Correlation vs Causation: Simple Examples for Reading Claims
Correlation describes a statistical relationship between variables. Causation means that a change in one contributes to producing a change in another. Observing two things move together does not, by itself, establish that one caused the other. The Australian Bureau of Statistics makes this distinction in its introductory guidance. Correlation and causation.
This matters whenever a headline moves from “people who do this have better results” to “do this to get better results.” The advice may be useful, but the observation alone may leave important questions unanswered.
The umbrella example
Imagine keeping a record of the days you arrive home with wet shoes. On many of those days, you are also carrying a wet umbrella. That does not establish that the umbrella made your shoes wet. Rain is a plausible shared explanation.
The example suggests a useful question: could another factor account for both observations? The aim is not to invent endless objections until every finding becomes unusable. It is to examine whether important alternative explanations have been addressed.
When the proposed cause follows the underlying difference
Consider another hypothetical example. Students who buy additional books receive higher grades. The books may help. But students who already have a strong interest in learning may also be more likely to buy them. Family resources and available study time could influence both.
“Buy these books to raise your grades” therefore goes beyond the original observation. It needs evidence about the effect of obtaining or using the books, including who used them and how.
Timing is useful to examine too. A proposed cause should precede its effect, but sequence alone is insufficient. If you reorganise your schedule on Monday and work more effectively on Tuesday, the new schedule may have helped. Tuesday may also have involved fewer demands.
Questions to ask about a study
Begin with the design. Did researchers observe existing behaviour, or compare an intervention with an alternative under planned conditions? Who participated? How were outcomes measured? Were the groups already different in ways that might affect the result?
Then consider the size and meaning of the difference. A small numerical change may not translate into a noticeable practical benefit. A finding in one setting also needs careful interpretation before being applied to every age group or circumstance.
Controlled experiments are important tools for investigating causes, but they are not possible or appropriate for every question. Multiple sources of observational evidence can also contribute to a strong causal case when alternative explanations are carefully addressed. A study design helps you interpret a claim; it does not guarantee that every conclusion drawn from it is sound.
Match the decision to the evidence
For a small, reversible experiment, such as moving a notebook to a more useful place, you can try a change and observe what happens without declaring a universal rule.
A consequential decision affecting health, finances or other people’s rights needs evidence and expertise proportionate to its effects.
When repeating a finding, choose language that preserves its strength: “was associated with,” “suggests a possible effect,” or “supports a causal relationship,” according to what the source establishes.
For the next striking claim you encounter, ask three questions: what was observed, what explanation was proposed, and what evidence connects the two? Those questions can make your reading more accurate without turning it into indiscriminate doubt.