Correlation vs Causation
Why two things moving together does not prove one causes the other, with examples.
Two things are correlated if they tend to change together. Correlation does not by itself show that one thing causes the other.
Possible explanations for a correlation
- A causes B.
- B causes A (reverse causation).
- A third factor causes both (a confounding variable).
- Coincidence, especially with small samples or many comparisons.
Example: ice cream and sunburn
Ice cream sales and sunburn cases rise together in summer. Eating ice cream does not cause sunburn. Hot, sunny weather is a third factor that increases both.
Example: reverse causation
People who carry umbrellas are more likely to be wet. The rain causes both the umbrella and the wetness.
How researchers test for causation
- Randomised controlled trials: participants are randomly assigned to groups, so other factors are, on average, balanced.
- Controlling for confounders in observational studies using statistical methods.
- Temporal order: the cause must come before the effect.
- Mechanism: a plausible way the cause could produce the effect.
- Replication: repeated results in different studies.
Questions to ask about any claim
- Is the claim about correlation or about cause?
- Could a third factor explain both?
- Could the direction be reversed?
- How big was the sample, and how were people chosen?
Common mistakes
- Concluding that A caused B from a graph alone.
- Saying "correlation proves nothing". Correlation can be a useful clue and a reason to investigate.
Quick recap
Correlation shows association. Causation needs further evidence from experimental design, controls and mechanism.