I am an
Assistant Professor of Economics and Education at Teachers College, Columbia University. My research uses randomized controlled trials (RCTs) to find low-cost, scalable interventions that improve education outcomes.
I am an Assistant Professor of Economics and Education at Teachers College, Columbia University. My research uses randomized controlled trials (RCTs) to find low-cost, scalable interventions that improve education outcomes.
Parent-Child Information Frictions and Human Capital Investment: Evidence from a Field Experiment
Investment (Revise and Resubmit, Journal of Political Economy)
This paper uses a field experiment to answer how information frictions between parents
and their children affect the inputs to human capital formation and how much reducing
these frictions can improve student achievement. I model the interaction between
parents and their child as a persuasion game with monitoring and incentives. I
show that parents have upwardly-biased beliefs about their child's effort, which
is associated with lower performance. In Los Angeles, a random sample of parents
was provided detailed information about their child's academic progress. More
information allows parents to induce more effort
from their children, which translates into significant gains in achievement.
However, additional information also changes parents' beliefs about their
child's effort, which spurs further parental monitoring. Relative to other
interventions, additional information to parents potentially produces
gains in achievement at a low cost.
Successful Schools and Risky Behaviors Among
(forthcoming, Pediatrics. with Mitchell Wong, Karen Coller, Rebecca Dudovitz, David Kennedy, Richard Buddin, Martin Shapiro, Sheryl Kataoka, Arleen Brown, Chi Hong Tseng, and Paul Chung)
This paper studies the long-run impact of a court-ordered desegregation ruling on education outcomes.
This ruling mandates that seven school districts, which serve higher-income, predominantly-white families,
accept a fixed number of minority elementary school students each year who apply to transfer from a nearby,
predominantly minority school district. The fixed number of slots are allocated to families via lottery.
The offer to transfer increases the number of students who enroll in college by 7 percentage points.
This result is driven by greater attendance to two-year and public colleges, though there are substantial
heterogeneous effects. Effects are substantially larger for Black and male students.
In theory, the publication of performance ratings may improve performance through reputation concerns or impede performance through embarrassment and stress. This paper uses school-district data and a regression discontinuity design to answer how consumers and employees respond to making performance information public. We find that higher-performing students sort into classrooms with highly-rated teachers. Teachers who were published do not perform better or worse than teachers who were not published, on average. This average effect is due to the heterogeneous impact of publication; highly-rated teachers perform worse following publication while low-rated teachers perform better. These results create two counter-veiling effects on the achievement gap between high and low-performing students. We find the net effect of making performance information public on test scores is zero, but the gap between high and low-performing students appears to close slightly as a result.
Research in Progress
Technology Adoption in Education: Usage, Spillovers and Student Achievement (Coming soon)
Previous research shows that significant information asymmetries can exist within
families and providing detailed information to parents about their child's academic
performance can significantly improve student achievement. Many school districts
accomplish the latter at scale via technology that places student information online
for parents. This paper uses a two-stage experiment across 59 schools and three
districts to study the adoption of this technology by parents along extensive and
intensive margins as well as spillovers and effects on student outcomes.
Adoption follows a typical S-shape; significant spillovers occur along
intensive but not extensive margins; and student achievement improves as a result.
The Impact of Tax Credit and Financial Aid
Information on College Outcomes: Evidence from a Large-Scale
Field Experiment (with Jeff Denning, Day Manoli, and Nick Turner)
An RCT designed to improve college enrollment and persistence for 200,000+ students in Texas by providing and re-framing education tax benefit and FAFSA information via the public-college application platform in the state of Texas.
The Effects of Defaults on Technology Adoption and Efficacy:
Evidence from a Field Experiment (with Todd Rogers)
Testing how defaults affect the take up and the efficacy of an automated-text message system alerting parents to their child's missing assignments, grades and absences.
Does Information on School Quality Impact Residential Choice? Evidence from a Nation-wide Field Experiment (with Eric Chan, Matt Hill and Heather Schwartz)
A nationwide-RCT adding school quality information onto low-income housing rental websites and adjusting default search frames to help 10,000+ families move to areas with better schools.
Engaging Parents as a Means to Address Educational and Health Behavioral Outcomes: Evidence from a Field Experiment (with Rebecca Dudovitz, Anne Escaron and Mitchell Wong)
Exploring whether engaging parents in their chid's education reduces teens' risky behaviors.
Estimations of a teacher's value added are founded on models of the education production function. In practice, teacher effects are often identified by assuming the effects are additively separable from observable and unobservable variables and that unobservable inputs and endowments correlated with teacher assignment are captured by prior test scores. These assumptions impose restrictions on how prior inputs and ability enter the model and eliminate heterogeneous teacher effects. We relax these restrictions by estimating teachers' value added using several semi-parametric methods. We document several facts: First, a teacher's value added varies significantly by students' initial performance. There is much greater within-teacher variation in value added than across teacher variation. Second, we find evidence of policy-relevant variation in teachers' value added by model specification. Our most flexible model finds at least 18% of teachers would be reclassified out of the lowest or highest quintiles. We find that a simple OLS specification that includes teacher fixed effects interacted with higher-order terms for lagged test scores approximates our most flexible semi-parametric model best, and that the models that allow for heterogeneity of the teacher effect tends to perform slightly better in out of sample predictions. We argue that, while the standard approach to estimating value added will be appropriate in most settings when the average marginal effect is desired, the interacted OLS model is more appropriate when the investigator is concerned with students at a certain achievement level (e.g., how well teachers perform with low-achieving students).