B7 · Data Literacy and Visualisation

Does One Thing Really Cause Another?

Two things are correlated when they tend to rise and fall together. Correlation does not prove that one causes the other. Often a third factor, sometimes called a hidden cause or confounder, affects both, such as summer heat raising both ice cream sales and fan use.

10 min Beginner Jump to the activity

What you need: Nothing. This page starts from the beginning, with no programming and no formulas.

Definition

In summer, ice cream sales go up, and so does the use of fans. Does buying ice cream make people switch on fans?

Two things are correlated when they tend to rise and fall together. Correlation does not prove that one causes the other. Often a third factor, sometimes called a hidden cause or confounder, affects both, such as summer heat raising both ice cream sales and fan use.

In simple words

When it's hot, people buy more ice cream. When it's hot, people also switch on fans. Ice cream and fans move together, but neither causes the other: the heat causes both.

Why it matters

News, ads and even AI predictions are full of patterns. If we mistake a pattern for a cause, we make bad decisions, such as banning ice cream to save electricity. Asking 'what else could explain this?' is one of the most useful thinking habits for reading data.

Key terms

Correlation
When two things tend to change together: both rising, or one rising as the other falls.
Cause
Something that makes another thing happen.
Third factor
A hidden cause that affects both things and makes them look connected; also called a confounder.
Coincidence
Two things moving together by chance, with no real link.
Fair test
Changing one thing on purpose while keeping everything else the same, to see whether it really causes a change.

Explained in detail

The ice cream and the fan

Suppose a shop records how many ice creams it sells each month, and the electricity office records how many hours fans run in a typical home. Put side by side, both rise from March to May and fall when the weather cools. Someone might say 'ice cream makes people use fans', or 'fans make people want ice cream'. Both are silly, because there is a third factor: temperature. Hot weather makes people want something cold and also makes them switch on fans. These made-up numbers for one town show the pattern:

MonthAverage temperatureIce creams soldFan hours a day
December21 °C1501
March29 °C3004
May38 °C90012
July, in the monsoon31 °C4507

Three explanations to consider

Whenever two things move together, consider three possibilities. One really causes the other: more rain causes higher river levels. A third factor causes both: heat causes ice cream sales and fan use. Or it is a coincidence: with enough data, some unrelated things will move together by chance. Statisticians sum this up in a short phrase: correlation is not causation.

How to check

Look for a third factor that could affect both. Then compare cases where that factor is the same: on days with the same temperature, does more ice cream still go with more fan use? If not, temperature explains the link. The strongest check is a fair test, where you change one thing on purpose and keep everything else the same, which is how new medicines are tested before they are approved.

Why it matters for AI

AI models find correlations in data; they do not know causes. A model might learn that people who buy umbrellas also buy raincoats, which is useful for suggesting products. But anyone using such a pattern to decide what causes what, for example to set a rule, needs to check for third factors first.

Headlines often blur this line with words such as 'linked to', 'boosts' or 'causes'. When you read one, ask whether the study changed one thing on purpose, or only noticed that two things happen together. The first kind of study can show a cause; the second can only show a link.

How it works

  1. Notice the two things that move together.
  2. Ask: could one really cause the other? How exactly?
  3. Ask: what third factor could affect both?
  4. Compare cases where the third factor is the same.
  5. If possible, run a fair test that changes only one thing.

Worked example

Ice cream sales and fan use both rise from March to May. Could ice cream cause fan use? There is no sensible way it could. Could fans cause ice cream sales? Also unlikely. What could affect both? Temperature: May is the hottest month and has the most ice cream and the most fan use, while December is the coolest and has the least of both. When the town compares only days with the same temperature, the link between ice cream and fans disappears. So the third factor, heat, explains the pattern.

Try it yourself

Find the hidden cause

Follow a detective who refuses to jump to conclusions.

Match the pattern to its likely third factor

What could be behind both?

  • Towns with more hospitals have more sick people
  • Children with bigger shoe sizes read better
  • More umbrellas sold, more traffic jams
  • Days with more swimming pool visitors have more sunburn

A new case: breakfast and marks

Think it through on paper.

  1. A survey finds that learners who eat breakfast get higher marks. Write down two third factors that could explain this.
  2. Describe a fair test that could check whether breakfast itself helps.
  3. Write one sentence a newspaper could use that does not claim more than the data shows.

Think about it

  • Why can't the survey alone prove that breakfast causes better marks?

Common mistakes

  • Assuming that because two things rise together, one causes the other.
  • Forgetting to look for a third factor such as weather, age or size.
  • Using a pattern found by AI as proof of a cause.

Check yourself

Question 1

1. Ice cream sales and fan use both rise in summer. What best explains it?

Question 2

2. What is a third factor?

Question 3

3. Which is the strongest way to check a cause?

Question 4

4. New case: in one town, more ice cream is sold on days with more drowning accidents. What should the town council do?

Summary

  • Correlation means two things change together; it does not prove that one causes the other.
  • Look for a third factor, such as heat, age or size, that could affect both.
  • Compare like with like, or run a fair test, before claiming a cause, and treat AI patterns the same way.

Finished reading?