Shifting Work Patterns with Generative AI (with Sonia Jaffe, Nicole Immorlica, and Christopher Stanton)
Conditionally accepted at American Economic Review: Insights
Coverage: Marginal Revolution
We present evidence on how generative AI changes the work patterns of knowledge workers using data from a 6-month-long, cross-industry, randomized field experiment. Half of the 7,137 workers in the study received access to a generative AI tool integrated into the applications they already used for emails, document creation, and meetings. We find that access to the AI tool during the first year of its release primarily impacted behaviors that workers could change independently and not behaviors that require coordination to change: workers who used the tool in more than half of the sample weeks spent 3.6 fewer hours, or 31% less time on email each week (intent to treat estimate is 1.3 hours) and completed documents moderately faster, but did not significantly change time spent in meetings.
Do Workforce Development Programs Bridge the Skills Gap? (with Lisa B. Kahn, Joanna Venator, and Michael Dalton)
Under review
Most U.S. states have workforce development programs that offer firms grants to train their own workers. We create unique data linkages between participating firms, employment, and vacancies to explore the determinants and consequences of such programs. Training grants are more prevalent in markets where firms face greater employee poaching risk, suggesting these programs help overcome a market failure in updating worker skills. After training, firms experience prolonged employment growth and downskilling in their job posts, relative to a matched control group. Training appears to help firms move toward optimal scale and expand opportunities for less skilled workers.
Self-Employment Dynamics and the Returns to Entrepreneurship (with Chris Stanton)
Coverage: USA Today
A third of men work for themselves at some point in their lives, but few remain self-employed for the bulk of their careers. We show that transitions between wage work and self-employment are consistent with a model of learning about comparative advantage in entrepreneurship. We characterize the distribution of an individual's potential entrepreneurial earnings as a function of earnings in wage work, with the goal of understanding which workers across the wage distribution can expect to benefit from experimentation with entrepreneurship. We find that only high-ability wage workers face a positive expected lifetime benefit from entrepreneurial entry, suggesting that policies to encourage entrepreneurship should be targeted to these workers.
A Meta-learner for Heterogeneous Effects in Difference-in-Differences (with Hui Lan, Haoge Chang, Vasilis Syrgkanis), accepted ICML 2025
A Causal AI Suite for Decision-Making (with Emre Kiciman, Darren Edge, Adam Foster, Agrin Hilmkil, Joel Jennings, Chao Ma, Robert Ness, Nick Pawlowski, Amit Sharma, Cheng Zhang), NeurIPS 2022 Workshop on Causality for Real-world Impact
Estimating the Long-term Effects of Novel Treatments (with Keith Battocchi, Maggie Hei, Greg Lewis, Miruna Oprescu and Vasilis Syrgkanis), Proceedings of the 34th Conference on Neural Information Processing Systems (NeurlPS), 2021
The Consequences of Academic Match between Students and Colleges (with Jeffrey Smith), Journal of Human Resources, 55, no. 3 (July 2020), 767-808
Coverage: Forbes
Risk and Return Tradeoffs in Lifetime Earnings, Journal of Labor Economics, 36, no. 4 (October 2018), 981-1021
The College Earnings Premium and Changes in College Enrollment: Testing Models of Expectation Formation, Labour Economics, 49 (December 2017), 84-94
Determinants of the Match between Student Ability and College Quality (with Jeffrey Smith), Journal of Labor Economics, 35, no. 1, (January 2017), 45-66
Preprint, Online Appendix, NBER version
Coverage: Inside Higher Ed, Chronicle
The Agentic Economy (with David Rothschild, Markus Mobius, Jake Hofman, Daniel Goldstein, Nicole Immorlica, Sonia Jaffe, Brendan Lucier, Aleksandrs Slivkins, and Matthew Vogel), accepted at Communications of the ACM. arXiv version
Early Impacts of M365 Copilot (with Sonia Jaffe, Sida Peng, Alexia Cambon), Microsoft whitepaper, 2025
Comment on Education and Innovation in The Role of Innovation and Entrepreneurship in Economic Growth, 2022.
Be careful when interpreting predictive models in search of causal insights (with Jacob LaRiviere, Scott Lundberg, Jonathan Roth, Vasilis Syrgkanis) Medium, 2021.
Why a Dollar Depreciation May Not Close the U.S. Trade Deficit (with Linda Goldberg), Federal Reserve Bank of New York Current Issues in Economics and Finance, 2007.
Reevaluating Causal Estimation Methods with Data from a Product Release (with Justin Young and Muthoni Ngatia)
Recent developments in causal machine learning methods have made it easier to estimate flexible relationships between confounders and treatments and outcomes, making unconfoundedness assumptions in causal analysis more palatable. How successful are these approaches in recovering ground truth baselines? In this paper we analyze a new data sample encompassing an experimental rollout of a new feature at a large technology company and a simultaneous sample of users who endogenously opted into the feature. We find that it is possible to recover ground truth causal effects, but only with careful choices in modeling. This builds on the observational causal literature stemming from LaLonde (1986), putting forth best practices that allow for more credible treatment effect estimation in modern, high-dimensional datasets.
The Role of Sorting and Skill Prices in the Evolution of the College Premium (with Greg Veramendi)