A classic example

November 24, 2000 5:17 PM

A classic example of Post Hoc Fallacy. Explanation inside.
posted by Steven Den Beste (9 comments total)

In the village are two clock towers. When the clock on the Methodist Church points to 3:00, the clock on the Baptist Church rings three times. Thus the Methodist church secretly controls the Baptist church clock.

Of course not. There's a correlation, and it's real. (Absent mechanical breakdown, the correlation is 1.0.) But both events are being caused by something else (the time of day); they aren't directly related to each other. Take one away and the other would continue to act just as it does.

So it is here. The last paragraph of the article gives away an alternative explanation:

"Since mentally ill people are over-represented in the lower socioeconomic groups, higher cigarette taxation could be an effective way of encouraging them to quit, she thinks."

Actually, consider it this way: There's a negative correlation between socio-economic class and smoking. (Not stated in the article, but it happens to be true. The less affluent you are, the more likely you are to smoke.)

There's a negative correlation between socio-economic class and mental illness. (Seems unlikely, but that's what they say. But assume it's true.) Then it's not unlikely that there should be a positive correlation between smoking and mental illness.

But that proves no relation between them; since each could be caused by the correlation with lower socioeconomic class. I would also expect a correlation with poor nutrition and with lower class housing, with less education and with ownership of older and cheaper cars, for exactly the same reason. The correlation would be real, but there would be no direct cause and effect.

The article spends a lot of time trying to figure out why the two are related, when it's not clear that they are related at all. Unfortunately, a great deal of modern epidemiology is infected with this flaw.
posted by Steven Den Beste at 5:29 PM on November 24, 2000

More fun with bad correlation studies, courtesy of mathmistakes.com.

posted by waxpancake at 6:25 PM on November 24, 2000

It's not so much "news", as the use of correlation for spin. We had a similar post hoc argument in the UK a few years back, when statistics suggested that the majority of a selection of crimes in London (ranging from assault to theft) were committed by young black men. It led to a hubbub amongst the police hierarchy and race relations group in the capital, but behind the tabloid spin that "young black men steal purses" lay a deeper truth: that that demographic is most afflicted by unemployment, social deprivation and other key indexes associated with crime.

Meaning that it can sometimes be rhetorically useful to point out that a perceived social problem afflicts a "deprived" group, to suggest the insidiousness of that problem. So the scientific argument is fallacious to the extent that it gets presented in the form "cigarette manufacturers target the mentally ill", but is effective to the extent that it isolates one of the many unfavourable correlations between socioeconomic status and smoking.
posted by holgate at 6:51 PM on November 24, 2000

I think this is just propoganda from the anti-smoking health Nazis, but I linked to it anyway!Someone print me up a t-shirt - "I'm mentally ill and I smoke tobacco!"
posted by Mr. skullhead at 7:59 PM on November 24, 2000

isn't it also true that older people are more likely both to smoke and have mental health problems?

posted by andrew cooke at 5:24 AM on November 25, 2000

I read somewhere that most smokers also consume bread, as well as dihydrogen monoxide. It's insidious...
posted by beth at 6:15 AM on November 25, 2000

So if you are mentally ill and don't smoke you are screwed, because we have no cure for you. If only had you smoked we could know why you are sick.
posted by john at 12:39 PM on November 25, 2000

Doesn't anyone find the numbers a bit suspicious? If the mentally ill constitute 44 percent of the market for cigarettes and they are only twice as likely to smoke as the general population, then either mentally ill individuals tend to smoke a huge amount of tobacco or a large percentage of americans are mentally ill. How broad is the definition of mental illness?
posted by rdr at 1:09 PM on November 25, 2000

rdr, if you're going to include the stressed-out and depressed, as this article suggested they were, you can get some pretty big percentages for mental illness at any given moment in time.
posted by dhartung at 5:05 PM on November 25, 2000