Research

Abstract

I consider an additive non-parametric regression model, in which one of its components is discontinuous whereas all others are continuous. A special case of this model is Regression Discontinuity (RD) Design. Empirically, researchers often include covariates other than the running variable in regression to guard against the misspecification or to increase the precision of the estimates. I propose a two-stage estimator, Spline-HLC estimator, that uses a B-Spline estimator for the continuous components of the regression at the first stage and uses the Hestenes local constant estimator (HLC), a general reflection kernel estimator I proposed in my dissertation, to estimate the jump discontinuity at the second stage. I provide the asymptotic properties of the proposed estimator and show that it is oracle efficient. A Monte Carlo study shows that when the additional regression components are not linear, the proposed estimator reduces the root mean square error by almost half of the additive local linear RD estimators that is commonly used in empirical research. Moreover, the finite sample performance of the proposed spline-HLC estimator is compared with other two-stage estimators that use marginal integration, instrumental variable, or back-fitting at the first stage and HLC at the second stage. An empirical example illustrates the applicability of the proposed estimator.


Publications

A New Estimator of a Jump Discontunity in Regression ( with C. Martins-Filho and Feng Yao) Economics Letters, September 2022

We proposed a new class of estimators for regression discontinuity (RD) designs, a popular econometric model for evaluating the effects of policy interventions. The insight is to use a mathematical result – the extension of Hestenes (1941) – to construct a class of RD estimators. The proposed estimators have simple analytical representations, desirable asymptotic properties and are computationally easy to implement. We provide asymptotic properties, extensive Monte Carlo results, and an illustrative empirical example to compare my estimators to commonly used local linear (LL) estimators.

I examine the impact of import and export tariff reductions on China's regional labor market in the wake of China’s 2001 accession to the WTO. Applying the specific factor model to intermediate goods markets and the labor demand theory to export goods markets, I found the growth of export industries mitigates the negative impact from increased import competition. In addition, a competitive intermediate goods market fostered by import tariff reductions positively enhances the comparative advantage in regions exposed to both import and export tariff reductions. By estimating a region and time fixed effect OLS model, I find that at the city-level, a 10% reduction in import tariff causes a 5.86% decrease in wages and a 7.14% decrease in employment. A 10% reduction in export tariff causes a 18.7% increase in wages and a 21.3% increase in employment.

I employ the specific factor model in intermediate good sectors to capture the negative impact of trade liberalization on import and Melitz’s (2003) heterogeneous firm model in final good sectors to capture the positive impact of trade liberalization on export through intermediate goods import. Using Chinese Customs and firm data, I empirically show that regions facing increasing globalization have better labor markets if they have fewer intermediate good industries that use entrenched specific factors, have more firms exporting by utilizing imported intermediate goods, and are closer to global markets.