Correlation is not causation
Two unrelated random series can track each other very closely. Press the button and find some.
Try it
Generate random data until two lines happen to match.
Two series of 40 random steps each, with no connection.
Simulated random walks. The search keeps drawing pairs until two of them correlate at 0.9 or more.
■ Series A ■ Series B
What is going on
A correlation can arise because A causes B, B causes A, a third thing causes both, or by pure coincidence. Series that trend over time are especially prone to coincidence, because anything that rises will correlate with anything else that rises.
When someone searches through many variables looking for a match, the best match found is almost guaranteed to look impressive, whatever the truth.
How to spot it
- A headline says “linked to” and then implies a cause.
- Two lines on one chart both rising over time.
- Ask: could a third factor drive both, and could the cause run the other way?
Chocolate and Nobel prizes
In 2012 Franz Messerli published a letter in the New England Journal of Medicine noting a strong correlation between a country’s chocolate consumption per head and its number of Nobel laureates per head. He suggested, tongue partly in cheek, that flavanols might boost cognition; the point most readers took away was how easily a correlation across countries suggests a story.
The series in the demo are simulated random walks with no relationship between them.