Distinguish correlation from causation
Examine what an association measures, consider alternative explanations and match your conclusion to the study’s method.
Follow the method ↓
The answer in 30 seconds
A persuasive chart can encourage an explanation that goes beyond the evidence. The aim is to distinguish an observation from a hypothesis and from a demonstrated effect.
Two variables moving together do not prove that one causes the other. Examine timing, shared influences, data collection and study design before using causal language.
Examples to adapt
Warm weather may increase both; their association does not show that buying ice cream makes people swim.
A relationship between city averages may not describe individual residents.
Follow the method
- 1Define the variables
Record units, population and period. Check whether the values describe individuals, places or averages. State the unit being observed: person, class, business or time period. A relationship between area averages does not automatically describe individuals living in those areas. Keep group sizes and the observation period alongside the finding.
- 2Describe the observation
State what the data shows before explaining why. Look for outliers and missing periods.
- 3Consider alternatives
A third factor may influence both variables. The direction of an effect may also be the reverse of your initial idea. Ask whether a third factor could influence both variables or whether the direction could be reversed. In a fictional example, greater activity may produce both more requests and more spending. Their joint change does not establish that spending alone causes the requests.
- 4Match the conclusion to the method
Read the study design and limitations. If the result is an association, use that term rather than causes or prevents.
Workshop: test an attractive explanation
Describe the observed relationship without using the word 'cause'. Then look for a factor affecting both variables, a time lag or selection bias. Compare the starting groups and ask whether the pattern appears in other data.
A correlation can help prediction without explaining why. A causal claim needs the study design, assumptions and limitations; a chart alone cannot establish it.
A closer look
A possible common cause
Fictional example: swimming-pool and ice-cream-shop attendance rise together over four periods. The table explores explanations without reporting a real observation.
| Element | Observation | Decision |
|---|---|---|
| Hypothesis 1 | The pool causes ice-cream sales | Joint movement does not establish this |
| Hypothesis 2 | Heat increases both kinds of attendance | Examine that variable and comparable periods |
| Hypothesis 3 | The observed periods were selected | Examine the full series and selection method |
A plausible explanation is not yet an established result. Separate what the data show from hypotheses requiring another study design.
Do more points prove the cause?
Not automatically. More observations can refine an association without removing a common cause, selection or another explanation.
A checklist to keep
Use these checks to record your observations. They are a reading aid, not an automatic assessment.
| Check | What to examine | Action |
|---|---|---|
| Observation | What changes together? | Describe |
| Scope | Who, where and when? | Define |
| Explanation | Other possible factors? | Examine |
| Conclusion | What does the method support? | Qualify |
Read the checklist directly in the table.
What to check
Statistical relationships depend on scope.
Compatible timing alone does not demonstrate an effect.
No linear correlation does not rule out every relationship.
Common questions
Does a strong correlation prove a cause?
No. Study design and competing explanations still matter.
Can I keep the hypothesis?
Yes, if you clearly describe it as something to test.
Sources and documentation
Documentation consulted on . Examples are illustrative; interfaces and results may change.