Statistics & Data
Why Ice Cream Is Not Killing Anyone (But Your Brain Thinks It Is)
For decades, researchers noticed that countries where people eat more chocolate win more Nobel Prizes — and the correlation is nearly perfect.
The Idea
Correlation measures how reliably two things move together. Causation means one of them is actually driving the other. The confusion between the two isn't a rookie mistake — it's a near-universal cognitive reflex, because human brains are pattern-completion machines that evolved to find causes fast, not accurately. When two things rise and fall together, the mind reaches for a story connecting them. This is usually wrong. The Nobel-chocolate relationship is explained by a third factor: national wealth. Richer countries consume more chocolate and also invest more in the research institutions that produce laureates. Wealth is the hidden driver — what statisticians call a confounding variable. The deeper problem is that correlation is genuinely useful. It is the first thing you look for when hunting for real causes. Researchers who discovered the link between smoking and lung cancer started with a correlation. The error isn't noticing the correlation — it's stopping there. To establish causation, you need more: a plausible mechanism, a dose-response relationship (more of X produces more of Y), the cause preceding the effect, and ideally a controlled experiment where one group is exposed and another is not. Without those checks, a correlation is a hypothesis dressed up as a conclusion. The world is full of variables that accidentally rhyme, and our instinct to narrate them into cause and effect is one of the most expensive cognitive habits we have.
In the World
In the early 1990s, a well-regarded study found that children who slept with a night-light were significantly more likely to develop myopia — short-sightedness — later in childhood. The finding was reported widely. Night-light sales dipped. Parents worried. Then a follow-up study looked more carefully at the data and found the real explanation: parents who are short-sighted are more likely to leave a light on in their child's room — possibly because they navigate better in low light themselves, or simply out of habit. And short-sighted parents are much more likely to have short-sighted children, because myopia is substantially heritable. The actual cause was genetic. The night-light was just along for the ride, statistically correlated with myopia without contributing a single dioptre to it. The original researchers had not been sloppy exactly — they had found a real correlation. But they had not asked whether a third variable might be quietly explaining both ends of it. A generation of parents lost sleep over a light that was entirely innocent. What makes this case instructive is that the corrective came not from new data but from asking a different question: what else might predict both of these things simultaneously? That question — simple, almost obvious once stated — is the one that separates a good data analyst from a dangerous one.
Why It Matters
Most of the decisions we make about health, money, relationships, and policy are based on pattern-matching that never gets subjected to this kind of scrutiny. Someone notices that people who drink a morning coffee tend to be more productive, and a belief calcifies. A company sees that customers who use their app more also spend more, and builds a strategy on the assumption that engagement is causing spending — when both might simply reflect a type of customer who was always going to spend. The practical move is not to distrust all correlations — that would make you useless. It is to hold them lightly and ask, reflexively, what third thing might be driving both. Who else is in this picture? What am I not measuring? Is there a plausible mechanism here, or am I just enchanted by the pattern? Training that question into a habit changes how you read news stories, interpret advice, and evaluate your own experience. It also makes you harder to mislead — which, in a world where data is increasingly used to sell conclusions, is not a small thing.
A Question to Ponder
What is one belief you hold about cause and effect in your own life that you have never actually tested — where you might just be watching two things move together and assuming one made the other happen?
Get a new one of these every morning.
Start learning with Thinkable