Forecasting Realised Volatility: Implied and GARCH Volatility in Bitcoin, Gold, Oil Markets

This paper investigates the predictive accuracy of implied and GARCH volatility models for bitcoin, gold and oil to determine whether (i) implied volatility is a reliable proxy for investors and (ii) bitcoin behaves differently from other commodities in terms of its volatility behaviour. Data analysis from January 2019 to December 2023 reveals that implied volatility underperforms GARCH (1,1) in all assets in terms of predictive accuracy. The difference between the estimated errors given by the implied volatility and by GARCH (1, 1) is the lowest in gold, showing that implied volatility predictions are more suitable in less volatile markets. The results are consistent with the yearly split sample data—aside from the additional fact that in times of turmoil, the predictive accuracy of both volatility models deteriorates significantly. The results further suggest that after a financial crisis, the predictive accuracy of the implied volatility falls while that of GARCH (1, 1) rises, particularly in the bitcoin market. Further analysis confirms that implied volatility (followed by GARCH volatility) is generally higher than realised volatility in all the three assets, except when the market is extremely volatile, which is consistent with previous findings for other assets. Taken together, these results can support investors in two key aspects. The first is that despite being widely used by investors, implied volatility is not the most accurate proxy available particularly when applied in volatile market conditions. The second is that implied volatility exhibits a further diminished level of predictive accuracy post financial crisis, specifically in bitcoin market.