Research

My dissertation, Essays in Methodology and Political Economy, develops two connected strands of research. Substantively, I study how labor market institutions and economic insecurity shape support for redistribution and political behavior. Methodologically, I develop statistical methods and research designs for settings in which observed data imperfectly measure theoretical quantities or provide no obvious counterfactual comparison. The dissertation consists of the three essays below.

Coarsened by Design: Measurement Error Bias from Grouped Variables in Interacted Regressions

Working paper, 2026

Social scientists theorize about continuous quantities such as income but often observe them only through a few ordered brackets, which they replace with category ranks or midpoints. I show that this practice induces a measurement error that can rescale a true interaction effect or fabricate a spurious one, and I develop a regularized interval regression with an empirical Bayes extension that corrects it. In six replications of published studies, the corrections change a published conclusion in five. Draft available upon request.

Labor Union Solidarity, Occupational Risks, and Demand for Redistribution in Europe

Working paper, 2026

Why do union members support more redistribution than their material interests alone would predict? Using ten rounds of the European Social Survey across twenty-six countries, I show that members support redistribution more than comparable non-members, that the difference is largest among the workers whose risk of unemployment is lowest, and that it is wider where unions organize low-income workers. Draft available upon request.

Identifying the Diffusion and Suppression of Political Naming in Turkey with a Matched Interrupted Time Series Design

Working paper, 2026

How does cultural behavior signal political alignment under authoritarian rule, and how does coercion suppress it? Using more than 28 million Turkish birth records from 1960 to 1990, I trace how naming newborns Deniz spread after the 1971 death sentence of the revolutionary Deniz Gezmiß and contracted after the 1980 military coup. A matched interrupted time series design builds the counterfactual from forty comparison names when no untreated region exists, and a Bayesian multilevel model maps where the name spread and where it was suppressed.

Download here