Research

"Firm-Bias, Skill-Bias, and Wage Inequality." (Appendix, Slides) R&R, IER.

Abstract. Changes to occupational and sectoral labor demand are thought to have been skill-biased, raising the skill premium and contributing to wage inequality. I show that these changes have also been firm-biased, impacting the distribution of firm-specific wage components such as shared rents. I characterize firm-bias empirically using German matched employer-employee data, and I quantify its 1993-2017 impact by structurally estimating a search-based model of occupational assignment with industry segmentation. While firm-bias is only marginally important in isolation, interactions with skill-bias - capturing demand's effect on the assortativity of labor markets - account for half of the rise in wage variance from occupational polarization and manufacturing decline. These interactions result in skill-bias being a poor overall predictor of wage outcomes, while policies that target firm rent-sharing can induce skill-biased demand shocks sufficiently strong as to reverse their effect on aggregate inequality. 

"What Drives Wage Sorting? Evidence From West Germany." (Appendix, Slides) Submitted.

Abstract. Many OECD labor markets are characterized by wage sorting: an observed but poorly understood association between high-earning individuals and high-paying employers. This paper leverages survey and administrative data to conduct a forensic analysis of West German wage sorting. I establish three main results. First, wage sorting is entirely a between-industry, between-occupation phenomenon; I find no evidence of assortative matching within markets. Second, wage sorting has strengthened over the 1993-2017 period due to (1) a shift in employment from low-skill, high-paying manufacturing sectors to low-skill, low-paying service industries, and (2) rising wages in skill-intensive jobs, which in all periods are associated with high-paying firms. Third, wage sorting is well-predicted by job characteristics related to innovation and IT. The source of these variables’ explanatory power is their ability to proxy for both employer size and worker skill, which in turn explain the preponderance of wage variance at the industry-occupation level but are themselves only weakly correlated. 

"Partial Job Automation." (Appendix, Slides) R&R, AEJ: Macro.

Abstract. I study the short-term and long-run effects of automation when jobs consist of multiple tasks. Using person-level survey data, I show that virtually all jobs contain both routine and non-routine tasks, and that computerization over 1979-2018 was associated with an increase in time spent on the latter. I develop an assignment-based model that captures these features. In the short-run, automation of a low-skill task endogenously generates labor market polarization. In the long-run, job losses from automation are at least partially reversed. I structurally estimate the model and demonstrate its ability to match historic wage and employment trends; implications for a displacement-minimizing tax scheme; and predictions regarding the timing and effects of AI adoption. 

Work in Progress

"The Adoption and Utilization of Technology." With Laurence Ales.

"The East German Wage Gap: TFP or Migration?" With Ashantha Ranasinghe.

"Measuring Task-Level Technological Exposure: A Language Model Approach."

"Missing Routine Work: Automation and the Life Cycle." With Natalie Duncombe.