Research

The transition from high school to college is crucial for children’s later life outcomes, yet little is known about the role of high school educational inputs in developing skills relevant to college coursework. I estimate high school teacher value added on ACT scores and evaluate the impacts of ACT score value added on college enrollment and college performance using administrative data from North Carolina. I find that ACT score value added varies substantially across 11th grade math and English teachers, both within and across high schools. Exposure to teachers with high ACT score value added increases on-time 4-year college enrollment and enrollment in selective 4-year colleges, decreases 2-year college enrollment, and improves 4-year college performance. A one standard deviation increase in math ACT score value added increases the likelihood of completing a 4-year college degree within 5 years of high school completion by 8%. Increased 4-year college enrollment explains 68% of the total college completion effect, while increased enrollment in selective 4-year colleges with high completion rates and improved college performance play smaller roles. My results suggest that high school teachers have significant scope to influence the accumulation of college-relevant skills.

Using administrative data from North Carolina, I study the decision to take multiple college admissions tests in the presence of universal ACT testing. I find that low-income students are less likely than their peers to take the SAT in addition to the state-mandated ACT, and that they improve upon their initial ACT scores by less when doing so. Taken together, these disparities decrease low-income students’ rankings in the test score distribution when evaluating students on their maximum score rather than their initial ACT score. I provide evidence that income gaps in multiple test-taking are partially driven by differential responses to the ability signal sent by a student’s initial ACT score.

Whether income support programs disincentivize marriage has long generated policy debate. We leverage data from a large-scale U.S.-based randomized controlled trial to examine the causal effects of a multi-year monthly unconditional cash transfer to mothers with low income on preregistered measures of marriage and family structure. The cash transfer increased marriage with children’s biological fathers within a year after birth, had no destabilizing effects on new or existing marriages, and did not increase single parenthood. We do not find that marriage resulted in additional contributions of income from the spouse, despite hypotheses that marriage will benefit children through increased household resources.

Household time and money allocations in response to income support programs vary across diverse family circumstances and preferences, yet such heterogeneous responses are not well understood. Using data from a large-scale, multisite, U.S.-based randomized controlled study, we examine heterogeneity in the effects of a monthly unconditional cash transfer on monetary and time investments in children. This study offers a novel opportunity to examine heterogeneous effects of a cash transfer by race and ethnicity, where receipt is independent of eligibility based on other demographic characteristics. The effects of the cash transfer on net household income, earnings, and household expenditures were similar for families irrespective of race or ethnicity, even given initial differences in family structure, government benefit receipt, and employment. However, effects on monetary and time investments in children differed. Latino families’ child-focused expenditures increased, equivalent to nearly one-third of the cash transfer, with no effect on maternal employment or time spent with children. Among Black families, maternal work hours decreased and time spent with children on early learning activities increased, with no effect on child-focused expenditures. Marginal propensities to consume child-specific goods from different income sources also varied: Estimates showed a higher marginal propensity to consume childspecific goods from government income than from maternal income among Latino families, and the opposite among Black families. Latino families’ responses to the unconditional cash transfer and to government income are consistent with the notion that signals regarding intended use of income influence spending decisions.

Government benefits have been found to reduce food insecurity, yet many U.S. households remain financially stretched to meet healthy food needs. We examine the impact of a monthly unconditional cash transfer on food related outcomes among families with low-income and young children from the Baby’s First Years’ randomized control study. The monthly unconditional cash had no impact on food insecurity, food expenditure, receipt of government food program benefits, or use of food pantries. The monthly cash transfer reduced the share of household income spent on food and increased spending on food not prepared at home.

Children represent the largest indirect beneficiaries of the U.S. social welfare system. Yet, many questions remain about the direct benefits of cash aid to children. The current understanding of the impacts of cash aid in the U.S. is drawn primarily from studies of in-kind benefits, tax credits, and conditional cash aid programs. A corresponding economics literature focuses on the role of income and the labor supply responses of parents, parenting skills, and early education as family investment mechanisms that reduce socioeconomic inequality in children’s well-being. In contrast to the U.S., dozens of low- to middle-income nations use direct cash aid—conditional or unconditional—as a central policy strategy, with demonstrated positive effects across a host of economic and health measures and selected aspects of children’s health and schooling. This paper reviews the economic research on U.S. safety net programs and cash aid to families with children and discusses what existing studies reveal about the impacts of cash aid on family investment mechanisms and children’s outcomes. We specifically highlight gaps in understanding the impacts of unconditional cash aid on children. We then review nine contemporary unconditional cash transfer programs and discuss their promise and limitations in filling the U.S.-based economic evidence gap. 

Despite unprecedented increases in women’s college attendance and labor force participation over the last century, large gender differences in pay persist even for graduates of elite universities. Can these gender differences in salary be explained by observed differences in graduates’ skills and preparation, such as choice of major? Using unique data from the University of Virginia Career Center, I apply linear Oaxaca-Blinder decomposition models as well as unconditional quantile decomposition methods to measure the extent to which differences in early career compensation can be explained by observable differences in productive characteristics. I find that the explained share varies greatly across the pay distribution. In particular, at the low end of the salary distribution, the gender pay gap is larger and is only partially accounted for by gender differences in college major choice, industry choice, and internship experience. At the middle of the pay distribution, the gender pay gap can be almost entirely attributed to observable gender differences in these characteristics. Interestingly, at the high end of the salary distribution, gender differences in college major choice, industry choice, and internship experience suggest that the gender pay gap should be even larger than what is observed in the data. My results suggest that female graduates in high-paying majors and industries do not encounter a “glass ceiling” at the beginning of their careers. Rather, female UVA graduates either receive preferential labor market treatment, are more competent than male peers in the same majors and industries, or demand compensating wage differentials to account for differences in preferences for certain high-paying positions.

Using rich administrative data from a large public university, we show that white, female, upperclassmen, and high-income students are less likely to withdraw from courses that get easier over the course of the semester, and that course withdrawal is correlated with subsequent college dropout even after conditioning on academic achievement. We explain these demographic differences using a dynamic discrete choice model of course withdrawal and college dropout with unobserved types. In particular, informed students are aware of changes in grade distributions between midterm and final exams and make their course withdrawal decision accordingly, while uninformed students mistakenly withdraw or fail to withdraw under the assumption that average course grades are the same across midterm and final exams. Mistakes in withdrawing from courses lead to long-term negative outcomes, in higher probability of dropout. We use counterfactual simulations to examine potential impacts of increasing the cost of withdrawing or equalizing information about grading practices. Our results have important policy implications, suggesting that closing information gaps about grading practices is a relatively low-cost intervention with potentially high returns.  

Over 15% of U.S. public school students receive special education services, with schools playing a crucial role in disability identification as mandated by the Individuals with Disabilities Education Act (IDEA). In this paper, we study the effects of students' third grade class rank on the likelihood that they are classified with a disability in future grades. Using administrative data on all public school students in North Carolina from 2009-2019, we leverage idiosyncratic variation in school-cohort achievement distributions to separately identify class rank effects from the effects of student test scores. We find that students ranked near the top or bottom of their third grade class are more likely to be classified with a disability in fourth grade, compared to students ranked near the middle. This relationship is primarily driven by learning disabilities and other health conditions, including Attention-Deficit/Hyperactivity Disorder (ADHD).