Abstract
This paper explores the predictive role of speculative sentiment on gold market volatility and its economic implications. Utilizing high frequency data for gold futures and speculative sentiment proxies for the gold and stock markets, we show that incorporating speculative sentiment in volatility models can improve volatility forecasts both in- and out-of-sample. While gold market sentiment helps improve the predictive accuracy of volatility models primarily in the short run, we show that cross-market speculative sentiment from the stock market has predictive ability at relatively longer forecast horizons. More importantly, incorporating speculative sentiment proxies in volatility models yields improved cumulative returns as well as Sharpe ratios for investors who allocate part of their portfolios in this traditional safe haven asset. Finally, examining the predictability of asymmetric risk patterns, we show that speculative sentiment in the gold market matters relatively more for good volatility forecasts and in the short and intermediate forecast horizons, while cross-market sentiment effect from the stock market generally applies to longer term forecasts for both the good and bad volatility metrics. Overall, our findings show that speculative sentiment contains significant incremental information over subsequent volatility patterns in gold, offering sizeable economic gains for a mean-variance investor in a practical investment setting.
Acknowledgments
We would like to thank the Editor and an anonymous referee for helpful comments. However, any remaining errors are solely ours.
Notes
1 The theoretical arguments that link short-term speculation to price deviations from fundamental values include the rational bubble model where bubbles emerge as the rational expectation equilibrium (e.g. Blanchard and Watson Citation1982) and the asymmetric information model where speculation induces prices to deviate from fundamental values due to noisy beliefs (e.g. Froot, Scharfstein, and Stein Citation1992).
2 Due to space considerations, we only report the estimated coefficients and not the standard errors (full results are available upon request).