Evaluating Switching GARCH Volatility Forecasts During the Recent Financial Crisis

Forecasting volatility is a fundamental topic in in both academic and applied financial economics. Different GARCH-specifications are by far the most popular model based approach used for this purpose. This thesis evaluates the forecast accuracy of some specific GARCH-models; GARCH, EGARCH, APGARCH and MRS-GARCH. The primary purpose of the essay is to investigate whether the more flexible two-regime MRS-GARCH model outperforms the more conventional one-regime GARCH models in a very volatile time period during the recent financial crises. The evaluation period stretches from the day when Lehman Brothers went bankrupt and one year ahead. Each model is evaluated using two indexes with different characteristics; the Standard & Poor 500 and the Bombay Sensex index. The result shows that the MRS-GARCH models are superior in predictive ability on S&P500 compared to the other tested models. Conversely, the overall relative performance accuracy of the BSE is less clear-cut since none of the tested models seem to perform particularly well. Generally, the results indicate that the MRS-GARCH provides better forecasts on S&P 500 compared to the other models and that no forecast can be distinguished as entirely superior on the BSE