November 21, 2026

Correlation and causality

Correlation describes a relationship: two variables move together in a certain way. Causality goes a step further: one variable causes a change in the other. That seems like a small difference, but it determines whether you only see an interesting pattern or can draw a conclusion that you can actually act upon.

Discover the difference between correlation and causality, with clear examples and practical tips to better assess causal claims.

That distinction is important in research, policy, health claims, and data visualization. A convincing graph, infographic, or animation can quickly make a connection seem credible, while evidence for cause and effect is still lacking. Do you want to interpret those patterns more carefully? Consider visualizing KPIs in context, so that relationships are not too quickly interpreted as cause and effect. Below, you will see how to recognize the difference, with practical examples and a simple way to better assess causal claims.

Why two patterns do not yet prove cause and effect

The fact that two variables rise or fall simultaneously does not automatically mean that one drives the other. Sometimes they both react to the same external factor, such as season, age, economic growth, or changing behavior. In other cases, they move along coincidentally, without any substantive relationship. A visible pattern is therefore primarily a signal that something *might* be going on, not immediate proof of cause and effect.

That distinction is crucial, just as with output, outcome, and impact. A correlation can be useful for predicting what is likely to happen, but that is different from knowing which button to press to change an outcome. For research, health information, and policy, this makes a big difference: those who confuse correlation and causality are more likely to draw incorrect conclusions and make ineffective decisions. Especially with strong data visualizations, a clear pattern can look convincing, while the substantiation for a causal claim remains too weak.

What correlation shows

  • Correlation means that two variables are statistically related. If one changes, the other relatively often changes along with it. However, this does not tell you why this happens, nor whether one factor influences the other.
  • With a positive correlation, two variables move in the same direction. A recognizable example is outdoor temperature and ice cream sales: when it gets warmer, sales often rise along with it.
  • With a negative correlation, variables move in opposite directions. Consider price and demand: in many markets, demand decreases when the price rises.
  • Sometimes there is no clear correlation. In that case, a change in one variable appears to have little or no predictive value for a change in the other.
  • Correlations are indeed useful. They help with pattern recognition, predictions, and formulating hypotheses that you can test more specifically later.
  • Moreover, a visible correlation can arise or appear stronger due to a third factor. Such confounding variables influence both measured variables simultaneously, making a direct relationship seem more plausible than it actually is.

What causality means in practice

Causality means that a change in factor A actually leads to a change in factor B. The relationship is then not only visible in the data, but also substantiated substantively and methodologically. Therefore, causal statements are much stronger than ordinary correlations and require more evidence than two lines moving simultaneously.

A well-known example is smoking and lung cancer: that is not just a pattern, but a causal link that is broadly substantiated. Similarly, with exercise and health, one looks at more than just correlation: increased physical activity can, for example, improve fitness. Such statements therefore require stronger evidence than a correlation that merely shows that two phenomena occur together.

How causality is investigated

  • The first step is to take alternative explanations seriously. Merely observing something in a dataset or graph is not enough if seasons, age, market growth, selection effects, or the lack of a representative sample may also play a role.
  • Controlled experiments often provide the strongest indications of causality. By deliberately treating groups differently and keeping other conditions as similar as possible, it becomes clearer whether factor A truly has an effect on factor B.
  • Longitudinal research examines the chronological order. A hypothesized cause must occur before the effect; if this does not happen, a causal explanation quickly falls away.
  • Researchers explicitly analyze confounding variables. For example, they correct for age, income, health, season, or other behavior to see if the association holds; visualizing data quality helps to identify measurement errors and gaps more quickly.
  • It is important to note that a statistical test, a high correlation coefficient, or a neat visualization does not in itself prove causality; with misleading graphs , that risk is even greater. A causal claim is usually made plausible by a combination of research method, timing, substantive logic, and context.

How you communicate your findings also matters. Formulate clearly, state limitations, and avoid unwarranted causality claims; this guide on presenting research results helps with that.

Make complex information clear, understandable, and convincing

Examples of correlation, causality, and spurious correlation

In a single overview, you can quickly see how differently the same type of data can be interpreted. It is not just about whether there is a correlation, but primarily about what explains that correlation.

ExampleType of relationshipWhat can explain the correlation?Correct interpretation
Smoking and lung cancerCausalityThis association cannot be adequately explained by chance alone. It has been found consistently and is consistent with what is known about harmful exposure.Assume a causal link. This is strong enough to base warnings, prevention, and policy on.
Search queries for “flu” and the number of flu casesCorrelation as a signalBoth move in line with the same underlying outbreak. The search volume does not cause the flu, but reflects what is happening in the population.Use to predict or monitor, do not present as a causal claim.
More screen time in the evening and poorer sleepCorrelation requiring further investigationPart of it may be causal, but stress, work rhythm, caffeine, or existing sleep problems can also influence the correlation.Investigate further before drawing a firm cause-and-effect conclusion.
Ice cream sales and drowningsSpurious correlationA third variable, warm summer weather, increases both. The simultaneous rise therefore says little about direct influence.Reject causality between these two variables themselves and first include the seasonal factor.
Rising mobile data usage and rising house prices over several yearsSpurious correlation due to trend similarityBoth series can increase in the same period, primarily because many societal developments change simultaneously over time.Do not explain causally. First correct for the time trend and then examine whether anything substantive remains.
Frequently Asked Questions

Is your question not listed? Feel free to ask.

Ask your question

Read also

All articles

Ready to take your A to AA?

Whether it concerns an ambition, product, process, technology, or idea: we make visible what makes it special and translate it into imagery that is clear, convincing, and impossible to ignore.

Check out our reviews on Trustpilot