Publications
Optimal model averaging for single-index models with divergent dimensions [paper] [supp]
This paper studies optimal model averaging for single-index models (SIMs), and provides the first theoretical results of averaging regularized estimators (such as Lasso) to deal with high dimensionality, which can be generalized to other models than SIMs.
Statistica Sinica, forthcoming [with Xinyu Zhang, Jiahui Zou and Guohua Zou]
Estimation of panel group structure models with structural breaks in group memberships and coefficients [paper] [supp] [code]
This paper considers a time-varying group pattern of heterogeneity in panel data models, and propose an estimation procedure to detect structural changes in the unknown group memberships and/or slope coefficients.
Journal of Econometrics (2023), 233: 45-65 [with Robin Lumsdaine and Ryo Okui]
Multi-dimensional latent group structures with heterogeneous distributions [paper] [supp]
Groups of units may differ not (only) in the conditional mean effects but (also) the quantile effects. We show that when the group structure is constant across different parts of distribution, using a composite quantile approach can improve clustering accuracy. We also allow cross-section fixed effects and slopes to exhibit distinct heterogeneity pattern, thus considering a multi-dimensional clustering.
Journal of Econometrics (2023), 233: 1-21 [with Heng Chen and Xuan Leng]
Too similar to combine? On negative weights in forecast combination [paper]
In forecast combination, when are the estimated optimal weights likely to be negative? How do we interpret combined forecasts with negative weights? Is it beneficial to restrict the weights to be positive? If yes, what is the best way to impose such constraints. This paper provides a comprehensive study on negative weights in combination by addressing these questions.
International Journal of Forecasting (2023), 39:18-38 [with Peter Radchenko and Andrey Vasnev]
Optimal model averaging for divergent-dimensional Poisson regressions [paper]
This paper proposes a Kullback-Leibler-distance-based model averaging method to address model uncertainty in Poisson regressions, allowing the dimension of covariates to increase with the sample size.
Econometric Reviews (2022), 41:775-805 [with Xinyu Zhang, Jiahui Zou and Guohua Zou]
Deleting unreported innovation [paper] [code]
This paper investigates the mechanism of firm unreported R&D. We document unreported innovation is systematically correlated with several firm, industry, and country characteristics, and show that unreported R&D are to some extent predictable. We assess various methods for dealing with unreported innovation.
Journal of Financial and Quantitative Analysis (2022), 57:2324-2354 [with Ping-Sheng Koh, David Reeb, Elvira Sojli, Wing Wah Tham]
(best paper award at Northern Finance Association Conference 2018)
Heterogeneous structural breaks in panel data models [paper] [supp] [code]
This paper considers heterogenous structural breaks in panel data models, allowing the number of breaks, timing, and size of these breaks to differ across groups of units. We employ a hybrid of K-means and adaptive group fused Lasso to jointly identify the latent group structure and detect the breaks.
Journal of Econometrics (2021), 220:447-473 [with Ryo Okui]
Panel threshold regressions with latent group structures [paper] [supp] [code]
This paper studies estimation of panel threshold regressions where slope coefficients and threshold parameters may exhibit latent group structures. We also consider inference issues, i.e., testing for the existing of (group-specific) threshold effects and for the homogeneity of threshold parameters across groups.
Journal of Econometrics (2020), 214:451-481 [with Ke Miao and Liangjun Su]
To pool or not to pool: What is a good strategy for parameter estimation and forecasting in panel regressions? [paper] [supp] [code]
In panel estimation and forecasting, the decision of whether to pool units involves a tradeoff between efficiency gains from pooling and bias due to heterogeneity. This paper proposes a pooling averaging estimator to makes an optimal tradeoff.
Journal of Applied Econometrics (2019), 34:724-745 [with Richard Paap and Xinyu Zhang]
Optimal model averaging estimation for partially linear models [paper] [supp]
This paper studies optimal model averaging for partially linear models with heteroscedasticity. Our techniques account for both covariate uncertainty (which covariates should be in the model) and structure uncertainty (whether a covariate should be in the linear or nonparametric component)
Statistica Sinica (2019), 29:693-718 [with Xinyu Zhang]
Sovereign credit risk, macroeconomic dynamics, and financial contagion: Evidence from Japan [paper]
We examine Japan’s sovereign credit risk, considering potential contagion from the global market and allowing for reverse causality between the risk and macro fundamentals. We find strong evidence of contagion and significant role of several credit events, such as 2011 Tohoku earthquake and rating cuts.
Macroeconomic Dynamics (2017), 21:2096-2120 [with Kan Ji and Zongxin Qian]
The forecast combination puzzle: A simple theoretical explanation [paper]
This paper concerns the forecast combination puzzle, namely why simple average often outperforms optimal weights in practice. We offer a theoretical explanation to this puzzle.
International Journal of Forecasting (2016), 32:754-762. [with Gerda Claeskens, Jan Magnus and Andrey Vasnev]
(list of most cited articles at IJF since 2016)
Weighted average least square prediction [paper]
This paper offers a variance estimator of a prediction, accounting for both model and error uncertainty. We argue that a small prediction variance is not always desirable if it ignores model uncertainty and thus is overoptimistic. Instead, an 'honest' variance estimator reflecting all relevant sources of uncertainty is more appropriate.
Econometric Reviews (2016), 35:1040-1074 [with Jan Magnus and Xinyu Zhang]
Concept-based Bayesian model averaging and growth empirics [paper] [supp] [code]
Many empirical analysis face the uncertainty which regressors to include, and how these regressors are measured, e.g., which measurement of education one should use when examining determinants of economic growth. This paper proposes a hierarchical model averaging method to address these uncertainties.
Oxford Bulletin of Economics & Statistics (2014), 76:874-897 [with Jan Magnus]
Natural resources, institutional quality, and economic growth in China [paper]
Using both a stock and a flow measure of resource abundance, we show that the effect of resource abundance on economic growth of Chinese provinces depends nonlinearly on institutional quality.
Environmental & Resource Economics (2014), 57:323-343 [with Kan Ji and Jan Magnus]
Working papers
Asymptotic properties of the synthetic control method (with Xiaomeng Zhang and Xinyu Zhang) [paper]
Time-varying group unobserved heterogeneity in finance (with Elvira Sojli and Wing Wah Tham) [paper]
(earlier version circulated as "Market-wide events and time fixed effects")
Old working papers
Dress-up contest: A dark side of fiscal decentralisation (with Ruixin Wang), 2014
Test for stock market contagion: A quantile regression approach (with S.Y. Park and N. Huang), 2010
My google scholar citations
Optimal model averaging for single-index models with divergent dimensions [paper] [supp]
This paper studies optimal model averaging for single-index models (SIMs), and provides the first theoretical results of averaging regularized estimators (such as Lasso) to deal with high dimensionality, which can be generalized to other models than SIMs.
Statistica Sinica, forthcoming [with Xinyu Zhang, Jiahui Zou and Guohua Zou]
Estimation of panel group structure models with structural breaks in group memberships and coefficients [paper] [supp] [code]
This paper considers a time-varying group pattern of heterogeneity in panel data models, and propose an estimation procedure to detect structural changes in the unknown group memberships and/or slope coefficients.
Journal of Econometrics (2023), 233: 45-65 [with Robin Lumsdaine and Ryo Okui]
Multi-dimensional latent group structures with heterogeneous distributions [paper] [supp]
Groups of units may differ not (only) in the conditional mean effects but (also) the quantile effects. We show that when the group structure is constant across different parts of distribution, using a composite quantile approach can improve clustering accuracy. We also allow cross-section fixed effects and slopes to exhibit distinct heterogeneity pattern, thus considering a multi-dimensional clustering.
Journal of Econometrics (2023), 233: 1-21 [with Heng Chen and Xuan Leng]
Too similar to combine? On negative weights in forecast combination [paper]
In forecast combination, when are the estimated optimal weights likely to be negative? How do we interpret combined forecasts with negative weights? Is it beneficial to restrict the weights to be positive? If yes, what is the best way to impose such constraints. This paper provides a comprehensive study on negative weights in combination by addressing these questions.
International Journal of Forecasting (2023), 39:18-38 [with Peter Radchenko and Andrey Vasnev]
Optimal model averaging for divergent-dimensional Poisson regressions [paper]
This paper proposes a Kullback-Leibler-distance-based model averaging method to address model uncertainty in Poisson regressions, allowing the dimension of covariates to increase with the sample size.
Econometric Reviews (2022), 41:775-805 [with Xinyu Zhang, Jiahui Zou and Guohua Zou]
Deleting unreported innovation [paper] [code]
This paper investigates the mechanism of firm unreported R&D. We document unreported innovation is systematically correlated with several firm, industry, and country characteristics, and show that unreported R&D are to some extent predictable. We assess various methods for dealing with unreported innovation.
Journal of Financial and Quantitative Analysis (2022), 57:2324-2354 [with Ping-Sheng Koh, David Reeb, Elvira Sojli, Wing Wah Tham]
(best paper award at Northern Finance Association Conference 2018)
Heterogeneous structural breaks in panel data models [paper] [supp] [code]
This paper considers heterogenous structural breaks in panel data models, allowing the number of breaks, timing, and size of these breaks to differ across groups of units. We employ a hybrid of K-means and adaptive group fused Lasso to jointly identify the latent group structure and detect the breaks.
Journal of Econometrics (2021), 220:447-473 [with Ryo Okui]
Panel threshold regressions with latent group structures [paper] [supp] [code]
This paper studies estimation of panel threshold regressions where slope coefficients and threshold parameters may exhibit latent group structures. We also consider inference issues, i.e., testing for the existing of (group-specific) threshold effects and for the homogeneity of threshold parameters across groups.
Journal of Econometrics (2020), 214:451-481 [with Ke Miao and Liangjun Su]
To pool or not to pool: What is a good strategy for parameter estimation and forecasting in panel regressions? [paper] [supp] [code]
In panel estimation and forecasting, the decision of whether to pool units involves a tradeoff between efficiency gains from pooling and bias due to heterogeneity. This paper proposes a pooling averaging estimator to makes an optimal tradeoff.
Journal of Applied Econometrics (2019), 34:724-745 [with Richard Paap and Xinyu Zhang]
Optimal model averaging estimation for partially linear models [paper] [supp]
This paper studies optimal model averaging for partially linear models with heteroscedasticity. Our techniques account for both covariate uncertainty (which covariates should be in the model) and structure uncertainty (whether a covariate should be in the linear or nonparametric component)
Statistica Sinica (2019), 29:693-718 [with Xinyu Zhang]
Sovereign credit risk, macroeconomic dynamics, and financial contagion: Evidence from Japan [paper]
We examine Japan’s sovereign credit risk, considering potential contagion from the global market and allowing for reverse causality between the risk and macro fundamentals. We find strong evidence of contagion and significant role of several credit events, such as 2011 Tohoku earthquake and rating cuts.
Macroeconomic Dynamics (2017), 21:2096-2120 [with Kan Ji and Zongxin Qian]
The forecast combination puzzle: A simple theoretical explanation [paper]
This paper concerns the forecast combination puzzle, namely why simple average often outperforms optimal weights in practice. We offer a theoretical explanation to this puzzle.
International Journal of Forecasting (2016), 32:754-762. [with Gerda Claeskens, Jan Magnus and Andrey Vasnev]
(list of most cited articles at IJF since 2016)
Weighted average least square prediction [paper]
This paper offers a variance estimator of a prediction, accounting for both model and error uncertainty. We argue that a small prediction variance is not always desirable if it ignores model uncertainty and thus is overoptimistic. Instead, an 'honest' variance estimator reflecting all relevant sources of uncertainty is more appropriate.
Econometric Reviews (2016), 35:1040-1074 [with Jan Magnus and Xinyu Zhang]
Concept-based Bayesian model averaging and growth empirics [paper] [supp] [code]
Many empirical analysis face the uncertainty which regressors to include, and how these regressors are measured, e.g., which measurement of education one should use when examining determinants of economic growth. This paper proposes a hierarchical model averaging method to address these uncertainties.
Oxford Bulletin of Economics & Statistics (2014), 76:874-897 [with Jan Magnus]
Natural resources, institutional quality, and economic growth in China [paper]
Using both a stock and a flow measure of resource abundance, we show that the effect of resource abundance on economic growth of Chinese provinces depends nonlinearly on institutional quality.
Environmental & Resource Economics (2014), 57:323-343 [with Kan Ji and Jan Magnus]
Working papers
Asymptotic properties of the synthetic control method (with Xiaomeng Zhang and Xinyu Zhang) [paper]
Time-varying group unobserved heterogeneity in finance (with Elvira Sojli and Wing Wah Tham) [paper]
(earlier version circulated as "Market-wide events and time fixed effects")
Old working papers
Dress-up contest: A dark side of fiscal decentralisation (with Ruixin Wang), 2014
Test for stock market contagion: A quantile regression approach (with S.Y. Park and N. Huang), 2010
My google scholar citations