Foreign aid, poor data, and the fragility of macroeconomic inference

The link between foreign aid and economic growth remains a controversial issue in the literature, and a large share of the disagreement could be explained by differences in the data employed. Using GDP data from three different versions of the Penn World Table and the World Development Indicators, I investigate the robustness of Juselius, Møller and Tarp (2014)'s (JMT) conclusions about long-run aid effectiveness. The analysis is carried out in two stages. First, I apply the same models as developed by JMT to the new datasets. Second, I re-specify the Cointegrated VAR models using the same criteria as JMT, but limit the analysis to the four most and least consistent countries respectively. The first exercise shows that results change in a significant manner in approximately 10 of the 36 countries examined. The second exercise shows that the if the models are re-specified for each country and dataset individually as a function of the data, this leads to more qualitative changes in the conclusions.