Title: Conditional heteroscedasticity models with time-varying parameters: Estimation and forecasting
Authors: Armin Pourkhanali - Monash university (Australia) [presenting]
jonathan Keith - Monash university (Australia)
Xibin Zhang - Monash University (Australia)
Abstract: The aim is to study the asymptotics and empirical relevance of time-varying generalized autoregressive conditional heteroscedasticity (GARCH) models, where time-varying parameters are approximated by different polynomials of the time variable with unknown orders. In comparison with some existing varying-coefficient GARCH models, our model provides a more flexible mechanism to capture time-varying dynamics of parameters using Chebyshev polynomials. We also investigate the asymptotic properties of these polynomials under mild conditions. Our approach is computationally feasible. The proposed estimation method is justified through Monte Carlo simulation studies and empirical studies. The proposed model is applied to modelling daily returns of the US Treasury bond with a sample period of 30 years, as well as modelling daily returns of the gold futures price. In addition, we compare the out-of-sample forecasting performance of the proposed model with a constant-coefficient GARCH model. The empirical findings support our time-varying GARCH models against its constant-parameter counterpart.