Taking the Temperature of Dr. Copper

Evaluating the use of the metal as a leading market indicator

Corrected 2021-08-30

An earlier copy of this article used data that was incorrect. The trailing returns were referred to rather than the forward returns, which showed a weak but statistically significant relationship. This new result shows no relationship.

Copper is affectionately called “Dr. Copper” by macro traders. It’s a raw material used in the manufacture of many products. If there is high demand for this input material, then the idea is that this demand is likely pulled by a high demand for the products it makes.

In this article I want to explore and quantify that relationship.

Bayesian Estimation Results

I compared the trailing return of nearby roll-adjusted copper futures with the forward return of the S&P 500 index and the Russell 2000 index. Using equal periods for trailing and forward return, the largest effect I found was for a 14 trading day period. The Russell 2000 was about twice as sensitive to copper than the S&P 500 and Dow indices.

Leave a comment

The difference of means image is the result of categorizing index returns as belonging to high or low copper return regimes and separately estimating their distributions.

The posterior distribution indicates that the Russell 2000 returns in a high copper return regime follow a distribution that is on average, the same as the returns in a low copper return regime. In other words: there’s no difference, on average.

That basically invalidates the narrative, but we can do better. Let’s dig deeper.

This scatter shows the same lack of relationship. If you traded on this narrative alone, you’d get yourself in trouble. As of this writing the trailing 14 day return in front month copper is about -8.8%; the regression above implies that the return in the Russell over the next 14 trading days should be between -14.8% and +17.2% with a dubious-looking 97% credibility. That credibility is based only on observed values: it’s probably wrong1.

Check the outliers around the current implied forward range; they are widow-makers. Trading based on this kind of regression is extremely dangerous.


On average, equity markets do not seem to follow copper. Also on average, the chamber of a revolver used in Russian roulette is empty. COMEX copper futures are physically settled contracts for future delivery of the metal. The price is affected by supply, demand, and financing factors. It is not a one-dimensional barometer for demand. Low supply would increase prices, and this would not be intrinsically bullish equities.

Treating it as harbinger of good or bad fortune is foolish. There is no free lunch; you’ll have to do your homework and dig into the supply/demand relationship to sharpen it into something useful.

Leave a comment


It doesn’t quite look it, but that credibility interval is set such that it will contain 97% of the observed data; I’m not just doing ±2σ.