Mean-Reversion in Commodity Futures Volatility: An Analysis of Daily Range-Based Stochastic Volatility Models

Abstract

We analyse the dynamic behavior of conditional volatility in commodity markets using a novel, manually collected dataset of daily price ranges over a time span of more than 140 years, which allows more precise daily volatility estimates than are otherwise prevalent in the commodity literature. We find that a one-factor range-based EGARCH-model (REGARCH) is not adequate to capture the very distinct long-run and short-run dynamic volatility components. While the long memory effect of volatility is numerically very small, it strongly affects the parameters of the short-run dynamics which become more stable and plausible in size. Moreover, long-run persistency in volatility shocks is practically unaffected after controlling for regimes which indicates that the stochastic movement of the long-run mean is not a statistical artefact. We also find that consistent with the theory of storage, long run volatility is positively related to lagged returns. Thus, asymmetry in volatility is not a short-run phenomenon.