“Correlation is cause” is a fallacy, a false argument. In fact, it is one of the most common errors in human reasoning. Correlation is used by science and misused by pseudoscience, racism and the news.
A correlation is when two measurements seem to go together. The more of A there is, the more there is of B. For example, in the 1970s some in the US said that violence on television made society more violent: people were watching more violent television (measure A) at the same time that the crime rate (measure B) was going up.
A correlation can be negative: the less there is of A, the more there is of B. For example, some people say that lack of money causes crime: the lower the average income (A) in a neighbourhood, the higher its crime rate (B).
The correlation between A and B can mean one of at least six things:
- A causes B – what you suspected all along turns out to be true.
- B causes A – it could be the other way round. Keeping with the television example, it could be that television is just “a mirror of society”. As US society became more violent, so did its television shows.
- A and B cause each other – television violence makes society more violent, which in turn makes its television shows more violent, which in turn makes ….
- C causes A and B – crime and television violence may have been caused by a third thing, like the number of men with military training after the Second World War, the Korean War and the Vietnam War. There might be other causes as well, like high levels of gun ownership.
- A is B – television violence and the crime rate might be measures of the same thing: US violence, or maybe something we do not have a word for yet.
- There is no causation – Few people still say that violence on television makes the US more violent, in part because the crime rate dropped in the 1990s and 2000s while the amount of violence on television did not. The correlation in the 1970s was spurious or accidental.
All of this assumes that A and B are properly measured and are what they seem to be. The crime rate, for example, might just be over policing or under policing. Or, an IQ test might be culturally biased. Etc.
In most cases correlations turn out to be accidental.
Correlations can point to important discoveries, but on their own they prove little.
“Correlation is cause” is a mark of pseudoscience. Science moves from facts to conclusions. Pseudoscience moves from conclusions to facts. If you believe something, it is easy enough to find facts that support it. That is how stereotypes work. It is called confirmation bias. And since the world has tons of accidental correlations, it is easy enough to find one to support your beliefs – and overlook those that do not.
– Abagond, 2015.
- White racist guide to Black pathologies