Working Papers
Does Coursework Matter? Uncovering the Role of Skills in the Returns to College
EdWorkingPaper: 26-1484 (04/2026)
Abstract
The continuing shift of the U.S. economy toward a high-skill base has increased the demand for college-educated workers. To understand how higher education prepares students for this evolving economy, a large body of literature in labor economics has focused on the causes and consequences of college enrollment, institutional selectivity, and major choice. Much less attention has been paid to a key dimension that shapes the skills students acquire in college—coursework. In this paper, I develop new methods to identify the skills taught in college courses and credibly estimate their causal impacts on labor market outcomes, helping to explain earnings variation among students within the same major. I first scrape and compile a new dataset of detailed course descriptions from Texas public universities. Using a large language model (GPT-4), I extract the skills students are likely to acquire from each course, focusing on two widely taught and consistently identifiable domains: quantitative and writing. I then link these course-level skill measures to Texas administrative records that track students’ educational histories and quarterly earnings. To estimate the returns to coursework-based skills, I implement an instrumental variables strategy that exploits variation in course offerings across cohorts within the same major. I find substantial early-career earnings returns to coursework-based quantitative skills, but no detectable returns to writing skills. These returns are especially large for underrepresented minority (URM) students and for students in less quantitatively intensive majors, suggesting that expanding access to quantitative coursework within majors may serve as a new lever for narrowing racial earnings gaps.
Meetings
(with David J. Deming, Katrine V. Løken, and Alexander Willén)
NBER Working Paper: 35706 (09/2026)
Press coverage
The Wall Street Journal
Bloomberg
The Economist
Business Insider
VoxEU (CEPR)
Charter
London Inc. Magazine
Abstract
Why do we have so many meetings? Few workplace features are so scorned, yet seemingly so necessary. This paper provides the first large-scale economic evidence on workplace meetings using an original survey of more than 9,000 workers linked to matched employer–employee administrative data from Norway. We show that meetings are both common and costly, consuming an average of 12 percent of work hours and 14 percent of firm wage bills. Planning, problem solving, information sharing, and project coordination account for the majority of meeting activity. High-paying and high-revenue firms devote more resources to meetings despite facing a substantially higher opportunity cost of employee time. Meeting frequency and intensity are positively related to worker wage growth. Workers in meeting-intensive firms report greater on-the-job learning, and interactions with more senior colleagues are associated with stronger wage growth, suggesting that knowledge transmission within firms is an important mechanism. Meetings are the broccoli of work – widely disliked, but probably good for us anyway.
The Division of Labor: Productivity, Coordination, and Learning
(with Katrine V. Løken, Mikko Silliman, and Alexander Willén)
NHH Discussion Paper (10/2026)
Abstract
The division of labor is fundamental to how firms organize production. Since Adam Smith, economists have recognized that dividing tasks more finely increases productivity through specialization but also raises coordination costs—a tradeoff formalized by Becker and Murphy (1992). Yet we have little systematic evidence on how this tradeoff operates within firms. Moreover, this classical perspective overlooks a second, dynamic tradeoff: firms not only produce output, but also develop a central input into future production—their workers’ human capital. A finer division of labor allows workers to develop greater expertise through repeated performance of narrowly defined tasks, but may also reduce opportunities to learn from coworkers by limiting interactions and knowledge exchange. We study both tradeoffs using population-wide Norwegian employer–employee records, linked workplace surveys, quasi-experimental variation in coworker availability, and randomized experiments with workers and managers. We find that more specialized firms exhibit higher productivity but also greater coordination costs. Their workers report more learning-by-doing but substantially less coworker learning, resulting in lower overall learning and slower wage growth. To establish the importance of coworker learning for human capital accumulation, we exploit extended coworker absences and show that losing access to colleagues reduces subsequent wage growth, particularly in occupations characterized by greater task overlap and coworker learning. To isolate the role of the division of labor from other workplace characteristics, we conduct randomized experiments that vary task allocation in otherwise identical workplaces. Workers and managers anticipate greater coordination demands, weaker coworker interactions, and lower skill accumulation under greater specialization. The experiments then ask workers to choose their preferred workplace and managers how they would organize production. Both favor greater task overlap, citing coworker interactions and learning opportunities as key reasons for their choices. Together, our findings show that the division of labor shapes not only the efficiency of current production but also the accumulation of human capital and future productive capacity.
Causal Inference in Staggered Adoption Panels: The Correct Comparison Set Depends on Your Intervention
(with Ian Lundberg)
arXiv Working Paper (10/2026)
Abstract
A powerful data structure for causal inference is staggered adoption: many units are observed over many time periods, during which some units adopt an irreversible treatment at various time points while others remain untreated. Popular methods that apply in this setting include difference in difference, fixed effects, matching, and synthetic control. All of these methods compare the future outcomes of treatment adopters and non-adopters to answer the question: for units who adopt treatment, what outcomes would have been realized if they had remained untreated? We show that this question actually hides two very different causal questions of interest. The first is what would have happened under a momentary intervention to prevent treatment at its factual adoption time. The second is what would have happened to a treated unit under a longitudinal ban preventing treatment at all periods until outcome measurement. We show how to answer each question under conditionally ignorable treatment adoption or under parallel trends. We highlight that common methods cannot answer the second question because this question requires a time-varying comparison set with time-varying weights.
Mobile Internet and Mental Health
(with Lipeng Chen)
Draft available upon request (06/2026)
Abstract
Has the expansion of mobile internet improved or worsened mental health? We study the rollout of 3G mobile internet across the United States by combining coverage data from 2007 to 2020 with the Behavioral Risk Factor Surveillance System. Exploiting geographic and temporal variation in 3G availability, we find that access to mobile internet significantly worsens mental health. A 10 percentage point increase in 3G coverage increases the number of poor mental health days by 0.03 per month (0.7 percent), with larger effects among younger adults and women. Evidence on mechanisms shows that 3G expansion increases smartphone use and mobile web access more than overall internet use, alongside increased frequent social media use and reductions in physical activity and social engagement.