Forecasting volatility in gold returns under the GARCH, IGARCH and FIGARCH frameworks: New evidence

• This paper examines the long memory property of volatility in gold returns.
• We use GARCH(1,1), IGARCH(1,1) and FIGARCH(1,dd,1) specifications.
• We use daily data from August 2, 1976 to February 6, 2015.
• FIGARCH(1,dd,1) appropriately describes the gold volatility process.
• FIGARCH(1,dd,1) is the best model to forecast volatility in gold returns.

This study employs three volatility models of the GARCH family to examine the volatility behavior of gold returns. Much of the literature on this topic suggests that gold plays a fundamental role as a hedge and safe haven against adverse market conditions, which is particularly relevant in periods of high volatility. This makes understanding gold volatility important for a number of theoretical and empirical applications, namely investment valuation, portfolio selection, risk management, monetary policy-making, futures and option pricing, hedging strategies and value-at-risk (VaR) policies (e.g. Baur and Lucey (2010)).We use daily data from August 2, 1976 to February 6, 2015 and divide the full sample into two periods: the in-sample period (August 2, 1976–October 24, 2008) is used to estimate model coefficients, while the out-of-sample period (October 27, 2008–February 6, 2015) is for forecasting purposes. Specifically, we employ the GARCH(1,1), IGARCH(1,1) and FIGARCH(1,dd,1) specifications.The results show that the FIGARCH(1,dd,1) is the best model to capture linear dependence in the conditional variance of the gold returns as given by the information criteria. It is also found to be the best model to forecast the volatility of gold returns.