Math 1 · Unit 10 · Lesson 5 · part 3 of 5

Lurking variables, and how cause is really shown

Coincidence is one reason.

Lurking variables, and how cause is really shown

Coincidence is one reason. There is a second, sneakier one: a third quantity that nobody graphed is driving both. It is called a lurking variable, a hidden third quantity that affects both of the ones you are looking at.

Larry and the cars

Larry reads that people who own more cars tend to live longer. He concludes: owning more cars makes you live longer, so he should buy another car.

The association is real, but the cars are not doing it. A lurking variable, like having more money, can lead to both: more cars, and also better food, housing, and health care.

Ice cream and sunburns

On days when more ice cream is sold, more people get sunburned.

Ice cream does not burn skin, and sunburns do not make people buy ice cream. The lurking variable is hot, sunny weather, which causes both.

Shoe size and reading

Among children, bigger shoe sizes go with higher reading scores.

Big feet do not teach you to read. The lurking variable is age: older children have bigger feet and have had more years of school.

So when two things move together, there are several possibilities to think through before saying one causes the other.

  1. 1Is it coincidence?A pattern that happened by chance, like the spelling words and the spiders.
  2. 2Is there a lurking variable?Something hidden that drives both, like weather, age, or money.
  3. 3Could it run the other way?Maybe the second thing causes the first.
  4. 4Is there a real reason?If there is a clear, sensible way one thing makes the other happen, causation is at least reasonable.
How do you actually show that one thing causes another?With a carefully designed experiment: a planned test in which people (or plants, or rats) are put into groups at random, one group gets the thing being tested and the other does not, and everything else is kept the same. Random groups spread the lurking variables evenly, so any difference at the end is caused by the one thing you changed. A strong r from data you just happened to collect cannot do that.

Where to go next