32. In Maddison’s footsteps (24-4-2020)
In Maddison’s footsteps (24-4-2020)
Written by: Jan Luiten van Zanden
Jutta Bolt and I finished a new, 2020 update of the Maddison project yesterday, after more than 3 years of hard work. It is, as you may know, a true ‘collaboratory’, in which 15 to 20 scholars work together to continue the great work by Maddison to quantify the evolution of the world economy in the past 2000 years. The goal is to publish a dataset of GDP, GDP per capita and population estimates covering all countries in the world, going back in time as much as is possible – in a few cases Maddison even included estimates of the Roman Empire. It is a highly delicate operation, because 20 specialists have different, sometimes conflicting ideas about the optimal strategy, and there are always tensions between the consistency of the entire system, and what individual scholars think and believe has happened to their country. They have often invested heavily in their series, and their approach, which is often the best that can be done for the country and period concerned. One scholar, who was confronted with an alternative, somewhat corrected series for his country, mailed that we were ‘robbing’ the history of his country from him. That was the right moment to reconsider.
One of the debates within the group is between economists, who prefer systematic, theoretically grounded solutions to, for example, the problem of deflation or the linking of various benchmarks, and the economic historians who have strong ideas about what is plausible and consistent with the historical evidence. To illustrate this, I have to tell you bit more about the purpose of this update. The Maddison dataset basically consists of an anchor, a year for which we have good estimates about the relative levels of GDP, and a large set of country series, sometimes going back to the Middle Ages, that are linked to this anchor. Maddison had created an anchor for 1990, which has however been criticized in the past. The WorldBank every so many years creates a new anchor (an ICP round resulting in PPP estimates). There was consensus that the 2005 anchor had not been good enough, but the 2011 anchor was considered state of the art. So the most important aim of the update was to relocate the anchor from 1990 to 2011 (and also to include the many new historical series that had been estimated by economic historians since the previous, 2014 update). But when we started to use the 2011 anchor, odd things happened. Linking the anchor with the historical time series resulted in estimates that economic historians could not accept. For example, Iraq became, in 1914, the richest country in the world, and many other oil producing countries became almost as wealthy. At the other extreme, a country like Peru had, if we were to believe these results, a GDP per capita which was during long periods below the ‘subsistence level’; this may happen occasionally – in times of civil war or famine – but no country can survive below this minimum for long (as Michalis can confirm).
The explanation of these strange results is interesting, we think (but we do not do a lot with this in the working paper). The fact is that Iraq, and other oil producing countries, have become rich without much economic growth – they simply profited from the rising price of oil. And when you do not grow a lot, and the anchor is in 2011, you become, paradoxically, increasingly wealthy going back in time (relative to other countries which do grow). The opposite happened with Peru (and many other relatively poor countries, but Peru is such a good example because thanks to painstaking research by the Peruvian economic historian Bruno Seminario there are really good series for this country available). Peru did grow, but did not achieve a comparable increase in GDP per capita, most likely because it saw the relative prices of its output and exports decline. And rapid growth, in combination with a 2011 anchor, in this case leads to below subsistence levels before 1950.
This outcome of this comparison is on the one hand a statistical problem, making it more difficult to create a consistent Maddison-type dataset, but also sheds light on the dynamics of global inequality. In the 1960s and 1970s similar changes in relative prices were studied by the members of the Dependencia-school in economics. One of them, Raúl Prebisch, argued that developing countries specializing on raw materials, were confronted by long term declines of their relative prices, as they imported industrial goods from developed countries of which prices were less affected by business cycles and maintained at a high level by monopolistic firms. What we found was somewhat different but does point to a comparable hidden redistribution of income in the world economy. It were the oil producing countries that profited most from trends in relative prices (until recently, I should perhaps add, because oil prices are even negative these days!), and oil-importing, raw materials exporting countries that fared worst.
So moving the anchor to 2011 did not improve the quality of historical estimates – at least that is how the economic historians saw it. We moved back to Maddison’s good-old 1990 anchor, after developing a number of tests to find out more systematically which anchor and anchoring-approach was doing best. In a way, Maddison’s approach was quite accurate, we can now confirm, and more accurate than the alternative approaches. We also think we managed to keep the group together, although not everybody is happy with all estimates. We were even thinking about organizing a workshop to celebrate the 10th anniversary of the Maddison-project (in Madrid), until a virus intervened.
Continue reading: The VOC and the Elephant (26-4-2020)
The website of the Maddison-project: