Research

Working papers

The Gryffindor Effect: the effect of manufactured identity on student high-stakes outcomes

I leverage a novel natural experiment in which incoming high school students are randomly assigned to be members of one of four groups, all of which share a common environment and common resources, to test whether manufactured identity can affect high-stakes student outcomes. This setting overcomes challenges to identification in applied work in identity economics: individual identities are often fixed over long periods of time or chosen endogenously. Using a dataset I collect from the yearbooks of a South African school students' LinkedIn and Instagram profiles, I find robust causal evidence that social identity plays an important role in determining student outcomes, including grades on high-stakes senior year exams, sports participation, tertiary education, labor market and social outcomes. I use text analysis to investigate the traits and behaviors groups emphasize to shed light on which group prescriptions drive outcomes in certain directions. 

Intergenerational Mobility in Cognitive Skills: New Evidence from India, with Arindam Nandi, Subha Mani, Jere Behrman & Ramanan Laxminarayan

Inequalities in schooling are known to persist across generations, but schooling does not always well reflect learning, especially in low- and middle-income countries. We collect and analyze unique Raven’s test score data from 4,514 parent-child pairs in India, providing the first evidence in an LMIC context on intergenerational mobility in learning—a better proxy for education and socioeconomic status than schooling attainment. We find that absolute intergenerational Raven’s test score mobility is low – only 33% of children have higher unadjusted test scores than their fathers. Relative mobility is also limited, with children whose fathers score in the bottom half of their generation, scoring on average at the 34th percentile. Notably, mobility in schooling is much higher than in learning, and substantial inequalities in learning prevail across population subgroups. Finally, we show that relying solely on mobility estimates based on schooling risks overstating educational and socioeconomic mobility by ignoring wide disparities in skill acquisition. 

Works in progress

Estimating Enumerator Amplifying Effects in Randomized Control Trials with Binary Treatment, with Antoine Deeb

We leverage random assignment of enumerators to respondents to develop an estimator for enumerator amplifying effects in Randomized Control Trials (RCTs) with binary treatment. Although enumerators are widely used in RCTs to implement and track interventions, researches often don't account them in their results, conclusions or generalizations. Our estimator and its standard errors provide researchers with a simple way to quantify and test for the presence of enumerator amplifying effects. 

The Effect of Automatic Express Lane Eligibility on Awareness of Own-Enrollment in Medicaid and CHIP for Children

I make use of a staggered difference-in-differences approach to estimate the effect of Automatic Express Lane Eligibility (ELE) on the number of low-income individuals who report that they are covered by Medicaid or some other government assistance program. Previous work finds that automatically enrolling individual's  increases enrollment in Medicaid and CHIP for children. Contrary to this, I find that  Automatic ELE has a negative effect on self-reported knowledge of program take-up. This result may be driven by sample composition, individuals that meet the 2008 income eligibility threshold for CHIP or Medicaid for children. These individuals tend to be low-income even among those typically eligible for Medicaid, which gives rise to two possible explanations: these individuals are being crowded out by higher income individuals making use of automatic ELE, or the automatic process makes it so they are unaware of their enrollment in Medicaid. These explanations are consistent with existing work that finds that ELE enrollees use healthcare services less than individuals enrolled through normal channels.