Enhancing Prediction by Incorporating Entropy Loss in Volatility Forecasting.

Area of Science:

• Quantitative Finance
• Econometrics
• Financial Modeling

Background:

• Accurate financial market forecasting is crucial for risk management and investment strategies.
• Heterogeneous Autoregressive (HAR) models are widely used for volatility forecasting.
• Evaluating different estimation techniques and horizons is essential for optimizing HAR model performance.

Purpose of the Study:

• To compare the forecasting accuracy of Heterogeneous Autoregressive (HAR) models using various estimation techniques and horizons.
• To identify the optimal estimation method and forecasting horizon for HAR models.
• To assess the impact of exogenous variables and enhanced model specifications on forecasting accuracy.

Main Methods:

• Examined three HAR-type models with five estimation techniques and four forecasting horizons.
• Utilized 5-minute intraday data for the Standard & Poor's 500 (SPX) index and the VIX index as exogenous variables.
• Employed Quasi-Likelihood (QLIKE), Mean Absolute Error (MAE), and Mean Squared Error (MSE) for performance evaluation.

Main Results:

• The Entropy Loss Function consistently yielded the best Quasi-Likelihood (QLIKE) results across all horizons, particularly the weekly horizon.
• The Robust Linear Model proved to be a competitive alternative, showing strong performance in Mean Absolute Error (MAE) and Mean Squared Error (MSE).
• Incorporating Realized Quarticity (HARQ model) and the VIX index significantly improved overall model forecasting accuracy.

Conclusions:

• Both the Entropy Loss Function and the Robust Linear Model demonstrate significant forecasting accuracy for HAR models.
• The choice of estimation technique and forecasting horizon critically impacts HAR model performance.
• Enhancing HAR models with informative lags and exogenous variables like VIX leads to improved predictive power.