Working Papers

Job Preferences, Labor Market Power, and Inequality
Job Market Paper; Updated: April 2026

This paper studies how a firm's labor market power shapes, and is shaped by, its workforce, and evaluates the consequences for wage inequality and welfare. Using matched worker-firm panel data from Norway (1995-2018), I develop, identify, and estimate an equilibrium model of the labor market where firms compete for workers who are heterogeneous in both skills and preferences over wages versus non-wage job amenities. I allow workers' wage-amenity trade-offs to correlate with skills, while also varying among equally skilled workers. When a firm adjusts its wages, its workforce composition shifts, and this in turn affects the shape of its labor supply curve. Thus, a firm's wage setting power varies based on which workers it employs. I use the model to draw inference about the incidence of labor market power within and across firms and the implications for wage inequality and sorting. Eliminating market power widens within‑firm skill premia while compressing wage differences between firms, leading to a 4% reduction in overall wage inequality and a 3.3 percentage-point (23% of the baseline) decline in the gender pay gap. Variation in wage setting power also generates substantial allocative inefficiency, with welfare losses from imperfect competition estimated to be 9.6% relative to the competitive benchmark.

Discrete Choice with Generalized Social Interactions
Revise and Resubmit at Econometrica (2nd Round); Supplemental Materials

This paper examines how individual identity influences group behavior through social interactions. I study a discrete choice model in which people are affected differently by different members of their network, conforming to the actions of some peers while deviating from the actions of others. Under this generalized framework, I explore what aggregate outcomes arise from noncooperative decisionmaking. I analyze uniqueness and stability of equilibria, and I characterize how negative spillovers impact social welfare. I then show how to take the model to data, introducing a novel identification strategy that leverages within-network variation in individual characteristics to account for unobserved network effects. I also show how to construct internal instruments to overcome the issue of measurement error, which is a primary source of endogeneity in models with incomplete information. Lastly, I apply my method to data from the large-scale education experiment Project STAR, where I find strong evidence that classroom peer effects differ by gender.

The Linear-in-Means Model with Heterogeneous Interactions
(with Magne Mogstad and Alex Torgovitsky)

We study peer effects in linear-in-means models with heterogeneous interaction effects. The classical linear-in-means model imposes strict homogeneity on the interaction effects, yielding testable implications that can be readily examined in data. We relax these restrictions to allow for both positive and negative interaction effects that vary within and across groups. This extension makes the linear-in-means model suited to study a wide range of economic behaviors in addition to peer effects, including joint labor supply decisions within households and strategic interactions among firms. We analyze what can and cannot be learned from frequently used OLS and IV estimands for linear-in-means models under heterogeneous interaction effects. While these estimands do not lead to point identification, they can still be used to draw inferences about key economic quantities. We apply these results to two economic applications: classroom peer effects in Kenyan primary schools and strategic pricing decisions among cocoa traders in Sierra Leone. In each application, we reject homogenous interaction effects. Yet, we still draw meaningful inferences about endogenous interactions and social multipliers while allowing for heterogeneous interaction effects.

Linear Supply and Demand in Heterogeneous Markets
(with L. Henry de Frahan, Ingvil Gaarder, Magne Mogstad, and Alex Torgovitsky)

We modify the classic linear supply and demand system to allow for the coefficients on price to be unobservable random variables that vary across heterogeneous markets. Known conditions for point identification place strong requirements on the available instruments. We show how to construct and estimate bounds on scalar target parameters that are valid for any type of instrument, or even with no instrument at all. Numerical simulations calibrated to a well-known data set show that the model is not point identified. However, the bounds can be remarkably informative even under limited instrument variation. We apply our approach to study the welfare effects of sales tax.