Observers have long noticed that the relationship between return shocks and equity price volatility is asymmetric. That is, positive shocks do not have a marked effect on volatility; negative shocks do. Fischer Black wrote on the subject more than 40 years ago.
Recently there has been a good deal of study on the relationship between shocks and volatility in the commodities markets. Evidence indicates that for at least some commodities there is an “inverted asymmetric effect,” that is, positive shocks have a marked volatility effects, negative shocks less so.
One of the hypotheses invented to explain the inverted asymmetric effect involves inventory. A sharp move downward in a commodity’s value (a negative shock) should lead to high inventory levels, which should in turn mean low prices and low volatility. But a sharp move upward may entail an exhaustion of inventories, which would mean high prices and—given the risk of unavailability—greater volatility. That is the view put forward for example by Chiarella, Kang, Nikitopoulos, and Tˆo (2016) in their study of the effects of shocks on the volatility of gold and oil futures.
More recently, Dirk G. Baur and Thomas Dimplf have moved this discussion forward with a paper that contends that the issue of inventory is less important than some of the earlier theorists believe.
Baur and Dimplf use a larger array of commodities and commodity indexes than earlier studies. Importantly, they look both at commodities that are consumed (such as oil or foodstuffs) and at commodities that are not consumed (such as gold). Surely an inventory effect would be more marked for the consumables than for the non-consumable commodities.
Baur and Dimplf also use four distinct measures to capture the direction and degree of asymmetry: simple correlation; a TGARCH model; a simple regression model; and an approach based on quantile autoregression.
GARCH, by way of reminder, is a Generalized Autoregression Conditional Heteroskedasticity. T (or Threshold) GARCH, on the other hand, refers to a rejiggering of GARCH proposed by Lawrence R. Glosten in 1993 precisely to distinguish negative and positive effects of the same cause.
The fourth measurement on the above list, quantile autoregression, is something developed by Baur and Dimplf in their earlier work. It “is based on the idea that extreme lower and upper quantile autocorrelation estimates are implicit range-based volatility estimates and determine the sign and magnitude of the negative asymmetry effect.”
In the new article, Baur and Dimplf break down their results into pre-crisis and post-crisis baskets: that is, data prior to 2007, and data subsequent to 2009.
Financialization of Commodities
Looking at the simplest metric correlation: the correlation between shocks and volatility for commodities as a group was +0.0124 before 2007. It has been -0.0281 since 2009. This suggests that there was an inverted asymmetric effect prior to the crisis, but there has been a more equity-like asymmetric effect since 2009. Breaking commodities down by type, agricultural, energy, and industrial commodities showed a positive correlation in the earlier period, a negative correlation later. The precious metals alone showed a negative (equity-like) correlation, a non-inverted asymmetry, in both the earlier and the latter periods.
I’ll skip forward here to quantile autoregression. This shows less variation between the full sample and the post-2009 estimates. Both sets point toward equity-like asymmetry, though the asymmetry becomes more marked since 2009 than it was before. The QAR only shows an inverted asymmetric effect for precious metals prior to 2007.
Even for the equity indices themselves, the classic asymmetric effect becomes more pronounced for the post-2009 data, especially by the TGARCH metric.
These authors conclude, then, that although there used to be an inverted asymmetry in the relationship between shocks and volatility in commodities, that is no longer the case. Why not? Because, Baur and Dimplf propose, “the financialization of commodity markets has changed this effect.” Commodity prices now serve as signals for the global economy and the international turnings of the business cycle. Price increases (even shocks) serve as a signal of a growing world economy and price decreases represent a signal of productivity losses in that world. The former correlates intuitively with low volatility and the latter with high volatility. That signaling overwhelms other considerations involving such matters as what is consumed and how large are the inventories.