Reconciling the GTAP Data Base: Where are the Big Changes?

Inaccuracies in the collection and compiling of data mean that data supplied by one country is rarely consistent with data supplied by another. Even within the same country, the same data collected from two alternative sources could vary due to differences in how the data is being collected, interpreted, classified, and valued, notwithstanding the differences caused by simple errors and omissions. Global economic analysis however, requires consistent and reconciled global data, and this reconciliation process is laden with judgments about the quality of the alternative data sets being reconciled. In the case of the GTAP Data Base (Narayanan, Aguiar et al. 2012) the externally collected trade, macro, protection and energy data supplied by international sources are considered superior to individual country data because it has typically gone through a standardization and balancing process. The extent to which this reconciliation process alters the country data however is not known. In this paper, we hope to shed some light on the extent to which the country data changes as a result of the GTAP construction process. In particular, we are interested in identifying where the largest changes occur as a result of this reconciliation process. Do changes mostly occur in particular countries where data are poorer; or do changes mostly occur in particular data across all countries. The answers to these questions can be used to help ascertain where resources might best be utilized to further improve the quality of global data. We find that there is some evidence that data from developing countries with weaker IO tables and less sectors do undergo more changes than those with more robust IO tables, the largest differences however occur in the sales shares due to differences between the trade data in the IO tables and the balanced trade dataset used in the GTAP database.