The analysis of bilateral trade flows features prominently in empirical research in international economics. Various different international statistical sources are available for researchers and commonly used. Unfortunately, the data happen to differ quite substantially across the different sources. It is the task of this project to identify those differences, quantify them, and track their origin and to demonstrate the consequences of differences in the data for estimation of fundamental relationships such as the gravity equation. We find the largest discrepancies in a comparison of UN and OECD databases to the IMF and Eurostat trade data. In the most extreme cases the differences to reported trade flows in other data sources amount to as much as 40 billion $ in measured export flows and to as much as 50 billion $ in bilateral "mirrored imports". Most importantly we find that these differences carry over to econometric results in applications of the gravity model, one of the workhorses of empirical trade research. Parameters of key variables such as log bilateral distance, common borders, common language, or a colonial relationship dummy variable vary substantially and do not even have a stable sign when using one database versus the other. Hence, heterogeneous reporting standards across data sources and the inhomogeneous sample coverage have a non-trivial impact on the quantifications of trade costs in empirical research.
bilateral trade flows statistical discrepancies data sources gravity model