Long-Run Determinants of Economic Growth: Putterman and Weil Revisited

How strongly is ‘deep’ history related to present GDP per capita, and how ‘deep’ are we talking? In a number of papers, Louis Putterman and collaborators claim that the relationship is strong and that several thousands of years of history are relevant. I replicate some of the work by Putterman and Weil (2010, Quarterly Journal of Economics), who found a “surprisingly” strong linear relationship between migration-adjusted state history during the first fifteen centuries of the Common Era and present income. I present four main empirical findings. First, by restoring dozens of countries mistakenly dropped from their regressions, I find the linear relationship is somewhat weaker than they reported. Second, I find that weighting the data to address the endogeneity of present-day borders—a concern raised but not addressed by Borcan, Olsson, and Putterman (2018)—causes the linear relationship to become markedly weaker or even nonexistent. Third, I conduct robustness checks that are suggested or implied but not pursued in Putterman’s work, generally finding a weaker linear relationship or no linear relationship. For example, there is no linear relationship when grouping countries into 11 world regions defined by Putterman and Weil themselves, nor is there a linear relationship when looking at sub-Saharan African countries alone. For another example, the linear relationship is weak when using GDP per capita for the year 1960 rather than the year 2000. Fourth, I show that had their data been correct, Putterman and Weil (2010) would have found a statistically significant quadratic relationship between their 1,500 years of state history and present income had they tested for it, which refutes the finding by Borcan, Olsson, and Putterman (2018) that a quadratic relationship only exists when using additional thousands of years of history. I then ask by what standard any of these results, whether in Putterman and Weil’s original presentation or herein revisited, can be deemed surprising; I propose one standard by which I find the results to be unsurprising. I close by calling attention to some problems with the “World Migration Matrix” data set constructed by Putterman and Weil (2010) and used in numerous studies.