Research
Links to some recent publications, forthcoming work and working papers
(please see my CV for a complete list of my publications and research in progress)
Hunger Pains: SNAP Timing, and Emergency Room Visits
(with John Gordanier and Chad Cotti)
The impact of poor nutrition has been established as an important determinant of health. It has also been demonstrated that the single monthly treatment of SNAP benefits leaves meaningful nutritional deficiencies in recipient households during the final weeks of the benefits cycle. Further, health related behaviors have been documented to be altered on the date of food stamp receipt. This project exploits highly detailed and linked administrative data on health care utilization of food stamp recipients and randomized food stamp receipt dates to allow us to measure the impact of food stamp treatment days and the low nutritional periods created by the SNAP benefits cycle on the likelihood of emergency department (ER) utilization among the Medicaid population. Our main finding is that among SNAP receiving individuals in the ER on a particular day, the share that received benefits on that day is 3.5% lower than would be expected. This effect is present across all age groups, although the magnitude is smallest for young children. Further, analysis of time-use data presents additional support for time use reallocation on receipt days, while an investigation of specific conditions is consistent with the ER utilization decline being related to a drop in less urgent conditions. Lastly, we find that for individuals 55 and over, the share of ER visits that comes from individuals that are past the third week of their SNAP benefit month, i.e. received benefits more than 21 days ago, is 1.5% larger than would be expected. This suggests that these older individuals are more likely to visit the ER late in the SNAP benefit cycle, which is consistent with increased food insecurity as a possible mechanism linking the food stamp benefits cycle to emergency care utilization. We find no such effect for younger individuals. Overall, there are several important policy implications of these findings, particularly among the Medicaid population.
(with Edward Frongillo, Christine E. Blake and Gabrielle Turner-McGrievy )
Published in Journal of Economic Behavior and Organization
In this study, we intervened in elementary schools on lunch entrée selection using some of the behavioral economic methods shown to be effective in earlier food choice studies. Unlike many earlier behavioral interventions, which were mostly done in controlled environments and smaller café type settings for one-off interactions, we conducted our interventions in a real-world environment in twelve elementary schools in one school district in South Carolina over nine school weeks. By increasing salience and prominence of the healthy entrée of the day through visual and verbal tools, we nudged students towards selecting healthier options in treatment schools. We estimated the treatment effects using a difference-in-differences setup, comparing changes in the share of students selecting nudged entrées during the treatment period relative to the shares before the treatment period in treatment and comparison schools. Our estimates show that the nudges are effective when present. They increase selection of the healthy option by thirteen to thirty-five percent on the days the entrée is treated. Effects disappear when the nudge is removed, however, and there is evidence for reduced effectiveness of nudges in repeat instances. There is no evidence of habit formation.
Economics of Education Review Volume 66, October 2018, Pages 40-50. (coauthored with John Gordanier and Chad Cotti)
The effect of poor nutrition has been established as an important determinant of learning and achievement among school-age children. Further, it has been shown that households’ receiving food stamps fail to smooth consumption over the benefit month and experience periods of meaningful nutritional deficiencies. This paper exploits detailed administrative data on standardized math tests scores and randomized food stamp receipt dates to allow us to measure the impact of these low nutritional periods on student performance. Our main results demonstrate that scores are notably lower when the exam falls near the end of the benefit cycle and when food stamps arrive on the 4 days immediately preceding the exam. While both male and female students experienced a similar penalty with receipt near the end of the cycle, the effect from receipt just prior to the exam appears to be partially explained by a large negative effect associated with weekend receipt, which coincides with the 4 days prior to the exam. Our results provide evidence that households do not sufficiently smooth consumption and that this has measurable effects on student performance.
Free Lunch for All! The Effect of the Community Eligibility Provision on Academic Outcomes
published in Economics of Education Review (coauthored with John Gordanier and Crystal Zhan and Breyon Wiliams)
We analyze the effect of the Community Eligibility Provision (CEP), a universal free-lunch program, on elementary and middle school students’ academic performance and attendance in the state of South Carolina. As part of the program, eligible schools can provide free lunches to all students, regardless of whether an individual student qualifies for free or reduced lunch. Using a difference-in-differences approach, we show that CEP leads to about 0.06 of a standard deviation increase in math test scores for elementary school students. We find smaller effects on reading scores and on middle school students. These effects also vary by student poverty, school poverty, and locality. In particular, we find students that were previously eligible for free lunches but not on other public assistance programs benefit the most from CEP. The results may suggest that the expansion of access to free lunch help improve students' academic outcomes.
Eat (and Drink) Better Tonight: Food Stamp Benefit Timing and Drunk Driving Fatalities
American Journal of Health Economics Fall 2016, Vol. 2, No. 4, Pages 511-534 (with John Gordanier and Chad Cotti)
This paper examines the relationship between the timing of food stamp benefits and daily alcohol-related fatal accidents. We exploit substantial exogenous variation in state food stamp distribution dates and enrollment numbers to estimate the relationship using binary outcome and count data frameworks. Our main result is that, in contrast to previous work on income receipt and mortality, alcohol-related accidents with fatalities are substantially lower on the date of food stamp receipt, and the result is largely driven by a same-day effect. Further, this effect is only present on weekdays. We find no effect of receipt on non-alcohol-related accidents. We hypothesize that this is possibly driven by families being more likely to eat at home on distribution days.
Does the Timing of Food Stamp Distribution Matter? A Panel‐Data Analysis of Monthly Purchasing Patterns of US Households
Health Economics Volume 26, Issue 11 November 2017 Pages 1380–1393 (with John Gordanier, Chad Cotti and Elena Castellari)
In this paper, we examine the relationship between the timing of food stamp receipt and purchasing patterns. We combine data on state distribution dates of food stamps with scanner data on a panel of households purchases tracked between 2004 and 2011. We find that purchases of a variety of goods are meaningfully higher on receipt days, consistent with previous work that suggests that recipients are very impatient. Additionally, and importantly, estimates indicate that when food stamp receipt days fall on weekends, total monthly purchases within the same households are affected. In particular, monthly purchases of beer are higher when food stamps are distributed on a weekend rather than in months where benefits are distributed on weekdays. For these households, total beer purchases are between 4 and 5% higher in those months. Among households ineligible for food stamps, no effect is identified. These results demonstrate that the ‘day‐of‐the‐week’ of SNAP treatment may have important impacts on household purchase habits. Copyright © 2016 John Wiley & Sons, Ltd.
Economics of Education Review Volume 62, February 2018, Pages 12-15 (with John Gordanier and Chad Cotti)
This paper examines the relationship between the start time and meeting frequency of college courses and the academic performance of students. Using administrative data from a large public university, we account for both student and instructor fixed effects. Consistent with a large literature, we find a positive time of day effect. That is, students earn higher grades in classes that start later. However, contrary to previous literature, we find students earn higher grades in classes with fewer meeting times when not accounting for instructor fixed effects. This effect is entirely explained by instructor sorting on course schedules. Instructors that assign higher grades, either due to quality of instruction or grade leniency, are more likely to meet twice a week rather than three times a week. Including instructor fixed effects, we find no difference in two-day a week classes and three-day a week classes. However, grades are lower in classes that meet just once a week.
Computer-based testing (CBT) is becoming an increasingly popular format of assessment in educational settings. If students face a digital divide in terms of access to computers at school and at home, CBT may exacerbate measured student achievement gaps. In this paper, we use the rollout of CBT in South Carolina starting in 2015 to investigate its effect on measured student performance. We link student-level test scores and poverty measures to the share of students taking CBT in a grade of a school and show that CBT has a significant negative impact on test scores of multiple subjects. The negative impact is not uniform across student subgroups but rather particularly large for students in poor households. There is little evidence that the effect fades as students and schools become more experienced with computerized testing. These results suggest that the testing mode change might have distributional consequences. However, we do find a smaller effect in schools where technology is more readily available, implying that school-level investments could mitigate the effect.
forthcoming in Industrial and Labor Relations Review (with John Addison and Liwen Chen)
We deploy a measure of occupational mismatch based on the discrepancy between the portfolio of skills required by an occupation and the array of abilities possessed by the worker for learning those skills. We report distinct gender differences in match quality and changes in match quality over the course of careers. We also show a substantial portion of the gender wage gap stems from match quality differences among the college educated. College-educated females are significantly more mismatched than males. Moreover, those individuals with children and in more flexible occupations tend to be more mismatched. Again, this is especially true of women. Cohort effects are also visible in the data: college-educated males of the younger cohort are worse off in terms of match quality compared to the older cohort, even as the new generation of women is doing better on average.
This paper mainly looks at the productivity loss of job displacement during recessions and assesses its longer-term impact on wages. We explore changes in match quality as a potential mechanism; that is, individuals who are displaced by recessions may be reemployed in jobs for which they are more mismatched during times of high unemployment. Using NLSY79 and NLSY97 data, we explore how workers’ match qualities change over the lifecycle. We further explore the role of age, education, gender, and labor market experience during the recessions using both NLSY and CPS data.
The Occupational Feminization of Wages
Industrial and Labor Relations Review Volume 71, Issue 1, January 2018, pp. 208-241 (with John T. Addison and Si Wang)
This article updates the 1995 study by Macpherson and Hirsch that used monthly Current Population Survey (CPS) data from 1973 to 1993 to examine the effects of occupational gender composition on earnings. In the updating process, the authors correct for biases in this data set that are attributable to the inclusion of imputed earners and the misreporting of occupation. They use CPS data from 1996 to 2010 to provide cross-sectional estimates of the impact of the feminization of occupations on wages, as well as its contribution to the gender wage gap. Longitudinal CPS data indicate that the negative effects of gender composition on earnings observed in cross-sectional data are lessened when researchers control for observed heterogeneity and are much reduced when controlling for unobserved heterogeneity. These findings are confirmed using much longer panels from the National Longitudinal Survey of Youth 1979 (NLSY79). Finally, the use of synthetic panels of aging cohorts suggests that wage penalties are largest for younger cohorts in predominantly female occupations.
The Role of Gender in Promotion and Pay over a Career
Journal of Human Capital, Vol. 8, No. 3, pp 280-317 Fall 2014 (with John T. Addison and Si Wang)
Using data from the NLSY79, this paper considers the role of gender in promotion and promotion-related earnings development over the course of a career. The raw data suggest reasonably favorable promotion outcomes for females over a career, but any such advantages are found to be confined to less educated females. Further, the strong returns to education in later career stemming from promotion-related earnings growth accrue solely to males. While consistent with fertility timing and choice on the part of educated females, this earnings result is not inconsistent with discrimination as well, reminiscent of findings from an earlier human capital literature.