Investing for value over the long run is not new. Many investors know about various long-term stock market models, such as the Cyclically Adjusted Price Earnings (CAPE) of Schiller or the Q-ratio of Tobin.
Last year the Vanguard Group did a study on the performance of a variety of models (see https://personal.vanguard.com/pdf/s338.pdf). The conclusion was that valuation-based models, such as CAPE, did the best. In fact, the models explained about 45% of the variation in stock market index prices. That’s a substantial investment edge. The catch is that the explanations only work at long-term time frames. The best fit was at a 10-year time horizon! Obviously, these are not trading models.
But could these models do the same thing for commodities? If so, they could help us adjust the percentage of commodities in our long-term investment portfolios. Indeed, there is a class of models that works well for this purpose. The catch is it is only valid for extended time frames.
As an illustration of this approach, let’s start with gold. The analytical theory behind the model is simple. First, consider the main purpose of gold as an investment vehicle: It is a store of value. Gold may not pay a dividend, but it also is not susceptible to government-driven inflation. So over the long run, the price of gold in dollars should rise to offset fiat currency inflation. Another way to think of this is to view gold as an alternate currency. Over the long run, its value should mean revert to purchasing power parity (PPP).
Over the shorter run, gold will go up and down depending on investor sentiment. In fact, it can do so quite a bit; just look at the gold market’s variation over the last 10 years. The key for the long-term commodity trader is to invest against these shorter-run moves and keep your eye on the long-term trend. This might lead us to suspect that when gold is trading below its PPP, it should be a long run buy — and, of course, vice versa.
The model itself is based on this theoretical base. The heart of it is in the scatter graph shown in “Gold vs. inflation” (below). For example, in January 2002 gold was trading at $281 per oz. That’s equivalent to $368 in 2013 dollars. In January 2012 gold was at $1,656; that’s equivalent to $1,688 in 2013 dollars. So the real return over the 10 years after 2002 was $1,688/$368 or 4.6 times. The series ends in July 2003 because that’s the last month for which a future 10-year return exists. Note that the axes are plotted on a ratio (logarithmic) scale. The fit on the model is quite good. The R-squared, a measure of how well our model explains variations in the dependent variable, is 82%. All the other standardized test statistics are fine.
Right now, with gold at about $1,330 (the red line on the graph), the model is forecasting a future 10-year annualized real return of 0.46%. So if you assume that inflation will continue to run at about 2% per year, gold in 2023 will only be at $750. Of course, the rationale for buying gold is that inflation will be much higher. The point that should be driven home by this model is that it has to be much higher to justify holding gold at current prices.