THE U.S.-CHINA BILATERAL TRADE BALANCE：It's Size and Determinants

　　by

　　Robert C. Feenstra

　　Dept. of Economics, Univ. of California, Davis

　　Haas School of Business, Univ. of California, Berkeley

　　and

　　National Bureau of Economic Research

　　Wen Hai

　　Dept of Economics, Fort Lewis College

　　and

　　China Centre for Economic Research, Beijing University

　　Wing T. Woo

　　Dept. of Economics, Univ. of California, Davis

　　Shunli Yao

　　Dept. of Economics, Univ. of California, Davis

　　May 1998

　　Paper presented at the UNDP-HIID Conference on China's Integrationotnithe Global Economy" January 15, 1998. We thank Ms. LI Yan, Ms. JIN Hongman and Ms. JIANG Xiaozhu from the Customs General Administration, People's Republic of China, for assisting with this study; and Jeffrey Sachs for his helpful comments when this project started.

　　Introduction

　　It is quite incredible that while the negotiations of China's accession to the World Trade Organisation (WTO) are greatly influenced by the deficit that the United States runs in its trade with China, the actual size of the US-China bilateral trade deficit is not actually known! The US puts the 1995 bilateral trade deficit to be $34 billion, while China puts it at $9 billion. If the US figure is correct, then China has the second highest bilateral deficit, after Japan whose bilateral trade deficit with the US is $59 billion. But if China's figure is correct, then the China bilateral trade deficit is lower than that the US bilateral trade deficits with Canada, Mexico, Germany and Taiwan.

　　Some analysts have interpreted the large US-China bilateral trade deficit as prima facie evidence of unacceptably high levels of protectionism in China, and have advocated stringent entry conditions for China's admissionotniWTO, even though China is in the poorest third of the world's economies. In response, supporters for easier entry conditions for China have emphasised other factors (e.g. the movement of low-skill, labor-intensive manufacturing industries to China from neighboring economies) for the recent widening of the bilateral trade deficit.

　　Because the overall trade balance of a country equals domestic saving minus domestic investment, the normal expectation is that low-income countries that have high rates of return to investments (e.g. China) should be borrowing from abroad to finance their development. However, since 1994, China has been running an overall trade surplus that is growing over time. This counter-intuitive phenomenon of a low-income China that is extending loans to the outside world has strengthened the impression of a mercantilist China intent focussed on export promotion. This impression has its origin in the great extent that China has increased its penetration of the US market over the last decade. Of the 34 categories of manufactured goods imported by the US, China was among the top 5 suppliers in 9 categories in 1995, up from 1 category in 1990 and none in 1985.

　　Our focus on the US-China bilateral trade balance is of course based on political economy considerations. In 1988, the U.S. Congress passed Super-301 legislation to combat "unfair" trading practices by foreign countries. Because the imposition of the retaliatory measures permitted by Super-301 can seriously rupture international political and economic relations, it is important that the criteria upon which retaliatory actions are made are accurately quantified. We focus on the US-China bilateral trade deficit because the bilateral trade deficit has been cited many times by U.S. politicians and by the U.S. Trade Representative as an important indicator of the existence of unfair trading practices. The large discrepancy between the U.S. and the Chinese estimates of the bilateral trade deficit doubtlessly makes decision-making about Super-301 actions more inexact.

　　This paper has two aims. The first is to reduce the range within which the true bilateral trade deficit lies. The second aim is to identify the determinants of the bilateral trade deficit, and offer an assessment of their relative importance. To anticipate our discussion, section 1 presents a smaller range of values for the bilateral trade deficit than in previous studies. We are able to achieve this because we devised a new estimation method that takes advantage of our access to detailed Chinese Customs data at the commodity-by-commodity level. The size of the discrepancy between our revised US figure and the revised Chinese figure is only 4 to 26 percent of the size of the discrepancy between the two official figures. The size of the average discrepancy between the revised figures is only 9 percent of the size of the average discrepancy between the official figures. For example, the revised US-China bilateral trade deficit is $15 billion to $20 billion in 1994, and $16 billion to $22 billion in 1995, compared to the official range of $8 billion to $30 billion, and $9 billion to $34 billion, respectively.

　　Section 2 of this paper points out that protectionism can influence the overall trade balance only if it can modify saving-investment behavior in the economy, and this requires wealth effects to be uncharacteristically strong and unusually fast to appear. The major cause of China's current account surplus is its high household saving rate. The high saving rate is, in turn, generated by China's demographic profile, the absence of social insurance for the bulk of the population, and the post-1978 appearance of investment-motivated saving in response to the scarcity of formal financial intermediation to finance the investment of the non-state sector.

　　The widening of the US-China bilateral trade deficit in recent years reflected many factors. In our opinion, the two chief factors are (i) macroeconomic forces in the US and China moving in opposite directions, causing their respective overall trade balance to move in opposite directions; and (ii) the accelerated relocation of production of US imports from East Asia to China.

　　1. Measuring the Bilateral Trade Deficit - The Role of Hong Kong

　　Of all the economic issues between the United States and China, none has the potential for greater confusion than the bilateral trade deficit. The official trade statistics of the United States and China have huge discrepancies. Much of the difference is due to, among others, the different treatment of Hong Kong's entrep?t trade by the two sides. In the 1988-95 period, on the average, over two-third of U.S. imports from China came through Hong Kong.

　　The resolution of the discrepancy between the U.S. and Chinese data rests large on two issues: first, the accounting of the Chinese goods shipped to the U.S. via Hong Kong; and second, the measurement of the value added by Hong Kong traders to these goods. The U.S. Commerce Department records the re-exports from Hong Kong as U.S. imports from China. Up until 1993, the Customs General Administrations of China recorded them as Chinese exports to Hong Kong, rather than to the U.S. Since 1993, China has gradually modified this approach to identify the final destination of its exports to Hong Kong, though its accounting of these goods remains incomplete.

　　The magnitudes of the Hong Kong re-export markups have been estimated in a number of studies, including Fung (1996), Fung and Lau (1996), Lardy (1994), Sung (1991), and West (1995). The markups cited in these studies come primarily from surveys conducted by Hong Kong agencies, and individual interviews. Since 1988, the Hong Kong Trade and Development Council has conducted several surveys on the magnitude of Hong Kong re-export markups among the local business community. These surveys have found the average markups around 14% for 1988 and 17% for 1991. A survey by Hong Kong Census and Statistics Department shows that the average markup for 1990 is 13.4%. In a report by the same agency to the GATT Informalpuorgof Experts on Export Statistics, the average markup is found to be 13% for 1988 and 25% for 1993. In Fung's interview for 1994 (reported in Fung, 1996; Fung and Lau, 1996), the markup for Chinese goods is reported at 25%. Besides showing that the average markup has increased over time, the surveys also indicate that the markup for the Chinese goods is well above the average markup for re-exports through third countries.

　　A recent report by the Joint Commission on Commerce and Trade (1996), a US-China intergovernmental agency, used disaggregate data on Hong Kong's import and re-export trade to estimate the markup for 1992 and 1993. This study (hereafter referred to as the "interagency report") finds that the average markup on Hong Kong re-exports of Chinese goods to the U.S. is 29 percent of the re-export value. It attempts to reconcile the differing trade statistics of the two countries using the markups and additional information on specific commodities.

　　The interagency report is a major advance in the estimation of the Hong Kong re-export markups, but it is vulnerable to two potential flaws. First, the import data used in the study only identifies the countries of origin and not the destination countries. This ignores that fact that there are three types of Hong Kong imports from China: imports that are retained in Hong Kong; re-exports to the U.S.; and re-exports to other countries. Therefore, the overall import unit-value may not reflect the actual unit-value for goods re-exported to the U.S. The distortion on the unit-value is more likely to be highest for goodserehwonly a small portion is re-exported to the U.S. Secondly, in the two datasets (Hong Kong imports and re-exports), there are some records that appear to suffer from measurement error, either because different unitserehwused for imports and re-exports, or because these transactions occurred in different calendar years. The estimates obtained for the markups are quite sensitive to these measurement errors.

　　In this part of the paper, we try to improve the markup estimation in these two respects, using both Hong Kong and Chinese trade statistics for the period 1988-1995. The Chinese trade data, provided by the Customs General Administration of China, contain information on destination countries for its exports via Hong Kong. The import unit-value calculated from this data should be more accurate than that obtained from the Hong Kong data alone. In addition, we construct a method to detect and eliminate records in the Hong Kong data that potentially suffer from measurement error, and we reconcile the results we obtain with other estimates of the markup. Finally, we use our estimates of the markups to correct the bilateral U.S.-China trade figures. For example, the bilateral deficit in 1995 is $33.8 billion according to official U.S. figures, but only $8.6 billion according to official Chinese figures. Using our markup estimates for the re-export activities in Hong Kong, we find that the actual trade deficit in 1995 is in the range of $15.6 to $21.6 billion.

　　The Role of Hong Kong in U.S.-China Trade

　　The entrep?t trade of Hong Kong has caused the bilateral US-China trade deficit estimated by the U.S. Dept of Commerce to be very different from that estimated by China's Customs authorities. In Table1, we contrast the values of eastbound and westbound trade between the U.S. and China, as reported by these two countries. Part A gives eastbound trade (i.e. China's exports and U.S. imports), and Part B gives westbound trade (U.S. exports and China's imports), while Part C computes the difference between westbound and eastbound trade to arrive at the U.S.-China trade balance. The information in the first column is obtained from the U.S. Department of Commerce, while that in the second to fourth columns is obtained from the Customs General Administration, People's Republic of China. We supplement this with information on Hong Kong re-exports in the last column, as reported by the Hong Kong Census and Statistics Office.

　　From Part C of Table 1, we can see that U.S.-China trade balance differs not only in the magnitude reported by the two countries, but even in its sign! The United States reports a trade deficit with China, which has increased about tenfold over the years 1988-1996, from about $3.5 billion to $39.5 billion. In contrast, China reports that the United States was running a trade surplus in the years 1988-1992, which turnedotnia deficit beginning in 1993. In 1996, the U.S. reported deficit with China of $39.5 billion compares to the Chinese reported value of the U.S. deficit of $10.5 billion, so that these two figures differ by $29 billion or a factor of three times. Clearly, the difference between these values is large enough that it risks misunderstanding between the countries on what the "true" value of the deficit might be.

　　The most important source of the different values for the trade deficit is the entrep?t trade of Hong Kong. The U.S. Department of Commerce calculates total Chinese exports to the US as the sum of (a) Chinese goods shipped directly to U.S., and (b) Hong Kong re-exports of goods from Chinese origin (i.e. Chinese goods shipped indirectly to U.S.). In other words, the value of all goods that originate in China are counted as Chinese exports to the United States. In contrast, the Chinese customs authorities calculate exports to the United States as consisting of component (a), and only those goods in component (b) whose final destination (the U.S.) is known at the time the product leaves China. It is often the case, however, that the Chinese exporter does not know the final destination of the good, so that many of the goods bound for the United States are not counted as such in the Chinese trade statistics; rather, these goods are treated as exports to Hong Kong.

　　Since 1993, the Chinese customs authorities have attempted to determine the final destination for goods exported to Hong Kong with greater accuracy, so as to improve its reported trade statistics. It is still the case, however, that many of the goods bound for the United States via Hong Kong are not recorded as such. This can be seen by breaking up the Chinese exports to the U.S.otnithose goods that are directly exported (in the third column of Table 1, Part A), and those goods that are exported via Hong Kong (in the fourth column). The latter can be compared to the value reported by the Hong Kong census authorities, in the last column, on the value of re-exports from China to the U.S. Thus, in 1988, China reported $705 million in exports to Hong Kong bound for the U.S., while Hong Kong reported $5.6 billion in re-exports to the U.S. that originated in China. These differ by a factor of eight times. In 1995, China reported $14.3 billion in exports to Hong Kong bound for the U.S., while Hong Kong reported $27.5 billion in re-exports to the U.S. that originated in China. These now differ by only a factor of two, though the difference in dollar values is still very large.

　　In comparison, for westbound trade in part B of Table 1, China used to report twice as many imports from the U.S. via Hong Kong as that country reported as re-exports ($2,538 versus $1,237 million in 1988). This discrepancy may be due to differing conventions in China and Hong Kong as to what constitutes a re-export: this term normally implies that there has been no "significant transformation" of a good, such as through a production process. If Hong Kong determined that there had been a transformation of a good imported from the U.S., and then sent to China, it would be counted as an actual export from Hong Kong, rather than a re-export. In this same instance, China could classify it as an import from the U.S. via Hong Kong, resulting in a higher import value, as shown in part B of Table 1. In any case, by 1995 the difference between the Chinese imports of U.S. goods via Hong Kong, and the Hong Kong re-exports, has been essentially eliminated.

　　There are several reasons why the value of Chinese goods sent to Hong Kong, and destined for the U.S., differs from the reported value of Hong Kong re-exports from China to the U.S. The first, which has already been mentioned, is that an exporter in China may not know the ultimate destination of a good when it is sent to Hong Kong. But even if this discrepancy did not occur, there is still a second reason. The value of Chinese exports to Hong Kong, which are destined for the U.S., represent the value of these goods when they leave China. In contrast, the Hong Kong re-exports from China to the United States represent the value of these goods when they leave Hong Kong. In other words, these two values differ by the markup or value-added in Hong Kong. Since it can be expected that traders in Hong Kong are providing various services to these goods, such as arranging for transportation and insurance, as well as identifying customers, the value-added in Hong Kong may be substantial. In 1995, for example, we saw in Table 1 that the Hong Kong re-exports from China to the U.S. were twice as large ($27.5 billion) as the Chinese exports to Hong Kong destined for the U.S. ($14.3 billion): some of this difference represents exports that are simply not recorded as such in China, but an equally large portion could represent the markup to the value of the goods in Hong Kong.

　　There are two other reasons why Hong Kong re-exports from China to the U.S. are larger than Chinese exports to Hong Kong: smuggling, and transfer pricing. Smuggling would lead to this discrepancy if it were recorded in Hong Kong but not in China. Since we will be making use of the Hong Kong re-export values in this study to correct the Chinese trade values, we will have taken smugglingotniaccount, provided that it is recorded in Hong Kong. (If it is not recorded in China or Hong Kong, but is recorded as an importotnithe U.S., then we are not able to correct for this). On transfer pricing, it is commonly asserted that many companies in China under-price their export invoices (with the help of the Hong Kong re-exporters) to transfer profits out of ChinaotniHong Kong. There are a whole array of reasons for this capital flight. Chinese managers use invoice under-pricing for "round-tripping"erehwthe transferred profits return to China under the guise of foreign investments inpuorgto reap the tax concessions offered to foreign enterprises and joint-venture companies. They use invoice under-pricing to diversify (internationalize) their companies' portfolios to reduce risks. Foreign partners of joint-ventures in China use invoice under-pricing to reduce the level of book profits which they share with the Chinese partners.

　　Given that there is some value-added in Hong Kong, that activity should be properly recorded as an export from Hong Kong - not from China. The U.S. Department of Commerce does not follow this practice, however. The United States counts the total value of goods from Hong Kong, that originated in China, as Chinese exports, so that it is implicitly ignoring the value added in Hong Kong. For this reason, the trade deficit with China reported by the U.S. is overstated. On the other hand, China is unable to count all of the goods leaving its country, and destined for the U.S. via Hong Kong, as an export to the United States. For this reason, the value of the U.S. trade deficit as reported by China is understated. Inpuorgto estimate the "true" value of the deficit, it is necessary to compute the value-added in Hong Kong on goods shipped from China to the U.S., and also in the reverse direction. Attributing this value-added as an export from Hong Kong, the discrepancy between the U.S. and Chinese magnitudes of the bilateral trade deficit can be substantially reduced. In the following section we proceed to estimate this value-added.

　　Markup on Hong Kong Re-exports to the United States

　　The value-added attributed to Hong Kong equals the markup on the prices of goods shipped through this entrep?t center, multiplied by the value of these goods. There are several estimates of the markup available, as summarized in Table 2. The interagency report from the Joint Commission on Commerce and Trade (1996) estimated the markup as 40.7% for 1992 and 1993. This markup is expressed as a percentage of the value of Chinese goods importedotniHong Kong. Alternatively, the markup can be expressed as a percentage of the value of goods when they are re-exported from Hong Kong, and we will consistently follow this practice. The 40.7% markup increases the value of the goods by a factor of 1.407, so that expressed in terms of the re-export value, the markup becomes 40.7/1.407=28.9%. This markup for 1992 and 1993 is shown in the first column of Table 2.

　　A second estimate of the markup has been made by the Hong Kong Census and Statistics Department, based on a survey of exporters. These markups are also expressed as a percentage of the re-export value, and are shown in the second column of Table 2. It can be seen that the markups reported by the Hong Kong survey is less than that estimated in the interagency report. This is no coincidence, but reflects underlying differences in the methodology used to estimate the markup. To explain these differences, we use the example illustrated in Figure 1.

　　Consider a product that is shipped from China to Hong Kong, and from there is re-exported to the United States and other locations. Suppose that the unit-value of this item when it arrives in Hong Kong is $1.00. This unit-value is the average over all units sent to Hong Kong, regardless of their final destination. It should be stressed that the "overall" unit-value of this type is all that is available in the actual Hong Kong import data. Because imports are collected by source country, but not by eventual destination, it is impossible to distinguish the unit-value of imports destined for the U.S. from those destined for elsewhere. It is precisely this limitation of the Hong Kong data that makes estimation of the markup difficult. To illustrate this, suppose that the goods destined for the U.S. are of higher quality, and have a unit-value of $1.10, while those destined for the rest of the world have a unit-value of $0.90: these unit-values are not observed in the Hong Kong data, however. Then to estimate the markup, one approach is to compare the "overall" unit-value of the Hong Kong imports from China ($1.00) with the unit-value of the Hong Kong re-exports from China to the U.S. ($1.50). This gives a markup of $0.50, or 33% when expressed relative to the re-export value of $1.50.

　　This calculation is labeled as Method A in Figure 1, and corresponds precisely to the calculation performed by the interagency Report, as shown in the first column of Table 2. That report computed the markup by comparing the "overall" unit-value of Hong Kong imports from China, to the unit-value of Hong Kong re-exports from China to the United States. We have made exactly the same calculation for a wider range of years, using the Hong Kong import and re-export data, and these results are reported in the third column of Table 2. We see that the markups range from 26.9% to 31.5% over 1988-1995. It is evident that this method will overstate the "true" markup if the Hong Kong imports from China that are destined for the United States are, on average, priced higher than those destined for other markets. In that case, the "overall" unit-value of Hong Kong imports from China is too low, so the markup obtained is to too high. To correct for this overstatement, we can consider an alternative calculation of the markup, which is labeled as Method B in Figure 1.

　　Method B compares the "overall" unit-value of Hong Kong import from China ($1.00) with the "overall" unit-value of Hong Kong re-exports from China to the world ($1.30). Note that both the goods entering Hong Kong and those leaving can be destined for any final market, so in this sense, the calculation is consistent. If the goods re-exported to the rest of the world are priced lower than those going to the United States, then this method gives an estimate of the markup that is less than that from Method A. For the values in Figure 1, Method B give a markup of $0.30, or 23% when expressed relative to the re-export value of $1.30. It turns out that Method B corresponds to the question that the Hong Kong Census asked exporters in their survey: namely, what is the average markup on all goods imported from China and re-exported (to anywhere in the world)? We have made this calculation using the Hong Kong data, resulting in the values shown in column four of Table 2. We find that Method B gives an estimated markup ranging from 19.6% to 23.4% over 1988-1995, which is less than that obtained from Method A in every year. Thus, the difference between the interagency estimates and those from the Hong Kong Census shows up equally well in our own estimates of the markups from Methods A and Method B.

　　This raises the question of whether it is possible to improve upon these approaches, and obtain another estimate of the markup, possibly lying in-between those from Methods A and B. To achieve this, we will have to rely on data beyond that of Hong Kong imports and re-exports since, as explained above, this data does not distinguish the Hong Kong imports from China that are destined for the United States from those that are destined for elsewhere. This means that the difference in prices of these imports, illustrated by $1.10 and $0.90 in Figure 1, cannot be measured from the Hong Kong data. Inpuorgto measure these prices, we rely instead on the Chinese export data from the General Customs Administration. Using that data, we can distinguish a unit-value for Chinese exports to Hong Kong, destined for the United States, from the unit-value for Chinese exports to Hong Kong that are destined for all other markets. It turns out that the goods destined for the U.S. tend to have higher prices than those destined elsewhere, as illustrated in Figure 1. By merging the China export data with the Hong Kong import and re-export data, we are therefore able to make a more accurate calculation of the markup, indicated by Method C in Figure 1.

　　In Method C, we use the Chinese export data to measure the unit-value of goods exported to Hong Kong and destined for the U.S. ($1.10). This is compared to the unit-value of Hong-Kong re-exports from China to the U.S. ($1.50), giving a markup of $0.40 or 27% when expressed relative to the re-export value of $1.50. In the final column of Table 2, we report the results from a calculation of this type, using both the Chinese and Hong Kong trade data. We see that Method C gives an estimate of the markup that varies between 22% and 28.7% over 1988-1995, and lies in between that obtained from Methods A and B in every year. Thus, we can treat Methods A and B as providing upper and lower-bounds, respectively, to the "true" markup, and view Method C as giving the preferred estimate. Our calculations have shown how the differences between the interagency report and the Hong Kong Census can reconciled, and improved upon, to obtain an estimate of the markup that lies in-between these two sources.

　　The same methods we have used above can be applied to measure the markup on Hong Kong re-exports to any other country. In Table 3 we report the results for re-exports from China to Germany, Japan, and the United Kingdom. In each case, we calculate the markups from methods A and C, since the results for method B - which gives the markup for re-exports from China to the entire world - do not depend on the country of destination. (The results for method B in Table 2 therefore are unaffected when the country of destination changes). It can be seen that the markups for re-exports to Germany are slightly larger than those for the United States, while the markups for Japan are larger still, and the markups to the United Kingdom are somewhat smaller. Overall, the differences with the markups obtained for the United States are not that great, especially as compared to the year-to-year fluctuation in the estimated markups for any country. Method C continues to give smaller estimates of the markup as compared to method A is virtually all cases, and it is our preferred estimate.

　　We note that the markup for re-exports to the U.S. estimated by Method C in Table 2 shows a broad decline over the 1988-94 period, going from 28 percent in 1988 to 25 percent in 1991, and then to 22 percent in 1994 (though increasing in 1995). This secular decline in markup is consistent with the competitive effects of more firms enteringotnithe re-export industry over time. However, this is only a conjecture because the markup for re-exports to Germany, Japan, and the United Kingdom appears to be, respectively, trendless, increasing over time, and decreasing slightly over time. In work under progress, we are determining how much each of the movements of the national average markup could be attributable to movements in the markup of particular products (i.e. markup is unstable but composition of products is stable), and to changes in the composition of re-exports (i.e. composition of products is unstable but individual markup is stable).

　　Markup on Hong Kong Re-exports from the U.S. to China

　　We turn next to the issue of U.S. exports to China that pass through Hong Kong. While there are rather substantial markups on the re-exports from China to the U.S. (eastbound trade), existing studies have found smaller markups on the re-exports from the U.S. to China (westbound trade). For example, the Hong Kong Census and Statistics Department finds from their survey that the markups on trade from the United States to China decline from 11.3% in 1990 to 5.7% in 1994. However, these figures actually refer to the Hong Kong re-export margin on all trade with country of origin other than China. For example, these margins include goods shipped from Japan to the U.S. via Hong Kong, as well as from Japan or the United States to China, etc. Thus, these figures are not an accurate estimate of the Hong Kong markup on re-exports from just the United States to China.

　　The interagency report prepared by the U.S. Department of Commerce and the Chinese General Customs Administration does not present markups on Hong Kong re-exports from the U.S. to China in 1992 and 1993, but focuses instead on the differences in trade values reported by these countries due to shipments of aircraft, and other factors. The reason why the interagency report omits the markups on westbound trade is because when they are computed according to the same method used for eastbound trade, negative markups are obtained. This is shown for various years in the first column of Table 4.

　　We see that from 1984-87 the markups are positive, but after that the markups become negative beginning in 1988. The more puzzling finding is that the post-1987 markups sometimes indicate a reduction in the re-export value, such as more than 100%, that is much too large to be believable. While mark-downs in the value of particular goods must occur in some instances, it is doubtful that they occur across most goods frequently, and by over 100 percent three times in the 1984-95 period. Our assessment is that the very large mark-downs reflect measurement errors in the data. In particular, since the unit-value for each traded item is constructed by dividing the value by the quantity, then any inaccuracy in the quantity (such as change in units when re-shipped) will result in measurement error in the unit-values.

　　In fact, measurement error was already found when we consider the re-export from China to the U.S. (eastbound trade), in Table 2. As indicated there, several of the markup estimates were made while omitting some outlying observations (i.e. particular SITC or harmonized system categories). A good example is the markup for 1994, using method A. When all the harmonized system categories are included, the markup obtained is 19.4%, as compared to 29.3% (reported in Table 2) after several observations are omitted. The difference between these two estimates is almost entirely explained by a single harmonized system category - HS 26090000, which is tin ores and concentrates. Evidence that its quantity is incorrectly measured in re-exports comes from the fact that the re-export quantity is 617 times greater than the quantity that Hong Kong imported from China in 1994! This almost surely reflects measurement error in the data, and when this single observation is omitted, the estimated markup becomes much closer to that obtained in surrounding years.

　　Whenever the re-export quantity is greater than the import quantity, we shall interpret this as evidence of measurement error, such as different units in measuring quantities. We compute the ratio QRATIO=(quantity re-exported through Hong Kong)/(quantity importedotniHong Kong) to alert us to the measurement problem. The storage of commodities would make the re-export quantity greater than the import quantity, so some values of QATIO greater than unity might still be acceptable. We therefore consider two criterion for eliminating outliers: first, omitting all harmonized system categories for which QRATIO>2; and, second, omitting all harmonized system categories for which QRATIO>1. The second criterion is stricter than the first, in the sense that more observations are omitted. (The actual number of observation deleted is reported in the Appendix, Tables A1 and A2).

　　The results for the Hong Kong re-exports from the United States to China are shown in Table 4. When the observations with QRATIO>2 are omitted, the extremely large and negative markups are eliminated, and most of the negative values occurring after 1988 are quite small. When the additional observations with QRATIO>1 are omitted, then the negative markups are eliminated almost entirely, except for 1994 and 1995. Based on this evidence, we conclude that more realistic markups are obtained when the outlying observations (as identified by QRATIO) are omitted. Generally, excluding these observation raises the markup. This follows because the observation with high values of QRATIO will necessarily have a low value for the markups on re-exports, so that excluding these erroneous observations will raise the overall markup.

　　Hence, our preferred value for the markups are those shown in the last columns of Table 4 (omitting all observations with QRATIO>1). In the Appendix, we also report the markups obtained on eastbound trade when outlying observation are excluded. These are generally higher than those reported in Table 2, indicating that even our preferred estimates in Table 2 (i.e. method C), may be an underestimate of the "true" markup.

　　The final columns of Table 4 indicate a decline in the markup over time. This decline could be partly due to more competition among traders in the re-sale of goods to China, and partly to due to exchange rate movements during this period. In particular, there is a very large fall in the markup from 20 percent in 1987 to 8 percent in 1988, which occurs simultaneously with an large percent depreciation of the Chinese Yuan against the U.S. dollar. This is illustrated in Figure 2,erehwwe plot the markups from Methods A and B (from Table 5, deleting observation with QRATIO>1) and the exchange rate between the Chinese yuan and the U.S. dollar. If many Hong Kong companies had signed yuan-denominated contracts in 1987 to deliver U.S.-made goods to China, then the large unexpected depreciation of the yuan against the U.S. dollar would naturally reduce the markup on re-exports. While contracts may explain the 1987-88 drop in the markup, it is surprising that the fall in the markup has apparently been permanent.

　　Revised Values for U.S.-China Trade Deficit

　　We can now use the estimated markups for goods shipped through Hong Kong to revise and reconcile the differing values for U.S.-China trade. The key principle is that the value-added on goods as they pass through Hong Kong should be attributed to Hong Kong, rather than treated as an export of some other country. This value-added is computed as the Hong Kong markup times the re-export value of the goods as they leave Hong Kong. Thus, for eastbound trade in part A of Table 5, the first column shows the value-added on Hong Kong re-exports of Chinese goods to the United States. This is obtained by using the Hong Kong value for these re-exports, multiplied by the markups obtained by method C in Table 2. In 1995, for example, the Hong Kong re-exported $27.5 billion in goods from China to the U.S. (Table 1A, column 5), of which we attribute $7 billion as the value-added in Hong Kong (Table 5A, column 1).

　　This value-added in Hong Kong should be deducted from the value that the United States reports as imports from China, which was $45.5 billion in 1995 (Table 1A, column 1), to obtain the revised value of U.S. imports of $38.3 billion (Table 5A, column 2). This revised figure therefore corrects for the policy of the U.S. Department of Commerce to attribute the value of all Chinese goods passing through Hong Kong en route to the United States as Chinese exports, thereby ignoring the value-added in Hong Kong. The value for Chinese exports to the U.S. also needs to be adjusted, to reflect the fact that many of these exports are simply not recorded. Instead of using the Chinese value for the exports to the United States via Hong Kong (Table 1A, column 4), we instead use the value reported by Hong Kong for Chinese re-exports to the U.S. Kong (Table 1A, column 5), less the value-added onto these goods in Hong Kong. This calculation yields the revised figure for Chinese exports to the United States of $30.8 billion in 1995 (Table 5A, column 3).

　　The discrepancy between the revised U.S. imports from China and the Chinese exports to the U.S. is now $7.5 billion. A small amount of this discrepancy reflects factors such as: differences in the geographic territories considered by the two countries (the U.S. includes Puerto Rico and the U.S. Virgin Islands as part of its customs territories, whereas China does not); the U.S. includes the value of repairs as imports whereas China does not include these as exports; and other factors identified in the interagency report. While the discrepancy is still sizable, it is much less than the original discrepancy of $20.3 in the reported value from each country. Thus, by properly attributed the value-added to Hong Kong, we have reduced to discrepancy in the U.S. and Chinese values for eastbound trade to about one-third of its original magnitude.

　　Revised values for westbound trade are considered in part B of Table 5. The Hong Kong value-added shown in the first column is computed using the simple average of the markups obtained from methods A and B in Table 4 (with QRATIO>1). Because these markups are all rather small, the precise figures which are used has little influence on the results. The revised value of U.S. exports shown in the second column is obtained from the published value of U.S. exports (Table 1B, column 1), and adding the value reported by Hong Kong for U.S. re-exports to the China (Table 1B, column 5), less the value-added onto these goods in Hong Kong. In 1995, this results in U.S. exports to China of $16.7 billion, as compared to the reported value of $11.8 billion. Taking the difference between these exports and the revised value of U.S. imports from China in 1995, we obtain the trade deficit of $21.6 billion in 1995 (shown in part C of Table ). This compares with a deficit of $33.8 billion using the published U.S. figure for 1995, so that our revised estimate is about two-thirds of its original value.

　　Turning to the Chinese data, the published import data from the U.S. is recorded on a c.i.f. (cost including freight) basis, whereas the U.S. export are recorded as f.a.s. (free along side), which does not include any transportation charges. Inpuorgto make these comparable, we multiply the original Chinese imports by 0.94, which offsets the inclusion of transportation charges. The revised value for Chinese imports from the U.S. is $15.2 billion in 1995 (Table 5B, column 3). Taking the difference with the revised value for U.S. exports to China we obtain a discrepancy of -$1.6 billion, with the U.S. exports being higher. This compares to a discrepancy of $4.4 billion in the published figures, with the U.S. exports being less than Chinese imports. Thus, the discrepancy is reduced to about one-third of its original magnitude.

　　Much of remaining difference on westbound trade reflects the treatment of aircraft exports to China, as well as car exports. The Chinese Customs authorities treat aircraft exports from the U.S. as a lease, and count only the value of the lease that year, whereas the United States counts the entire value of the aircraft as an export. In addition, cars broughtotniChina by foreigners for personal use are not included in published trade statistics, whereas the U.S. would include these as exports, which also helps to explain why the revised U.S. exports are higher.

　　Taking the difference between the revised value of Chinese imports and exports with the U.S., we obtain another estimate of the U.S.-China trade deficit shown in the second column of Table 5, part C: the revised Chinese numbers give a deficit of $15.6 billion in 1995, as compared to $21.6 billion from the revised U.S. figures. These differ by $5 billion, while the original data had a difference of $25.2 billion for 1995 in the trade deficits reported in Table 1, part C. Figure 3 graphs the two official estimates and our revised estimates. Thus, proper attribution of trade flows through Hong Kong has therefore tremendously reduced the discrepancy in the U.S. and Chinese values of the trade deficit, to one-fifth of its original size. It is hoped that these calculations will prove useful to official statistical agencies in both countries, so that the trade figures reported by each will be in closer correspondence, and can therefore contribute to improved understanding of the bilateral trade situation.

　　2. Explaining the Bilateral Trade Deficit

　　We begin by laying out the determination of the overall current account (CA) position, within which the overall trade balance is the most important component. Broadly speaking, the primary determinants of the overall trade balance are macroeconomic and structural in nature, and they work through saving and investment behavior. This can be seen from the following decomposition:

　　CA = (S - I)private + (S - I)SOE + (S - I)govt

　　So, for a given pegged value of the exchange rate, contractionary fiscal policy increases CA directly by increasing (S - I)govt; and contractionary monetary policy increases CA by decreasing the investment spending of private and state-owned enterprises (SOEs).

　　Any structural factor that increases Sprivate without increasing Iprivate correspondingly would raise CA. This is the situation in China after the initiation of economic deregulation in 1978. Past central planning, closure to international trade, and promotion of local self-sufficiency created many highly profitable production niches and arbitrage possibilities, but private businesses cannot undertake the required investment immediately because they cannot get loans either from the wholly state-owned banking monopoly or from off-shore foreign financial institutions. Because the state banks do not channel much of the private savings to private investors, the only way that a private Chinese businessman can invest is to accumulate enough savings to start his own workshop-factory. This is a feasible option because the threshold size for the average Chinese enterprise is small. In an economy that is open to trade but closed to private capital movements, the failure of the domestic banks to match the available private saving flow with desired private investment creates the twin phenomena of a high private saving rate, and a persistent CA surplus.

　　Investment-motivated saving is only one of the important factors behind China's high household saving rate. Two other important factors are the low coverage of social insurance programs, and the demographic profile of the population. The absence of pensions and medical insurance for most of the population necessitates a higher saving rate to prepare for retirement, and health emergency. A large proportion of China's labor force is now in the phase of the life-cycle to begin saving seriously in anticipation for retirement. The state policy of keeping the size of family small has reduced the number of dependent children, enabling the parents to save more.

　　An appreciation of the exchange rate should reduce CA by lowering the private saving rate because the rise in the purchasing power of domestic wealth would reduce the amount needed to be saved inpuorgto make the required purchase. Investment spending could also increase because the decline in the price of imported capital goods would allow more investment projects to be implemented sooner.

　　The ability of import liberalization to reduce a CA surplus requires that: (a) it can to generate the same wealth effects as an appreciation of the currency; and/or (b) the resulting investment boom in the export sector and nontraded goods sector be greater than the collapse in investment in the import-competing sector ; and/or (c) saving to decline substantially because of previous intertemporal substitution,erehwconsumers, in the past, had expected an eventual decline in the prices of imports, and postponed their consumption. The import liberalization would hence release the pent-up demand for imports.

　　The uncertainty over the efficacy of import liberalization to change the overall trade balance significantly in a sustained way can also be seen in terms of sectoral production. Import liberalization would reduce the importable sector, and increase imports; but the resulting flow of scarce resources (particularly capital)otnithe exportable sector would increase exports, and the final result may be an unchanged overall trade balance, particularly if the level of aggregate demand is kept constant.

　　The US-China Bilateral Trade Deficit

　　Table 6 shows that the bilateral trade balance swung from a surplus of $3 billion in 1980 to a deficit of $40 billion in 1996. This reversal of the bilateral trade balance is in line with the opposite movements in the overall trade balance of the two countries.

　　The US overall trade deficit increased from $23 billion in 1980 to $170 billion in 1996, or, equivalently, from 0.8 percent of GDP to 2.2 percent of GDP. In this period, the private saving rate fell as personal consumption increased from 63 percent of GDP to 68 percent of GDP. It is hence quite natural that of the 25 largest US trade partners, 18 of them ran surpluses in their trade with the US in 1996 compared to only 8 countries in 1980.

　　Just as US saving-investment behavior was widening the US overall trade deficit, macroeconomic forces within China were reducing China's overall trade deficit. China's consolidated budget deficit (formal government budget deficit plus central bank financing of SOE losses) amounted to about 6 percent of GDP for the last two years, which is down from the 8 percent of the previous three years. Monetary policy has also been tight. It is hence not surprising that China's overall trade balance has gone from a deficit of $11 billion in 1993 to a surplus of $20 billion in 1996,erehwwere -2 percent of GDP and 2 percent of GDP respectively.

　　Hence lies the first important determinant of the bilateral trade balance: the opposite movements of macroeconomic forces in the two countries, reinforced by demographic trends and by China's reforms. To a first approximation, the widening bilateral trade deficit reflected the saving slow down in the US (that has been pulling in foreign resources to finance capital formation) and the surge in investment-motivated saving in China (that was necessary to compensate for the low level of financial intermediation available to the increasingly liberalized non-state sector).

　　Table 6 also shows the second important determinant of the US-China bilateral trade balance: the shifting of some of the production of US imports from other countries to China. The result is that a large part of the bilateral trade deficits that the US had with the newly-industrialised countries of Hong Kong, Singapore, South Korea and Taiwan (NIC-4) in the 1980s was transferred to US-China bilateral trade in the 1990s. (The most massive relocation of production to China is the case of Hong Kong. US-Hong Kong bilateral trade balance went from a deficit of $2 billion in 1980 to a surplus of $4 billion in 1996.)

　　It is difficult to say that Southeast Asia's share of the US market has been reduced by the emergence of China as a trading state. While Malaysia, Indonesia, Philippines and Thailand (the ASEAN-4) have seen their collective share of US overall trade deficit falling from 19 percent in 1980 to 12 percent in 1996, it is important to note that ASEAN-4's share rose throughout 1985-96 subperiod when China really began serious integrationotnithe global economy.

　　Table 7 confirms the view that the developments in US-China bilateral trade deficit over the last decade mirrored opposite developments in US trade with NIC-4. NIC-4's share of US imports dropped 4.0 percentage points over the 1988-96 period, while China's share rose 4.6 percentage points in the same period. The point that China's penetration of the US market has come largely at the expense of its industrialised Asian neighbors is vividly seen in Table 8 which reports the top 5 suppliers by selected commodities in US imports. By 1995, China has displaced South Korea as one of the top 5 suppliers of "apparel and other textile products" and "leather and leather products"; displaced Japan as a top 5 supplier of "household appliances" and "other manufacturing"; and displaced Taiwan as a top 5 supplier of "miscellaneous plastic products," "other paper and allied products," and "stone, clay, concrete and gypsum"; and displaced Indonesia as a top 5 supplier of "lumber, wood, and furniture."

　　Concluding Remarks

　　We found that proper adjustment for value added in Hong Kong on China's exports going to the US reduced (on the average) 91 percent of the discrepancy between the official US and official China estimates of the US-China bilateral trade balance. Our revised estimates for the US data reduced the official US estimate of the bilateral trade deficit by a third on the average. The same adjustments on the official Chinese estimates of the US-China bilateral trade balance converted the surpluses of 1988-92 to deficits, and magnified the deficits of 1993-95 by a factor of two.

　　The increase in the bilateral trade deficits in the 1990s reflects the stance of macroeconomic policies and structural conditions in both countries. The United States had an investment boom that was sucking in foreign funds from high-saving countries like China. If the post-1978 surge in investment-motivated saving in China is responsible for some of the rise in the household saving rate, then the provision of more financial intermediation to facilitate investment by the non-state sector will help to reduce China's overall trade surplus, and attenuate the wrong image of China as a mercantilist state. As the Chinese financial system is relatively backward compared to that of the United States, the granting of permission to more US financial institutions to operate in China will not only improve financial intermediation but will also provide a learning opportunity for Chinese financial institutions.

　　Increased trade benefits both countries, and the rest of the world. It is therefore important for the future growth of the world trading system that agreement be reached soon on the conditions of China's accessionotniWTO.

　　References

　　Fung, K. C., 1996: Accounting for Chinese Trade: Some National and Regional Considerations, NBER Working Paper 5595.

　　Fung, K. C. and Lawrence Lau, 1996: The China-United States Bilateral Trade Balance: How Big Is It Really?, Asia/Pacific Research Center, Institute for International Studies, Stanford University.

　　Joint Commission on Commerce and Trade, Trade and Investment Working Group, 1996,"Report of the 'Trade Statistics Subgroup'," Washington DC.

　　Lardy, Nicholas R., 1994, China in the World Economy, Institute for International Economics, Washington DC.

　　Liu. Liang-Yn and Wing Thye Woo, 1994, "Saving Behaviour under Imperfect Financial Markets and the Current Account Consequences," Economic Journal ,May 1994.

　　Sachs, Jeffrey, 1982, "The Current Account in the Macroeconomic Process," Scandinavian Journal of Economics.

　　Sheffrin, Steve and Wing Thye Woo, 1990, "Present Value Tests of an Intertemporal Model of the Current Account," Journal of International Economics, November.

　　Sung, Yun-Wing, 1991, The China-Hong Kong Connection: The Key to China's Open Door Policy, Cambridge University Press.

　　West, Loraine A., 1995, Reconciling China's Trade Statistics, IPC Staff Paper No. 76, International Program Center, Population Division, U.S. Bureau of the Census.