Essays on time varying parameter models

Abstract

Time-varying Parameter (TVP) models have become increasingly popular among macroeconomic researchers, as they are able to capture structural changes in the relationships between economic variables. This dissertation explores three topics related to time-varying regressions. Chapter 1 focuses on time varying parameter version of Panel Vector Autoregressive (PVAR), which allow considering endogenous the variable in the system, panel-data aspects and admitting individual heterogeneity. We propose Bayesian method to estimate PVAR model featuring time variation in parameters and stochastic volatility. The TVP-PVAR is then applied to model global economic uncertainty shocks from multicountries. Chapter 2 proposes efficient Bayesian method to estimate factor model with time varying factors loadings in large data set. In Chapter 3, a semiparametric version of a time varying regression with a subset of fixed coefficients is developed and its associated theoretical properties are established. Monte Carlo studies and empirical exercises are conducted to support the potential use of these models.