April 20, 2017
Some U.S. stock market investors are getting worried about the price of iron ore in China. This week’s chart helps to show why.
It is definitely an intriguing chart, and a relationship I had not explored before. I have come across a large number of interesting intermarket relationships like this one, and it is always fun to find (or be shown) a new one. But not all of them have merit. So what I’d like to do is use this one to show you how I typically like to contemplate a new relationship that I encounter.
So my first question is about whether the relationship is a durable one. The answer, it turns out in this case, is no. The two have only recently fallen into this apparent correlation. Here is a longer term look:
Prior to early 2015, there did not seem to be any relationship at all. Since that point, they do seem to be correlating. But perhaps what we have is more of a leading indication relationship. Here is that same chart, with the iron ore futures price plot shifted forward:
I played around with different offset periods, and got the best looking alignment of the recent data with a 29 trading day forward offset of the iron ore price data. I do not have an explanation for that period; it is just what seems to work the best. And this adjustment still leaves the period before early 2015 showing hardly any correlation at all.
As an aside, I am often asked about whether I have investigated the Pearson’s Correlation Coefficient for a particular relationship. A lot of technical analysts like to employ that tool, because it is very easy to get a result in Excel or other programs. But statisticians know that Pearson’s is not a tool that is designed for time series data. It is better for analyzing attributes of a population. And there are better tools for time series data, but they are harder to use.
Pearson’s Correlation Coefficient can also get fooled by trends in the data. See more about that topic in this article from 2010: Correlations May Not Be What They Seem.
The last part of the answer about correlation coefficients is that I usually do not need a number to tell me whether data are correlated. It is usually pretty obvious from looking at the chart. The eye can get fooled, it’s true, and so one should be mindful of that when doing any chart analysis. But if there is something there in a relationship, or if there is not, it is usually pretty obvious.
Coming back to the relationship under discussion, this difference in behavior before and after 2015 leaves us with a difficult question: Is this a durable relationship, or just a spurious correlation that the stock market happens to have fallen into just recently? That matters a lot as we contemplate the recent sharp drop in iron ore prices. If this is a relationship which can be believed, then that sharp drop says bad things about the future for the stock market.
To contemplate that point further, here is a chart that zooms in on that 29TD forward offset relationship:
The correlation between the SP500 and this forward-shifted plot of iron ore prices which seems to have worked pretty well in 2015 and 2016 seems to now be breaking down. So on that basis, I am not worried about the stock market repeating the recent plunge in iron ore prices. But I do plan to keep watching it.
Subscribe to Tom McClellan’s free weekly Chart In Focus email.
|Aug 06, 2010
Correlations May Not Be What They Seem
|Dec 10, 2016
Bonds and Gold in Unusual Correlation
|May 07, 2015
Eurodollar COT’s Leading Indication