Drinking To Prosperity and Freedom [Updated]

The sound, taste, and look of freedom and prosperity.
A Bavarian barmaid serves beer during Oktoberfest.
Source.

There is absolutely no reason in the world why I should be asking about the relationship between alcohol consumption and prosperity, human rights, and democracy.

On the other hand, there's data! Thanks to the WTO, it's actually possible to do the sort of social science that horrifies PTJ but which involves numbers and statistics. Ergo, it's science.

So, I start with a simple question. Why do countries vary in their consumption of alcohol? We might think that there are two competing theories. The first is that alcohol consumption is largely culturally determined; the second is that alcohol consumption is a normal good, subject to the sort of demand and supply curves that affect all other goods.

From the first, we might deduce further that more politically liberal countries will see higher alcohol consumption; similarly, countries with more conservative religious traditions--which I operationalize, largely on the back of the envelope, as percent Muslim--will drink less. From the second, we deduce that richer countries will drink more.

(There will be no broad conclusions here, as I largely just wanted to have some experience merging datasets and playing around with some packages in Stata and R.)

We start with exploratory graphs. [UPDATE: I added a second plot, which I think is a little nicer.]

Plot shows alcohol consumption plotted against per-capita GDP.
Alcohol consumption statistics come from the World Health Organization.
Polity Score and Percent Muslim from the Quality of Government dataset.

Plot shows alcohol consumption plotted against per-capita GDP.
Polity Score and Region from the Quality of Government dataset.

The overall relationship seems clear. The richer countries are, the more likely they are to consume high amounts of alcohol. By contrast, the more Muslim a country is, the less likely it is to consume alcohol. (Note Nigeria at 10L/year and Bahrain at about 3.5L/year.) Note also that there is a pretty big cultural element here

But enough with these descriptives! Let's try some blatantly irresponsible modeling. We use OLS throughout.

	(1)	(2)	(3)	(4)
	AlcoholLitres	AlcoholLitres	AlcoholLitres	AlcoholLitres
Religion: Percent Muslim	-0.0474***	-0.0521***	-0.0535***	-0.0523***
	(0.00631)	(0.00798)	(0.00797)	(0.00594)
Revised Combined Polity Score	0.113*	0.0318
	(0.0464)	(0.0560)
Real GDP per capita (Constant Prices: Chain series)	0.000121***	0.000133***	0.000120***	0.000113***
	(0.0000268)	(0.0000390)	(0.0000320)	(0.0000233)
Gini (mean)		-0.0914**	-0.0884**
		(0.0302)	(0.0283)
Democracy			0.292	1.104
			(0.617)	(0.562)
Constant	4.534***	8.789***	8.799***	4.522***
	(0.466)	(1.630)	(1.550)	(0.486)
Observations	150	135	143	173
Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001

What we see here is obvious support for the idea that alcohol consumption is responsive to wealth and to cultural factors. Note, however, that Polity scores (higher scores = more democratic) and Democracy (a dichotomous score created by Jim Vreeland et al) have no effect once we begin controlling for economic inequality. Interestingly, the relationship between a country's Gini (income inequality score) means that the more unequal a country, the less likely it is to consume.

	(1)	(2)	(3)	(4)
	AlcoholLitres	AlcoholLitres	AlcoholLitres	AlcoholLitres
Religion:  Percent Muslim	-0.0429***	-0.0410***	-0.0516***	-0.0513***
	(0.00747)	(0.00625)	(0.00836)	(0.00785)
Democracy	0.362	-0.366	0.125
	(0.683)	(0.742)	(0.735)
Real GDP per capita (Constant Prices: Chain series)	0.000142**	0.0000773***	0.0000612
	(0.0000499)	(0.0000222)	(0.0000341)
Average Schooling Years (Total)	0.161
	(0.183)
Personal Autonomy and Individual Rights		0.357***
		(0.102)
Human Development Index			5.403*	5.141*
			(2.245)	(2.089)
Gini index (inequality measure)			-0.0779**	-0.0752**
			(0.0281)	(0.0277)
GDP/Capita PPP in Constant USD				0.0000785*
				(0.0000343)
Constant	2.894**	2.045*	5.405**	5.463**
	(1.100)	(0.817)	(2.042)	(2.034)
Observations	98	173	122	122
Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 We now proceed to different models. The broad takeaway here is that the more developed and the more equal a country's income distribution (controlling for other factors), the more alcohol it tends to consume. This is just exploratory--I'm pretty much just screwing around with models--but I'm interested in this Gini index finding, which was unexpected. Well, one of the obvious implications of the second chart is that there are some fairly strong regional effects going on. So let's adjust our models to account for the fact that Europe is full of a bunch of rich, egalitarian, atheist and Christian beer and wine swillers.

	(1)	(2)	(3)
	AlcoholLitres	AlcoholLitres	AlcoholLitres
Gini index (inequality measure)	0.00316	0.000111	0.0660
	(0.926)	(0.997)	(0.167)
GDP/Capita PPP in Constant USD	0.000122**	0.0000849*	0.0000714
	(0.007)	(0.047)	(0.074)
Religion: Muslim	-0.0481***	-0.0427***	-0.0494**
	(0.000)	(0.000)	(0.002)
1b. Eastern Europe	0	0	0
	(.)	(.)	(.)
2. Latin America	-4.950***	-4.702***	-7.588***
	(0.000)	(0.000)	(0.000)
3. MENA	-4.921***	-5.251***	-6.581***
	(0.000)	(0.000)	(0.000)
4.Sub-Saharan Africa	-4.377***	-3.447**	-5.338***
	(0.000)	(0.008)	(0.000)
5.Western Europe/North America	-2.431*	-1.856	-3.175**
	(0.045)	(0.137)	(0.009)
6. East Asia	-4.116**	-3.571*	-4.059*
	(0.004)	(0.048)	(0.043)
7. South-East Asia	-6.325***	-5.813***	-9.148***
	(0.000)	(0.000)	(0.000)
8. South Asia	-7.067***	-7.009***	-7.178***
	(0.000)	(0.000)	(0.000)
9. The Pacific	-7.850**	-5.883***
	(0.004)	(0.001)
10. The Caribbean	-2.207	-2.024	-5.826***
	(0.117)	(0.292)	(0.000)
Improved Water Source (% of Population)		0.0382+
		(0.083)
Happiness			5.219+
			(0.079)
Constant	8.911***	5.694*	4.351
	(0.000)	(0.012)	(0.110)
Observations	122	115	83
p-values in parentheses + p < 0.10,* p < 0.05, ** p < 0.01, *** p < 0.001

With region controls, the Gini finding disappears. On the other hand, we can play around with more fun covariates--such as access to improved water source, which we suspect is linked to GDP per capita but not strictly. In contrast to our expectation, the more improved water a country has, the more it consumes alcohol. (Clearly people in countries with unimproved water are getting their potable liquids from sources unexpected in the eighteenth century.) On the other hand, the happier a country, the more likely it is to drink.