Forecasting daily dynamic hedge ratios of indian index future contracts

Abstract

Forecasting plays a crucial role in determining the direction of future trends and in making necessary investment decisions. This research presents the forecasting performance of three multivariate GARCH models: SGARCH, EGARCH, and GJR-GARCH based on Gaussian and Student’s t-distribution. The forecasting ability of the models is evaluated on the basis of forecasting performance measures: MAE, SSE, MSE, and RMSE. This is done by examining the hedged portfolios of three indices of NSE: NIFTY50, BANKNIFTY, and NIFTYIT. Daily data from Jan 2006 to Dec 2017 is taken and forecasts are conducted using out of sample data from Jan 2016-Dec 2017. Minimum mean square error (MMSE) forecasting method is used to generate conditional variance and covariance forecasts which in turn generate hedge ratios and corresponding hedged portfolio. Minimum variance hedge ratio framework of Ederington (1979) is used for hedging. The in-sample analysis shows that SGARCH with both the distribution performed better than the other models while out-of-sample analysis provides mixed results. EGARCH model assigns the lowest hedge ratio to NIFTY50 and BANKNIFTY while SGARCH model assigns the lowest hedge ratio to NIFTYIT. Forecasting performance measures show the least value for SGARCH and EGARCH model. In future these models are able to reduce maximum risk from the spot market. The results of this research has important implications for financial decision and policy makers.