Correlation Is Not Causation By: Gabriel PotterMBA, AIFA® 2014.05.14

We’re stepping away from a direct discussion of fiduciary concerns, economics, or market commentary this week.  I was directed to a website, which I will introduce to you in a moment, that perfectly encapsulates an old argument.   I thought I should share it with you.

Any statistician will tell you the following truth:  correlation is not causation.  In other words, just because two events, A & B,  occur at the same time, it doesn’t mean that A causes B.  B might cause A.  Both A & B may in fact by caused by another event C.  Finally, it could just be a complete coincidence with no relationship between events A & B other than random chance. 

We refer to this principle frequently.  For example, we have discussed Richard Rolls’ 1987 research paper which showed that  the price of butter was correlated to the market returns of the S&P 500 for no reason other than random chance.  There was no real correlation.  It was just a dumb luck, and the so-called relationship broke down immediately the following year, leaving some other arbitrary data set the dubious honor of being most correlated to the S&P 500 for that particular year.

However, market pundits often try to “overfit the data” and derive meaning from uncorrelated events.  For example, you can give a technical analyst a stock-price chart whose up and down movements is determined at random, simply by flipping a coin, and the analyst will still try to describe the characteristics of the stock.  (Famed statistician Nate Silver performed this very experiment; you can read about his work in the book, “The Signal and the Noise”).

Here’s the website what I wanted to show you:

This website includes a series of completely unrelated data sets which, due to random chance, experience high correlation.  For example, the per capita consumption of cheese relates to the number of people who died becoming tangled in their bed sheets with a very high 94.7%.  Sometimes, the datasets present a kind-of logic For example, the days of sunlight in California are inversely related to precipitation in California.  More often, there is no logic to these connections and it really is just random chance. 

Gabriel Potter

Gabriel is a Senior Investment Research Associate at Westminster Consulting, where he is responsible for designing strategic asset allocations and conducts proprietary market research.

An avid writer, Gabriel manages the firm’s blog and has been published in the Journal of Compensation and Benefits,...

More about Gabriel Potter
Sign up for our Newsletter
Sign up for our Newsletter