Confusing Correlation and Causation

Statistics has a very famous sentence: 'Correlation does not imply causation.'

Concluding that 'one is the cause of the other' from the mere observation that two variables rise and fall together is a logical error. This mistake is called the 'questionable cause' fallacy.

For example, ice cream sales and the number of water-play accidents really do rise together. But ice cream doesn't cause the accidents. Both simply move together according to a common cause — 'hot weather.'

Like the 'hot weather' that was the true cause in the ice cream example, a factor that affects both variables and, on the surface, makes them look related is called a confounding variable.

Miss the confounder and two utterly unrelated indicators look like causation. So in statistics, you always first suspect 'isn't there a hidden third variable behind this relationship?'

The same trap exists in investing. Even if some economic indicator and stock prices look like they move together, in fact a common background like 'the overall economy' or 'interest rates' may be moving both.

Even if there is a real causal link between A and B, there's the mistake of getting the direction backward. This is called reverse causation.

For example, seeing the correlation 'the longer the study time, the higher the test score,' it's easy to think studying raises the score. But it could be that students who score well get motivated and study longer. That is, data alone can't tell which is cause and which is effect.

Ultimately, when a correlation appears between two indicators, there are at least three explanations: (1) A really is the cause of B, (2) conversely, B is the cause of A (reverse causation), or (3) a third variable moves both, or it's just chance.

Claims like 'when this indicator rises, stock prices rise too' are usually mistaking correlation for causation. The fact that two values moved together in the past is no guarantee that one will pull the other up going forward.

Especially with more data, you can find any number of indicator pairs that moved together by chance (see the next article, 'Spurious Correlation'). So before trying to predict the future based on some correlation, it's important to build the habit of first weighing the possibility of confounders, reverse causation, and chance.

This site does not predict future prices from a particular indicator. Instead, it shows — without hiding maximum drawdown and loss periods — what results actually occurred in the past when you held a good asset for a long time.