What is a Correlation vs. Causation Fallacy?

Start by reading this out loud to your kid.

Every afternoon at three, the mail truck turns onto our street, and every afternoon at three, Biscuit the beagle hurls himself at the window barking like the fate of the house depends on it. Without fail, the truck drives away within a minute or two. Biscuit is completely convinced he's the reason the evil truck leaves.

Ask your kid why Biscuit's reasoning is flawed. You're listening for something like: the truck was going to leave either way, it's finishing its stop on the route, not responding to the bark. You might also hear that it's the truck's arrival, not its departure, that sets Biscuit off in the first place. That's true, but it explains why he barks, not why the truck leaves.

Biscuit is ignoring a confounding variable: something that influences both the supposed cause and the supposed effect. Here, the two things he links together, his barking and the truck pulling away, are both caused by the same thing. The mail route. The schedule brings the truck to the curb, which sets him off, and that same schedule sends the truck off a minute or two later. Neither the barking nor the departure causes the other.

In 2012, a cardiologist named Franz Messerli went looking for something to do with a rainy afternoon. Toying with data, he plotted how much chocolate different countries eat against how many Nobel Prizes they've won and turned up a correlation: more chocolate, more laureates. What unsettled him was the p-value. By the same standard that vouched for his own 800-odd published papers, this nonsense relationship looked airtight. That was the point worth making. He wrote it up for the New England Journal of Medicine, in their Occasional Notes section, joking that the U.S. would need an extra 275 million pounds of chocolate a year to produce one more Nobel winner.

The joke didn't fully land, at least not in the headlines. For example, CBS ran it under "Eating chocolate may help you win Nobel Prize." Now the article beneath actually hedged: it told readers to take the finding with a grain of salt, flagged that it had been published as a note rather than a peer-reviewed study, and quoted a skeptic. But none of that made it into the headline, and unfortunately a lot of people only look at the headline (researchers at Penn State who went through more than 35 million public Facebook posts found that links get shared without a single click to actually see the article over 75% of the time - people just share based on the headline).

A skeptical reader seeing those headlines should have run the claim through the options above:

Reverse causation. Could winning a Nobel Prize come first? Maybe it creates a national celebration that bumps up chocolate sales for a season?

Common cause. Wealth is the obvious candidate. Rich countries can afford both a lot of chocolate and a lot of funded research. Chocolate isn't doing anything; money is buying both.

Coincidence. Test enough snack foods against Nobel counts — cheese, coffee, pretzels — and something is going to correlate with something, purely by chance.

Settling which one is true takes more than a chart. It takes careful, controlled research. In this case, it's almost certainly wealth.