Correlation and Causation

What correlation and causation actually is

Correlation describes a pattern. When two things tend to rise and fall together, they’re correlated. As ice cream sales go up, so do sunburns. Causation is a stronger claim. It says one thing actually produces the other. The challenge is that a correlation, by itself, can’t tell you whether there’s a cause behind it.

In the ice cream example, ice cream doesn’t cause sunburn. Hot, sunny weather drives both. That hidden third factor is why correlation so often gets mistaken for cause, and why “these two things go together” is a much weaker statement than “this one causes that one.”

How it works

Two things can move together for several reasons. One might cause the other. The relationship might run in the opposite direction from what you’d guess. A third factor might be driving both. Or the pattern might be a coincidence that won’t hold up in the next dataset.

To move from “these are correlated” to “this causes that,” researchers need stronger evidence. A randomized controlled trial is the most powerful tool here, because randomly assigning people to groups breaks the link to hidden third factors. When that isn’t possible, researchers use careful designs and statistics to rule out alternatives, but those approaches are harder to trust than a clean experiment.

What it isn’t

A correlation isn’t proof of cause, no matter how strong it looks. It also isn’t worthless. Correlations are often the first clue that points researchers toward a question worth studying. The error is treating the clue as the conclusion.

It also isn’t always obvious which direction a relationship runs. People who exercise less may be more depressed, but depression can also sap the energy to exercise. A correlation can’t sort out which is driving which.

Related terms you’ll see next

Bias is one reason a correlation can be misleading. A randomized controlled trial is the standard way to test whether a relationship is actually causal. A p-value can tell you a correlation is unlikely to be chance, but still can’t prove cause. An effect size tells you how strong the relationship is.

Why it matters when you read about mental health

Mental health headlines are full of correlations dressed up as causes. “People who use social media more are more anxious” sounds like a cause, but it could run either way, or be driven by something else entirely. Knowing the difference keeps you from over-reading a study. When you see a claim that one thing causes another, the useful question is whether the study was actually built to show cause, or just spotted a pattern.

<blockquote>
  <p><strong>Correlation and Causation:</strong> Correlation means two things tend to move together. Causation means one actually makes the other happen. A correlation alone doesn't prove cause.</p>
  <p>&mdash; <a href="https://shrinktionary.com/terms/correlation-and-causation/">Shrinktionary</a>, a plain-English mental health dictionary, medically reviewed by a board certified psychiatrist.</p>
</blockquote>