The Quantile-Heterogeneous Autoregressive Model of Realized Volatility : New Evidence from Commodity Markets

We provide a comprehensive view on volatility dynamics in precious metals and crude oil markets. Using high-frequency futures data, we construct realized volatilities and estimate (Quantile) Heterogeneous Autoregressive models for the daily volatility of Gold, Silver and Crude Oil futures. We model realized volatility as a linear function of lagged realized volatility measured over different time resolutions to explicitly account for the potentially heterogeneous impact of market participants with different trading motives and investment horizons. Using quantile regression allows us to identify potential non-linearities and asymmetries in the short-, mid- and long-term autoregressive dynamics with respect to different levels of current volatility. We document considerable changes in the relative importance of short-, mid-, and long-term volatility components under varying market conditions. The patterns that we identify are remarkably similar across the three assets. Specifically, past daily and monthly volatility have a strong impact on today's volatility, when current volatility is low (lower quantiles of the volatility distribution). The effect of past weekly volatility, however, increases distinctly from lower to higher quantiles of the conditional volatility distribution. The results might indicate considerable investor attention shifts and changes in the proportions of traders with different time horizons