Seminar Vol. 159
Title: A New Approach to Test Predictability in Quantile Regressions with Persistent Predictors
Speaker: Haiqiang Chen, Xiamen University
Time: May 22th, 2019 13:30–15:00
Venue: Conference Room 106B, Zhonghui Building (IESR, JNU College of Economics)
About the speaker:
Haiqiang Chen is a Professor at The Wang Yanan Institute for Studies in Economics, Xiamen University. He is also Associate Director of the Key Laboratory of Econometrics, and Director of Data Science and Decision Consulting Center at Xiamen University. Haiqiang Chen's main research interests are Financial Econometrics, Time Series Econometrics, and Financial Economics. He has published widely in such international journals as Econometric Theory, Journal of International Money and Finance, Journal of Empirical Finance, and many more. Haiqiang Chen received his Ph.D. in Economics from Cornell University in 2011.
Abstract:
For predictive quantile regressions with highly persistent regressors, the traditional test statistics based on least squared estimators lose their validity and their limiting distribution relies on the unknown persistence parameters of predictors. This paper proposes a novel econometric method to offer a robust inference theory across all types of persistent regressors. Particularly, we construct a weighted estimator based on a quantile regression with an auxiliary regressor, which is generated as a combination of an exogenous simulated nonstationary process and a bounded transformation of the original regressor. Under some mild conditions, we show that the self-normalized test statistics based on the weighted estimator converge to a standard normal or Chisq distribution. Comparing to the existing approach, our method could reach the local power under the optimal rate T with non-stationary predictors and √T with stationary predictors respectively. Moreover, the method can be easily generalized to multiple regressors with mixed persistence degree. Simulations and empirical studies are provided to demonstrate the effectiveness of the newly proposed approach. The heterogeneous predictability of US stock returns at different quantile levels is reexamined.
