One reason many investors find correlations so seductive is they offer the illusion of certainty. The idea that ‘if X happens, the market will do Y’ allows people to believe they possess some sort of map or manual for the world of investment and yet, as The Value Perspective has argued before in articles such as Tangled up with blue and Jellybean Trilogy 2, correlation is nothing without causation.
Investors who believe otherwise should visit the entertaining Spurious connections site, where they will find examples of highly correlated, if wholly unrelated, relationships such as the divorce rate in Maine and the US’s per-capita consumption of margarine. Any investor who still believes otherwise after that might perhaps consider taking up a different activity, such as the ‘Guess the correlation’ game here.
Alternatively, they could just stick with The Value Perspective as we highlight a new academic paper, Channelling Fisher, in which Alwyn Young, an economics professor from the London School of Economics, explores to what extent the statistically significant conclusions of scientific experiments published in some major economics journals were down to nothing more than random chance.
Young tested a total of 2003 regressions across 53 experimental papers from the American Economics Association by taking every sample that had been deemed statistically significant and then randomly generating them again himself many times over. This so-called ‘randomised inference’ approach found more than half the supposedly statistically significant conclusions did not hold up at all.
There are, of course, plenty of reasons why someone might look to prove their point by zeroing in on data correlations even if they are not statistically significant – for example, research money, the acclaim of one’s peers and the simple desire to be seen to be right. This being The Value Perspective, we would also throw the behavioural finance sin of ‘confirmation bias’ into the mix.
Still, as we suggested at the start of this piece, it is not just kudos-hungry or grant-seeking academics who can be inclined to see correlations where none actually exists and the ‘correlation du jour’ now gripping the investment world is the supposed relationship between the plunging price of oil and the similarly downward trajectory of world equity markets.
The fact is, however, were you to look at the data over a rather longer period than the last 18 months or so, you would see that the relationship between equity markets and the oil price is sometimes positively correlated and sometimes negatively correlated. Sometimes the relationship is very strong, sometimes it is very weak.
As it happens, where there is a huge oil-related correlation – and once again we know this courtesy of the Spurious connections site – is between total crude oil imports into the US and that country’s per-capita consumption of chicken. Yes, that is laughable but really no more so, thanks to the absence of causation, than linking the fortunes of oil and equities.
The problem is, the more that is written about perceived correlations, the more investors are likely to take causation as read and ‘backfill’ a suitable story. In our technological age, an extra consideration here are the investors who build computer algorithms to try and take advantage of correlations – even if there is no causation, if a correlation is acted upon enough, it can ultimately become self-fulfilling.
It is human nature to want certainty and correlations play into that desire, grabbing investors’ attention for short periods of time. Here on The Value Perspective, however, we would argue they would do better to take a longer-tern view and focus on fundamentals such as valuation and balance sheet strength. After all, real certainty is a commodity that has always been in short supply in investment.