Research
Identification and Estimation of Generalized Selection Models, with an Application to Texas Direct
[Abstract]This paper studies models in which the observability of people's outcomes depends on their economic choices. In these models, the outcomes resulting from a given treatment are observed only for those who chose to participate in that treatment through utility maximization. The proposed method enables the identification and estimation of general models with such choice-dependent outcomes while relaxing some of the identifying restrictions commonly used by existing tools, providing a simple framework for studying counterfactual choice-specific outcomes for the population in a way that reflects the available information. The generality of our method allows it to be applied even when data limitations, such as limited or censored instrumental variables, preclude researchers from fully determining how people’s decisions affect the observability of their outcomes. The methods are applied to evaluate the effect of Texas Direct, a major community college transfer initiative. Using Texas administrative data, we find that the benefits of completing the Texas Direct curricula components for students’ post-transfer outcomes depend on how students accumulate and apply credits from each curriculum.
Publications
Variation in English learners' programmatic experiences and postsecondary trajectories
Language and postsecondary trajectories: How "ever-English learner" status predicts college outcomes
Developmental education reform and student outcomes at California community colleges: The role of transfer-level course enrollment
[Abstract] California's Assembly Bill 705 was intended to increase student success by maximizing access to transfer-level math and English coursework in the state's community colleges. These transfer-level courses are broadly comparable to college-level courses in other states, though California's placement reform mandated that most students have direct access to transferable (rather than associate degree-applicable) courses. Although prior research illustrates that the reform increased access, few studies have examined transfer-level course enrollment during the initial rollout — when colleges retained discretion over placement — and how that enrollment predicts students' short- and long-term outcomes. Using statewide administrative data and inverse probability weighted regression adjustment, we examined which students enrolled in transfer-level courses and whether such enrollment predicts academic progress up to 4 years later. Our findings offer insights into placement reform implementation, the role of early access to transfer-level coursework in shaping subsequent academic momentum, and the implications of uneven access during discretionary implementation.
Working Papers
Two-sample two-stage least squares estimation of nonparametric models
[Abstract] This paper investigates nonparametric instrumental variable estimation in a data combination context, in which one dataset contains the outcome and instrumental variables, and the other contains the explanatory and instrumental variables. We develop an estimator for structural functions, derive its uniform convergence rate, and introduce a two-sample bootstrap method for uniform inference. The proposed estimator is equipped with procedures for adaptive estimation and for constructing honest, adaptive uniform confidence bands, thereby enhancing its practical value. Our simulations demonstrate that the proposed procedure enables estimation and inference from two incomplete datasets that are comparable to its full-variable benchmark. We apply our findings to examine the relationship between healthcare expenditures and household budgets among the elderly population in the U.S., combining data from the Medical Expenditure Panel Survey and the Consumer Expenditure Survey. We find that universal Medicare coverage does not eliminate meaningful variation in healthcare consumption according to household characteristics and spending capacity among the U.S. elderly population.
Expanding access, advancing outcomes: Evidence from California's developmental education reform
[Abstract] California's AB705 reformed developmental education in the nation's largest community college system. The policy required colleges to maximize the probability that students complete transfer-level coursework within one year and to use high school academic measures for placement. Yet, AB705's impacts on college completion and transfer remain unclear. Using student-level statewide administrative data, we examine its effects on course completion, degree attainment, and vertical transfer. We find that following AB705, students with the lowest academic readiness — the policy's target — experienced greater gains in transfer-level credit accumulation, credential attainment, and university transfer within four years than their better-prepared peers.
enesis of the Mississippi Miracle: State Reading Reforms and K-3 Reading Skills, 2013-201
[Abstract]Mississippi’s reading reforms have attracted national attention, but discussion has focused almost exclusively on the rise in 4th grade scores on the National Assessment of Educational Progress (NAEP) between 2013 and 2019. Commentators have debated whether that rise was due to the science of reading, preschool, or a “third-grade gate” test that kept low-scoring from reaching 4th grade and taking the NAEP. We present new evidence from data on K-3 test scores, teacher training and coaching in the science of reading, preschool, grade retention, and interviews with 27 teachers. We find that reading scores improved earlier than previously noticed, rising more than 0.3 SD in kindergarten by 2016-17, then fading through grades 1-2 before resurging in grade 3. Seventy percent of teachers received science-of-reading training within 2 years, and schools where more teachers received training showed significantly higher value-added to kindergarten test scores (approximately 0.14 SD higher at the average training level of 70 percent). Schools that received literacy support, which included coaches, reported slightly larger increases in kindergarten test scores (0.08 SD). Teachers who did not receive training learned about the science of reading through their peers. Preschool attendance did not increase and could not have caused the rise in scores. Grade reten- tion at the third-grade gate accounts for at most a quarter of the fourth-grade score increase. Together, the results suggest that Mississippi’s gains were real but began earlier, and depended on more elements of the reform, than the fourth grade NAEP trend alone would suggest.
Work in Progress
Single-Crossing, Information Ordering, and Identification in Choice under Uncertainty
[Abstract] We are interested in testing whether economic agents make rational choices that allow them to maximize their returns when payoff realizations are contingent upon an uncertain state of the world, using data consisting of observed choices and payoff realizations that take values in a finite, ordered set. Applying a theoretically grounded model of choice under uncertainty, we show how the assumption of a Bayesian decision maker with single crossing utility, which we label “Monotone Bayesian Expected Utility decision-maker (MBEU-DM), provides identifying restrictions on the distribution of utility payoffs, as well as the conditions under which it yields testable implications of the hypothesis. We illustrate our findings by testing the optimality of course-load decisions among community college students in Texas using a statewide, population-level longitudinal dataset.