I became somewhat peripherally involved in the debate on long-term trends in global poverty that is raging these days in the WebSphere, prompted first by some very strong claims by Steven Pinker and Bill Gates, and then by an equally strong rebuttal by Jason Hickel.
Branko Milanović is an economist specialised in development and inequality.
Branko will be holding a talk in our series “Economics beyond the Swabian hausfrau” on 1 April 2019 in Berlin at 7 pm in the Monarch Bar, Skalitzer Str. 134, 10999 Berlin (U Bahn Kottbusser Tor). The subect will be “”Recent trends in global income distribution and their political Implications”
Cross-posted with kind permission from Branko Milanovic’s blog Global Inequality
Max Roser whose work is to be strongly commended became somewhat of a lateral casualty in this debate because it was his graph (based essentially on the work of Martin Ravallion and the World Bank) that Pinker and Gates quoted and Hickel disagreed with. Joe Hassel and Max Roser have now written a detailed explanation of the data they used to create the graph and how others (and themselves) have calculated global poverty over the long-term. So, to be clear, I find Martin’s and Max’s work absolutely crucial and indispensable for the understanding of long-term trends in this and a number of other dimensions. But I disagree with the excessively political and strident tone in which this work has been promoted by Steven Pinker (and more recently by Bill Gates) and their disregard of many caveats which come with some of these numbers.
As I mentioned in my tweet, there are (in my opinion) at least four such caveats that Hickel rightly brings out explicitly. (He uses other arguments too, but I do not comment on them.)
First, Maddison and Maddison-project data which are the only game in town for both global poverty and global inequality and which I use in my work too, do tend, like all GDP calculations, to overestimate increase in real income where there is a transition from activities that are not commodified to the same activities becoming commodified. GDP, as is well-known, is geared to measure mostly monetized activity. At the time of industrialization, as well as today during the ICT revolution, such underestimation is likely to be significant. It is odd that people today would question this—while we are going through a similar period of increased commodification and a rising share of activities that are now marketable but hitherto were not. Until Airbnb and Uber came along, your hosting friends of friends or driving them to the airport was not part of GDP. Nowadays, it is because you will be paid for such services. (The same is true for home activities that used to be performed without monetary compensation mostly by women—and which, at some point, become commercialized.)
Even more dramatic were, as Hickel points out, the changes that occurred during the Industrial Revolution. Many activities performed within households became monetized while people were often physically chased away from, or disposed of, land, water and other rights that they enjoyed for free. I do not need to go into too many examples there—just take the enclosures, or the land dispossession of Africans. This was not solely a transfer of wealth, but seriously reduced income of those who had the right to the usus fructus of land, water or other resources. Their reduced access to the actual goods and services was not recorded in any income statistics. It is thus reasonable to think that both GDP growth rates and the decline in poverty were overestimated.
Second, income distribution data for the 19th century that we all use come almost entirely from the 2002 seminal paper by Francois Bourguignon and Christian Morrisson (B-M). There are two more recent papers, one by van Zanden, Baten, Foldvari and van Leeuwen and one by myself that used a somewhat different methodology (that is, more diversified sources) in order to check the robustness of B-M findings. Both conclude that B-M results stand, but in both papers (van Zanden et al. and Milanovic) the number and/or reliability of these new sources are extremely limited. (I use social tables to estimate countries’ distributions in the 19th century. But the number of social tables that we have is very limited; both in terms of country and temporal coverage.)
Furthermore, Morrisson’s original distributions, while made available by the author, are unsourced. So, one cannot tell if they are right or wrong. In addition, even if individual country distributions were right, many of them are made to represent a vast variety of countries (say, Colombia, Peru and Venezuela; or Cote d’Ivoire, Ghana, and Kenya; or “45 Asian countries”; or “37 African countries” all have the same distributions’) because B-M divide the world into 33 “regions”, simply because information from most of countries is lacking.
Fragility of such distributions has a particularly strong effect on poverty numbers. It affects inequality somewhat less because, from other (fragmentary) sources we know what are the ranges within which inequality moves. But we know that less for poverty. The bottom line is that income distributions for the 19th century are, to put it mildly, fragile.
Third, Hickel questions the use of $PPP 1.90 absolute poverty line. There is a huge debate on this, and I will not enter it—but it suffices to look at the critiques made by Thomas Pogge and Sanjay Reddy (regarding the underestimation of the price level faced by the poor), large degree of arbitrariness with which the poverty line of, at first $PPP1, and now $PPP 1.90, was drawn (see e.g. Angus Deaton here), or more recently the methodological questioning of the World Bank approach by Bob Allen (here). Hickel simply mentions these issues. They are real and they should not be ignored.
Fourth, Hickel makes a more philosophical point that economists (unlike say, anthropologists or historians) are less well-equipped to deal with: human costs of the Industrial Revolution, from England to the forced labor (and probably ten million dead) in the Congo and Java, to Bengal famine (more than 10 million dead) to Soviet collectivization (more than 5 million dead) and China’s Great Leap Forward (about 20 million dead). The dead enter our calculations only to the extent that their deaths affect the estimated life expectancy. (And in the B-M paper there is an attempt to calculate global inequality over the past two centuries taking into account also changes in life expectancy). But, otherwise, so far as poverty calculations are concerned, the deaths have the perverse effect of reducing the population and increasing per capita output (since the marginal output of those who die as forced laborers or from famine is zero or close to zero). Jason is right to bring up this point.
The effect of this last point is ambiguous though. While it would—were we somehow able to account for it—increase the costs of industrialization and make the gains, compared to the pre-industrialization era, less, it would on the other hand improve the relative position of the present with respect to the era of industrialization—simply because such massive famines do not occur today, or occur less frequently (e.g. North Korea, and before that in Ethiopia).
To conclude. In my opinion, Jason Hickel had brought up several valid issues that most economists acknowledge as well (and have actually frequently written about). However, others, once a graph is created, tend to use the results less scrupulously or carefully in order to make political points. This is why bringing these issues to the fore is valuable and should be encouraged—and not shot down.