What Correlation Implying Causation Means in the Real World

People use the "correlation does not equal causation" retort a lot without quite knowing its significance. All over the web, it's a stock debating tool, a line of defense that neither side puts much weight into, a rhetorical device otherwise unsupported.

I'm not proud of it, but I just dropped a Facebook friend for making a stupid anti-transgenic foods argument. It's the one about how the release of GM food products into the market correlates to increases in gastrointestinal disease and obesity, and therefore… evidence. I stuffed my correlation ≠ causation retort (and a side note about how calories and rapidly changing calorie density in foods is in real-life tied to obesity increases, not GMOs) and unfollowed, just because I didn't want to deal with it. That's a lot like ceding the argument, unfortunately.

Humans love correlation. We love correlation because we love stories, narratives: this happened, leading to this, and next should be this other thing. We look for the forms of stories in the world, and a story is roughly the opposite of coincidence, which is things just happening together because time is just a substance of many layers, a stack of happenings.

Some different things are really the same thing, yes, but many aren't the same thing or are part of other things that we can't as readily see. It's more than coincidence just not being any fun, it doesn't allow us to make predictions about the future, which is what stories are all about: seeing the future and the things in it that might kill us. Coincidence doesn't help do that.

But coincidence happens. Depending somewhat on the information resolution one has, there is much more coincidence out there than there are causes and effects (the simplest explanation …). Correlation as causation, the notion that two or more events are connected because they share some trend over time, hits human brains even harder because we don't consider correlation ever unless we're already looking for causes and effects.

Seldom do we note that thing A and thing B both increased over some period unless we were already considering them as interrelated. We don't see enough of the world's noise, in other words.

The graphs above and below are of noise. Two things that we have no reason to seek out cause and effect between, correlated sometimes just enough to notice but sometimes matching perfectly, close enough to shock and alarm.

Stats fan Tyler Vigen put them together, presumably with help from a script, and you can take your own lesson from them. The graphs are presented on his website Spurious Correlations without comment, just numbers.

I might suggest the winning lesson is just about that noise, and how easily noise can be manipulated without much effort. And the next time someone declares "correlation does not imply causation" just refer back to this little lesson in the vast world of correlation without meaning.

You get the idea, but there are plenty more where these came from.