Published and Forthcoming
How Much Should we Trust Estimates of Firm Effects and Worker Sorting? (with Stéphane Bonhomme, Thibaut Lamadon, Elena Manresa, Magne Mogstad and Bradley Setzler)
Many studies use matched employer-employee data to estimate a statistical model of earnings determination with worker and firm fixed effects. Estimates based on this model have produced influential yet controversial conclusions. The objective of this paper is to assess the sensitivity of these conclusions to the biases that arise because of limited mobility of workers across firms. We use employer-employee data from the United States and several European countries while taking advantage of both fixed effects and random effects methods for bias correction. We find that limited mobility bias is severe and that bias correction is important.
For code cf. here.
The Two Faces of Worker Specialization (with Zsófia Barany)
We study how worker specialization—the distance between a worker’s skill set and those prevalent in the labor market—shapes employment outcomes. Using US and French data, we first document that specialized jobs are characterized by asymmetric skill profiles and a scarcity of nearby employment opportunities. We incorporate these features into a random search model with multidimensional skills, mismatch penalties and skill complementarity. We show that specialization lowers job-finding rates due to a lack of suitable jobs, but raises re-employment wages via improved productivity. Empirical evidence from displaced workers in both countries confirms these predictions. Our findings reconcile competing views in the literature by showing that specialization entails trade-offs and is neither uniformly beneficial nor harmful.
Earlier version with random search model solved with Neural Networks.
Life-Cycle Wage Growth and Firm Productivity in Developed and Developing Economies (with Juan Munoz)
Life-cycle wage growth varies substantially across countries. We study the firm productivity distribution's role in a random-search model with learning and labor-market frictions, estimated for Brazil, Colombia, and the United States. Brazil has a long left tail of low-paying firms, and Colombia has a highly compressed firm ladder. Replacing each country's firm distribution with that of the United States closes most of the cross-country gap in wage growth, raising Brazil's relative wage growth from 73 to 103 percent and Colombia's from 52 to 115 percent. The results are robust to alternative distributional assumptions and additional sensitivity checks.
Forthcoming at IER
Working Papers
Wage Bargaining and Wage Posting Firms (with Jean-Marc Robin)
This paper studies the coexistence of wage-posting and wage-bargaining firms. We embed two protocols in a common job-ladder model in which posting firms offer wage–tenure contracts and matching firms renegotiate against outside offers, and develop a maximum-likelihood classification procedure treating wage-setting regimes as latent. Using a German employer survey to benchmark the classification and Austrian matched employer–employee data for estimation, we obtain consistent evidence across settings. Bargaining firms account for about 23% in Austria and 24% in Germany; they occupy higher rungs of the ladder, exhibit greater within-firm wage dispersion, employ a more skill-intensive workforce, and concentrate in services. The model rationalizes wage cuts on job-to-job moves as transitions from posting to bargaining firms in exchange for future renegotiation-driven growth. Allowing posting firms to offer wage–tenure contracts is quantitatively essential: a no-growth specification inflates the implied share to 34% by misattributing deterministic wage growth to bargaining heterogeneity.
Revise and Resubmit at The Review of Economic Studies.
The Role of Workers in Knowledge Diffusion Across Firms (with Anders Akerman)
We study how worker mobility contributes to aggregate productivity growth by diffusing knowledge across firms. We develop a multi-worker-firm framework with random search and on-the-job mobility in which firm productivity improves through in-house innovation and through hiring workers from other firms. Using Swedish matched employer--employee and product-level data, together with event-study evidence based on shift-share international trade shocks, we show that the extent and direction of worker mobility affect firm productivity. We estimate that mobility-driven knowledge diffusion accounts for over 44% of growth in average firm productivity. Our results suggest that lower worker mobility, and more mobility from less productive firms, can depress growth in average firm productivity.
Under review.
Lost Knowledge and New Products: Event-Study Evidence from Worker Deaths (with Anders Akerman)
Do worker-embodied knowledge stocks shape firms’ ability to expand product scope? We exploit unexpected worker deaths as plausibly exogenous shocks to firm knowledge, using Swedish linked employer–employee data to recover workers’ prior product experience from their former employers. Event-study estimates show that each year of prior product-specific experience lost to a worker death lowers the on-impact probability of entering the corresponding product by about 0.13 percentage points, implying a roughly 0.7 percentage point decline — about 27 percent of the baseline rate — for a worker with average prior experience. The effect peaks in the first post-event year before attenuating. Static estimates confirm that cumulative lost experience predicts lower subsequent entry and substantially lower product-level sales. A descriptive decomposition indicates that the experience-scaled component accounts for about 92 percent of the sales decline and about 58 percent of the entry decline, with the latter interpreted more cautiously because entry is rare. Effects are roughly twice as large for university-educated workers, markedly attenuated for core products, and suggestive of greater exposure among high-skill firms. A back-of-envelope calculation implies an expected loss of roughly $70,000 per exposed worker for a product line with $10 million in net entry value, providing a benchmark for codification, retention, and redundancy investments.
Revise and Resubmit at Labour Economics
Separations on the Job Ladder (new version with Lewin Nolden)
We document a U-shaped relationship between within-job wage changes and separations in administrative data from Italy, Germany, and Austria. To interpret this relationship, we estimate a random-search model with worker and firm heterogeneity, learning, persistent match productivity, and wage renegotiation, using Italian data. For Italy, matching the U-shape attributes the decline in wage volatility primarily to learning and implies a 14.5-year half-life for match productivity. Imperfect information lowers output by 0.229 percent through misallocated mobility. Omitting the U-shape moments attributes the volatility decline to negotiation history, shortens the half-life to 3.5 years, and reduces the output loss to 0.073 percent.
Under review.
Workplace Amenities and the Gender Pay Gap (with Yana Gallen, Nabanita Datta Gupta and Kristian Stamp Hedeager)
We study how workplace amenities shape the gender wage gap by linking Danish firm-level amenity data to administrative earnings records and an incentivized survey of workers' willingness to pay for job attributes. We document gender differences in hours, schedules, and commuting distance that widen after parenthood. Mothers move toward firms with lower wage premia, but this sorting largely disappears after controlling for firm amenities. Survey evidence shows that women especially dislike evening and on-call work and value proximity to home and schedule flexibility. Incorporating these non-wage amenities reduces the gender gap from 21% in earnings to 16% in total compensation.
Under review. Presented at NBER SI 2026 Gender in the Economy.
Preserving Jobs, Freezing Dynamics? Employment Effects of Conditional Inheritance Tax Relief (with Charlotte Bartels, Theresa Buehrle, Gedeão Locks and Duncan Roth)
We analyze the impact of tying preferential tax treatment of family firms undergoing succession to an employment preservation clause (EPC), which requires the new owners to maintain a specified fraction of the firm's pre-transfer wage bill. Exploiting the 2016 extension of the German EPC to firms with 6--20 employees and linked ORBIS-IAB employer-employee data, we show that exposed firms forgo employment growth and shift the workforce toward cheaper labor. The composition shift begins before the transfer, while pre-transfer headcount remains unchanged. After the transfer, an employment gap opens: within five years, exposed firms are about 4 percent smaller than comparable non-family transfers and pay about 4 percent lower average wages, with no detectable change in revenue or productivity. These intention-to-treat estimates suggest that employment preservation clauses secure wage-bill stability at the cost of slower growth in succeeding firms.
Self-employment is widespread in Brazil, where informal businesses rarely borrow. We use Brazilian panel data to estimate a labor-market model in which workers save out of wages to finance business entry, and wages respond to their self-employment options. The estimates imply that two-thirds of labor-market entrants cannot finance entry, while wage adjustment retains employees who can. Targeted grants expand self-employment and raise welfare. Better savings opportunities raise welfare while reducing self-employment, as financial assets become more attractive than business ownership. Relaxing firms' collateral constraints has little effect. Financial development need not promote entrepreneurship to improve workers' welfare.
Work in Progress
The Option Value of Distance: Wages and Commuting in a Spatial Labor Market (with Elena Mattana).
What is the effect of labor-market access on wages? We build a spatial search model with commuting and wage bargaining in which distance has two opposing effects on pay. While longer commutes raise wages through compensation for disamenities, they may also reduce the option value of search by lowering the arrival rate and quality of outside offers, which weakens bargaining and can depress wages. A key implication is heterogeneity: when workers can choose location upon unemployment (untied), location choice partially offsets the option-value channel; when location is constrained (tied), wages respond more strongly to changes in access. We bring the model to Danish administrative data for Greater Copenhagen using high-resolution commuting-network measures and quasi-experimental changes in cycling connectivity that alter effective travel times for job stayers. The estimates support a small but significant option-value component in wage gradients.
Recipient of AUFF NOVA grant and SAB 2023 grant.
Acqui-Hires (with Pauline Corblet). Recipient of SAB 2026 grant
Worker Specialization and the Aggregate Economy