Lesson 6 of 7 · 16 min
Association and causation
Explain why an association between two variables, categorical or numerical, does not by itself show cause, and identify coincidence or another variable as possible explanations.
What you will be able to do
- Explain that an observed association does not necessarily mean a causal relationship.
- Identify a confounding variable that could explain an association, and explain how.
- Recognise when an association could be coincidence.
- Describe what kind of study could establish cause.
Over a Gold Coast summer, the days with the highest ice-cream sales are also the days with the most surf rescues. The association is strong and positive. Should the council ban ice-cream to keep swimmers safe?
Of course not. Both variables go up on hot days: more people buy ice-cream, and more people go swimming. The heat drives both. Ice-cream has nothing to do with the rescues.
A confounding variable (hot weather) is associated with both variables, producing an association between them without one causing the other.
Key idea
Association does not imply causation
An observed association between two variables — numerical (a scatterplot, ) or categorical (a two-way table) — does not necessarily mean that one causes the other. However strong it is, an association on its own cannot show cause.
A causal relationship means that changing one variable actually produces a change in the other.
Across 30 Queensland towns, the number of pharmacies and the number of births each year have a strong positive association, . Which conclusion is valid?
When you find an association, there are several possible explanations. The syllabus asks you to identify and communicate two non-causal ones.
| Explanation | What it means | Example |
|---|---|---|
| Confounding variable (the influence of another variable) | A third variable is associated with both variables and produces the association. | Shoe size and reading ability in primary-school children are both driven by age. |
| Coincidence | The association has arisen by chance, with no real link. Likely with small data sets, or when many pairs of variables are compared. | Over ten years, a football team's goals and the number of rainy days in Perth. |
| Causation | A change in one variable produces a change in the other. | Hours of sunshine and energy from solar panels. |
Worked example
Explain an association with a confounding variable
At a Townsville primary school, the shoe size and reading score of 60 students from Years 1 to 6 have a strong positive association, . A parent suggests that buying children bigger shoes would improve their reading. Explain why this conclusion is not justified, and give a possible non-causal explanation for the association.
Step 1: Acknowledge the association
There is a strong positive linear association (): students with larger shoe sizes tend to have higher reading scores.
Step 2: State the principle
An association does not necessarily mean that one variable causes the other.
Step 3: Identify a confounding variable
Age (or year level). Older students have bigger feet, and older students have had more years of reading practice, so they read better.
A good non-causal explanation names a variable and says how it is linked to both variables in the association.
Step 4: Conclude
Age explains the association, so changing shoe size would not change reading scores. The parent's conclusion is not justified.
Try it yourself first, then check the working one step at a time.
Answer
Association does not imply causation. Age is a confounding variable: older children have both bigger feet and better reading, which produces the association.
Method
Describe the association and quote the statistic.
State that association does not necessarily imply causation.
Name a specific confounding variable and explain its link to both variables.
Watch out for
Saying "there could be another variable" without naming and explaining one.
Tip
Time can be the other variable
Two quantities that both grow over the years — the number of Queenslanders with smartphones and the number of cafés in Brisbane, say — will be strongly associated. Population growth and changing lifestyles over time drive both. Be suspicious of any association between two things measured year by year.
For a sample of Queensland bushfires, the number of fire trucks sent and the area burnt have a strong positive association. Which is the most plausible non-causal explanation?
Worked example
Association and causation in a two-way table
A survey of 400 Queensland adults asked whether they own a dog and whether they exercise at least three times a week.
| Exercise 3+ times a week | Don't | Total | |
|---|---|---|---|
| Own a dog | 93 | 57 | 150 |
| Don't own a dog | 103 | 147 | 250 |
| Total | 196 | 204 | 400 |
A pet-food company claims: "Owning a dog makes you exercise more." Use percentages to describe the association, then evaluate the claim.
Step 1: Percentage within the explanatory variable
Dog ownership is explanatory, so percentage the rows: of dog owners exercise regularly, compared with of non-owners.
Step 2: Describe the association
The percentages differ substantially (62% versus 41.2%), so there is an association: dog owners are more likely to exercise regularly.
Step 3: Association is not causation
This is a survey, not an experiment, so it shows association only.
Step 4: Give non-causal explanations
Another variable could explain it. People who are already active are more likely to choose to own a dog; people with houses and yards (or more free time) may be more likely both to own a dog and to exercise.
Step 5: Evaluate the claim
The data supports an association but not the claim that owning a dog causes more exercise. That would need a study that controls for these other variables, such as an experiment.
Try it yourself first, then check the working one step at a time.
Answer
62% of dog owners exercise regularly versus 41.2% of non-owners, so there is an association. But the survey can't show cause: active people may be more likely to get a dog, and other variables (housing, free time) may affect both. The claim is not justified by this data.
Method
Calculate percentages within each category of the explanatory variable.
Compare them to describe the association.
State that association does not imply causation and give specific non-causal explanations.
Conclude on the claim.
Watch out for
Comparing counts (93 versus 103) instead of percentages.
Accepting the claim because the difference in percentages is large.
Note
How do you establish cause?
With a controlled experiment. Randomly allocate people (or plots, or days) to groups, change only the explanatory variable, and keep everything else the same. Random allocation balances confounding variables across the groups, so a difference in the response can be attributed to the explanatory variable. Data that is just observed, like most of the data in this topic, can show association but not cause.
Common mistake
Naming a confounder without explaining it
"It could be another variable" earns little. Name a specific variable and explain how it is linked to both variables in the association: "Hot weather — on hotter days more people buy ice-cream and more people swim, so there are more rescues."
Over the years 2014 to 2023, the number of goals a Brisbane soccer team scored each season and the number of rainy days in Perth each year have . What is the most likely explanation?
Exam tip
Answering a causation question
- Acknowledge the association (quote or the percentages).
- State that association does not necessarily imply causation.
- Give a specific non-causal explanation in context — a named confounding variable and how it affects both, or coincidence.
- If asked, say how cause could be established: a controlled experiment.
Summary
An association — categorical or numerical, weak or strong — does not necessarily mean one variable causes the other.
A confounding variable is associated with both variables and can produce the association.
Coincidence can produce an association by chance, especially with few data points.
Name the specific variable and explain how it affects both.
Cause is established by a controlled experiment, not by observed data alone.
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.