A1993
Title: Extracting predictability: A time-varying factor-augmented approach under mixed persistence
Authors: Tingting Cheng - Nankai University (China) [presenting]
Jiti Gao - Monash University (Australia)
Yayi Yan - Shanghai University of Finance and Economics (China)
Xuanbin Yang - Nankai University (China)
Abstract: A time-varying factor-augmented forecasting model is developed to extract predictable variation in the target variable when predictors exhibit mixed persistence. The framework accommodates stationary, nonstationary, or mixed observable and latent predictors, and allows latent factors to follow either static or dynamic factor structures, thereby nesting many classical and new examples as special cases. A sieve estimation procedure for the time-varying coefficients is developed, corresponding large-sample properties are established, and prediction intervals for the forecasts are constructed. Simulation studies show that the proposed methods perform well in finite samples. In an empirical application to U.S. inflation forecasting, the proposed model performs favorably relative to competing benchmarks. Further evidence indicates that latent factors remain informative even when inflation exhibits strong autoregressive dynamics, since they absorb predictable variation that would otherwise be left in the forecast error.
