What correlation measures

Correlation studies whether two variables move together, in which direction, and how strongly.

• Positive correlation: the variables move in the same direction. When income rises, consumption rises; on hotter days, more ice-cream is sold.
• Negative correlation: they move in opposite directions. When the price of apples rises, people buy fewer apples.

But moving together does not prove that one causes the other. Two kinds of trap:

• Coincidence: some variables move together purely by chance. The book's examples: the arrival of migratory birds and the birth rate in a place, or a child's shoe size and pocket money. Neither causes the other.
• A third variable: two variables may both depend on something else. Ice-cream sales and deaths by drowning both rise in summer. Ice-cream does not cause drowning; hot weather increases both (more people eat ice-cream, and more people go swimming).

So correlation measures covariation (how the variables vary together), not causation.