The Difference Between Correlation and Causation

"It’s correlation, not causation!" You've probably heard this phrase before; it's practically a cliché at this point. But clichés become clichés because they're true, and this one is pretty critical for understanding how scientific evidence actually works. Misunderstanding the difference between correlation and causation is one of the most common errors in how research gets interpreted by even scientists themselves.

What does correlation mean?

A correlation is a statistical relationship between two variables. When one tends to go up, the other tends to go up (or down) as well. Ice cream sales and drowning rates, for example, are positively correlated: as one increases, so does the other. Shoe size and reading ability in children are also correlated.

These correlations are real in a mathematical sense, but none of them imply that one thing is causing the other.

What does causation mean?

Causation means that one thing directly produces another. Smoking causes lung cancer. The causal chain involves specific biological mechanisms: carcinogens in tobacco smoke damage DNA in lung cells in ways that lead to malignant growth. It's not just that smokers tend to get lung cancer more often, there are actual processes by which smoking induces cancer.

Establishing causation is much harder than establishing correlation, and it requires more than just observational data. The gold standard for establishing causation in medicine and social science is a randomized controlled trial — a study in which participants are randomly assigned to a treatment or control group, which controls for confounding variables (other factors that might explain the relationship).

Why do correlations fool us?

Correlations can appear for several reasons that have nothing to do with causation. The ice cream and drowning example is a classic case of a confounding variable: both are caused by a third factor (hot summer weather) that makes both go up at the same time. Another common pattern is reverse causation: the direction of the relationship is backwards from what we assumed. Studies have shown that depressed people tend to exercise less, but does lack of exercise cause depression, or does depression cause people to exercise less?

When you read about a study, especially an observational study or a survey, your first question should always be: does this establish correlation or causation? Most headlines that say "X causes Y" are based on correlational data.