Research

This paper shows how to use a randomized saturation experimental design to identify and estimate causal effects in the presence of social interactions—one person’s treatment may affect another’s outcome–and one-sided non-compliance–subjects can only be offered treatment, not compelled to take it up. Two distinct causal effects are of interest in this setting: direct effects quantify how a person’s own treatment changes her outcome, while indirect effects quantify how her peers’ treatments change her outcome. We consider the case in which social interactions occur only within known groups, and take-up decisions do not depend on peers’ offers. In this setting we point identify local average treatment effects, both direct and indirect, in a flexible random coefficients model that allows for both heterogenous treatment effects and endogeneous selection into treatment. We go on to propose a feasible estimator that is consistent and asymptotically normal as the number and size of groups increases.

Presented at: Econometric Society World Congress 2020, The Philadelphia Fed, the 2018 IAAE Annual Conference, and the 2018 SEA Annual Meetings. 

Following the seminal paper by Altonji and Segal (1996), empirical studies have widely embraced equal or diagonal weighting in minimum distance estimation to mitigate the finite-sample bias caused by sampling errors in the weighting matrix. This paper introduces a new weighting scheme that combines cross-fitting and regularized weighting matrix estimation. We also provide a new cross-fitting standard error, applying cross-fitting to estimate the asymptotic variance. In a many-moment asymptotic framework, we demonstrate the effectiveness of cross-fitting in eliminating a first-order asymptotic bias due to weighting matrix sampling errors. Additionally, we demonstrate that some economic models in the earnings dynamics literature meet certain sparsity conditions, ensuring that the proposed regularized weighting matrix behaves similarly to the oracle weighting matrix for these applications. Extensive simulation studies based on the earnings dynamics literature validate the superiority of our approach over commonly employed alternative weighting schemes. 

Experimenters often collect baseline data to study heterogeneity. I propose the first valid confidence intervals for the VCATE, the treatment effect variance explained by observables. Conventional approaches yield incorrect coverage when the VCATE is zero. As a result, practitioners could be prone to detect heterogeneity even when none exists. The reason why coverage worsens at the boundary is that all efficient estimators have a locally-degenerate influence function and may not be asymptotically normal. I solve the problem for a broad class of multistep estimators with a predictive first stage. My confidence intervals account for higher-order terms in the limiting distribution and are fast to compute. I also find new connections between the VCATE and the problem of deciding whom to treat. The gains of targeting treatment are (sharply) bounded by half the square root of the VCATE. Finally, I document excellent performance in simulation and reanalyze an experiment from Malawi.

I establish primitive conditions for unconfoundedness in a coherent model that features heterogeneous treatment effects, spillovers, selection-on-observables, and network formation. I identify average partial effects under minimal exchangeability conditions. If social interactions are also anonymous, I derive a three-dimensional network propensity score, characterize its support conditions, relate it to recent work on network pseudo-metrics, and study extensions. I propose a two-step semiparametric estimator for a random coefficients model which is consistent and asymptotically normal as the number and size of the networks grows. I apply my estimator to a political participation intervention in Uganda and a microfinance application in India. 

Using a wide sample of countries during the period 1974-2004, we instrument leader exits and financial crises to assess the causal effect each has on the other.  We find that leader exits due to scheduled elections and term limits raise the probability of a banking crisis in the same year by 9% and that of a twin crisis by 7.6%. These effects are highly significant statistically, robust, and confined primarily to presidential regimes.  In contrast, for financial crises instrumented with determinants from early warning models, only sovereign defaults appear to induce the exit of national leaders.