My research explores the persistence of intergenerational inequality by race, ethnicity, and class, specifically within the U.S. context. Although all children in the United States are guaranteed access to public K-12 education, race, ethnicity, and income are some of the largest predictors of academic performance. Academic achievement is the foundation of children's later-life success, associated with outcomes such as college attendance and earnings. Therefore, policies that affect children's academic achievement have long-term consequences. I am interested in how out-of-school policies that create, or ease, stress for families affect the academic outcomes of children. I explore these questions using quantitative methods, with an emphasis on causal inference.

Immigration Enforcement and Student Achievement

My dissertation, which has been funded by competitive awards from the Russell Sage and Horowitz Foundations, examines the effects of immigration enforcement policies on student outcomes, including absences, academic achievement, and disciplinary incidents. Compared to the number of immigrants removed from the U.S. interior in the early 2000s, between 2008 and 2011, nearly twice as many immigrants were removed from the U.S. interior per year. The foundation of my research in this area is the premise that immigration enforcement affects not only immigrants but also their families. The Pew Hispanic Center estimates that 5.1 million children living in the U.S. have at least one unauthorized parent, who is potentially a target of immigration enforcement. Although not all children with an unauthorized parent experience a parental removal due to immigration enforcement, the larger group of children with an unauthorized parent likely experience increases in stress and fear as immigration enforcement intensifies.

I use quasi-experimental methods to isolate the effects of immigration enforcement on children's academic achievement, examining this question in two ways. First, I focus on average academic achievement across U.S. counties. Second, I focus on student-level academic achievement within one state, North Carolina.

To examine the impact of immigration enforcement on county-level outcomes, I use the staggered rollout of Secure Communities, a biometric sharing program ultimately activated in every U.S. county. I match information on Secure Communities with newly available measures of average county achievement from the Stanford Education Data Archives (SEDA). I find that the activation of Secure Communities decreased test scores for both Hispanic and non-Hispanic black students. Similarly, I find that increases in removals due to Secure Communities are associated with decreased achievement for both Hispanic and non-Hispanic black students in ELA. This paper is currently under review and also available as a working paper.

My second paper uses data on individual-level student outcomes from North Carolina. As the source of variation, I use applications for 287(g) programs, which are partnerships between local law enforcement and ICE. In North Carolina, nine counties established 287(g) programs, whereas another 15 counties applied to participate. Using a triple difference strategy, I compare outcomes for different groups of students in these two sets of counties over time, disaggregating students by race/ethnicity and lifetime identification as limited English proficient (LEP).

Food Instability and Student Outcomes

Over the past 25 years, poor households have become poorer relative to wealthy households, as well as less income stable. Anna Gassman-Pines and I take advantage of a a natural experiment in North Carolina to analyze how recency of Supplemental Nutrition Assistance Program (SNAP) benefit transfer affects students' test scores (published in American Educational Research Journal) and disciplinary incidents. We find that student achievement varies based on how recently students received a SNAP transfer. The relationship between time since SNAP transfer and test scores is curvilinear, with test scores peaking in the third week post transfer (shown below). Our findings suggest that increasing the generosity of the U.S. social safety net would improve student achievement, either by decreasing family stress or increasing families' access to nutritious foods or both (op-ed and additional press coverage).

Teacher Preparation Programs

Good teachers matter. But most beginning teachers are not very good. Understanding how to train beginning teachers has therefore been a goal of many interested in education policy. Indeed, sixteen states hold teacher preparation programs (TPPs) accountable for the test scores of their trainees' students. However, even though the desire to improve teacher training is laudable, this practice may be faulty. 

  1. With Jane Lincove, Cynthia Osborne, and Nick Mills, we find that, even though teachers from independent non-profits TPPs outperform teachers from other TPPs, these trainees teach only in very select areas of Texas (published in The Journal of Teacher Education).

  2. With Paul von Hippel, Jane Lincove, Cynthia Osborne, and Nick Mills, we find that it is very difficult to distinguish between TPPs in Texas - differences are small, and estimates are noisy (published in Economics of Education Review). As shown below, few TPPs appear to be significantly different from the average when we correct for multiple testing, and the distribution of TPP effects very closely resembles the null distribution - the distribution of what estimates would look like under homogeneity (if estimates differed only due to random estimation error).

  3. Paul von Hippel and I apply similar statistical techniques from our Texas study to previous studies in five other states. We find, like in Texas, that detectable differences between teachers from different TPPs are negligible (published in Economics of Education Review; article in Education Next; additional press coverage). We also release a Stata program (ssc install caterpillar) that allows users to replicate our methods, as well as easily calculate null distributions more generally.

In blue are value-added estimates for every teacher preparation program in Texas, along with 95% confidence intervals for those estimates. In pink is the null distribution, or the pattern estimates should follow if programs were actually identical and ranked on estimation error alone.

In blue are value-added estimates for every teacher preparation program in Texas, along with 95% confidence intervals for those estimates. In pink is the null distribution, or the pattern estimates should follow if programs were actually identical and ranked on estimation error alone.