We propose a methodology for gravity models which overcomes weakness of institutional distance measures identified in international business literature (Shenkar, 2001). Our methodology is based on combination of the distance approach with cluster analysis of states according to their institutional level. The methodology reflects the critiques concerning symmetry, linearity and "discordance illusion" of traditional distance specification of institutional variables within gravity models and tries to bridge the gap between those two streams of literature. We also use the methodology to test the "middlemen hypothesis" theorising that intermediate corrupt countries can mediate the trade flows between low and high corrupt groups because they possess skills to succeed on both markets. To present the methodology and examine the hypothesis we estimate micro-founded augmented gravity model for bilateral exports of 131 countries within period 1995-2013. The model is estimated also on disaggregated data. The results confirm that the proposed methodology overcomes discussed weaknesses especially when we estimate the model on disaggregated data because we find significant heterogeneity between sectors. We conclude that when we are interested in causal claims about the impacts of institutions on trade we should abandon traditional distance specifications. However, we reject the middlemen hypothesis.
gravity model corruption institutional distance middlemen hypothesis