Rebasing Maddison: relative prices and the shape of long run global development
Jutta Bolt, Robert Inklaar, Herman de Jong and Jan Luiten van Zanden
The aim of this paper is twofold: first we present an updated version of the Maddison project database which includes new historical income estimates for 10 countries. Second, we discuss various methods for reconstructing the shape of long-run global development. We compare the Maddison dataset (making use of PPPs from 1990) with two alternative datasets, one based on the new 2011 ICP round and one based on a multiple benchmarks. We develop a new approach to test the accuracy (or the impact of the cumulative biases) of these three ways for reconstructing global development. We start by collecting independent historical benchmarks of countries’ relative GDP levels from the literature. Next we compare the relative performance of countries during the 19th century and early 20th century as given by the alternative datasets based on the different benchmarks approaches, to the relative performance of countries as indicated by the newly created dataset of independent benchmarks. This way we aim to see which dataset (benchmark) ‘predicts’ these independent relative output levels more accurately. Further, we analyse how the indirect approach as suggested by Leandro Prados de la Escosura (2000) to estimate relative incomes ‘perform’ in this context. Finally, we explore how often these alternatives ‘predict’ highly implausible outcomes, such as GDP levels below subsistence, and to what extend these alternatives result in really different stories of global development. We find that although the alternative benchmark approaches entail a level shift in GDP, the differences between approaches is not extreme and does not rewrite long run comparative global development for the far majority of the countries included. Additionally, the biases in historical income series resulting from the extrapolation method, i.e. from using fixed contemporary prices for converting historical series into one common currency, turn out to be far less severe than often suggested. The alternative benchmark approaches employed in this paper result in comparative income levels relatively close to the independent benchmark estimates available in the literature.