ON THE (IN)ACCURACY OF ECONOMIC OBSERVATIONS - AN ASSESSMENT OF TRENDS IN THE RELIABILITY OF INTERNATIONAL-TRADE STATISTICS

Previous investigations which evaluated the accuracy of trade statistics were generally based on the partner country approach, i.e., a given country's reported exports (imports) were compared with a partner's reported imports (exports) with the objective of identifying major value differences and the reasons for these discrepancies. The present study employs a new methodology that focuses on the internal consistency of trade statistics. Four key questions are addressed: (1) are the data reported under different classification schemes (like SITC Revision 1 and 2) consistent at points where there are perfect correspondences between product groups: (2) are trade totals consistent with the sum of values reported for lower level products; (3) is total trade consistent with the sum of trade reported with individual partner countries; and (4) are trade values compiled by one reporting organization (United Nations) consistent with those of an alternative (IMF) agency. By performing these consistency tests annually from the early 1960s this study is able to assess trends in the reliability of trade data. The tests results should be extremely disquieting for data users. Differences of 100 percent (or more) are observed in some countries' (supposedly directly comparable) SITC Revision 1 and 2 trade totals with the average discrepancy for all developing countries' exports exceeding 60 percent in the early 1980s. Second, total trade figures are frequently found to be inconsistent with trade data for component products (i.e., the sum of trade in (say) three- or four-digit SITC products is often considerably more, or less, than a country's reported trade) and such discrepancies persist in some country's data for more than 20 years. Third, there has been a clear deterioration in reporting the origin or destination of trade since 1980 as over 30 percent of developing countries' imports or exports are not properly allocated. Fourth, different international agencies sometimes report trade values for specific countries that differ by 100 percent or more. Such discrepancies appear especially important for the least developed countries, but there is evidence that the differences for all countries widened over the last decade.