Research

This paper examines how Low-Income Housing Tax Credit (LIHTC) development shapes neighborhood employment, combining project records with administrative employment data and population estimates from 2010 to 2019. Using a staggered event study with propensity score matching within counties, I find that jobs held by neighborhood residents increase by approximately 3 percent despite little net population growth, while the employment to population ratio rises gradually after placement. Larger ratio gains accompany more intensive housing construction, and higher income neighborhoods experience larger employment gains despite broadly comparable housing additions. To examine where these gains occur, I distinguish jobs at nearby and more distant workplaces. Nearby jobs account for the larger share of the estimated increase in jobs per resident, although both categories grow at similar proportional rates. Employment growth thus extends across destinations, with little estimated change in the nearby employment share or average commuting distance. Evidence of employment growth is more consistent for jobs held by residents than for jobs located within the neighborhood itself.

This paper investigates whether master’s degrees protect college graduates from the adverse effects of entering the labor market during economic downturns. Using data from the National Survey of College Graduates covering 2000-2018, I exploit variation in national unemployment rates across graduation cohorts to identify how initial labor market conditions affect earnings trajectories differently for bachelor’s versus master’s degree holders. I find that while bachelor’s graduates experience persistent earnings penalties from graduating during recessions—with a one percentage point increase in unemployment reducing earnings by approximately 2% in the early career years and remaining negative even 12-14 years after graduation—master’s degree holders face substantially smaller penalties throughout their careers. Economic conditions also influence educational pathways, with recessions significantly increasing the likelihood of field switching in graduate education, particularly among Social Sciences (7.8%), Engineering (4.5%), and Biological Sciences (4.1%) graduates per percentage point rise in unemployment.

How has exposure to generative AI changed workers' labor supply and earnings, and did the shock spill over to their spouses? I study both questions using monthly CPS data on 2.13 million married-couple observations from 2015 to 2026, comparing exposed and unexposed occupations before and after the release of ChatGPT in November 2022. Given the strong correlation between AI exposure and teleworkability, I estimate effects separately within teleworkable and non-teleworkable occupations. I further test whether each effect emerges after ChatGPT rather than during COVID. Three patterns emerge. First, exposed husbands in non-teleworkable jobs reduce their usual weekly hours by about 0.3 hours (roughly 0.75%) after ChatGPT. The decline begins after the release and is not driven by any single occupation. Second, exposed wives earn about 4% less per hour in both teleworkable and non-teleworkable jobs. This gap, however, opened during COVID and therefore cannot be attributed to ChatGPT. Third, husbands of exposed wives in teleworkable jobs work about 0.3 more hours per week, consistent with an added-worker response. Roughly half of this estimate, however, reflects movement in the comparison group, which consists largely of teachers. Spillovers on spouses' labor force participation and employment are null throughout. These results suggest that many apparent AI effects in post-2022 data reflect telework shifts or trends that predate ChatGPT.