New research by Putterman and Weil based on a new large dataset of the world's 165 largest countries suggests that inequalities within countries is explained by the past histories of populations more than by ethno-linguistic differences and linguistic distance. In other words, it appears to suggest the impact of history on GDP flows through families, not nations. What matters is the history of the people who live in a country today, more than the history of the country itself.
Ancestors and incomes: More on the roots of world inequality, Louis Putterman David N. Weil, VOX EU
In our last Vox column, we described a new data set that allows us to identify what proportion of the ancestors of people who live today in the world’s 165 largest countries lived within the current borders of those same present-day countries in the year 1500.
Migration matrix and historical
A key finding from our research with the migration matrix concerns persistence in income differences. We find that the long historical persistence of capacity for generating or capturing income is considerably stronger than has been suggested by earlier studies – studies that made cruder corrections for the great population movements of the past half-millennium (or no correction at all). In this, our findings are in line with other studies that find centuries-long persistence in the impact of other growth determinants. Studies by Bockstette, Chanda and Putterman (2002), Hibbs and Olsson (2004) and Comin, Easterly and Gong (2007) indicate that the long-standing presence of state-level polities, the early advent of agriculture, and early advancement with respect to technologies ranging from writing to metallurgy, can explain an appreciable part of today’s global variation in country incomes.
We find similar results for state history and agricultural onset variables in our large cross-country sample, but with an important difference. We show that the impacts of these variables on today’s incomes are twice as large when correcting for the origin of nations’ current populations. That is, when we use the weighted average values of the countries of origin of today’s country populations instead of unadjusted country values. Our country-origin adjustment also allows us to explain three times more of the variance of country income than is possible with unadjusted figures.
In other words, it seems that the impact of history on GDP flows through families, not nations. What matters is the history of the people who live in a country today, more than the history of the country itself. More precisely, state history and year of agricultural transition explain under 10% of the variance in today’s country incomes; corrected for post-Columbian population shifts, either variable can explain about a quarter of that variance. Accounting for population origins in those countries having large non-indigenous population shares, like the countries of the Americas, Oceania, Taiwan and Singapore, provides a substantially more accurate window into the persistence of past advantages.
Within country income variation
If past history of populations has this much power to explain current country incomes, can it also help to explain the incomes of population sub-groups within countries? Do the different histories of people of African, Asian and European ancestry in the US, those of Indian and Polynesian ancestry in Fiji, or those of Chinese and Malayan ancestry in Singapore, help to explain socio-economic stratification within those countries? We explored the issue using both cross-country regression techniques and individual country case studies.
In our regression analysis, we calculated for each country the weighted average standard deviation (among its subgroups of different country origin) of both our state history indicator and our years since agricultural transition variable, and we entered each of these variables in a separate regression, along with the log of current income, to predict current within-country income inequality as measured by the Gini coefficient. Both standard deviation measures turn out to be highly significant predictors of the Gini, accounting (along with log of income) for about a third of inequality’s cross-country variation. In contrast, a conventional ethnic fractionalisation measure, due to Alesina et al. (2003), explains (also along with income) only 12% of the variance of countries’ inequality levels. We also tried using the cultural diversity measure of Fearon (2003), which is based on estimates of “linguistic distance” (roughly, degrees of separation on a tree representing the evolution of the world’s spoken languages), in some regressions. That measure turned out to explain (along with income) less than 10% of the cross-country variation in income inequality, compared to the 31% explained by the standard deviation of state history.
Perhaps the Alesina et al. and Fearon measures explain less of the variation in within-country inequality because the factors they account for are inherently less relevant to groups’ income levels than are the early development histories of populations’ ancestors. To give their measures a fair trial, we decided to explore how much they too might be improved by accounting for post-1500 migration.
We created a counterpart to Alesina et al.’s ethnic fractionalisation variable that measures the degree of fractionalisation among the ancestors of each country’s current population, rather than among countries’ current ethno-linguistic groups. The two sets of measures differ in their treatment of groups such as the mestizos or mulattos of Central and South America or Cape Verde. Such groups are treated as ethnically homogeneous by measures like Alesina et al.’s, but as having ethnically heterogeneous ancestry, in our counterpart measure. We did the same with Fearon’s cultural diversity measure, creating an indicator that looks at the linguistic distances between people’s ancestors rather than their linguistic distances today.
Our new measure, for instance, accounts for the very large linguistic distance between the Spanish and the Amerindian ancestors of a typical Spanish-speaking Mexican of today and for the smaller linguistic distance between the Spanish and the Italian ancestors of a typical Spanish-speaking Argentine of today, rather than treating such Mexicans and Argentines as being of zero linguistic distance since they now all speak Spanish.
It also works for the Alesina and Fearson measures
When we substituted the ancestry-amended counterparts of the Alesina et al. and Fearon measures for the originals, we found that the new variables are far more statistically significant predictors of inequality. The substitutions double or more than double the regressions’ explained variances. It seems remarkable and somewhat surprising that to bear in mind that Mexicans (to use that example one last time) have both Spanish and Amerindian ancestors even if those once separate lineages have long ago inter-mixed, doubles our ability to predict their country’s level of income inequality today.
These last two “successes” for the migration matrix do nothing to diminish the importance of our early development indicators, however. We added the standard deviation of state history back into the inequality regressions that include the original and the modified Alesina et al. and Fearon measures. The state history variable remained highly significant, whereas even the modified Alesina et al. and Fearon measures were no longer significant predictors of countries’ inequality levels once standard deviation of state history was included. Thus, the degree of difference in levels of early development as proxied by state history dominates both ethno-linguistic difference and linguistic distance as a predictor of inequality.
A natural explanation of the link between variation of early state history and inequality of income would be that descendants of people with high scores for early development such as Chinese and Italians tend to have higher incomes in whatever countries they are found in than do descendants of people with lower scores for early development such as Kikuyus and Polynesians. The bigger are the differences in early development, the bigger are the differences in incomes. But is there any direct evidence of this at the within-country level? We decided to investigate by searching for information on the identities, origins, and incomes of the main ethnic groups in the ten countries in our sample in which the weighted average standard deviation of state history is highest. We also looked at the US, an ethnically diverse country for which relatively good data exists, which ranks sixteenth in the sample. Once again, our findings support the presence of remarkable persistence of early advantages within human groups.
In Fiji, people of Indian origin tend to have higher incomes than indigenous Fijians, while the small percentage of the population having Chinese or European ancestry has the highest average income, the same ordering as for ancestral state history. In Guyana, the large population group of East Indian origin tend to have higher incomes than those of African origin whose incomes are in turn higher than those of Amerindian origin, again in line with state history rankings. In the five Latin American countries included, people of mainly European origin have higher average incomes than those of mixed European and Amerindian origin who in turn have higher incomes than those of mainly Amerindian origin. All in all, in nine of the eleven
countries considered, the socio-economic ordering of groups perfectly dovetails with that of their average state history values, and in the other two countries, the broad pattern is also consistent but with small exceptions (such as higher income for US Asians than Whites despite a slightly lower average state history for the former).
The role of slavery
A troubling aspect of the case studies is that slavery, indentured servitude, colonial regimes, and legal discrimination played large parts in shaping economic stratification in almost every country looked at.
It is difficult to know, for example, whether Indians would occupy a middle economic stratum between whites and blacks in South Africa but for the political variables at play. Whether human capabilities forged in these three populations during the millennia before they were thrown together on the canvas of South Africa would have led to a similar pattern of socio-economic differentiation without the outright exercise of power is something that we cannot tell from our analysis. But what we can conclude is that the world at the beginning of the 21st century displays remarkable
persistence of economic success for populations as they existed in 1500, and that this persistence is no less evident at the sub-national than at the national level.
Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat, and Romain Wacziarg, 2003. "Fractionalization," Journal of Economic Growth, vol. 8(2), pages 155-94, June.
Bockstette, Valerie, Areendam Chanda and Louis Putterman (2002). “States and Markets: The Advantage of an Early Start,” Journal of Economic Growth 7: 347-69.
Comin, Diego, William Easterly, and Erick Gong, “Was the Wealth of Nations determined in 1000 BC?” mimeo, 2006.
Fearon, James, D., “Ethnic Structure and Cultural Diversity by Country,” Journal of Economic Growth, 8:2, June 2003, 195-222.
Hibbs, Douglas A., and Ola Olsson, 2004, “Geography, biogeography, and why some countries are rich and others are poor,” Proceedings of the National Academy of Sciences 101: 3715-3720.
Putterman, Louis and David N. Weil, 2008, “Post-1500 Population Flows and the Long Run Determinants of Economic Growth and Inequality,” NBER Working Paper 14448.