a novel bi-clustering method to recover these objects empirically and uses these results to structurally estimate a dynamic spatial search model with two-sided sorting. I find that regional sorting of firms is more pronounced than regional sorting of workers and the former is an important determinant of workers' job ladders and lifetime values. Compensating differentials between regions are large, driven in part by better labor market outcomes in rich places. The model allows me to consider the redistributive effects of spatial policy, which I find to be strong.

We develop an equilibrium model to study the dynamic adjustment of a frictional labor market to aggregate shifts in the demand and supply of a job amenity. When preferences for the amenity are heterogeneous in the population, and its availability is heterogeneous across jobs, labor reallocation ensues. The defining traits of such reallocation (a rise in vacancies and job-to-job transitions, a fall in matching efficiency and in relative wages of jobs supplying the amenity) closely resemble those observed in the post-pandemic U.S. labor market in the aftermath of the shift to remote work. A version of the model calibrated to the U.S. experience matches the data well with shocks of plausible magnitude. Cross-sectional and survey data from various sources offer support for this mechanism.

This paper explores the role of selection in shaping the dynamics of unemployment during recoveries. A matching model with many-to-many matching and permanent worker heterogeneity delivers such selection and generates recovery unemployment dynamics that mirror the data closely. In line with empirical evidence, the model predicts that, during a recession, firms become more selective and job finding rates decline more for less productive, unemployed workers. This reinforces negative composition effects and creates a feedback loop, which slows down the recovery. I find empirical support for the cyclicality of job seeker quality implied by the model in data from the NLSY.

We study a general equilibrium model of the labor market in which agents slowly learn about their suitability for jobs. Our model reproduces desirable features of the data, many of which standard models fail to replicate. We explore how, in such an environment, asymmetric information can lead to substantial misallocation. We calibrate our model to US data and quantify the welfare loss arising from misallocation due to informational frictions. The tractability of the model allows us to explore the responsiveness of wages and employment to an aggregate shock. We find that wage rigidity arises endogenously because of protracted learning, and in line with the data, the model is able to generate a larger and more persistent employment response.

We propose a new identification strategy to assess the efficacy of macroprudential measures. We use a novel instrumental variable based on the idea that a politically sensitive macroprudential measure is more likely to be implemented if a politically independent institution, such as a central bank, is in charge. Our results show that borrower-based macroprudential measures have had a strong and statistically significant dampening effect on credit growth in the European Union. 

Information frictions and matching efficiency