This study provides a comprehensive investigation of statistical discrepancies in generally accepted international trade figures published by the IMF between 1948 and 1994. We calculated export over- and underestimation for each country and all of their bilateral trading partners. By keeping totals for over- and underestimation in separate categories we avoided the cancellation effect of aggregating positive and negative discrepancies among partner countries. In general, the results show a significant improvement in the quality of trade data over time. However for many countries, relatively large discrepancies still exist that defy technical explanations, such as the CIF-FOB margins. Also, because export over- and underestimation coexist for most of the countries at varying degrees, use of the aggregate sum of the discrepancy might disguise the actual magnitude of the problem. A significant difference exists in the relative magnitude and dispersion of trade discrepancies between OECD countries and non-OECD countries. Trend analysis suggests that the accuracy of trade data is improving at a faster rate in the non-OECD than in the OECD countries.