Working Papers

“Regression Discontinuity Design: Identification and Estimation of Treatment Effects with Multiple Selection Biases”, 2007 (JOB-MARKET PAPER) [Supplemental Web Appendices]

Previous work on the regression discontinuity (RD) design has emphasized identification and estimation of an effect at the selection threshold (discontinuity). Focusing on the so-called "fuzzy" RD design, this paper examines identification and estimation of the average treatment effect (ATE) under various forms of selection bias --- selection on the observables, selection on the unobservables, and selection based on heterogeneity in the effects of the treatment. Easy to implement estimators that are root-N consistent and asymptotically normal are derived. They allow for general functional forms for the selection biases and imply specification tests for the plausibility of the statistical assumptions. This paper also investigates the trade-off between efficiency and bias in estimating the average treatment effect (and average effects local to the discontinuity) when the effects covary with the observables and the unobservables. The theoretical results leverage the dual nature of the RD design --- both the "borderline experiment" provided near the threshold and the strong and valid exclusion restriction provided in the selection equation for the choice of treatment. This point is demonstrated through Monte-Carlo experiments and empirical applications.

“Identification and Estimation of Social Interaction-Based Models: A Changes-in-Changes Approach with an Application to Adolescent Substance Use”, 2007

This paper outlines a method for detecting and assessing the strength of social interactions through a changes-in-changes design. The proposed approach is based on a linear-in-means model and aims to resolve the “reflection problem”, unobserved heterogeneities and endogenous group formation that plague identification of social interactions. Using longitudinal data from Add Health with rarely collected information on peer group’s composition, we explore an exogenous variation in peer’s drug use induced by a “mover friend” that occurs between Add Health’s survey periods. This quasi-experiment shares a similar nature of a policy intervention of removing drug-user friends from a peer group. Such treatment-control group differences together with changes over time form the basis of our changes-in-changes design. Our study confirms a strong endogenous effect, which in turn motivates a “social multiplier”, both of which are large enough to be relevant and are well worth attention to policy makers, researchers, health-care providers and educators for better understanding of how to protect young people and secure our future.

“‘Rebel without a Cause’: A Social Interactions-Based Analysis of Youth’s Truancy Behavior”, 2005

Recent studies on youths’ truancy regard truancy as a rational choice. We frame the issue of truancy behavior using a social interactions-based rational choice model, which synthesizes the sociological viewpoint on truancy (as a kid’s culture of rebellion) and the economic rationale for truancy (as a rational choice). Our study of this rational decision-making process with social interactions provides a theoretical foundation for explanations of why students may be trapped into an “educational poverty”, whether “social multipliers” exist, and how such multiplier effects induced by social contexts can become a way out of the “truancy trap”. The corresponding empirical analysis, employing the EM algorithm, avoids the “reflection problem” typical of linear-in-means social interactions models, and allows of the possibility to distinguish among endogenous, contextual and correlated effects. Identification of these structural parameters provides empirical counterparts of “truancy traps” and “social multipliers”, which bear rich policy implications. Using data from the Los Angeles Family and Neighborhood Survey (L.A. FANS), this study empirically identifies strong peer influence in teenagers’ truancy behavior. The traditional viewpoint on truancy as a personal deficit or problem might be misleading. More responsibilities may be required from schools, families and neighborhoods, to prevent truancy from becoming deeply rooted due to the conformity effect, and to reduce the threat of the “truancy traps”.

“A Simple Lagrange Multiplier F-test for Multivariate Regression Models”, 2005 (with T. K. M. Beatty and J. T. LaFrance)

This paper proposes a straightforward, easy to implement, approximate F-test for testing restrictions in multivariate regression models. We derive the asymptotics for the test statistic and investigate its finite sample properties through a series of Monte Carlo experiments. Both theory suggests and simulations confirm that this approach results in improved inferences relative to the leading alternative.