Enhancing stock volatility prediction with the AO-GARCH-MIDAS model

Area of Science:

• * Econometrics
• * Financial Modeling
• * Time Series Analysis

Background:

• * Outliers in financial data introduce errors and bias, degrading volatility forecast precision.
• * Existing models may not adequately address the impact of outliers on volatility estimation.
• * Accurate volatility forecasting is crucial for risk management and investment strategies.

Purpose of the Study:

• * To introduce a novel outlier correction method for GARCH-MIDAS models.
• * To develop an Additional Outliers corrected GARCH-MIDAS (AO-GARCH-MIDAS) model.
• * To evaluate the performance and robustness of the proposed AO-GARCH-MIDAS model.

Main Methods:

• * Correcting additive outliers using a weighting method, replacing original outliers with corrected values.
• * Re-estimating model parameters using the modified return series.
• * Applying the new approach to Generalized Autoregressive Conditional Heteroskedasticity Mixed Data Sampling (GARCH-MIDAS) models.

Main Results:

• * The AO-GARCH-MIDAS model demonstrated consistent superiority across all evaluation criteria compared to standard models.
• * Outlier correction significantly reduces distortion in volatility estimates, enhancing model adaptability.
• * GARCH-MIDAS models show superior predictive power over GARCH models, with realized volatility being a key predictor.

Conclusions:

• * The proposed AO-GARCH-MIDAS model effectively mitigates outlier effects, leading to improved volatility forecast accuracy.
• * Outlier correction is essential for robust financial modeling and reliable volatility predictions.
• * Realized volatility provides greater predictive information than other low-frequency factors for GARCH-MIDAS models.