Spurious Correlation

A spurious correlation is a case where two variables appear statistically associated but actually have no causal relationship. The cause is usually one of two things: pure chance, or an invisible third factor moving both.

Tyler Vigen, a Harvard Law School graduate, showed this amusingly with a project called 'Spurious Correlations.' He collected plenty of cases where two completely unrelated datasets showed surprisingly high correlation.

The most famous example is 'the number of films Nicolas Cage appeared in each year' and 'the number of people who drowned in swimming pools that year.' The correlation coefficient of the two values comes out fairly high, at about 0.67. But there's no way Cage's films cause drowning accidents.

Another example is a graph where 'cheese consumption' moves together with 'the number of people who died tangled in their bedsheets.'

The key lesson is this. There are countless data series in the world, and pick any two of them and you can find plenty of pairs that happen to move similarly by chance. The mere fact that a correlation is high doesn't prove a relationship.

Investing has similar superstitions. A representative one is the 'Super Bowl indicator.' It's the folk belief that if a team from a certain league (the NFC) wins the American football championship game, the stock market rises that year, and if a team from the other league (the AFC) wins, it falls.

Surprisingly, its hit rate was very high early on. The first 11 times all 11 were correct, and it was reportedly correct in 30 of 31 times (about 97%) through the late 1990s. But since 2000, the hit rate has plunged to about 46%. That's no different from a coin toss.

There is no causation whatsoever between football game results and stock prices. Experts see this as a clear spurious correlation — pure coincidence.

What's frightening is that a pattern can hold for decades and still turn out to be coincidence. The fact that it was correct for a long time does not become evidence that it's a 'real law.'

In investing, when you meet promotion that some indicator 'worked this well in the past,' you should first suspect whether that relationship has a plausible causal mechanism, or whether it's just a coincidence caught while digging through data.

This site does not provide such superstitious predictions. Instead, using actual price data, it shows — without hiding maximum drawdown and loss periods — the results of holding a good asset for a long time.