Direct comparison
Correlation vs causation
Correlation means two variables tend to vary together; causation means a change in one variable actually produces a change in the other. A correlation can exist without any causal link.
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
Side-by-side comparison
| Dimension | Correlation | Causation |
|---|---|---|
| What it means | Two variables tend to change together — a statistical association. | A change in one variable directly produces a change in the other. |
| What it establishes | That a relationship or pattern exists, and its strength and direction. | That one variable is a genuine cause of the other’s change. |
| Direction | Symmetric — X correlates with Y exactly as Y correlates with X. | Directional — the cause precedes and produces the effect. |
| Third variables | A hidden confounder can create the association between the two. | Confounders must be controlled or ruled out before a cause is claimed. |
| How it is established | Observational data; a correlation coefficient or scatterplot. | Controlled experiment, ideally a randomised controlled trial (RCT). |
| Temporal precedence | Not required — the variables simply co-vary. | Required — the cause must come before the effect in time. |
| Reverse causation | Cannot tell which variable, if either, came first. | Specifies the direction, ruling out the effect causing the cause. |
| Classic example | Ice-cream sales rise with drowning deaths (both driven by hot weather). | A drug lowers blood pressure in a randomised, placebo-controlled trial. |
| Strength of claim | A weaker claim — describes a pattern, not a mechanism. | A stronger claim — requires converging evidence and explanation. |
Why correlation does not imply causation
Two variables can correlate for reasons other than one causing the other. There may be a confounding third variable that drives both — hot weather raises ice-cream sales and swimming (hence drownings), creating a spurious link between ice cream and drowning. The direction may be reversed, with the supposed effect actually causing the supposed cause. Or the correlation may be pure coincidence, especially when many variables are compared. Bradford Hill’s criteria — including strength, consistency, temporality, a dose–response gradient and plausibility — offer a structured way to weigh whether an observed association is likely to be causal.
Common questions
FAQ
Why does correlation not imply causation?+
Because an association between two variables can have other explanations. A hidden third variable (a confounder) may cause both; the causal direction may be reversed; or the pattern may be coincidence. Without controlling these alternatives — usually through a randomised experiment — a correlation alone cannot show that one variable produces the other.
How can causation be established?+
The strongest evidence comes from controlled experiments, especially randomised controlled trials, where random assignment balances confounders so a difference in outcome can be attributed to the manipulated cause. Where experiments are impossible, researchers draw on criteria such as temporal precedence, a dose–response relationship, consistency across studies and plausible mechanism — broadly, the Bradford Hill criteria.
What is a spurious correlation?+
A spurious correlation is an association between two variables that are not causally linked, produced either by coincidence or by a third variable affecting both. The classic example is ice-cream sales correlating with drowning deaths: hot weather independently increases both, so the two move together without one causing the other.
