Correlation/cause confusion

"The arts have their value in society. You look at the Royal Albert Hall on Proms night. How many of those people are going to mug old ladies on the way home? Not many – they’ve got more important things to worry about."
Prunella Scales, Big Issue in the North , 1996

It is indeed highly unlikely that anyone returning from a classical music concert is going to mug someone along the way. But such a person is also probably more likely than the average member of the population to embezzle funds from their company the day after. Are we then to conclude that listening to Mozart will make you less likely to mug someone but more likely to fiddle the books?

The conclusion would seem at best premature and at worst simply absurd. The problem is a straightforward confusion of correlations and causes.

For example, Britons who regularly eat plantains are also more likely to regularly eat sweet potatoes. Does that mean eating one causes people to eat the other? Of course not.

The explanation of the correlation is a third factor: plantains and sweet potatoes are both staples of the Caribbean diet. Coming from a family of Caribbean origin is thus the causal factor which explains the correlation between consumption of both foodstuffs.

Similarly, it remains true that audiences at classical music concerts tend to be middle class, a social group not prone to mugging but with a virtual monopoly on corporate fraud. It is far more likely that these broader facts about social position do more to explain the lack of muggers in prom audiences than any morally improving quality in the music.

Even if, as a matter of fact, the music does affect morality, the important point is that the mere correlation of reduced criminality and listening to orchestras does not show that it does.

Nonetheless, is the existence of a correlation evidence that there is some causal story to be told that links the two, as is the case with both my examples? Often there is such a story to be told, but little is to be gained by telling it. The setting of the sun may explain both the closing of a flower and the locking of the park gates by the keeper, but the two effects are still caused by two very different mechanisms and have no deep connection.

Similarly, the 9 o’clock train leaves at the same time as the 9 o’clock radio news bulletin starts. (Well, perhaps not on Britain ‘s creaking railways.) But to say that both are caused by it being 9 o’clock is surely an error: times are just the wrong kind of thing to be causes. The mere sharing of a common causal factor does not provide a causal link between two such events.

Leaping from correlation to cause does seem to come naturally to us, however, perhaps because, as David Hume argued , ultimately regular correlations of a certain sort are the only evidence that there is such a thing as causation at all. (Many claim he went further and argued that causation was just a form of exceptionless correlation.) But even Hume would agree that not just any correlation points to a cause.

News pages are full of reports of correlations where it is implied that there is some causal link. Consider, for example, what causal links might be assumed from the following findings and what other explanations are possible. The links take you to a report of each of the findings. No rash assumptions are made (tabloid papers almost certainly would not have been as careful in their reporting) but nor is the full range of plausible explanations always considered. I would expect that Butterflies and Wheels readers would be more likely than others to work out what these areā€¦

Comments are closed.