The study employs different types ofrelatively novel Markov regime switching GARCH (MRS-GARCH) models along with standard GARCH to identify better models to forecast the conditional volatility of Dow Jones Islamic Market World Index (DJIM). Several statistical and risk-management-based loss functions are employed to evaluate the out-of-sample volatility and Value-at-Risk (VaR) forecast from these models. Our empirical results show that although there is no single model - either of standard GARCH and MRS-GARCH type - is consistently outperforming the others if both the statistical and the risk-management loss functions are considered, overall different asymmetric Markov regime switching GARCH models, namely MRS-EGARCH andMRS-GJR-type models, perform better than any single regime GARCH specification.