Abstract
This article provides the in-sample estimation and evaluates the out-of-sample conditional mean and volatility forecast performance of the conventional Generalized Autoregressive Conditional Heteroscedasticity (GARCH), Asymmetric Power Autoregressive Conditional Heteroscedasticity (APARCH) and the benchmark RiskMetrics model on the US real estate finance data for the pre-crisis and post-crisis periods in 2008. The empirical results show that the RiskMetrics model performed satisfactorily in the in-sample estimation but poorly in the out-of-sample forecast. For the post-crisis out-of-sample forecasts, all models naturally performed poorly in conditional mean and volatility forecast.
Acknowledgements
The author would like to thank the Editor, Mark Taylor, and two anonymous referees for their invaluable comments on the earlier draft of the article. Research funding from the City University of Hong Kong (Strategic Research Grant numbers 7002523 and 7008129) is gratefully acknowledged. The two research assistants, Siyang Ye and Douglas K. T. Wong, have provided useful research inputs. The author alone is responsible for the errors found in the article.