Research

Publication

Where Are the Workers? From Great Resignation to Quiet Quitting
with Jinhyeok Park and Yongseok Shin
paper

Abstract To better understand the tight post-pandemic labor market in the US, we decompose the decline in aggregate hours worked into extensive margin changes (fewer people working) and intensive margin changes (workers working fewer hours). Although the preexisting trend of lower labor force participation, especially by young men without a bachelor's degree, accounts for some of the decline in aggregate hours, the intensive margin accounts for more than half of the decline between 2019 and 2022. The decline in hours among workers was larger for men than women. Among men, the decline was larger for those with a bachelor's degree than those with less education, for prime-age workers than older workers, and also for those who already worked long hours and had high earnings. The reduction in workers' hours can explain why the labor market is even tighter than what is expected at the current levels of unemployment and labor force participation.

Work in Progress

Worker Learning and Management
Draft available upon request

Abstract

This paper shows that managers matter not only as supervisors of current production but also as developers of talent: they help develop future managers. Using German linked employer-employee data, I document two facts about workers’ paths into management that point to this role. First, workers exposed to higher-quality managers earn persistently higher wages once they themselves become managers, even after switching firms, consistent with the accumulation of portable managerial skill. Second, higher manager quality is associated with a lower short-run probability that a worker transitions into management. To reconcile these patterns, I develop a quantitative model with three key features: on-the-job search, worker-to-manager transitions, and endogenous coaching choices by managers. Workers search while employed and receive offers from other managers or to become managers themselves. Managers allocate time between production and coaching, where coaching lowers current output but raises workers’ future managerial skill. When workers transition into management, accumulated skill carries over and affects their managerial wages.

In the model, exposure to a high-quality manager raises both current match value and the option value of learning. Workers therefore delay entry into management until they draw sufficiently attractive offers, but earn higher wages once they do transition. Calibrating the model to post-transition wage profiles and transition hazards, I show that the model can jointly rationalize both patterns. The calibrated model allows a comparison between the coaching managers choose and the allocation that maximizes discounted output. The gap arises because managers choose coaching but capture only part of its return and workers carry the skill they acquire into renegotiated wages, outside offers, and their own managerial careers. As a result, the model implies that equilibrium coaching falls short of the level that maximizes discounted output.


Hours and Business cycles
with Alexander Bick and Yongseok Shin

Abstract We study intensive- and extensive-margin adjustments in hours over the cycle across heterogeneous worker groups in the U.S. and selected European economies. We document systematic heterogeneity by gender, age, and education, and analyze how shifting workforce composition shapes aggregate labor fluctuations. We then develop a model with both margins of adjustment and ex-ante worker heterogeneity to assess how increasing flexibility in hours (via technology and evolving norms) alters extensive-margin responses.

Recovering Historical Inflation Expectations Using Large Language Model (BERT)
with Miguel Faria-e-Castro, Fernando Leibovici and Yongseok Shin

Abstract We fine-tune a domain-adapted BERT on Federal Reserve speeches and FOMC minutes to predict contemporaneous survey- and market-based inflation expectations, then apply it to earlier archives to backcast a unified historical series with uncertainty bands. By exploiting BERT's bidirectional context, the model interprets forward-looking language that simple sentiment or n-gram indexes miss. Out-of-sample tests and event validations around regime shifts show the text signal is incremental to standard controls. The resulting series enables new evidence on expectation formation across eras, sharper identification in policy rules and VARs, and communication counterfactuals that separate information from guidance. We provide horizon-specific estimates, code, and phrase-level attributions to support replication and policy use.