You've heard the phrase: correlation does not imply causation. It's the most repeated warning in statistics — and the most ignored. A correlation tells you two variables move together; it says nothing, by itself, about whether one causes the other.

Why two things can correlate without one causing the other

  1. A confounder (lurking variable). Ice-cream sales correlate with drowning deaths. Neither causes the other — hot weather drives both. The confounder creates the link.
  2. Reverse causation. You find that people who exercise are healthier. Does exercise cause health, or do healthier people find it easier to exercise? The arrow might point the other way.
  3. Coincidence. With enough variables, some will correlate by pure chance. The number of films Nicolas Cage appears in has correlated with pool drownings. Nobody thinks that's real — it's a spurious correlation.
  4. Selection / sampling effects. How the data was collected can manufacture a correlation that doesn't exist in the wider population.

The rule of thumb: a correlation is a clue, not a conclusion. It tells you a relationship is worth investigating — it doesn't tell you which way the arrow points, or whether there's an arrow at all.

What it actually takes to show causation

How this connects to your analysis

When you run a correlation or regression and get a significant result (small p-value), you've shown the relationship is unlikely to be pure chance — but "not chance" is not the same as "causal." Significance rules out coincidence; it does not rule out confounders or reverse causation. Always ask what else could explain this? before claiming cause.

Quick example. A study finds a strong positive correlation (r = 0.7) between hours of sleep and exam scores. Tempting headline: "sleep boosts grades." But conscientious students might both sleep more and study more — conscientiousness is the confounder. Without an experiment, you can't separate them.

Measure the correlation first

Correlation Calculator →Get Pearson/Spearman r with a p-value and scatter plot — then reason carefully about cause.

Related reading: Interpreting p-values · Effect size explained · Choosing a statistical test