Hours Worked and Lifetime Earnings Inequality, 2025, with Blandin & Rogerson
PDF, BibTeX, OTE Blog Post, VoxEU Column, Media Coverage: Wall Street Journal
Abstract: We document large differences in lifetime hours of work using data from the NLSY79 and argue that these differences are an important source of inequality in lifetime earnings. To establish this we develop and calibrate a rich heterogeneous agent model of labor supply and human capital accumulation that allows for heterogeneity in preferences for work, initial human capital and learning ability, as well as idiosyncratic shocks to human capital throughout the life-cycle. Our calibrated model implies that almost 20 percent of the variance in lifetime earnings is accounted for by differences in lifetime hours of work, with over 90 percent of this effect due to heterogeneity in preferences. Higher lifetime hours contribute to lifetime earnings via two channels: a direct channel (more hours spent in production at given productivity) and a human capital channel (more hours spent investing in human capital, which increases future productivity). Roughly one-half of the effect of lifetime hours on lifetime earnings is due to the human capital channel. Higher lifetime hours are also an important source of upward earnings mobility over the life-cycle for many workers.
The Rapid Adoption of Generative AI, 2025, with Blandin & Deming
PDF, BibTeX, OTE Blog Post I and II, VoxEU Column, RPS Website, NBER Digest
Abstract: Generative artificial intelligence (AI) is a potentially important new technology, but its impact on the economy depends on the speed and intensity of adoption. This paper reports results from a series of nationally representative U.S. surveys of generative AI use at work and at home. As of late 2024, nearly 40% of the U.S. population age 18-64 uses generative AI. Among employed respondents, 23% used generative AI for work at least once in the previous week: 9% used it every workday, and 14% on some but not all workdays. Relative to each technology's first mass-market product launch, work adoption of generative AI has been as fast as the personal computer (PC), and overall adoption has been faster than either PCs or the internet. Generative AI and PCs have very similar early work adoption patterns by education, occupation, and other characteristics. Between 1 and 5% of all work hours are currently assisted by generative AI, and respondents report time savings equivalent to 1.4% of total work hours. This suggests that substantial productivity gains from generative AI are possible.
Work from Home and Interstate Migration, 2024, with Blandin, Mertens, & Rubinton
PDF, BibTeX, OTE Blog Post, RPS Website
Abstract: Interstate migration by working-age adults in the US declined substantially during the Great Recession and remained subdued through 2019. We document that interstate migration rose sharply following the 2020 Covid-19 outbreak, nearly recovering to pre-Great recession levels, and provide evidence that this reversal was primarily driven by the rise in work from home (WFH). Before the pandemic, interstate migration by WFH workers was consistently 50% higher than for commuters. Since the Covid-19 outbreak, this migration gap persisted while the WFH share tripled. Using quasi-panel data and plausibly exogenous changes in employer WFH policies, we address concerns about omitted variables or reverse causality and conclude that access to WFH induces greater interstate migration. An aggregate accounting exercise suggests that over half of the rise in interstate migration since 2019 can be accounted for by the rise in the WFH share. Moreover, both actual WFH and pre-pandemic WFH potential, based on occupation shares, can account for a sizable share of cross-state variation in migration.
Real-Time Labor Market Estimates During the 2020 Coronavirus Outbreak, 2021 (dormant), with Blandin
Overview: Shortly after the COVID shock hit the US, we started running a Current Population Survey (CPS)-style online survey (initially twice a month, between October 2020 and June 2021 once a month). Between May 2020 and June 2021 the survey was run in collaboration with the Federal Reserve Bank of Dallas. The goal was twofold: first, to provide real-time data on the state of the labor market in the US; second, to gather extra information that the CPS and many other datasets do not have, or that will only become available to researchers with a significant time lag. With every new data release, we published an update on the state of the US labor market. The PDF linked above was the last of these reports, all other reports including media coverage are available on the project website. For the surveys covering the same reference week as the CPS, we released a forecast about three weeks before the release of the Employment Situation Report by the BLS. Therefore in analogy to the CPS, we called the survey the Real-Time Population Survey. We used the additional questions in our survey to address questions of interest, both for academic research and for public policies. We published two academic papers based on the RPS in the Review of Economic Dynamics and the American Economic Journal: Macroeconomics. We are also using the data in our recent working paper on interstate migration, and Laura Pilossoph and Jane Ryngaert added some questions to our survey for their research on job search, wages, and inflation.