The economics book that has launched a thousand blog posts, Thomas Piketty’s Capital in the Twenty-First Country, tells a grand story of inequality past and present. One would expect that a book on global inequality would have much to say about development. However, the book has limited relevance for the developing world, and the empirical data he marshals for developing countries is weak.
Piketty’s central story is that convergence in the developed world and slower population growth will leave us with a permanently modest economic growth rate (g). Coupled with a constant return to wealth (r), concentration of capital ownership, and high rates of savings among the wealthy, the low g leads to rising wealth inequality over a longish run—something like the second half of the 20th century.
A low-g future for the developed world is a mostly uncontroversial assumption. (He assumes future GDP per capita growth of 1.2 percent for the U.S.) But Piketty draws conclusions for the world as a whole, and we are a long way from global convergence. As Branko Milanovic noted in his review , catch-up growth could fend off Piketty’s inequality dystopia for some time.
Piketty projects a worldwide growth slowdown after 2050, and his technical appendix includes the region-specific projected growth rates of output per capita behind this calculation. Here is a portion of those figures:
Here is the path of output per capita under these growth rates:
These projections are detailed here . Under these assumptions, India and sub-Saharan Africa will see a drop in growth rates to 2.0% after 2050, when their income per person is about one-tenth that in the U.S, and then a further drop to 1.5% in 2070. Piketty is effectively assuming that convergence will never take place for India and SSA, which will constitute half the world population in 2100, according to UN projections.
Piketty says “historical experience suggests that the principal mechanism for convergence at the international level as well as the domestic level is the diffusion of knowledge.” Why then, would not convergence continue longer than he projects? A sustained high g (and thus a lower likelihood that r>g) would make Piketty’s thesis of less immediate relevance for developing countries.
Piketty might say that, nonetheless, the same process will kick in eventually, whether in 2050 or 2150. Fair enough, but the uncertainty in any long-run prediction explodes to dizzying levels as we stretch out the horizon from a few decades to more than a century. Many events good and bad—catastrophic climate change, labor-augmenting technological innovation, alien invasion, etc.—could come along to disrupt the dynamics Piketty foresees.
Apart from provocative prediction, Capital’s other main contribution is its impressive massing of income and wealth data. A central finding is the U-shape pattern of income inequality in the United States, Britain, Canada, and Australia. Inequality fell from its stratospheric heights of the 1930s in the post-war period and then rose again starting around 1980. Similar patterns hold for Japan and Continental Europe, but with a less pronounced recent rise.
For developing countries, Piketty harvests the fruit of his World Top Incomes Database  (WTID), built up using tax records from countries around the world. Here’s a version of the book’s Figure 9.9 , created using the data files on Piketty’s website .
Although Piketty does not draw this conclusion, the graph seems to say that the same U-shaped evolution is taking place in the developing world. (This was indeed the interpretation of the New Yorker’s John Cassidy .) Particularly striking is the apparent evidence that inequality has risen dramatically since 2000. This is different from the story told by household survey data , which shows a decline in within-country inequality on average since 2000.
Surveys typically understate high-end incomes, because of underreporting and poor coverage of wealthier households, so survey data could mislead. However, the evidence of a widespread upturn in inequality in developing countries is weaker than the graph suggests:
- Much of the recent apparent upward trend for India is based on the post-2000 change, but the last value in WTID is from 1999. In the technical appendix , Piketty writes ”the data for India and China for the recent period have to be taken very carefully” and explains that to get the 2011 India data point, he used “an average estimate of different ‘executive pay’ - type surveys.”
- The figure shows an upward spike in the top 1 percent’s share in Indonesia 2004-2010. But the last value in the WTID for Indonesia is from 2004, and Piketty gives no other source for the 2010 value. The relevant table in the technical appendix (sheet TS9.5 of this file ) refers to a separate “Details” sheet which is not in the file.
- Likewise, the most recent China value is not found in WTID. Piketty says the China data are especially weak as “they are based on data from surveys and severely underestimate the highest incomes.” He writes, “I thus add a 40% increase in these series, which seems plausible given the under-estimation of the Chinese income tax revenues.” Comparing the book’s figures to the WTID series, I found  that he actually scaled up the China figures by 50%, not 40%. As I understand the source paper , these are actually estimates for the top 1 percent of the urban population, over a period when the country’s urban population share roughly doubled.
- The apparent decades-long climb in the top 1 percent’s share in Argentina between 1973 and 1997 is the product of connecting two estimates far apart in time and thus of less certain comparability.
In this chart the only individual country with a clear U pattern is South Africa, and the data is too sparse to indicate a general rise in inequality since 2000.
This is all to say that we are still ignorant about trends in the high income share in developing countries. We definitely need more creative approaches to understand the evolution of inequality in developing countries. What we have so far from tax records is not enough to tell a substantially different trend story from what we see in household surveys.
- inequality 
- Global Economy 
- South Asia 
- Middle East and North Africa 
- Latin America & Caribbean 
- Europe and Central Asia 
- East Asia and Pacific 
- Africa 
- United States 
- United Kingdom 
- South Africa 
- Indonesia 
- India 
- Colombia 
- China 
- Canada 
- Australia 
- Argentina