Understanding the Difference Between Association and Causation

On days when many people carry umbrellas, there are more traffic delays. Nobody concludes that umbrellas slow down cars. Rain does both. The reasoning is obvious when the subject is weather and much easier to lose when the subject is food, sleep or exercise, and the sentence reads “people who do X have less of Y”.

Two words, two different claims

The Australian Bureau of Statistics (ABS) gives clear definitions. Two things are correlated, or associated, when their values change together: as one goes up, the other tends to go up, or down. Causation means that one event is the result of the other.

The ABS notes that the two are easy to tell apart in theory, while in practice it is far harder to establish cause and effect than to establish a correlation. A correlation is still useful, because it shows where further research into causes is worth doing. It is a starting point for a question.

Four ordinary reasons two things move together

1. A third factor drives both

The ABS uses ice cream and sunscreen: sales of both rise in summer, and the season is the shared cause. Neither product sells the other. Researchers call the hidden third factor a confounder.

Two more illustrations of the same logic, made up for this article: in a primary school, children with bigger shoes would read better, because older children have bigger feet and more years of school. And if a survey found that children in homes with more books did better at school, those homes might also differ in income, time and many other ways.

2. The arrow points the other way

Imagine a survey finds that people who own running shoes are fitter than people who do not. Perhaps the shoes helped. Or perhaps people who already run buy running shoes. When two things are measured at the same moment, you cannot tell which came first. This is known as reverse causation.

3. Chance

Compare enough things and some will line up by accident. A town’s library visits might track its bus delays for three years running and then stop. The US National Cancer Institute (NCI) lists chance as one of the possible explanations for an association seen in a study.

4. Who ends up being counted

Suppose a gym surveys its members and finds that those who have attended for five years are very satisfied. The people who were dissatisfied left long ago and are not in the survey. The pattern comes from who was available to be measured.

What makes a causal claim stronger

None of this means causes can never be known. They can, but it takes more than one pattern.

A controlled comparison

The ABS describes controlled studies as the most effective way of establishing causality. People are split into groups that are comparable in almost every way; the groups receive different treatments, for example a new medicine or a placebo; and the outcomes are compared. If the groups really were alike at the start, a clear difference at the end may be caused by the treatment. In medicine, the groups are usually formed at random so that hidden third factors are spread evenly between them. The US National Center for Complementary and Integrative Health (NCCIH) says clinical trials, which are carried out with people, give the clearest information on whether a treatment or lifestyle change is effective.

Careful observation over time, when a trial is not possible

Some questions cannot ethically be tested by experiment. Nobody can assign people to a harmful exposure. The ABS explains that observational studies are used in these cases: researchers follow groups and record their behaviour and outcomes over time. The link between smoking and lung cancer, which the ABS gives as an example of a causal relationship, is of this kind.

Many studies pointing the same way, and a mechanism

The NCI explains that scientists become more confident in a link when many studies show a similar association and there is a plausible biological explanation for it. The NCCIH makes a related point about systematic reviews and meta-analyses, which examine multiple studies on one topic: when many studies reach the same conclusion, the result is more likely to be reliable.

Even so, the ABS describes the role of observational studies modestly: they provide statistical information to add to the other sources needed to establish whether a causal link exists. Establishing causality is a judgement built from several kinds of information.

Listen to the verbs

Careful writers choose their words to match the evidence. “Linked to”, “associated with”, “tied to” and “more common among” describe a pattern. “Causes”, “prevents”, “leads to”, “protects against” and “reduces” claim an effect. When a headline uses the second kind of word about the first kind of study, it has gone beyond the evidence. Our guide to reading a health news headline turns this into a checklist.

Applying it to your own notes

The same caution is useful with personal records. If a sleep diary shows worse nights after late coffees, that is an association in a sample of one. A third factor may be involved: late coffees may happen on late working days, and the work may be what keeps you awake. The pattern is still worth having, because it tells you what to look at more closely or what to raise with a clinician.

Three mistakes to avoid

• Treating a pattern as proof. This is the classic error.
• Dismissing every pattern. “Correlation is not causation” is not a way to wave evidence away. Much well-established health guidance began with observed associations that were later confirmed by many studies of different kinds.
• Assuming a real cause is a big one. A genuine effect can be small. To see how large it is, you need the numbers: see relative and absolute risk.

The NCI observes that news coverage of association studies can lead to misunderstandings. A useful habit, on meeting any “X is linked to Y”, is to ask the four questions in order: could a third factor explain it, could it run the other way, could it be chance, and who was counted? If a finding still stands after that, and other studies agree, it deserves attention. Whether it should change anything for you personally is a question for a health professional who knows your circumstances.