Journal Articles: 1. Statistical inference for panel dynamic simultaneous equations models (with Cheng Hsiao), Journal of Econometrics 189, 383–396, 2015. online appendix 2. Asymptotic theory for linear diffusion processes under alternative sampling schemes (with Jun Yu), Economics Letters 128, 1-5, 2015. 3. A Stein-like estimator for linear panel data models (with Yun Wang and Yonghui Zhang), Economics Letters141, 156-161, 2016. 4. Asymptotic distribution of quasi-maximum likelihood estimation of dynamic panels using long difference transformation when both N and T are large (with Cheng Hsiao), Statistical Methods and Applications 25, 675-683, 2016. 5. Common Correlated Effects Estimation of Unbalanced Panel Data Models with Cross-Sectional Dependence (with Yonghui Zhang), Journal of Economic Theory and Econometrics 27, 25-45, 2016. 6. Panel Kink Regression with an unknown Threshold (with Yonghui Zhang and Li Jiang), Economics Letters 157, 116-121, 2017. 7. Many IVs estimation of dynamic panel regression models with measurement errors (With Nayoung Lee and Roger Moon), Journal of Econometrics200, 251-259, 2017. 8. First Difference or Forward Demeaning: Implications for the Method of Moments Estimators (with Cheng Hsiao), Econometric Reviews 36, 883-897, 2017. 9. To Pool or Not to Pool: Revisited, (with M.Hashem Pesaran), Oxford Bulletin of Economics and Statistics 80, 185-217, 2018. 10. Estimation of time-invariant effects in static panel data models (with M.Hashem Pesaran), Econometric Reviews 37, 1137-1171, 2018. Stata command to implement the FEF and FEF-IV estimators can be found here (Credit to Yui Law), help file of the command can be found here, a detailed description of the Stata command can be found here. 11. Binary choice model with interactive effects (with Sen Xue and Tao Yang), Economic Modelling70, 338-350, 2018. 12. Incidental parameters, initial conditions and sample size in statistical inference for dynamic panel data models (with Cheng Hsiao), Journal of Econometrics207, 114-128. 13. JIVE for Panel Dynamic Simultaneous Equations Models (with Cheng Hsiao), Econometric Theory 34, 1325-1369, 2018. 14. Estimation for time-invariant effects in dynamic panel models with application to income dynamics (with Yonghui Zhang), Econometrics and Statistics 9, 62-77, 2019. 15. Panel Parametric, Semi-parametric and Nonparametric Construction of Counterfactuals (with Cheng Hsiao), Journal of Applied Econometrics34, 463-481, 2019. 16. Identification and estimation in panel models with overspecified number of groups (with Ruiqi Liu, Zuofeng Shang and Yonghui Zhang), Journal of Econometrics 215, 574-590, 2020. 17. Partially Linear Functional-Coefficient Dynamic Panel Data Models: Sieve Estimation and Specification Testing (with Yonghui Zhang), Accepted for publication at Econometric Reviews. Stata command to implement the 2SLS sieve estimation can be found here (Credit to Kerry Du). A detailed description of the Stata command can be found here. 18. Forecasting equity index volatility by measuring the linkage among component stocks (with Yue Qiu, Tian Xie and Jun Yu), Accepted for publication atJournal of Financial Econometrics.
Book Chapters: 1. Correction for the Asymptotical Bias of the Arellano-Bond type GMM Estimation of Dynamic Panel Models (with Yonghui Zhang), Advances in Econometrics Volume 41, Essays in honor of Cheng Hsiao, 2020.
1. GMM and IV Estimation of Dynamic Panel Models with Heterogeneous Trend (with Shiyun Cao, Yonghui Zhang) 2. Specification Tests for Time-Varying Coefficient Panel Data Models (with Alev Atak and Yonghui Zhang) 3. Estimation and Inference of Treatment Effects using a New Panel Data Approach: Measuring the Impact of US SYG Law (with Huayan Geng) 4. Transformed Estimator for Panel Interactive Effects Models (with Cheng Hsiao and Zhentao Shi) 5. Estimation of Dynamic Panel Data Models with Interactive Effects: Quasi-differencing over Time or Pairwise? (with Cheng Hsiao) 6. Factor Dimension Determination for Panel Interactive Effects Models (with Cheng Hsiao and Yimeng Xie) Abstract: Since the estimation of the slope coefficients and the common factors are mutually dependent, we suggest a recursively iterating procedure and a computationally simple projection strategy for implementing the Bai and Ng's (2002) information criterion. We show that both approaches lead to consistently selected factor dimension when both sample cross-sectional dimension N and time series dimension T are large. Monte Carlo simulations show that both the recursively iterating procedure and projection approach work remarkably well. 7. Panel Data Approach for Measuring the Average Treatment Effects with Multiple Treated Units: To Aggregate or Not (with Cheng Hsiao and Yan Shen) Abstract: We consider aggregation methods in terms of the equal weight, weight derived from the eigenvector that corresponds to the largest or smallest eigenvalues of the covariance matrix of the underlying micro-data. We show that the statistical properties of the aggregated data not only depends on the weighting method, but also depends on the normalization conditions. Different weighting schemes also shed light on the issue of "to aggregate or not" for panel data analysis of average treatment effects (ATE) when there are multiple treated units over time. Monte Carlo studies are conducted to illustrate the issues involved with different aggregation methods. We also consider changes in China's peer-to-peer (P2P) outstanding loan balance to a policy change using different aggregation methods. 8. Statistical Inference for the Low Dimensional Parameters in the Presence of High-Dimensional Data: An Orthogonal Projection (with Cheng Hsiao) Abstract: We consider the estimation and statistical inference for low dimensional parameters for a model with many covariates. Instead of using usual machine learning/big data approach to select important covariates, we suggest a computationally simple orthogonal projection approach that has the advantage that (i) bypass the steps of selecting relevant important control variables, hence avoid selection mistakes, and (ii) is easy to conduct statistical inference. Monte Carlo simulations are conducted to investigate the finite sample performance of the proposed estimator, and the double/debiased estimator of Belloni et. al (2014) and Chernozhukov et. al (2018).