Lognormal trickery tends to smooth out price charts. That’s one of the irreverent thoughts that came to mind as I read a recent study of one of the perennial arguments of gold bugs has stood up to empirical testing.
Three German scholars, Christian Pierdzioch, Marian Rose, and Sebastian Rohloff, argue on the basis of “Bayesian additive regression trees,” a machine-learning algorithm, that “gold and silver investments are a strong hedge against exchange rate movements of the dollar vis-à-vis the yen, the euro, the pound, the Canadian dollar, and the Australian dollar.” This lends direct support to one such perennial argument.
The same scholars look also at the ability of such investments to act as a hedge against U.S. stock market downturns, also a conventional claim of gold bugs. Here, too, the gold bugs turn out to be right, although the findings are different for silver. “Silver returns, in contrast [to gold], show no systematic response to stock-market movements, indicating that silver investments are not a strong hedge against stock-market movements.”
Another natural use of these scholars’ algorithm, once they had given it its test runs, turned out to be a comparison of gold and silver on the one hand to palladium and platinum on the other. It turns out that the latter two metals have exchange-hedging value only with respect to the movements of the Canadian and Australian dollars. Those two nations’ dollars are classic commodity currencies, that is, currencies heavily dependent on the export of raw materials by the issuing country.
Palladium and platinum are worse than useless as hedges against stock market movements, that is, they are pro-cyclical, “possibly reflecting fluctuations in industrial demand at business-cycle frequencies,” the authors suggest.
Actually, more than half of the world’s supply of palladium, and nearly one-third of its newly mined platinum gets directed toward the production of catalytic converters. That the fortunes of the producers of raw materials for automobiles would prove pro-cyclical seems quite intuitive.
The reaction of all four of these metals to the commodity market movements embodied in the GSCI is “procyclical but nonlinear.”
Log Normal and Positive Numbers
The study involved a sample period running from January 1999 to December 2014. Gold prices rose steadily through most of this period, with only a very mild downturn at the time of the global financial crisis, though they have been trending downward, slowly, since late 2011.
Well … note that I speak here easily of “mild” downturns and a “slow” movement. In part at least this is a testament to the authors’ use of log normalization, a statistical technique that smooths out the charts. More technically speaking, using logarithms converts a multiplicative scatter into an additive scatter, and makes formerly lop-sided charts look symmetrical. This is said to be appropriate for use on a data base with only positive numbers. The price of gold, obviously, isn’t going to become negative. Also it is hard to imagine a world in which gold has a value of 0 – which leaves us with only the positive part of the number line. Now I can always say that at least once I explained that point.
But back to the findings. Here are three more to take us home:
- “Gold, thus, seems to be to a larger extent [than the three other metals] a financial asset that is somewhat decoupled from price developments in commodity markets.”
- The hedging value of gold and silver in the currency-exchange world is asymmetric. They are a better hedge in times of U.S. dollar devaluation than in times of dollar appreciation
- There are no clear-cut patters at all linking the prices of these metals to the corporate bond spread, the term spread, or the VIX.