Research

Papers 

Sensitivity Analysis of Treatment Effects with Endogenously Censored Duration Outcome

Abstract: With a non-randomly censored duration outcome, we perform sensitivity analyses on various treatment effect parameters when the dependence between the event time and the censoring variable is modeled by a family of Archimedean copulas. Bounds of policy effects are characterized as smooth functionals of the {copula graphic estimands} that satisfy an index sufficiency condition.  We then provide an estimation procedure and establish uniform inference theories for the proposed semiparametric estimators. Confidence bands are constructed using multiplier bootstrap. The estimators demonstrate good finite sample properties in Monte Carlo simulations.  The methodology is applied to study the efficacy of treatment protocols for acute lymphoblastic leukaemia. 

Covariate Distribution Balance via Propensity Scores

(with Pedro Sant’Anna and Xiaojun Song, Journal of Applied Econometrics, 2022)

Abstract: This paper proposes new estimators for the propensity score that aim to maximize the covariate distribution balance among different treatment groups. Heuristically, our proposed procedure attempts to estimate a propensity score model by making the underlying covariate distribution of different treatment groups as close to each other as possible. Our estimators are data-driven and can be used to estimate different treatment effect parameters under different identifying assumptions, including unconfoundedness and local treatment effects. We derive the asymptotic properties of inverse probability weighted estimators for the average, distributional, and quantile treatment effects based on the proposed propensity score estimator and illustrate their finite sample performance via Monte Carlo simulations and an empirical application. We analyze the effect of 401(k) retirement plans on asset accumulation. Our findings are comparable to those by existing estimation procedures. However, our proposed estimators tend to improve upon covariate distributional imbalances.

Two Sample Unconditional Quantile Effect

(with Atsushi Inoue and Tong Li, R&R at Journal of Econometrics)

Abstract: This paper proposes a new framework to evaluate unconditional quantile effects (UQE) in a data combination model.  The UQE measures the effect of a marginal counterfactual change in the unconditional distribution of a covariate on quantiles of the unconditional distribution of a target outcome. Under rank similarity and conditional independence assumptions, we provide a set of identification results for UQEs when the target covariate is continuously distributed and when it is discrete, respectively.  Based on these identification results, we propose semiparametric estimators and establish their large sample properties. Applying our method to a variant of Mincer's earnings function, we study the counterfactual quantile effect of actual work experience on income.

Difference-in-Differences with Compositional Changes

(with Pedro Sant’Anna, R&R at Journal of Econometrics)

Abstract: This paper studies difference-in-differences (DiD) setups with repeated cross-sectional data and potential compositional changes across time periods. We start the analysis by deriving the efficient influence function and the semiparametric efficiency bound for the average treatment effect on the treated (ATT). We propose nonparametric estimators that attain the semiparametric efficiency bound under mild rate conditions for the estimators of the nuisance functions, a type of rate doubly-robust (DR) property. When one uses parametric methods to estimate these nuisance functions, we show that our estimator enjoys a DR consistency property. We also document a trade-off related to compositional changes: we derive the asymptotic bias of DR DiD estimators that erroneously rule out compositional changes and also the loss of efficiency when one fails to rule out compositional changes correctly. Based on these trade-offs, we propose nonparametric Hausman-type tests for compositional changes. The finite sample performance of the proposed DiD tools is examined by means of Monte Carlo experiments and an empirical application. A new uniform stochastic expansion of the local polynomial logit estimator is provided as a by-product of our analysis, which can be of independent interest. 

Work in Progress

Markov Chain Monte Carlo Based Inference for Dynamic Discrete Choice Models

(with Atsushi Inoue, Tong Li, and Tatsushi Oka)

Moment Condition Approach for Dynamic Network Formation with Degree Heterogeneity (with Martin Schmitz)