Forecasting the volatility of the FTSE Bursa Malaysia KLCI Futures Index using the heterogeneous autoregressive (HAR-RV) model, vs. forecasts derived from GARCH(1,1) and exponential weighted moving average (EWM).
We will forecast the one-day ahead volatility of the FTSE Bursa Malaysia KLCI Futures Index (more info: https://www.bursamalaysia.com/trade/our_products_services/derivatives/equity_derivatives/ftse_bursa_malaysia_klci_futures) using the Heterogeneous AutoRegressive (HAR-RV) model. The backtest is performed on a rolling out-of-sample manner. You can read more about the HAR-RV model from Corsi, F. (2004) A Simple Long Memory Model of Realized Volatility available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=626064.
We will also compare the forecast with those generated by GARCH(1,1) as per Hansen, P.R. (2005) A Forecast Comparison of Volatility Model: Does Anything Beat a GARCH(1,1)? available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=264571. We also included volatility derived from exponential weighted moving average (EWM) of various time horizon for good measure.
Taking a look at the R2, the HAR-RV model appears to outperform other metrices. To be fair the prediction accuracy of the GARCH model relies heavily on the choice of time horizon. The simple EWM model also appears to have decent predictive ability, at least on a shorter time horizon.
