Abstract
This paper investigates the predictive power of different volatility index (VIX) terms for future realized volatility (RV). Using the heterogeneous autoregressive model of realized volatility (HAR-RV), we analyse the relationship between various terms of VIX and their corresponding future RVs across short-term and long-term horizons, incorporating both in-sample and out-of-sample evaluations. While the 1-day VIX (VIX1D) is the best predictor of future 1-day RV, we find that future RVs do not always align with their respective terms of VIX. Furthermore, the best predictors for 1-week, 9-day, and 2-week future RVs closely align with their corresponding maturities, whereas predictors for longer maturities show weaker and less consistent alignment. Even though all terms of VIX exhibit some bias in forecasting, including VIX in HAR-RV model, regardless of the term, still improves predictive performance for future RV.