Research
David Deming's early work studied the long-run effects of childhood education programs. In a 2009 study of Head Start, he compared siblings in the same families and found that the program's effects on test scores faded within a few years, while its effects on high school graduation, college attendance, and other young-adult outcomes persisted. Using admissions lotteries in the Charlotte-Mecklenburg school district, he showed that winning admission to a preferred school reduced later criminal activity among high-risk youth, and, with Justine Hastings, Thomas Kane, and Douglas Staiger, that attending a higher-quality school increased college enrollment and degree completion.
With Claudia Goldin and Lawrence Katz, Deming documented that for-profit colleges enrolled a larger share of disadvantaged and older students and had higher completion rates in short certificate programs, but that their students carried more debt, defaulted more often, and earned less than comparable students at public and nonprofit institutions. In a field experiment that submitted fictitious résumés to online job postings, he and coauthors found that a business bachelor's degree from a for-profit online institution was about 22 percent less likely to receive a callback than one from a nonselective public institution. With Raj Chetty and John Friedman, he studied admissions and outcomes at highly selective private colleges using linked admissions, attendance, and tax records. They found that applicants from families in the top one percent of the income distribution were more than twice as likely as middle-class applicants with comparable test scores to attend one of these colleges, and that attending one instead of a flagship public university raised a student's chances of reaching the top one percent of earnings by 50 percent, nearly doubled the chances of attending an elite graduate school, and almost tripled the chances of working at a prestigious firm. When Harvard announced in April 2024 that it would again require standardized test scores in admissions, after similar decisions at Dartmouth, Yale, and Brown, it cited findings by Chetty, Deming, and Friedman that test scores helped identify promising students from less-resourced high schools, and Deming argued that the tests' wide availability made requiring them "the fairest admissions policy for disadvantaged applicants".
Deming is known for research on the growing labor-market value of social skills. In a 2017 paper, he showed that between 1980 and 2012, jobs requiring high levels of social interaction grew as a share of U.S. employment and that wage returns to social skills rose, and he proposed a model of team production in which social skills lower the cost of coordinating with coworkers. The Society of Labor Economists cited this work in awarding him the 2022 Sherwin Rosen Prize. With Ben Weidmann, he designed an experiment to measure individual contributions to group performance, finding that some people consistently improve the performance of teams they join and that this ability is predicted by social intelligence rather than cognitive test scores. With Kadeem Noray, he found that high starting salaries in applied STEM fields flatten with experience because required job skills change quickly and earlier training loses value. He summarized this body of work in a 2022 survey, "Four Facts about Human Capital."
His recent research concerns management, artificial intelligence, and work. In a laboratory experiment with Ben Weidmann and other coauthors, he found that a one-standard-deviation increase in manager quality raised team performance by about 0.2 standard deviations, that people who volunteered to manage performed worse than randomly assigned managers, and that economic decision-making skill predicted managerial performance. With Alexander Bick and Adam Blandin, he fielded the first nationally representative U.S. survey of generative AI use, which found that by late 2024, 45 percent of adults aged 18 to 64 had used generative AI and 27 percent of employed respondents had used it for work in the previous week, a faster rate of adoption than either the personal computer or the internet. With Christopher Ong and Lawrence Summers, he examined occupational change in the United States since 1880 and found that labor-market churn from 1990 to 2017 was slower than in earlier eras, while trends since 2010, including rapid growth in STEM employment and a decline in retail jobs, may mark the beginning of AI-driven disruption.
