Correlation Is Not Causation. So What Is It?

Ice cream sales and drownings both rise in summer. Nobody blames the ice cream. So when two things move together, what should you actually ask?

Correlation is a clue, not a verdict. Here are four explanations worth checking, roughly in order.

1. A causes B

Sometimes the obvious story is true. Rain really does cause wet sidewalks. It just should not be the only story you consider.

2. B causes A

The arrow may point the other way. Busy fire crews are found at bad fires, but they do not cause them.

3. Something else causes both

This is the ice cream case. Hot weather sends people to both the freezer aisle and the lake. A hidden driver like this is called a confounder.

4. Coincidence

Compare enough unrelated lists and some will line up by chance, especially over short periods. If you search hard for a pattern, you should expect to find one.

How careful people sort it out

The strongest tool is a controlled experiment, where chance decides who gets what. When that is impossible, look for a plausible mechanism, a sensible time order, and the same result turning up in independent data. None of that is as catchy as the slogan, but it is what the slogan is for.