VOLATILITY SPILLOVER BETWEEN SPOT AND FUTURES MARKET OF HIGHLY TRADED COMMODITIES IN INDIA: The DCC-GARCH Approach

Abstract

This study intends to examine the volatility spillover effects and measure the time-varying correlations between futures and spot prices of thirteen highly traded commodities traded on Multi Commodity Exchange (MCX) of India. The research uses Exponential GARCH proposed by Nelson (1991) to explore the direction and magnitude of spillover effects between futures and spot commodity market and employs Dynamic Conditional Correlation (DCC) GARCH proposed by Engle (2002) to demonstrate the time varying conditional correlation between heteroscedastic coefficients of the futures and spot markets. Empirical results show that significant and asymmetric bi-directional volatility spillover effects exist in case of most of the selected commodities, even though, the magnitude of volatility spillover is found larger in the direction from futures market to spot market. The dynamic correlation between the conditional variance of the spot and future markets is found to be significant in case of all the commodities except Silver and Copper. It proves that significant volatility spillover effect is present between spot and futures markets of selected commodities. Understanding of volatility transmission and interrelationship between spot and futures commodity market will help investors make right investment decisions, portfolio optimization and financial risk management. Policy makers and regulators can use this knowledge in planning and implementing appropriate regulatory framework. Much of the earlier research focuses on inter market volatility spillover taking into consideration two or more different financial markets. This study focuses on intra market volatility spillover by studying the interactions of spot-futures prices of commodities. Also, considering the time-varying nature of conditional correlations, this study employs EGARCH and multivariate GARCH (DCC) to capture the volatility spillover effects instead of univariate GARCH or standard linear VAR models.