Higher Education
Lightning Round SessionSaturday, Jan. 3, 2026 10:15 AM - 12:15 PM (EST)
- Chair: Peter Arcidiacono, Duke University
Peer Institution Networks, Test-Optional Admission Policies, and STEM Major Completions
Abstract
Higher education institutions differ substantially in ways that may be difficult to observe. This complicates the estimation of causal effects, which requires like-for-like comparisons of treated and control units. To address this problem, I construct a network of private colleges and universities based on their self-reported peer institutions and detect communities within it using an algorithm from the network science literature. Community members serve as controls for each other when estimating the effects of pre-pandemic test-optional policy adoption. This approach drastically attenuates the severe pre-trends that arise when all untreated units serve as controls, suggesting the satisfaction of a parallel counterfactual trends assumption that is conditional on community membership. Policy adoption causes the first and third quartiles of the reported SAT score distribution to increase by 14 and 22 points, respectively, under-represented minority and federal grant recipient enrollment to increase by 5 to 10 percent, and the share of graduates majoring in a STEM field to decrease by 5 percent. Effects vary by community. For the community of Elite National Liberal Arts Colleges, policy adoption causes a precisely estimated 8 (23) percent decline in the share of all (under-represented minority) graduates with a STEM major.When Harry Met Sally: Coeducation and Startup Formation
Abstract
We study how gender integration in higher education affects entrepreneurial entry and subsequent startup quality. Leveraging staggered coeducation reforms at 91 previously all-male U.S. universities between 1942 and 1995, and linking historical enrollment records with resume data, we find that exposure to female peers increases male students’ likelihood of becoming entrepreneurs. Three mechanisms drive this effect: cross-gender interactions that expand human and social capital, improved household risk-sharing, and sorting into masculine-perceived fields. The resulting startups survive longer, but show similar employment sizes and grow more gradually, consistent with a rise in sustainable, necessity-driven entrepreneurship rather than high-growth, transformative ventures. Overall, our findings suggest that the gender environment of higher education has lasting economic consequences by altering both the quantity and nature of entrepreneurial activity.Migration Restrictions, College Choices, and Spatial Skill Sorting
Abstract
College education is widely regarded as a pathway to local labor markets, particularly where migration frictions limit labor mobility. This paper examines how such frictions shape college choices in China, where mobility is constrained by both formal migration restrictions and informal barriers. Using a national administrative dataset on four-year college admissions from 2005 to 2011, I show that relaxing migration restrictions through hukou reforms enabled colleges in reformed cities to attract higher-quality students. The largest gains occurred in colleges located in economically more developed cities relative to students' origins, consistent with the mechanism of improved local labor market prospects. Counterfactual analysis based on a college choice model indicates that easing migration restrictions in major cities strengthens the sorting of stronger students into treated colleges and raises aggregate welfare, though the gains are unevenly distributed. Welfare increases further when students can freely access the highest-paying labor markets. These results highlight the role of both formal and informal migration frictions in shaping spatial skill sorting and welfare.The Effect of Optional Practice Training on Domestic and Foreign Graduates' Labor Market Outcomes
Abstract
Policymakers have emphasized the importance of foreign STEM college graduates to the U.S. economy.However, few studies have documented the spillover effects of the growing number of foreign STEM students, particularly their impact on domestic students’ labor market choices. To help fill this gap, this paper examines the effects of Optional Practical Training (OPT)—a temporary work permit that allows F-1 students in the U.S. to gain practical work experience after completing their studies. The duration of OPT was extended for STEM students twice, in 2008 and 2016. Using these two OPT extensions, I analyze their effects on (1) the likelihood that domestic college graduates choose STEM occupations, and (2) the starting salaries of both foreign and domestic college graduates. Based on the Current Population Survey (CPS) and Integrated Postsecondary Education Data System (IPEDS), I conducted a nationwide difference-in-differences analysis with a continuous treatment variable. The results suggest that the likelihood of domestic college graduates choosing STEM occupations declined by 3% after 2008 and by 2.4% after 2016. Domestic college graduates also had a decline in starting wages by 3.5% after the 2008 extension and 0.8% after the 2016 extension. Foreign college graduates experienced a growth of starting wage by 2.6% after the 2016 OPT extension. This paper contributes to the broader immigration and labor economics literature by providing causal evidence on the spillover effects of skilled migration policies, explicitly revealing how OPT extensions shape occupational decisions and wage outcomes of domestic STEM college graduates.
Hard Problems: Young Math Talents in the U.S. and Their Choices of Higher Education
Abstract
Mathematics competitions like the USA Mathematical Olympiad (USAMO) and the William Lowell Putnam Mathematical Competition (Putnam) are seen as early-stage talent selections, providing winners access to prestigious institutions and career advantages. However, their long-term impact on academic trajectories remains unclear. This study examines whether such competitions effectively predict future academic decisions and how receiving different levels of honors varies in these effects. Using a unique dataset of USAMO and Putnam awardees from 1985 to 2010, collected through both manual research and automated matching, we analyze their higher education and career paths. Nearly half of USAMO awardees obtained a Ph.D. inSTEM majors or Economics, compared to around 21% of Putnam awardees. Furthermore, USAMO gold medalists were significantly more likely than Honorable Mention recipients to pursue mathematics-related fields and remain in academia. These results contribute to discussions on talent screening through competitive challenges, suggesting that while high-level mathematics competitions identify strong academic candidates, psychological factors and institutional influences shape long-term career outcomes.
Exposure to Artificial Intelligence in Higher Education
Abstract
I use natural language processing to extract verb-object pairs from the full corpus of U.S. utility patents and from millions of course descriptions across more than 800 U.S. colleges and universities. Following Webb (2020), I construct technology exposure scores for individual college majors and institutions by comparing the task language in course descriptions to that found in patents related to three classes of technology: robotics, software, and artificial intelligence. This task-based approach measures overlap between what students learn to do in college (e.g., "analyze data," "solve problems," or "write essays") and the tasks performed by emerging technologies. Preliminary results suggest that exposure to artificial intelligence is substantially higher than to robotics or software. The median major exhibits software exposure comparable to the 40th percentile of Webb’s occupational exposure distribution for software, whereas the median major's AI exposure is comparable to the 68th percentile of the occupational exposure distribution for AI. The most AI-exposed fields—such as Statistics and Cybersecurity—rank comparably to the 90th percentile of the occupational exposure distribution. Teaching-oriented universities tend to offer more AI-exposed courses than selective research institutions, largely because they offer less-advanced courses that overlap with the current capabilities of AI. Causal analysis shows no evidence of course enrollment or offerings shifting away from AI-exposed content following the release of ChatGPT—in fact, exposure has slightly increased since the pandemic.Cap-and-Apply: Unintended Consequences of College Application Policy in South Korea
Abstract
Starting in the 2013 academic year, the South Korean government implemented a policy limiting students to a maximum of six college applications. This paper examines the impact of this application cap on matching quality and the socioeconomic gap in college prestige. Using college-level administrative data, I find that matching quality decreases as second-tier colleges are able to attract more desirable students following the introduction of the cap. This finding aligns with theoretical predictions from a simplified model in which colleges compete for top applicants, and applicants exhibit ability noise. Moreover, I extend the model to incorporate application constraints influenced by socioeconomic status (SES). The theoretical framework suggests that the cap reduces the socioeconomic gap: after the cap, the number of students from lower socioeconomic backgrounds attending more prestigious colleges increases, as the constraints primarily limit applicants from higher socioeconomic groups. Empirical analysis supports these theoretical predictions. These findings may provide valuable policy insights for the U.S. higher education market.The Effect of Diversity Statements in Faculty Hiring
Abstract
In recent years, U.S. universities have adopted various initiatives to promote diversity, equity, and inclusion (DEI), including the increasingly common requirement that faculty job applicants submit diversity statements—1–2 page documents outlining their views, contributions, and plans related to DEI.The use of diversity statements is controversial. Proponents argue that they help universities identify candidates who will foster a more inclusive environment. Critics argue that diversity statements may serve as political litmus tests, favoring candidates whose political views align with those of existing faculty, often resulting in a progressive bias. Others question whether diversity statements have any meaningful impact at all, as they may be ignored by hiring committees or encourage insincere responses.
In this paper, we use panel data methods to estimate the effects of diversity statements on student representation and faculty hiring outcomes. We compile a novel dataset combining job postings from online boards, faculty hiring data from Academic Analytics, and student composition data from IPEDS, covering U.S. economics and political science departments from 2014 to 2023. We also measure faculty political ideologies using political donations data.
We find little effect of diversity statements. Specifically, diversity statements do not affect student underrepresentation, except by reducing the number of Asian graduates. We find no significant changes in the demographics of the hired faculty, contradicting the view that they are used as a tool to prefer candidates from underrepresented groups. Lastly, we find no change in the degree of political involvement or left-right orientation of hired faculty. Collectively, these results suggest that diversity statements have little effect on student outcomes or on who is hired.
The Impact of Dropping Standardized Testing Requirements on Applicants, Admits, and Enrollees in Higher Education
Abstract
Concerns about the validity and utility of standardized testing in higher education admissions have been growing for years. The COVID-19 pandemic accelerated the trend of schools de-emphasizing test scores in determining whether a student would be admitted. This creates a setting well-suited for empirical evaluation of how dropping testing requirements impacts the relative number of applicants, admits, and enrollees at schools that implemented the change. We employed a staggered difference-in-differences method to assess the relative impact of moving from requiring standardized tests to not requiring them, compared to a control cohort that never adopted test-optional policies. Data were obtained from the Integrated Postsecondary Education Data System (IPEDS) and spanned the fall 2010 to 2023 application cycles. Our sample included 1,122 higher education institutions, with 15,708 school-year observations. Within three years of dropping testing requirements, applications increased by 11.7 percent, while admissions rose by 18.7 percent during the same period. These results are statistically significant at the 95% confidence level and persisted for several years after the change. However, enrollments did not change significantly after test-optional policies were adopted. The use of standardized testing in admission decisions remains contentious. While some universities commend test-optional policies for enhancing equitable access, others are reversing their decisions to drop testing, arguing that their exclusion diminishes equitable access. Our study indicates that universities that deemphasized standardized testing requirements experienced increased applications and admissions, although this did not increase enrollment.JEL Classifications
- I2 - Education and Research Institutions