Research

Job market paper

Mental health disorders are particularly prevalent among those in the criminal justice system and may be a contributing factor in recidivism. Using North Carolina court cases from 1994 to 2009, this paper evaluates how mandated mental health treatment as a term of probation impacts the likelihood that individuals return to the criminal justice system. I use random variation in judge assignment to compare those who were required to seek weekly mental health counseling to those who were not. The main findings are that being assigned to seek mental health treatment decreases the likelihood of three-year recidivism by about 12 percentage points, or 36 percent. This effect persists over time, and is similar among various types of individuals on probation. In addition, I show that mental health treatment operates distinctly from drug addiction interventions in a multiple-treatment framework. I provide evidence that mental health treatment's longer-term effectiveness is strongest among more financially-advantaged probationers, consistent with this setting, in which the cost of mandated treatment is shouldered by offenders. Finally, conservative calculations result in a 5:1 benefit-to-cost ratio which suggests that the treatment-induced decrease in future crime would be more than sufficient to offset the costs of treatment.

[For those who don't have access to Google Drive, the paper can also be found at http://arxiv.org/abs/2212.06736]

with Jake Blackwood, Cindy Cunningham, Matthew Dey, Lucia Foster, Cheryl Grim, John Haltiwanger, Sabrina Wulff Pabilonia, Jay Stewart, Cody Tuttle, and Zoltan Wolf

Technology, Productivity, and Economic Growth, forthcoming. NBER Working Paper 30620.

An important gap in most empirical studies of establishment-level productivity is the limited information about workers’ characteristics and their tasks. Skill-adjusted labor input measures have been shown to be important for aggregate productivity measurement. Moreover, the theoretical literature on differences in production technologies across businesses increasingly emphasizes the task content of production. Our ultimate objective is to open this black box of tasks and skills at the establishment-level by combining establishment-level data on occupations from the Bureau of Labor Statistics (BLS) with a restricted-access establishment-level productivity dataset created by the BLS-Census Bureau Collaborative Micro-productivity Project. We take a first step toward this objective by exploring the conceptual, specification, and measurement issues to be confronted. We provide suggestive empirical analysis of the relationship between within-industry dispersion in productivity and tasks and skills. We find that within-industry productivity dispersion is strongly positively related to within-industry task/skill dispersion. 

with Barbara Downs, Lucia Foster, and Danielle Sandler

CES Working Paper 23-25.

Existing work has shown that the entry of a child into a household results in a large and sustained increase in the earnings gap between male and female partners in opposite-sex couples. Potential reasons for this include work-life preferences, comparative advantage over earnings, and gender norms. We expand this analysis of the child penalty to examine earnings of individuals in same-sex couples in the U.S. around the time their first child enters the household. Using linked survey and administrative data and event-study methodology, we confirm earlier work finding a child penalty for women in opposite-sex couples. We find this is true even when the female partner is the primary earner pre-parenthood, lending support to the importance of gender norms in opposite-sex couples. By contrast, in both female and male same-sex couples, earnings changes associated with child entry differ by the relative pre-parenthood earnings of the partners: secondary earners see an increase in earnings, while on average the earnings of primary and equal earners remain relatively constant. While this finding seems supportive of a norm related to equality within same-sex couples, transition analysis suggests a more complicated story. 

Since 2001, 32 states have implemented policies that require all public high schools to administer either the ACT or SAT to juniors during the school day free of charge, making that aspect of the college application less costly on both time and financial dimensions. I evaluate these policies using American Community Surveys (ACS) from 2000 to 2019 and exploiting variation in policy implementation across state and time. I augment ACS data with the Census Master Address File to precisely identify the state in which individuals took the exam during high school. I find, across all specifications, that increased access to standardized college entrance exams has no effect on subsequent college attendance. Access to the exam also does not shift students between public and private colleges, nor does it shift students between two-year and four-year degree schools. This evidence suggests that ability to take a college entrance exam is not one of the binding constraints for students on the margin of going to college. 

with Jeff Denning, Nolan Pope, and Merrill Warnick

Teachers have discretion over how they map student achievement into grades, and their leniency in doing so may positively or negatively impact student achievement. In this paper we construct two measures of grading leniency: "mean grade inflation" which measures how much higher grades are than would be expected, and "passing grade inflation" which measures leniency in receiving a passing grade. We show that these measures represent related but distinct grading practices of teachers. Grading leniency is not very correlated with other well-established teacher characteristics such as test score and non-cognitive value-added, which suggests that teachers may face tradeoffs in classroom practices. We show that more lenient teachers reduce performance on tests in subsequent years, and that leniency also has persistent effects, decreasing some students' likelihood of taking the SAT and graduating. However, while mean grade inflation negatively affects outcomes, we find that passing grade inflation is positively associated with grade progression, especially for lower-performing students.

with Ken Coriale and Ethan Kaplan

We use conditional random assignment of roommates at the University of Maryland to estimate the causal impact of social connections on political behavior. Using university data on roommate assignments matched to voter registration data, we compare counterfactual assignments of roommates to those that actually occurred. We develop new methods to estimate exact bounds on treatment effects to account for a partially missing independent variable. We find evidence of homophilic preference formation as roommates are more likely to both be registered with the same party and to both vote, in comparison to counterfactual roommate pairs randomly not selected.

The Earned Income Tax Credit (EITC) is the largest cash-transfer program in the United States for low-income families with children. Although it has been extensively studied, most of the focus of the literature has been on single mothers, for whom the EITC provides unambiguously positive incentive to work. In contrast, the direction of the labor supply response of married mothers depends on several factors including the income of her spouse. Eissa and Hoynes (2004) showed that expansions between 1984 and 1996 resulted in a decrease in the labor supply of married mothers. I revisit married mothers in the context of the 1975 introduction of the EITC. I replicate and reconcile conflicting results from recent analyses (Bastian 2022, Kleven 2019), highlighting heterogeneity in the response by marital status as a main driver of those conflicting results. I then turn my focus to married women, finding across all specifications that the 1975 EITC increased married mothers’ employment by about 3.7 percentage points. To understand why this finding contrasts with the effect of later EITC expansions, I show that, compared to the 1990s, married women during this period were much more likely to be in the phase-in range, for which theory predicts an increase in labor supply.

work in progress; with Guido Kuersteiner