Thursday, July 05, 2012

Alcohol and suicide

Using WHO data, I calculated the cross-national correlation between annual alcohol consumption and suicide rates. For 101 countries, it is .58--a strong assocation. Countries with high levels of consumption have much more suicide. Eastern Europe illustrates this well. The average person there drinks roughly 15 liters of (pure) alcohol per year, and suicide rates are two or three times the average.

Such a finding supports the socially conservative view that drinking immoderately (or at all) needs to be discouraged.

7 comments:

Josh Wexler said...

"Such a finding supports the socially conservative view that drinking immoderately (or at all) needs to be discouraged."

Weakly, I'd think. This is a Pearson correlation, right? You seem pretty satisfied that this says something strong about the causal arrow, but it's just a correlation. It's not too hard to think of covariables that would increase both rates of alcohol consumption and suicide, even if the two have a very weak casual relationship in the general population. The heritability of Depression is something like .5. Would it surprise you if the group of people who are depressives had both greater suicidal ideation and a tendency to drink a lot?

You could respond that drinking should be discouraged because it will likely further heighten depressives' suicide rates. But:

1. Your correlation doesn't really evidence that.

2. The magnitude of the effect might be relatively small (though it wouldn't surprise me to learn otherwise)

3. This is no reason for haranguing the rest of the population who is just living it up.

4. Besides the pleasure that drinking brings to the drunk, I think it's safe to say we'd be culturally poorer if not for the writing and music and art that was produced under the influence.

5. Speaking as a depressive, I wouldn't consider my suicide to be of much, if any, negative value. If one reaches the point where life is too discomfiting, suicide is a blessed relief to be counted on the putative side of the ledger. Nanny me, if you must, about the evils of a ramos gin fizz, but I'll be damned if you're going to tell me I have to live against my will.

Really enjoy the blog. -Wexler

Black Death said...

"Such a finding supports the socially conservative view that drinking immoderately (or at all) needs to be discouraged."

....

The data do not support this conclusion. First of all, the data show correlation, not causation. It could be that alcohol abuse is an epiphenomenon in persons who commit suicide, i.e., those who are going to kill themselves are also prone to drink too much, but the two things aren't related. I think that's probably not the case, but the data allow no definite conclusions in this regard.

Secondly, the data do not address the degree of individual alcohol consumption as related to suicide. In order to show that moderate drinking was causally related to the incidence of suicide, some sort of dose-effect study would be necessary. It could be (and probably is) that suicide is much more common in alcoholics than among temperate drinkers, but, again, the data are lacking. Of course, one could always make the argument that all alcoholics start out as moderate drinkers, but, once again, we do not know that the alcohol abuse is not an epiphenomenon. Anyway, once we tried to deal with the serious problem of alcohol abuse by punishing all the drinkers. It was called Prohibition. Don't think it worked out to well.

Anonymous said...

"The data do not support this conclusion."

Of course they support it. They lend support to the idea. True, they don't demonstrate it beyond a reasonable doubt, but they provide good reason for suspecting that a causal relationship may exist. For a blog post, that's pretty good work.

Cennbeorc

Anonymous said...

Heh. I'd bet you'd find a lot more positive correlations between alcohol consumption and various variables...

Also, not sure about not drinking at all being very socially conservative - seems restricted to weird US religious conservatives, rather than global normal conservatives.

Mihoda said...

It seems that these single-regressor estimates leave a lot of room for omitted variable bias.

Have you run multivariate models?

Mihoda said...

It seems that these single-regressor estimates leave a lot of room for omitted variable bias.

Have you run multivariate models?

S said...

Black Death is correct; I'll say it more strongly: this post is idiotic.

The correlation exists because alcohol numbs emotional pain -- but the author's conclusion is precisely analogous to blaming pain killers for headaches.