Friday, January 30, 2009

Influence Of Government Size On Income Distribution

I have been reading Philip A. Klein's Economics Confronts the Economy (Edward Elgar, 2006). He presents data in two tables that I thought, when combined, suggest relationships. So I drew the regression lines below. The data on income distribution is from 1998. Total government expenditures as a percentage of GDP is from 1999. Government expenditures include, for example state and local spending in the United States. The graphed data are for the United States, and, in order of decreasing percentage of government spending as a percentage of Gross Domestic Product, Sweden, Denmark, Belgium, Finland, France, Austria, Italy, Netherlands, Norway, Canada, Germany, New Zealand, the United Kingdom, Spain, Ireland, Australia, and Japan. Apparently, in advanced industrial democracies, as government spending directs a greater percentage of resources, the poorest have relatively more income and the richest have relatively less. One can see why the richest would like lackeys to fight the latter tendency.

Figure 1: Share of Income in Lowest 10% Among Developed Economies


Figure 2: Share of Income in Highest 10% Among Developed Economies

I haven't finished Klein's book, but I think I'll note two points I find of interest. He argues that when economists advocate for positive analysis of existing economies, they implicitly accept the status quo. This is a value judgement that could be contested. Secondly, when economists limit normative theory to Pareto-improving recommendations, they restrict themselves from commenting on such matters as income distribution and the quality of life of the vast population. (Moving along one of the regression lines probably makes some worse off.)

7 comments:

Luis Enrique said...

Apologies if this is a stupid question - but does your govt size and income data include transfers or not? That is, does a government show up as 'large' on your plot, if it's taking money via taxation (most likely progressive taxation) and giving it to the poor as income support? If so, it's hardly a surprise that the greater extent to which governments redistribute income, the relatively better off the poor are.

Or are you trying to suggest that there's something about the structure of economies with small government that, even net of transfers, the poor are worse off? I can well believe it (there are all sorts of correlates of government size that you'd expect to influence income distribution, net of transfers), but it's a different sort of claim - can you clarify?

Luis Enrique said...

umm, confusing myself - I think I mean "before transfers", in second para above, not "net of transfers"

Anonymous said...

@Luis: It could be argued, that most governent expenditure represents(paid by progressive tax or not) a wealth transfer, because it provides a good "for free" which in turn obsoletes a private expenditure for the same good.

Assume for example, a government would introduce free health care for everyone, rendering private health insurance unnecessary. If health care has a high utility, then it follows that your wage now generates less utility than before. You will contract your work supply or demand higher wages to compensate for the lower utility against the rigour of work. So I'd expect wages to rise somehow, by an amount correlated to the price of the abolished insurance. This should favour lower incomes and hence improve income equality.

Just an idea..
- Marco

YouNotSneaky! said...

As Luis says it does seem important whether this is post tax/transfer or pre. My guess you can get the same sign of the regression coefficient for both but one (pre tax/transfer) is gonna be weaker than the other.

Also causality matters, particularly the influence of omitted variables. (This is also why I'm pretty sure the relationship will hold for both pre and post tax/transfer). Basically for the OECD the level of income is positively correlated with measures of inequality (Gini, but probably shares too). Richer countries are more unequal. Also, for the OECD, the share of government in income is negatively correlated with the level of income.

So you take the (-) correlation between G/Y and y, and the (+) correlation between y and inequality and you put'em together and you get a negative correlation between inequality and government share.

But, at least for the pre tax/transfer data this would not imply that government (or its share) has any kind of casual impact on inequality. It could be that there's just a structural - independent of government - relationship between inequality and average incomes and also that government consumption lowers incomes and then the relationship documented in this post just falls out as an aside.

Robert Vienneau said...

Klein's data on income distribution is for selected countries from the World Development Report 2000/2001, Table 5. The World Bank, in turn, gets this data for high income countries from the Luxembourg Income Study database. In the technical notes, the WDR mentions these are percentiles of population, not households, and are based on surveys of households. Since no clarification is given, would you not think that this is distribution after government transfers?

Klein's data on total government expenditures from the OECD.

I don't know that I believe YouNotSneaky's facts. They sound like the dubious stuff some like to tell themselves. I would expect, based on my offhand knowledge of the postwar golden age, for countries with more equally distributed income to grow faster and, eventually, to be richer. In fact, I can check this with these countries.

YouNotSneaky! said...

Then check. Income per capita vs. pre-transfer Gini. If I'm wrong then I'll happily admit that I'm wrong.

(I think I actually did a post on it way way way back when)

YouNotSneaky! said...

Also a bit weird is that when I do the regression for share of lowest 10% vs. government share, using data from here (CIA factbook data):

http://www.nationmaster.com/graph/eco_sha_of_hou_inc_low_10-share-household-income-lowest-10

I do get a positive relationship but with an Rsqr of .07 (slope coefficient of .0382) with a measly t stat of about 1, which means it's really zero.

The difference seems to be in the government shares (which I got from PWT) which are much lower than the ones you have. Still, there's a big big difference between an Rsqr of .07 and of .59 (and quite honestly an Rsqr of .59 for this type of relationship is suspiciously high)