Working Papers

         The Effect of Software Adoption on Skill Demand, with Bledi Taska

We study how firms change their demand for workers of different skill sets when they adopt additional software varieties. We construct software adoption events from job posting data compiled by Lightcast and use a latent variable strategy to estimate the causal impacts of these adoptions. After an adoption, analytic and social skill requirements for the job using the software increase by 0.8 and 1.1 percentage points, respectively, and the number of vacancies also rises by 30%. We then embed these effects in an equilibrium model of software adoption and occupation sorting amongst white-collar occupations, and find that falling software prices increase inequality both within and between occupations. The upskilling effects of software drive the increase in within-occupation inequality by restricting labor from moving to higher wage software jobs

         Is Affirmative Action Still Effective in the 21st Century?, with Noriko Amano and Julian Aramburu

We study Executive Order 11246, an employment-based affirmative action policy targeted at firms holding contracts with the federal government. We find this policy to be ineffective in the 21st century, contrary to positive effects found for the late 1900s (Miller (2017)). Our novel dataset combines data on federal contract acquisition and enforcement with US linked employer-employee Census data 2000–2014.

Work in Progress

How Pay Information Impacts Job Search , with Alice Gindin

Job ads increasingly include wage or salary ranges, partly in response to new transparency laws in Colorado and other states. This project examines how pay transparency in job ads influences equilibrium wage distributions, with a focus on the role of workers’ beliefs. We develop a model of job search under imperfect wage information, which we will estimate using survey data on worker beliefs before wage disclosure, supplemented by job-posting data from Lightcast.

Job-level Skill Mismatch over the Business Cycle, with Jumi Kim

We examine skill mismatch between workers and the skill requirements of their jobs over the business cycle. To do this, we create a matched employer-employee dataset that links workers’ job profiles from Lightcast to the skill requirements of the relevant job postings at the time the worker was hired.

Generative AI and Learning, with German Reyes

We ran a lab experiment with approximately 200 college students to understand the impact of generative AI on learning. Participants were randomly assigned to complete a learning task either with or without access to generative AI. In a second session, all students completed a follow-up task without any external resources. We are studying how AI use while learning affected performance in the second session, and the underlying mechanisms through which generative AI impacts learning.

In a related study, we survey college students to document their generative AI usage patterns and perceptions.