Research

Automation, immigration, and climate crisis are redrawing who gains and who loses in advanced economies. In turn, they shape what voters, workers, and employers ask of the state. My research studies what people exposed to these pressures want from the welfare state, both the kind of social and economic protection they favor and whom they think it should cover.

What people want often depends on who is losing and why. People react differently to the same job loss when the worker is a woman instead of a man, or when the job goes to a migrant worker instead of a machine. I am interested in these judgments about who deserves protection and in how they carry into support for social policy. More recently, I have turned to the politics of AI infrastructure, where the costs of technological change fall on communities as well as on workers.

I begin from material interest, which I often approach through occupation. The work people do shapes the risks they face, and those risks shape the policies they support. However, material interest is only half the story. People who face the same risk often ask for different things, because their social identity, the institutions around them, and the way a policy is framed all affect how economic risk turns into political demands. Sorting these out takes more than one kind of evidence, so I work across methods: survey analysis, experiments, computational approaches to text, and case studies.

Left: The Leader of the Luddites, 1812, via Wikimedia Commons.
Right: Alfred Stieglitz, The Steerage, 1907, via Wikimedia Commons.

Working Papers

Saving His Job, Not Hers? - Selective Protection in Automation-Driven Job Loss (Soohyun Cho and Jaewook Lee)

– Awarded SASE 2025 Early Career Workshop
Abstract

Existing studies link automation risk to demand for social protection but pay less attention to whether responses depend on whose jobs are at risk. Does displacement affecting women generate the same support for government intervention as displacement affecting men? We argue that expectations casting men as primary earners lead citizens to view men’s job losses as more consequential, producing gender-selective demand for protection. Using a pre-registered survey experiment in South Korea, we find that male displacement reduces opposition to an automation tax relative to female displacement and increases support for it among respondents who deny ongoing gender inequality. Neither vignette shifts support for unemployment benefits or retraining. Respondents also perceive laid-off men as primary earners more often than laid-off women, a gap concentrated among the same respondents and present among women who hold these attitudes. These findings suggest that gendered evaluations of economic loss shape demand for particular government interventions.

Teaching Machines to Read Occupations: Introducing the OPUS Dataset of Occupational Partisanship (Scott Patterson, Krzysztof Pelc, and Jaewook Lee)

Abstract

Many of the richest social data come as open-ended, write-in responses. Occupational data are a prime example, yet the difficulty of coding such free-text entries has long limited research. We address this challenge by fine-tuning a large language model (LLM) to classify millions of write-in occupation fields into a standardized taxonomy. Treating classification as a text-generation task, we supplement standard occupational descriptions with a large corpus of alternative job titles, generating synthetic data for infrequent occupations. The resulting classifier matches or exceeds the accuracy of the US government’s benchmark tool, especially in the long tail of rare occupations. We use this method to construct the Occupational Partisanship in the United States (OPUS) dataset, which assigns partisan scores to nearly all occupations in the O*NET standard classification. Two findings illustrate its potential: (i) occupation predicts partisan identity as well as, or better than, county of residence; and (ii) majority-female occupations lean consistently Democratic. Our approach provides a low-cost, generalizable framework for classifying write-in data and a new empirical lens on the political structure of the American labor market.

Migrants, Machines, and the Legitimacy of Job Displacement: Comparative Public Opinion in Institutional Context (Jaewook Lee)

Abstract

A growing literature connects automation to immigration politics, but less is known about how citizens compare immigration and automation directly as sources of worker displacement. This paper examines whether citizens judge the same economic outcome differently when job loss is attributed to migrant workers rather than labor-saving technology. Using a pre-registered vignette experiment in the United Kingdom and Sweden, I find that immigration-led displacement is consistently evaluated more negatively than a generic layoff and automation-driven displacement, and that this asymmetry partly carries into policy demand. An offshoring benchmark — which also involves foreign labor but without local worker presence — shows no such penalty, suggesting that the reaction is not driven by foreign labor per se. At the country level, Swedish respondents differentiate more sharply between sources than British respondents regarding workplace-norm and regulatory dimensions. The findings suggest that the legitimacy of economic adjustment is therefore source-dependent, and the displacement linked to migration may carry a political penalty that automation does not attract.

Governing Labor Scarcity: A Comparative Map of Public Preferences across Six Policy Instruments (Jaewook Lee)

Abstract

Population aging is reshaping the agenda of the welfare state, turning labor scarcity into a problem that cuts across family, migration, working time, and the governance of workplace technology. Yet little is known about how citizens evaluate this menu of responses when asked to weigh the instruments side by side. Using the 2024 OECD Risks That Matter survey across 27 countries, this paper offers a comparative map of public receptivity to six labor shortage policy instruments. Support follows a hierarchy that is broadly consistent across countries: technology adoption, activation of women and underrepresented groups, and moving part-time workers into full-time hours attract broad support, migration is consistently the least popular, and longer working lives and pronatalism occupy a contested middle. Demographic and socioeconomic differences are concentrated on the instruments most consequential for each group, though more clearly for some pairings than others. The findings speak to the political feasibility of alternative social policy responses to demographic change.


Work-in-Progress (Selected)

Beyond Party Lines? - Local Opposition to the AI Data Center Buildout in the United States

Using an original tracker of 231 county-level opposition movements, this study shows that resistance to AI data centers emerges in red and blue communities alike. A pre-registered survey experiment finds that partisan leadership shapes willingness to participate, while bipartisan coalitions attract as much support as same-party efforts.

The Electoral Costs of Computing: Data Centers and the Politics of Environmental Footprint
– Awarded APSA 2025 Centennial Center Research Grant

Combining geocoded data center locations with county-level 2024 election results, this study finds that exposure is associated with greater Green/protest voting. An original survey experiment shows that environmental and economic cost frames increase willingness to vote against pro-data-center candidates and support for ending tax breaks across party lines.

Uncertain Returns: Automation Risk and the Politics of Social Investment (with Jiyeong Jeon)

We argue that uncertainty about the effectiveness of reskilling in an age of rapid AI advancement drives workers to favor immediate compensation over long-term social investment. Using cross-national survey data, the study highlights how declining confidence in re-employment prospects under automation reshapes support for welfare policies.

European Competitiveness and Attitudes Towards EU Fiscal Integration (with Lars van Doorn and Olaf van Vliet)

This paper investigates how narratives about European economic competitiveness shape public preferences for EU-level fiscal integration. Through survey experiments across Europe, it shows that priming the need for building Europe-wide competitiveness increases support for fiscal coordination— especially among respondents in manufacturing sectors who are most likely to benefit from it.

When Growth Breeds Exclusion: Macroeconomic Prospects and Welfare Chauvinism in Europe (with Lars van Doorn and Olaf van Vliet)

Drawing on a novel survey experiment across thirteen European countries, this paper tests whether economic growth or decline affects exclusionary welfare preferences. We find that personal economic optimism can paradoxically heighten welfare chauvinism, especially among socioeconomically disadvantaged respondents.

Occupational Asymmetry in American Electoral Campaign Finance (with Scott Patterson)

This study uses machine-learning classification of Federal Election Commission contribution data to map occupational inequalities in campaign finance. It reveals that donations are overwhelmingly concentrated among high-income professionals, underscoring the occupational roots of political inequality in the United States.