Title: Quantile regression models with factor-augmented predictors and time-varying factor loadings
Authors: Yiguo Sun - University of Guelph (Canada)
Alev Atak - City University London (United Kingdom) [presenting]
Yonghui Zhang - Renmin University of China (China)
Abstract: A semiparametric quantile regression model with factor-augmented predictors and time-varying factor loadings is developed, where the time-varying factor loadings is allowed to change across quantile regressions at different probability masses while taking the latent factors fixed. We propose a two-stage procedure. In the first step, we simultaneously estimate the latent factors and time-varying factor loadings, defined by a nonparametric smooth function, using a local version of the principal component method. In the second step, we develop our quantile regression model with factor-augmented predictors that is derived in the first step. The proposed method extracts and combines distributional information across different semiparametric quantile regression models. Results of Monte Carlo simulations demonstrate that the proposed criterion performs well in a wide range of situations. We also apply our model to investigate a U.S. macroeconomic data set and find strong evidence of heterogeneity in dynamic responses.