Sunday, February 09, 2014

Booze and bloody murder

Does the demon drink make people more likely to kill each other? The answers are ambiguous to say the least.

I looked up tables of adult (age 15+) annual alcohol consumption per capita here, and intentional homicides per 100,000 population here. Sifting out countries where data was not available under both headings, I was left with 184 nations for the purposes of statistical correlation.

The range runs from 1.0 (perfect correlation, so that as one figure increases so does the other, in every case) to -1.0 (perfectly negative correlation, so that as one increases the other reduces).

Overall, the correlation between the two factors for this list of countries is, surprisingly, -0.1. That is, virtually no connection at all.

What if we are more selective in our survey?

Boozer countries

If we look at the top 30 countries by alcohol consumption, ranging from Moldova's 18.22 litres of pure acohol down to Spain's 11.62 litres (UK: 13.37 litres), there is a significantly negative correlation with homicide: -0.29.

Yet when we narrow down further to the top 10 toping nations, there appears to be a positive correlation: 0.48. Having said that, within those ten countries the level of alcohol intake is pretty similar: 18.22 to 15.11 litres; whereas the murder rate varies widely, from 0.7 to 7.5. Perhaps all this shows is that with too small a sample you get erratic results.

Killer countries

The list of the 30 most homicidal countries starts with Honduras (91.6 murders per 100,000) and finishes with the Democratic Republic of the Congo (21.7 murders per 100,000). The annual alcohol intake ranges from Guinea's 0.36 litres of pure alcohol up to Uganda's 11.93 litres. In this violent subset of countries, the relationship between drink and killing is pretty much random: 0.09.

But narrowing down further to the top 10 most homicidal nations, starting with Honduras again but finishing at the Bahamas (36.6 murders per 100,000), we see that their range of alcohol consumption(3.61 - 9.43 litres) is about the same as the variance in the murder rate (i.e. max = about 2.5 * min), yet the two factors are negatively correlated: -0.39. (Too small a sample, again?)


The conclusion is that there is no definite conclusion, though we can suspect that other (perhaps political-economic and social) factors may have a more direct influence on the propensity to kill, than the average quantity of alcohol consumed.

If you had a "session" last night, you can at least endure your thick head today with a fairly clear conscience on that score.

If you want to know a bit more, here are the four lists, followed by the long list of 184 countries. Compare the UK with the USA, for example!

All original material is copyright of its author. Fair use permitted. Contact via comment. Unless indicated otherwise, all internet links accessed at time of writing. Nothing here should be taken as personal advice, financial or otherwise. No liability is accepted for third-party content, whether incorporated in or linked to this blog; or for unintentional error and inaccuracy. The blog author may have, or intend to change, a personal position in any stock or other kind of investment mentioned.

1 comment:

Paddington said...

This explains why things such as correlation cannot be done legitimately once one has looked at the data. At that point, one has to go Bayesian.