Figuring out various income inequalities: what can they tell us?

The recently released World Bank data on national accounts (GDP per capita, national consumption, CPIs, exports and imports etc.) give us an opportunity to update calculations of international inequality. Things are not always very obvious and I will have to explain some methodological choices too. Using the classification that I introduced in my “Worlds Apart” (2005), I will review three types of international and global inequalities.

Let’s start with the simplest one. Take all countries’ GDPs per capita (all expressed in mutually and over-time comparable 2005 international dollars –based on 2011 International Comparison Project) and calculate Gini across them. (Ignore whether they are small or populous countries.) The number of countries varies as there are more countries in the world today than in 1952 when our series begins, so best would be to look at the red line only after 1980. After that date the number of countries is about constant, and today’s independent countries (say, Ukraine, Macedonia, Slovakia, Eritrea) are considered as separate countries even then, when they were parts of larger wholes. (USSR, Yugoslavia etc. produced the data for their constituent units like the US produces GDP data for its states).

What do we see in the red line? Increasing Gini between mid-1980s and 2000, implying a divergence of country mean incomes, and then, after the turn of the century, a convergence. The main reasons for the convergence are faster growth rates in Africa, Asia and Latin America than in rich countries (especially after 2007). (People who studied growth economics can easily recognize here the story of unconditional convergence or divergence calculated using Gini rather than a regression.)

So that would be the end of the story if each country had the same population. But obviously they do not. For world inequality, convergence of China and Chad, India and Israel, are not equivalent. If Chinese GDP per capita converges to that of the rich countries, this is obviously going to matter more. To see how much look at the blue line where we use countries’ GDPs per capita as before but weigh them now by populations. The striking thing is that after 1978, precisely when Chinese growth picks up, the blue line starts to go down. At first slowly and then more and more precipitously.  After about 2000, Gini seems to be in a free fall. When we tease out the data a little bit more, we find that up to 2000, the entire “job” of inequality reduction was done by China. Without China, the blue line would have gone up. But after 2000, even if we drop China out, the population-weighted inequality goes down: it is driven down by the fast growth of India (and also Indonesia, Vietnam etc.). This is why we now have two big engines of international income reduction: China and India. However, as China becomes richer it may not play that role for very long. Today, China’s GDP per capita is almost exactly equal to the world’s average, but India’s is at less than ½ of world average. So in the near future, India’s growth rate compared to the world would be of paramount importance.

This would be the end of the story if within each country all citizens had the same income (i.e., there were no within-national inequalities). Note that the blue line implicitly assumes this: that all Chinese have the mean income of China, all Americans the mean income of the United States etc. Obviously, this is not true. Moreover we know that income inequality in most countries has gone up. When we try to look at inequality across all citizens of the world (global inequality shown by the green dots) we leave the world of national accounts because that world cannot give us data on individual incomes and move to the world of household surveys. (From US national accounts, I can learn that the average value of output produced in the US annually is $53,000. But I have no idea what is the income of the top 1% or of the bottom decile.  For that I need to move to the world of household surveys.)

It should be obvious that the green dots must lie above the blue line. It should also be obvious that the more important the within-national inequality, the greater would be the gap between the green dots and the blue line. This is indeed what we see: global inequality, calculated using the same international dollars, starts by being around Gini of 0.7, and then, pulled down by the blue line, begin its downward slide ending at around Gini of 0.63.

Movements in global inequality today reflect two forces: a big one of Asia’s convergence that brings mean incomes of China, India, Vietnam, Indonesia etc.  closer to the rich world, and a smaller, but important, force of within-national widening of inequalities. If in the future Asian convergence ceases, and within–national inequalities go up, the green dots will move back towards Gini of 0.7. But if alternatively convergence continues, expands to Africa, and within-national inequalities cease to grow, the green dots will keep on going down.

This would be the of the story were it not for another problem. Our household surveys tend to underestimate top incomes (the proverbial top 1%). And the question can then be asked: perhaps we underestimate the height of the green dots by not accounting fully for the super-rich. Christoph Lakner andI have tried, the best we could, to adjust for that, assuming country-specific underestimates, and raising, in a Pareto fashion, incomes of the top 10%. That’s how we got the high orange dots.  

Now the situation is not as splendid as before: global inequality is down compared to the 1980s, but the decrease is more modest: around 3.5 Gini points rather than almost 7. Could the adjustment overturn the entire decrease and produce an increase in global inequality? It is possible but unlikely—because the strength of Asian convergence is so big, and affects so many people (almost 3 billion) that even very strong increase in within-national inequalities cannot fully offset this gain. But we need more and better data to say that with certainty, and indeed one of the topics that several people are working on right now is how to combine household survey data (that are very good for 90-95% of national distributions)  with fiscal data that are better for top 1%, and possibly even top 5%.

So, stay tuned!