Despite massive surveys, tax records, etc., there are often spirited debates about how unequal incomes and wealth are in modern societies. Knowing whether we are more or less equal than pre-industrial societies, with these sparse data, is a much tougher problem. Milanovic et al estimate Gini coefficients for 30 ancient societies (from Rome to pre-independence India) using “social tables”, which are roughly lists of the standard of living of various classes. Since most pre-industrial societies are dominated by class divisions, inter-class inequality is their predominant form of inequality. By assuming that the richest member of one class is exactly as rich as the poorest member of the next highest class, the authors can provide a rough bound on Ginis even allowing for that limited amount of inter-class inequality. This data generally shows that modern and ancient societies had roughly similar Ginis.
That alone would make for a fairly boring paper. The real contribution here is to note that Ginis are not a very good measure of inequality for poor societies. The literature on measuring inequality is actually really interesting. I went through Cowell’s inequality book a few years back for a project at the Fed; that book has an exhaustive description of the benefits and drawbacks of Ginis, percentile ratios, Atkinson indices, etc. I seem to recall that back in the 70s there were a bunch of impossibility theorems for inequality indices satisfying various properties, but I can’t recall a citation for these.
The basic problem is that if we want to use “inequality” as a proxy for “highly unequal power and exploitation between groups”, then Ginis will by definition be low for poor countries, no matter how exploitative. Consider a society with total income 1100, 1001 members, one of whom is in the top class, and the rest of whom who live at a minimum subsistence level ($1 per year). The wealthiest member will have $100 – that is, he will literally extract every dollar beyond that which the laborers need to stay alive. The Gini coefficient will be very low for such a society, however. An alternative definition of inequality is an extraction ratio: define the subsistence level of income, calculate the maximum Gini coefficient if all but the top class of people get subsistence income and the top class get the rest, then calculate the ratio of the actual Gini to its maximum.
The extraction ratios of pre-industrial societies are quite a bit higher than those of modern states. Developed countries today have extraction ratios of 30-50%, while poorer countries are slightly higher (56 in Brazil, 61 in South Africa, etc.). The lowest extraction ratio in the pre-industrial sample is 55%, and the average is 75%. This result comes even though there is basically no difference in Gini coefficients between modern and preindustrial societies. Such a theoretical apparatus can help avoid some paradoxes. For instance, the Engerman-Sokoloff thesis, that Latin America’s low level of development today is a result of institutions derived from high inequality in the colonial era, appears to be contradicted by the near-identical Ginis of 18th century Europe and Latin America. But since Latin America was poorer than Northwest Europe in that era, near-identical Ginis mean Latin America had a higher extraction ratio, and therefore were in some sense more exploitative. Interesting.
http://www.santafe.edu/media/workingpapers/09-07-022.pdf (WP version – Final version, with more readable slides, in the December 2010 Economic Journal)