Evaluating monthly volatility forecasts using proxies at different frequencies

- This paper analyzes the forecastability of the EuroStoxx 50 index monthly returns.
- Punctual and aggregated monthly forecasts obtained from GARCH models using returns at different frequencies.
- The models performance is measured with respect to alternative proxies for the unobserved volatility.
- Monthly volatility proxies are obtained using mixed data sampling methods using daily frequency.
- Robust results found across proxies support multi-step-ahead GARCH-type forecasts.

This paper analyses the forecastability of stock returns monthly volatility. The forecast obtained from GARCH and AGARCH models with Normal and Student's t errors are evaluated with respect to proxies for the unobserved volatility obtained through sampling at different frequencies. It is found that aggregation of daily multi-step ahead GARCH-type forecasts provide rather accurate predictions of monthly volatility.