What is an example of correlation but not causation?
Ice cream and sunscreen sales rise and fall together across the year, yet neither causes the other. Hot weather drives both. That shared third factor, called a confounder, is the most common reason correlation gets mistaken for causation.
Suppose a shopkeeper tracks two product lines for a full year. Sunscreen sells best in summer and worst in winter. Ice cream follows the same pattern, month after month. Plot the two sales figures together and they move in tandem. It would be easy to conclude that buying sunscreen makes people buy ice cream, or the reverse. Neither is true. The Australian Bureau of Statistics uses precisely this pairing to show how two variables can move together systematically while having no direct relationship at all. The shared driver is the season: hotter weather sees more people wearing sunscreen and more people eating ice cream. The correlation is real. The causation is not.
[1]Why the mix-up happens
A correlation is a number describing the size and direction of a relationship between variables. It says nothing about which way influence runs, or whether influence exists at all. Research literature identifies three recurring reasons a correlation appears without a causal link. First, chance: two trends can coincide for no deeper reason, as when rising incidence of breast cancer or liver disease tracks rising rates of joint replacement surgery. Second, reverse causality: the supposed effect actually precedes the supposed cause, as when poor underlying health before surgery leads people to stop drinking, so abstainers show the highest mortality risk rather than abstinence harming health. Third, confounding: a hidden third variable influences both sides, as with the weather behind sunscreen and ice cream. Confounding is described in the orthopaedic research literature as the most common and well-known reason for the mistake.
[1][2]A comparison worth holding in mind
Contrast two statements about smoking. Smoking correlates with alcoholism, but the correlation does not mean smoking causes alcoholism. Smoking also raises the risk of lung cancer, and that one is a causal claim, supported by controlled study designs where comparable groups receive different exposures and their outcomes are compared. Same substance, two different relationships. The lesson is that the same pair of variables can sit on either side of the line, and only the study design, not the strength of the correlation, tells you which side you are on.
[1][2]A short decision process
- Ask what else could move both variables. Look for a shared driver like season, age, income or underlying health status.
- Check the direction. Could the outcome be producing the exposure rather than the other way round?
- Ask whether the pattern could be coincidence, especially with short time spans or small samples.
- Look at how the claim was studied. Controlled studies comparing matched groups are described as the most effective way to establish causation; observational studies can only add supporting evidence.
- Remember the limitation: a correlation coefficient measures association, usually of the linear kind, and cannot by itself establish cause and effect.
One caution: the Australian Bureau of Statistics and the orthopaedic research literature explain the concept and the general methods, but they do not cover every statistical technique, and applying these ideas to a specific study, especially in medicine, benefits from expert statistical review. As a next step, take a headline claiming that something causes something else and run it through the five questions above. If you cannot name a plausible confounder or rule out reverse causality from the information given, treat the claim as an association, not a cause.
[1][2]Sources and dates
- https://www.abs.gov.au/statistics/understanding-statistics/statistical-terms-and-concepts/correlation-and-causation · accessed 2026-10-18
- https://pmc.ncbi.nlm.nih.gov/articles/PMC10010939/ · accessed 2026-10-18
- https://thelongerlook.com/ · accessed 2026-10-18