The Colonial Origins of Comparative Development: A Solution to the Debate on Settler Mortality Rates

Raphael Anton Auer



JEL classification
N10, O11, O57, P16, P51

Comparative Development, Growth, Institutions, Colonialism, Property Rights, Mortality


I address David Albouy's (2006) critique of the data constructed by Daron Acemoglu, Simon Johnson and James Robinson (2001). The contribution of this paper is to instrument for settler mortality rates that are collected from historical sources - and that may be measured with error - with a geographic model of the determinants of disease. I first establish that my instruments are significant predictors of mortality and are otherwise excludable to institutions. Among other things, the excludability is established by a falsification exercise, in which I document that the geographic potential for mortality strongly affected institutions in former colonies, yet it had no effect on institutions in the rest of the world. This differential effect settler mortality had on development can only be rationalized by the early institution building hypothesis that Acemoglu et al. argue for. I next repeat the analysis of Acemoglu et al. instrumenting for the historical mortality rate with its geographic projection. The instrumented mortality rate is a highly significant predictor of institutional quality. Moreover, this result is true when instrumenting for either the original data or the revised mortality series of Albouy. This result is also true when accounting for the population that the historical data was sampled from. Turning to the instrumental variable estimations, I show that also the relation between institutions and income is highly significant and that the associated importance of institutions for international income differences is substantial. Again this finding is true when using either of the two historical series and also when accounting for the population that the historical data was sampled from. I thus conclude that the empirical results presented in Acemoglu et al. indeed reflect their early institution building hypothesis rather than measurement error.