Forecasting volatility in Asian financial markets: evidence from recursive and rolling window methods

Area of Science:

• Financial Econometrics
• Time Series Analysis
• Stock Market Volatility Modeling

Background:

• Asian stock markets exhibit volatility clustering, leverage effects, and persistence.
• Accurate volatility forecasting is crucial for risk management and investment decisions.
• Existing GARCH-type models need evaluation for their predictive accuracy in diverse market conditions.

Purpose of the Study:

• To assess the out-of-sample predictive performance of various GARCH models.
• To compare symmetric (GARCH) versus asymmetric (GARCH-M, EGARCH, TGARCH, PGARCH) models.
• To investigate the impact of different time frames, forecasting methods (recursive vs. rolling), and error measures on model selection.

Main Methods:

• Empirical analysis of ten Asian stock markets.
• Application and comparison of GARCH, GARCH-M, EGARCH, TGARCH, and PGARCH models.
• Utilized three time frames, two window methods (recursive and rolling), and five comparison measures, validated by the Diebold-Mariano test.

Main Results:

• Asymmetric models, particularly EGARCH, outperform symmetric models in higher time frames.
• Symmetric GARCH models are superior in lower time frames.
• The recursive method favors linear GARCH models, while the rolling method penalizes them; model performance is sensitive to the chosen error statistic.

Conclusions:

• GARCH-type models demonstrate adaptability to Asian stock index volatility.
• Model selection is critical and depends on the time frame, forecasting method, and evaluation metric.
• EGARCH and symmetric GARCH models offer valuable forecasting capabilities under specific conditions.