Knowledge spillovers : on the impact of genetic distance and data revisions

This paper assesses the robustness of the empirical results in Ertur and Koch (2007), who develop a model, which accounts for technological interdependence among countries through spatial externalities. The original version models interdependence via an interaction matrix based on geographic distance. In contrast, in this paper, data on genetic distance, defined as the time since two populations have shared a common ancestor, from Spolaore and Wacziarg (2009) is used to construct the interaction matrix. It is found that, whereas in the original model indirect spillovers from capital investment were insignificant, using genetic distance, these spillovers now have a significant effect on steady-state income per worker. However, the version of the model with an interaction matrix based on genetic distance implies an implausibly large capital share of income. In addition, the model is subjected to a further series of robustness checks. The original version relies on data from Penn World Table (PWT) Version 6.1. More recent versions are currently available, and the data has been extensively revised (Johnson et al., 2013). It is shown that results are in general not robust across different versions of the PWT. Furthermore, the estimation results are highly sensitive both to the measure used to model interaction between countries (genetic or geographic distance) and to the specific functional form on which the weights in the interaction matrix are based.