Discrepancies in International Data: An Application to China-Hong Kong Entrepot Trade

Before discussing discrepancies in international data, it is essential to talk about the sources of the data. The principal source for global trade statistics is the United Nations (UN). That agency relies on the trade statistics reported by its member countries. The obvious discrepancy in these reports is that what country A says it sends to country B is not the same as what country B says it receives from A. Such differences in what the exporting and importing countries report is not just a feature of developing countries: these discrepancies are pervasive in industrial countries, as well. For example, since 1989 the United States has abandoned its attempt to keep track of exports to Canada and now just relies on Canadian data on its imports from the United States (see Robert Feenstra, 1997 table 2). Prior to this action, the United States underreported its exports to Canada by more than 10 percent, as compared to the Canadian value. This illustrates the general principle that the information collected by the importer, which is often collecting tariff revenues and therefore has an incentive to record imports accurately, is usually viewed as more accurate than that collected by the exporter. There are, then, two problems with the UN data: it contains different information from the exporting and importing countries, and it is expensive, at least by the standards of lowbudget academic work. Progress on the first of these problems has been made by Statistics Canada, in the World Trade Database (WTDB). That agency obtains the UN trade data and then attempts to resolve the differences between the exporters and importers, providing a single number for the value of each disaggregate good sent from country A to B. These data are distributed on CD-ROM but are still fairly expensive by academic standards. To make progress on the second problem, the Statistics Canada data set for an early set of years (1970-1992) is distributed by the National Bureau of Economic Research (NBER) for a nominal fee, also in CD-ROM format (see Feenstra et al., 1997). In view of the widespread use of these data (distributed by the UN, Statistics Canada, or the NBER), it is worth asking how well the Statistics Canada database resolves the differing reports by the exporters and importers. To take a particularly important example, consider U.S. trade with China. In 1996, the United States reported a deficit with China of $39.5 billion, while the Chinese reported a value of the U.S. deficit of $10.5 billion; these two figures differ by $29 billion, or a factor of three times. At the heart of these differing values is the treatment of goods shipped through Hong Kong, which remains a separate entity from China for customs purposes. To what extent does Statistics Canada resolve the differences between these reported trade values, to arrive at some credible value of bilateral trade? Unfortunately, this issue is not resolved at all. The values given in the Statistics Canada WTDB for Chinese exports to the United States are very close to those reported by China, and are much less than those reported by the United States. This is contrary to the general principle, stated above, of treating the exporter's record of trade as less accurate than the importer's record. On the other hand, the U.S. official figures for imports from China are too high, because they include the * Feenstra: Department of Economnics, University of California, Davis, CA 95616, Haas School of Business, University of Califomia-Berkeley, and National Bureau of Economic Research; Hai: Department of Economics, Fort Lewis College, and China Centre for Economic Research, Beijing University; Woo and Yao: Department of Economics, University of Califomia-Davis. We thank Li Yan, Jin Hongman, and Jiang Xiaozhu from the Customs General Administration, People's Republic of China, for assisting with the study from which the results reported here are drawn.