Forecasting Market Volatility: A Comparative Study of ARIMA and GARCH Models on the VIX and S&P 500

This undergraduate research paper investigates the comparative forecasting performance of two time-series models, ARIMA (AutoRegressive Integrated Moving Average) and GARCH (Generalized Autoregressive Conditional Heteroskedasticity),in modeling market volatility. The ARIMA model is applied to the CBOE Volatility Index (VIX) to capture implied volatility dynamics, while the GARCH model is used on S&P 500 index returns to estimate realized volatility. Despite targeting different but related volatility metrics, the study evaluates the models’ outputs using RMSE and correlation to assess predictive alignment across different market conditions. The results suggest that ARIMA, when applied to structured indicators like the VIX, may offer superior forecast accuracy relative to GARCH in certain volatility regimes.