Friday, September 15, 2006

Correlation and Causality

They just don't seem to get it. Just because events A and B occur together does not mean that A causes B. This Yahoo story claims that alcohol use helps boost income. And what do they cite as evidence? The fact that alcohol drinkers earn more than teetotalers.

Did it ever occur to them that the causality -- if any -- could run in the reverse direction? Or that there is no direct causality at all but merely common drivers like education, opportunity and social strata and norms?

What would be interesting is if they were to study the set of heavy drinkers and segment them. How much would poor undergrads and celebrity has-beens distort the picture? I wonder.

(The story is based on a published paper in a journal. I don't know if the paper has stronger evidence but somehow I doubt it when the words "libertarian thinktank" and "contradicted....Harvard School of Public Health" show up in the article.)

Update: The paper (pdf link) does a better job than I gave it credit for, controlling for age and religion among other things. But its methodology would still appear to be flawed. It seems to think that the result is implied if social drinkers earn more than people who drink alone!