We are the first to document that Supplemental Nutrition Assistance Program (SNAP) caseworker behavior impacts program receipt, likely due to differing levels of helpfulness in navigating the complicated application process. We use the conditional random assignment of caseworkers as an instrument for SNAP receipt to assess the impact of SNAP on work decisions. Two-thirds of SNAP applicants do not work before applying and experience no change in work if granted SNAP. Those working beforehand decrease work temporarily. The canonical, static labor supply model cannot fully explain these results, but taking account of other reasons individuals do not work can.
In 2007, the Supreme Court declared race-conscious school admissions unconstitutional. This paper provides the first evaluation of a related federal mandate where a school district was forced to adopt a race-blind lottery system for its magnet schools. I explore the impact of the dramatic increase in racial segregation resulting from the mandate. More segregated schools spend less per-pupil, enroll lower achieving students, employ lower value-added teachers, and perpetuate "white flight'' out of the district. Ultimately, segregation arising from mandated race-blind admissions causes student achievement and college attendance rates to decline.
Selected Works in Progress
How Place and Need Intersect: How Lack of Retail and Enrollment Offices Deters Nutrition Assistance Take-Up (with Marianne Bitler and Sonya R. Porter)
The Supplemental Nutrition Assistance Program (SNAP) is central to the US safety net. SNAP is the only universal program that is part of the safety net. Surprisingly, a sizable portion of the eligible population does not claim benefits. From 2010 to 2013, SNAP participation rates among those eligible ranged between 72% to 86% with take-up rates among the elderly as low as 33%. Despite the size and importance of SNAP, we have an incomplete understanding of the determinants of participation, particularly among rural, impoverished populations. Specifically, we understand very little about the causal role that geographic barriers play in determining program participation across rural America.
This paper provides a comprehensive assessment of whether travel distances from place of residence to SNAP-authorized retail stores and SNAP program offices impacts program participation. Data limitations have prevented researchers from fully exploring the impact of travel distance on nutrition program participation. Existing research uses coarse proxies for travel distance and surveys with documented underreported program participation or, relies on a short cross-section as in the case of FoodAPS, which combines high-frequency national surveys with administrative data on SNAP participation. We overcome these challenges using novel administrative SNAP and WIC data for every applicant and recipient across 23 states over two decades, allowing us to explore these topics at an unprecedented level of breadth, depth, and accuracy. These data include administrative records on program participation and geocoded residence, which permit us to precisely measure the household-specific travel distance to the closest program office/clinic or grocer as well as program participation each month. We merge these records with the location of every SNAP-authorized food retailer in the country and every SNAP office as well as with administrative TANF and HUD program records, which allow us to determine cross-program participation.
We leverage variation in the timing of openings/closings among grocers and SNAP offices in a panel design that allows for heterogeneous treatment effects across groups and time. Preliminary findings show that decreases in travel distance to SNAP offices meaningfully increases SNAP participation among low income households, improving targeting. We find little systematic evidence that travel distance to SNAP-authorized grocers impacts program participation. Taken together, we conclude that geographic barriers play a meaningful role in accessing the safety net across rural America.
Funding: Institute for Poverty Research
Nutrition Assistance Eligibility and Labor Supply (with Marianne Bitler and Jonathan Rothbaum)
Incomplete program take-up during a crisis: Evidence from the COVID-19 shock in one U.S. state (with Marianne Bitler, Danea Horn, and Nathan Seegert), International Tax and Public Finance, 2022.
In the U.S., means-tested cash, in-kind assistance, and social insurance are part of a patchwork safety net, often run with substantial involvement of state and local governments. Take-up–participation among eligible persons–in this system is incomplete. A large literature points to both neo-classical and behavioral science explanations for low take-up. In this paper, we explore the response of the safety net to COVID-19 using newly-collected survey data from one U.S. state—Utah. The rich Utah data ask about income and demographics as well as use of three social safety net programs which collectively provided a large share of relief spending: the Unemployment Insurance program, a social insurance program providing workers who lose their jobs with payments; the Supplemental Nutrition Assistance Program, which provides benefit cards for purchasing unprepared food at retailers; and Economic Impact Payments, which provided relatively universal relief payments to individuals. The data do not suffice to determine eligibility for all of the programs, so we focus on participation per capita. These data also collect information on several measures of hardship and why individuals did not receive any of the 3 programs. We test for explanations that differentiate need, lack of information, transactions costs/administrative burden, stigma, and lack of eligibility. We use measures of hardship to assess targeting. We find that lack of knowledge as well as difficulty applying, and stigma in the UI program each play a role as reasons for not participating in the programs.
Rent-Seeking through Collective Bargaining: Teachers Unions and Education Production (with Stéphane Lavertu and Corbin Miller), Economics of Education Review, 2021.
We explore how teachers unions affect education production by comparing outcomes between districts allocating new tax revenue amidst collective bargaining negotiations and districts allocating tax revenue well before. Districts facing union pressure increase teacher salaries and benefits, spend down reserves, and experience no student achievement gains. Conversely, districts facing less pressure hire more teachers (instead of increasing compensation) and realize significant student achievement gains. We interpret these results as causal evidence of the negative impact of teacher rent seeking on education production, as the timing of district tax elections relative to collective bargaining appears to be as good as random.
Working for Your Bread: The Labor Supply Effects of SNAP (with Marianne Bitler and Jonathan Rothbaum), American Economic Review, Papers & Proceedings, 2021.
The Supplemental Nutrition Assistance Program (SNAP), the only universal US means-tested safety net program, has a low benefit-reduction rate. Thus, many SNAP recipients are working. We apply recent methods to study whether there is evidence of moral hazard among SNAP recipients. We see if individuals respond to incentives in SNAP eligibility by bunching near kink points in the budget set. While this responsiveness has been shown for various taxes and tax credits, little work has examined responsiveness of safety net program participants to kinks in their eligibility formulae. We use novel administrative data on eligibility determination and find little evidence of responsiveness around these kinks.
Funding: W.E. Upjohn Institute; Tufts/UConn RIDGE
Is There Still Son Preference in the United States? (with Francine D. Blau, Lawrence M. Kahn, Peter Brummund, and Miriam Larson-Koester), The Journal of Population Economics, 2020.
In this paper, we use 2008–2013 American Community Survey data to update and further probe evidence on son preference in the USA. In light of the substantial increase in immigration, we examine this question separately for natives and immigrants. Dahl and Moretti (Review of Economic Studies 75, 1085-1120, 2008) found earlier evidence consistent with son preference in that having a female first child raised fertility and increased the probability that the family was living without a father. We find that for our more recent period, having a female first child still raises the likelihood of living without a father, but is instead associated with lower fertility, particularly for natives. Thus, by the 2008–2013 period, any apparent son preference in fertility decisions appears to have been outweighed by factors such as cost concerns in raising girls or increased female bargaining power. In contrast, some evidence for son preference in fertility persists among immigrants. Immigrant families that have a female first child have significantly higher fertility and are more likely to be living without a father (though not significantly so). Further, gender inequity in source countries is associated with son preference in fertility among immigrants. For both first- and second-generation immigrants, the impact of a female first-born child on fertility is more pronounced for immigrants from source countries with less gender equity. Finally, we find no evidence of sex selection for the general population of natives and immigrants, suggesting that it does not provide an alternative mechanism to account for the disappearance of a positive fertility effect for natives.
Media: NY Times, The Atlantic, VOX
Government Privatization and Democracy: The Case of Charter Schools In Ohio (with Vladimir Kogan, Stéphane Lavertu, and Zachary Peskowitz), Journal of Politics, 2020.
Governments around the world have privatized public services in the name of efficiency and citizen empowerment, but some argue that privatization could also affect citizen participation in democratic governance. We explore this possibility by estimating the impact of charter schools (which are publicly funded but privately operated) on school district elections. The analysis indicates that the enrollment of district students in charter schools reduced the number of votes cast in district school board contests and, correspondingly, reduced turnout in the odd-year elections in which those contests are held. This impact is concentrated in districts that serve low-achieving, impoverished, and minority students, leading to a modest decline in the share of voters in those districts who are black and who have children. There is little evidence that charter school expansion affected the outcomes of school board elections or turnout in other elections.
The Effect of Charter Competition on Unionized District Revenues and Resource Allocation. Journal of Public Economics, 2018.
This study examines the impact of competition due to charter school entry on the level of revenues and the composition of expenditures within traditional public school districts (TPSDs). I leverage a policy change affecting the location and timing of charter entry to account for endogenous charter competition. TPSDs respond to competition by allocating resources away from instructional and other expenditures toward new capital construction. Using teacher contracts, I show that collectively bargained salaries are largely unresponsive to competition and that declines in instructional spending are primarily due to decreases in the number of employed teachers. Competition depresses appraised housing valuations, in turn causing TPSDs to lose property tax revenues resulting in a decline in overall spending.
Task-Specific Experience and Task-Specific Talent: Decomposing the Productivity of High School Teachers (with Richard Mansfield) Journal of Public Economics, 2016.
We use administrative panel data to decompose worker performance into components relating to general talent, task-specific talent, general experience, and task-specific experience. We consider the context of high school teachers, in which tasks consist of teaching particular subjects in particular tracks. Using the timing of changes in the subjects and difficulty levels to which teachers are assigned to provide identifying variation, we show that a substantial part of the productivity gains to teacher experience are actually subject-specific. Similarly, while three-quarters of the variance in the permanent component of productivity among teachers is portable across subjects and levels, there exist non-trivial subject-specific and level-specific components. Counterfactual simulations suggest that maximizing the test-score contribution of task-specific experience and task-specific talent can increase student performance by as much as .04 test score standard deviations relative to random assignment of teachers to classrooms.
Partially Adaptive Estimation of Interval Censored Regression Models (with James McDonald), Computational Economics, 2013.
Several valuable data sources, including the census and National Longitudinal Survey of Youth, include data measured using interval responses. Many empirical studies attempt estimation by assuming the data correspond to the interval midpoints and then use OLS or maximum likelihood assuming normality. Stata performs maximum likelihood estimates (MLE) under the assumption of normality, allowing for intra-group variation. In the presence of heteroskedasticity or distributional misspecification, these estimates are inconsistent. In this paper we focus on an estimation procedure that helps prevent distributional misspecification for interval censored data. We explore the application of partially adaptive estimation, which builds on the MLE framework with families of flexible parametric probability density functions which include the normal as a limiting case. These methods are used to estimate determinants associated with household expenditures based on US Census data. Monte Carlo Simulations are performed to compare the relative efficiency of the different methods of estimation. We find that the flexible nature of our proposed partially adaptive estimation technique significantly reduces estimator bias and improves efficiency in the presence of distributional misspecification.