77260 THE WORLD BANK ECONOMIC REVIEW, VOL. 12, NO. 1: 1-28 Does Mercosur's Trade Performance Raise Concerns about the Effects of Regional Trade Arrangements? Alexander J. Yeats This study employs a new methodological approach that measures production effi- ciency in changing trade patterns. It shows that the most rapidly growing products in Mercosur's intratrade generally are goods in which members do not have a compara- tive advantage and have not been able to export competitively to outside markets. This is consistent with substantial trade diversion within the arrangement. Mercosur's dis- criminatory tariffs against nonmembers, which are four to six times higher than those in arrangements such as the European Union, European Free Trade Area, or North American Free Trade Agreement, are likely the cause. Recent further increases in Mercosur's tariffs against nonmembers are likely to exacerbate the magnitude of trade diversion. The recent proliferation of regional trade arrangements among countries raises several legitimate concerns. Primary among these is the fear that the new region- alism will divert attention from the multilateral negotiation process that GATT (General Agreement on Tariffs and Trade, now the World Trade Organization) employs to reduce international trade barriers. Second, some regional trade ar- rangements may raise trade barriers against nonmembers, which could seriously undermine the achievements of the GATT. Third, the discriminatory trade barri- ers incorporated within regional arrangements may have undesirable effects. That is, they may cause the sales of members to displace those of more efficient third countries, thereby denying both consumers and producers access to lower- cost and superior-quality goods. However, assessments of the influence of re- gional arrangements have been hampered by a lack of appropriate and reliable empirical procedures for evaluating their actual effects. This investigation attempts to determine whether, when used jointly, two indexes, which measure the regional orientation of exports and revealed com- parative advantage, provide insights concerning the extent to which a re- gional trade arrangement may distort trade from patterns expected on the basis of efficiency conditions and comparative advantage. As a case study, the analysis focuses on the countries of Mercosur, perhaps the most impor- Alexander J. Yeats is with the Development Research Group at the World Bank. The author would like to thank L. Alan Winters and Maurice Schiff for many helpful comments on an earlier version of this article. © 1998 The International Bank for Reconstruction and Development/THE WORLD BANK 2 THE WORLD BANK ECONOMIC REVIEW, VOL. 12, NO. 1 tant recent arrangement involving developing countries. In addition, Mercosur's discriminatory tariffs on third-country suppliers are approximately four to six times higher than those in arrangements among industrial coun- tries like the European Union, European Free Trade Area (EFTA), or NAFTA (North American Free Trade Agreement). Given the magnitude of Mercosur's departure from universal most-favored-nation tariff treatment, this study looks at the trade-diverting and trade-creating potential of the arrangement's dis- criminatory trade barriers. In this respect, it differs from related analyses that focus on welfare effects. Mercosur was established under the Treaty of Asuncion, signed on 26 March 1991 by the presidents of Argentina, Brazil, Paraguay, and Uruguay. Under the terms of the treaty, staged reductions of tariffs against trade among members were to begin in June 1991 with the objective of removing all tariffs on this exchange by the end of 1994. Most of this study's analysis begins in 1988 be- cause Argentina and Brazil were already implementing some preferential sectoral trade arrangements by then. By starting with 1988, the study includes the effects of the strong multilateral liberalizations that Mercosur countries initiated around that period. Section I analyzes the trade statistics of Mercosur countries to determine if the direction and composition of their trade changed significantly during the period when the agreement was being implemented. Unless sizable changes occurred, concerns that trade was actually diverted from more-efficient pro- ducers would appear to be groundless. Section II examines the characteristics of a new index that measures the changing regional orientation of exports of specific goods and then uses the index to analyze recent shifts in the direc- tion of trade. It compares this index with a measure of revealed comparative advantage to show how the two can be employed jointly for analyzing potential inefficiencies in trade patterns. The section analyzes data drawn from United Nations sources to determine whether the changing regional orientation of Mercosur's trade during the period when the agreement was being implemented was consistent with member countries' comparative ad- vantage. Section HI analyzes supplemental information on Mercosur coun- tries' tariffs and nontariff barriers to determine how they might have influ- enced trade patterns. Section IV closes the study with an overall assessment of the findings and their implications for further research on issues relating to regionalism. An important qualifying point should be noted at the outset. The analysis in this article focuses on only one aspect of Mercosur—its static trade effects. Many commentators see other benefits stemming from the agreement such as political cooperation, enhanced negotiating power, better credibility for the members' general economic reform programs, the possibility of achieving otherwise unat- tainable economies of scale in production, and dynamic gains in trade. These are legitimate objectives that, if achieved, could offset distortionary trade effects. In the present context, however, nothing can be said about them. Yeats 3 I. RECENT TRENDS IN MERCOSUR TRADE Available data indicate that the trade patterns of member countries have changed significantly since the formation of Mercosur. For example, table 1 provides sum- mary statistics on Mercosur's intratrade and on exports to destinations such as the countries in the OECD (Organization for Economic Cooperation and Development), NAFTA, or OECD Europe. These data are shown for selected periods from 1979-81 to 1994 in order to help determine when the trend changed. This analysis is re- stricted because data with which to explore trade patterns were available only up to and including 1994. This is sufficient to identify the effects of preferential trade policies. Some preferences were included in the Argentina-Brazil sectoral agree- ments at the end of the 1980s, and widespread preferences were introduced in the transition period for Mercosur starting in June 1991. This exercise clearly sheds no light on events since 1994, such as the further progress toward a common external tariff or any progress in the nontrade policy aspects of the agreement. These remain on the agenda for further research. The earlier intervals in table 1 are three-year periods in order to reduce the influence of any annual irregular variations in trade statistics, such as those that might accompany significant fluctuations in commodity prices. Statistics for 1993 and 1994 are shown separately in order to reflect the recent influence of Mercosur on trade flows. A more detailed analysis of the annual trade data used in the construction of table 1 strongly suggests that 1991 was the year in which intratrade became significantly more important. In June of 1991 Mercosur be- gan to implement discriminatory tariff preferences on intratrade. The figures reported in table 1 show the increasing relative importance of Mercosur markets for all four member countries. For example, in 1984—86 less than 10 percent of Argentina's exports went to Mercosur countries compared with 30 percent by 1994. Although the 1984-86 level was lower (about 5 per- cent), a threefold increase also occurred for Brazil's exports to Mercosur (almost 14 percent in 1994), while Uruguay's share of exports rose almost 20 percentage points (reaching almost 47 percent in 1994). For all member countries taken together, the 1994 share of exports to Mercosur (reaching 20 percent) was al- most three times the corresponding 1984-86 level. The data in table 1 reveal two key trends in the direction of Mercosur's ex- ports over the last decade: intra-Mercosur trade became significantly more im- portant at the expense of trade with countries in NAFTA, which remained stable or declined slightly, and of trade with Europe, which declined. Statistics on mem- ber countries' imports also reflect the major increase in the relative importance of trade between Mercosur members. From 1984-86 to 1994 the share of im- ports originating in member countries rose from 11 to 20 percent. The reorien- tation of Mercosur's trade toward member countries over this full interval, or during the 1990s, was far greater than that in any other regional arrangement, including EFTA, the European Union, the Association of Southeast Asian Na- tions, and the Canadian-United States Free Trade Arrangement. Table 1. Trade Destinations for the Mercosur Countries, 1979-81 to 1994 World Percentage of total exports (millions of OECD NAFTA Non-OECD Mercosur Exporter dollars) countries countries United States Europe countries countries Argentina 1979-81 8,322.7 45.0 11.4 8.7 32.5 55.0 13.4 1984-86 7,785.2 44.9 14.4 11.2 28.3 55.1 9.5 1990-92 12,187.1 49.5 14.5 11.7 33.3 50.5 16.8 1993 13,114.4 43.6 11.9 9.7 29.3 56.3 28.1 1994 15,803.3 40.2 13.1 10.9 25.6 59.8 30.4 Brazil 1979-81 19,556.3 58.2 21.6 18.0 32.7 41.8 8.1 1984-86 25,008.6 64.7 29.8 27.2 29.1 35.3 4.7 1990-92 32,987.8 63.6 25.2 21.4 32.8 36.4 7.8 1993 38,679.4 56.1 24.5 20.7 27.6 43.9 13.9 1994 43,355.2 57.4 24.2 20.6 29.0 42.6 13.7 Paraguay 1979-81 303.6 52.7 6.6 5.7 41.3 47.3 38.4 1984-86 290.3 48.8 3.7 3.6 43.7 51.2 36.3 1990-92 784.1 41.9 5.0 4.7 36.8 58.1 37.6 1993 725.2 45.8 8.2 7.3 38.0 54.2 39.6 1994 816.8 36.3 7.6 7.0 29.1 63.7 52.0 Uruguay 1979-81 1,021.0 44.0 9.8 8.5 33.6 56.0 30.6 1984-86 953.3 41.7 14.9 13.5 25.1 58.3 28.4 1990-92 1,629.9 38.8 13.1 10.1 26.4 61.2 34.7 1993 1,603.3 32.6 12.3 9.2 21.9 67.3 41.2 1994 1,918.1 30.4 10.1 6.8 21.7 69.6 46.7 Mercosur 1979-81 29,203.6 53.9 18.2 14.9 32.7 46.1 10.7 1984-86 34,037.4 59.5 25.6 23.0 28.9 40.6 6.7 1990-92 47,588.9 58.8 21.7 18.3 32.8 41.2 11.5 1993 54,122.2 52.2 20.9 17.6 28.0 47.8 18.5 1994 61,893.3 51.8 20.7 17.5 27.9 48.1 19.5 Note: For countries included in OECD, Europe, and non-OECD, see World Bank (1992: 40). NAFTA includes the United States, Canada, and Mexico. Source: Compiled from United Nations Comtrade records. Yeats 5 Given that major shifts occurred in the direction of members' trade toward Mercosur, what products were most important in this exchange, and how did the composition of exports change? Table 2 lists the value of Mercosur mem- bers' exports to one another and indicates the shares for several broad product groups. On average, in 1994 about 63 percent of Mercosur's intratrade con- sisted of manufactures (about 15 percentage points higher than their share in the region's global exports), with Brazil, as expected due to its relative size, having a major influence on the overall average. More than 81 percent of Brazil's exports to Mercosur in 1994 consisted of manufactured goods, almost double the corre- sponding share for Argentina and more than four times that for Paraguay. Table 2 also documents the overall importance of the transport and machinery group (Standard International Trade Classification [srrc] 7) in Mercosur's intratrade; these goods comprised almost one-third of total trade in 1994. From table 2, the second largest product category in intratrade, namely foods and feeds, by 1994 accounted for about one-quarter of the goods traded within Mercosur (their share in the region's global exports was about 36 percent). The importance of this category declined after the early 1980s. Agricultural materi- als and ores, minerals, and nonferrous metals also declined in relative impor- tance. Mineral fuels was the only product group, in addition to manufactures, that increased its relative share. In short, manufactures provided the catalyst for the increase in Mercosur's intratrade, with transport and machinery products being the most dynamic subsector within this group. Trade intensity indexes can provide additional insights into the nature and im- portance of secular changes in bilateral trade flows such as those occurring for Mercosur. The intensity of trade refers to a tendency for two countries to trade more or less heavily with each other based on factors such as their global importance in world exports and imports. The measure has been used since the 1960s in numerous analyses of the direction and level of international trade. For illustrative examples, see Kojima (1964), Drysdale and Garnaut (1982), and Anderson (1983). Specifi- cally, these indexes can highlight the relative importance of (seemingly minor) changes in trade between countries that have relatively small shares in global trade. If the trade intensity index takes a value above unity, the countries have greater bilateral trade than would be expected based on the partner's share in world trade. When computed for a single point in time, the measure is of limited utility because it does not incorporate the influence of factors such as distance and languages on trade. However, analysis of changes in these indexes over time can show whether two countries are experiencing an increased or decreased tendency to trade with each other. In the case of Mercosur, the magnitude of the change in this index can pro- vide a useful yardstick for assessing the importance of the expansion of intratrade. The intensity of trade index (7,y) is defined for country fs exports to country j as the share of i's exports going to / (X,y/ X,) relative to the share of ;'s imports in world imports (Mw). That is, (1) Table 2. Product Composition of Mercosur Countries' Intratrade, 1979-81 to 1994 All items Percentage of total exports (millions of Food and Agricultural Ores and Mineral All Transport and Exporter dollars) feeds materials metals fuels manufactures machinery' Argentina 1979-81 1,112.2 51.3 0.8 1.1 9.3 37.5 14.8 1984-86 739.3 48.7 1.8 1.5 11.8 36.2 10.5 1990-92 2,045.5 44.1 1.5 1.2 9.0 44.2 16.4 1993 3,684.0 34.3 0.8 0.7 18.4 45.7 24.1 1994 4,803.2 36.0 2.6 0.7 16.3 44.3 23.6 Brazil 1979-81 1,588.1 14.4 3.9 5.0 5.9 70.8 34.6 1984-86 1,175.4 13.1 3.3 9.1 4.9 69.5 24.2 1990-92 2,564.2 9.7 1.2 5.5 1.4 82.1 38.6 1993 5,393.7 10.6 1.1 3.2 3.1 81.7 40.8 1994 5,920.0 11.2 0.9 2.8 3.6 81.3 39.6 Paraguay 1979-81 116.7 36.2 45.7 0.2 0.3 17.6 n.a. 1984-86 105.4 48.8 46.1 n.a. n.a. 5.1 n.a. 1990-92 295.0 32.8 56.2 0.2 0.6 10.5 0.3 1993 287.3 23.2 58.9 0.5 0.6 16.9 1.0 1994 424.8 41.2 38.8 0.5 0.5 19.1 1.7 Uruguay 1979-81 312.2 49.5 1.7 1.8 n.a. 46.4 9.1 1984-86 270.8 51.0 4.4 0.4 1.0 41.8 5.3 1990-92 565.6 45.0 3.3 0.9 0.1 50.4 6.9 1993 661.3 39.7 2.4 0.7 0.1 56.8 18.4 1994 895.7 39.1 2.0 0.8 0.0 57.7 23.9 Mercosur 1979-81 3,129.2 31.8 4.1 3.1 6.3 54.6 23.7 1984-86 2,290.9 30.7 4.9 5.2 6.4 52.5 16.4 1990-92 5,482.0 27.4 4.5 3.1 4.1 60.8 25.0 1993 10,026.3 21.6 2.7 2.0 8.5 65.0 32.1 1994 12,043.6 24.2 3.0 1.7 8.3 62.6 30.7 n.a. Not applicable. a. Subset of all manufactures. Source: Computed from United Nations Comtrade records. Yeats 7 Brown (1947) was one of the first to use this index. Some analyses have netted out country j's imports from global imports in the denominator of equation 1 in recognition of the fact that a country cannot trade with itself. This is not true, however, for a regional block of countries. Even netting out intratrade, given the small size of Mercosur in global trade, the adjustment would produce only small changes in the index values reported. Table 3 reports trade intensity ratios that were computed using United Na- tions Series D Comtrade statistics for Mercosur countries' trade with one an- other and with NAFTA members for select periods from the late 1970s to 1994. The index documents the increased intensity of trade between Mercosur mem- bers. For example, in 1994 Argentina's intensity index for trade with Brazil was 39.2, which was more than five times its corresponding level in 1979-81. Brazil's index for trade with Argentina more than doubled over the same period. With the exception of Paraguay, where trade data are of poor quality, the bilateral intensity index ratios for Mercosur intratrade were markedly higher in 1994 than in any previous period. These results strongly reinforce the impression provided by tables 1 and 2 that exports were reoriented toward regional markets. A key question is whether these changes are consistent with efficiency and the true comparative advantage of member countries. In contrast to the pattern for Mercosur, no similar results occurred for trade with Canada and the United States. With one exception (Brazil's exports to the United States), the index for exports to both Canada and the United States were below unity. Although Mercosur members' trade with Mexico produced index values above unity, the numbers contained considerable volatility (for example, Argentina's index was 6.0 in 1979-81 and 1.5 in 1994), and lower values were generally recorded in 1994 than in the previous periods. The key message is that Mercosur members are typically becoming more dependent on trade with one another and less dependent on trade with NAFTA countries. n. ASSESSING THE EFFECTS OF REGIONAL TRADE ARRANGEMENTS Analysts often consider the effects of regional trade arrangements by fo- cusing on changes in import shares (see, for example, Sapir 1992 on the European Union). Although useful, this approach fails to address issues of efficiency in production. This article extends such analyses with several new measures based on exports to investigate whether the changes in trade were consistent with member countries' current comparative advantage. Did the increased intra-Mercosur trade take place in sectors where Mercosur coun- tries had evidenced an ability to compete in markets where they were not shielded by discriminatory trade arrangements? One way to address this ques- tion would be to determine whether Mercosur was also able to export the fastest-growing products in intratrade to third countries. In other words, did the exchange in these goods meet the test of the marketplace? Indexes avail- able for addressing this question include measures of global market perfor- THE WORLD BANK ECONOMIC REVIEW, VOL. 12, NO. Table 3. Trade Intensity Ratios for Mercosur Countries in Intratrade and for Trade with North America, 1979-81 to 1994 Trading partner Mercosur country North American country Exporter Argentina Brazil Paraguay Uruguay Canada Mexico United States Argentina 1979-81 7.4 86.8 27.5 0.3 6.0 0.7 1986-88 14.0 40.5 49.2 0.2 2.8 0.8 1994 39.2 92.7 89.4 0.1 1.5 0.7 Brazil 1979-81 9.0 80.7 19.0 0.6 3.2 1.4 1986-88 9.9 56.0 31.6 0.6 1.4 1.7 1994 21.7 71.5 36.7 0.3 2.1 1.3 Paraguay 1979-81 41.7 11.2 45.1 0.0 3.4 0.5 1986-88 48.6 41.9 48.8 0.0 0.2 0.3 1994 25.2 67.2 27.7 0.0 0.4 0.4 Uruguay 1979-81 22.7 14.6 52.2 0.4 0.6 0.7 1986-88 36.0 33.1 22.2 0.2 1.7 0.8 1994 44.7 43.7 36.8 0.2 2.1 0.4 Note: The trade intensity index (/#) is defined for country fs exports to country / as the share of Ts exports going to /- (X , / X() relative to the share of ps imports (M.) in world imports {MJ). That is, / = (X^l X) I (M(/ M.J. If the trade intensity index takes a value above unity, the countries have greater bilateral trade than would be expected based on the partner's share in world trade. Source: Computed from United Nations Comtrade Series D statistics. mance and trade orientation and measures based on the factor (labor and capital) intensities of different products. The regional orientation index (R;) for Mercosur exports of product / is de- fined as (2) /= [VX,r]/[xo//X,o]xl00 where x^ and xO] represent the value of exports of / in Mercosur's intratrade and in trade with third countries, respectively. Similarly, Xfr and Xto reflect the total value of member countries' exports within and outside the arrangement. The index value ranges between zero and infinity, with a value of unity indicating the same tendency to export the good to members and nonmembers. Increasing values above 1 indicate a greater tendency to export to regional markets. Several specific points should be noted with regard to the properties of this index. First, the index conveys only limited information about trade patterns if computed for a single point in time. Various factors, such as comparative ad- vantage, transport costs, or trade barriers in alternative markets, determine the geographic orientation of trade. However, intertemporal comparisons of this index over relatively short periods can provide useful information about changes in the geographic pattern of trade. Second, in the short to medium term, changes in comparative advantage, transport costs, or relative tastes should be minimal, Yeats 9 so factors such as differential changes in trade barriers (such as those that ac- companied the formation of Mercosur) would influence changes in the index value more heavily. Third, examined in isolation, the percentage changes in ex- ports of different goods within a regional arrangement do not indicate the changes in demand for products in third markets. For example, products with the high- est growth rates within Mercosur could move away from the region if exports to third markets were growing even faster. The regional orientation index does not suffer from this defect. In addition to the regional orientation measure, a second index reflects re- vealed comparative advantage (RCA) and can be computed for each country in the arrangement and for each product traded. This measure (C;) is defined as (3) C; = [x o/ /XJ/[x^/X*] where x ^- and X * represent world exports of product / and total world exports exclusive of the regional trade between member countries, respectively. The in- dex value ranges between zero and infinity; values above unity indicate that the country has a revealed comparative advantage in the product. The index ex- cludes regional trade in order to reflect more accurately the capacity of Mercosur members to compete evenly in markets where discriminatory trade arrangements do not provide an unnatural edge. Analysts generally compute RCA indexes only for processed goods or manu- factures because trade in agricultural products is distorted by export incentives and trade barriers that are likely to obscure whether a country has a real com- parative advantage or disadvantage in these products. The present analysis does not attempt to derive revealed comparative advantage indexes for agricultural products and other primary commodities. As such, it is based on 128 three-digit srrc products that include all manufactured goods as well as a number of pro- cessed foodstuffs and processed raw materials. However, the composition of Brazilian (and therefore Mercosur's) manufactures exports may still be distorted by this country's long history of providing export subsidies. Direct comparisons of the regional orientation and RCA indexes suggest the extent to which Mercosur distorted exports from the patterns consistent with comparative advantage. Although the two indexes do not measure import diver- sion directly, they provide closely related information that can be used to infer whether the additional trade generated by Mercosur was primarily in products in which Mercosur countries had low enough costs to be competitive in third markets. If the Mercosur countries were not competitive in those products, then the additional trade within Mercosur could have been replaced by more efficient outside suppliers. The issue is essentially whether regional trade ar- rangements foster high-cost imports at the expense of low-cost ones. Traditional calculations analyze the displacement of imports from nonpartners by those from partners, implicitly comparing partner and nonpartner costs by their relative competitiveness in the regional market before the regional trade arrangement. The supplementary view developed here makes inferences about high and low 10 THE WORLD BANK ECONOMIC REVIEW, VOL. 12, NO. 1 costs by implicitly comparing the relative competitiveness of partner and nonpartner goods in world markets. This study tests the use of these indexes in this context using actual trade data. Additional data provide useful supplementary information concerning changes in the composition and direction of trade that accompanied the forma- tion of Mercosur. Table 4 lists the 30 three-digit SITC processed products that met two separate criteria. First, they recorded at least $250,000 in intratrade in 1988 (this lower limit was set to prevent the tests from being biased by marginal products, while 1988 was selected as a period prior to the implementation— and likely the anticipation—of major preferences). Second, they registered the highest 1988-94 compound annual growth rates in trade among Mercosur members. Table 4 provides the values for the regional orientation index and the relative factor intensity in production for the 30 product groups. The regional orienta- tion index for 1988 and the rate of change in the index over 1988-94 indicate the extent to which a reorientation of trade toward the region contributed to each product's dynamism. The factor intensity index is drawn from World Bank (1992) and takes a value of 100 for products whose labor intensity is average relative to that of all manufacturing activity. Increasing values above this level identify goods that are more capital-intensive in production (conversely, decreas- ing values below 100 identify products that are more labor-intensive). The factor intensity for industry/ (L;) is defined as (4) L;.= [ ( V , / N , ) / ( V ( / N , ) ] x l 0 0 where V; and V, represent value added in industry / and in all manufacturing in the United States, respectively, while N ; and N, represent the number of workers in the industry and in all manufacturing activity, respectively. Studies by the National Bureau of Economic Research show that products manufactured by labor-intensive processes in the United States are also manufactured by rela- tively labor-intensive processes in other countries, although the levels of use may differ (Lary 1968). On this basis, Lary justifies computing the index with data from the U.S. Census of Manufactures. There is an inverse relation between the numeric value of the index and the labor intensity of a given product. That is, the lower the numeric value, the higher the labor intensity. See Yeats (1989) for details concerning computation of the index and the results reported in table 4. Much of the dynamism in the intratrade of the goods in table 4 was associ- ated with a shift in the regional orientation of exports toward Mercosur. For example, in 1988 the trade-weighted share of these products within Mercosur was approximately 49 percent higher than that for other destinations, as re- flected in a value of 1.49 for the regional orientation index. By 1994 this index value rose more than twofold to 4.55. Road motor vehicles (srrc 732) played a major role in this overall shift, as intra-Mercosur trade in these goods increased by a factor of more than 10, in value terms from $207 million in 1988 to more than $2.1 billion in 1994. Of these 30 dynamic products, 24 recorded a shift in Yeats 11 the regional orientation index toward Mercosur, and the indexes of half of the 30 products more than doubled. High tariffs and nontariff measures may have constrained exports to third markets, particularly for agricultural products. In these cases, values for the regional orientation index could have risen because of restrictions in third mar- kets and not because of the more favorable tariff treatment of member countries in interblock trade. There is reason to believe, however, that this situation oc- curs infrequently. Records of the World Bank and United Nations Conference on Trade and Development (UNCTAD) show that most of the products listed in table 4 do not encounter major OECD restrictions (with the exception of pro- cessed foodstuffs). In addition, Mercosur provides sizable trade preferences on intratrade. It is also possible that idiosyncrasies in demand patterns and in the ability to produce certain varieties of goods made trade between Mercosur coun- tries increase disproportionately fast as a result of the most-favored-nation lib- eralization. This is sufficiently unlikely, however; the burden of proof must lie with advocates of this view. Although the results are somewhat mixed, another important point is that the dynamic products generally consist of goods that are relatively capital-intensive in fabrication. Overall, the 30 items listed in table 4 had a factor intensity index ratio of 118, which indicates that they were 18 percent more capital-intensive than average for all manufacturing activity. In contrast, Yeats (1989) employs these same data and determines that the recent capital intensity of exports for Hong Kong, the Republic of Korea, and Taiwan (China) were 20 to 25 percent below average, while Singapore's ratio was about 2 percent higher than average. The high capital intensity of Mercosur's exports is troubling because it raises the question of how capital-intensive goods from one developing country to another can compete with similar exports from industrial countries in the absence of discriminatory trade measures. Comparative advantage explanations of the composition of trade between industrial and developing countries generally focus on factor proportions. That is, countries with a relative abundance of low-cost labor should export labor- intensive products to countries where capital is relatively abundant. Empirical tests by Lary (1968), Tuong and Yeats (1980), and Yeats (1989) confirm the accuracy of factor proportions as a predictor of trade flows. Theory is less con- clusive in explaining the composition of trade between developing countries. However, although countries may import capital-intensive goods from indus- trial countries and labor-intensive goods from developing countries, nearly all models suggest that exports will be concentrated in one part of the factor inten- sity spectrum. Donges (1987) concludes that discriminatory trade arrangements among developing countries foster this exchange. Havrylyshyn (1987) and Havrylyshyn and Wolf (1987) also indicate that domestic and trade policy dis- tortions promote these exports. Table 5 provides a different perspective on the shifts in Mercosur's intratrade and the efficiency implications of these changes. The table presents the 30 prod- Table 4. Dynamic Products in the Intratrade of Mercosur Countries, 1988-94 Exports to Mercosur Regional orientation index*1 Factor (thousands of dollars) Percentage point intensity Trade growth Commodity' 1988 1994 1988 change, 1988-94 index* rate, 1988-94 Nonalcoholic beverages (111) 349 26,238 2.35 46.12 162 1.05 Tobacco manufactures (122) 2,032 112,681 0.95 0.91 210 0.95 Articles of plastic (893) 4,225 95,535 12.45 -7.52 86 0.68 Alcoholic beverages (112) 4,137 81,671 1.87 4.61 182 0.64 Perfumes and cosmetics (553) 4,766 86,282 5.22 8.16 160 0.62 Furniture (821) 3,972 66,213 1.15 -0.10 60 0.60 Iron and steel castings (679) 287 3,696 0.30 0.41 145 0.53 Nonelectric power machinery (711) 25,140 290,687 0.40 0.79 105 0.50 Nonmotor road vehicles (733) 3,118 35,854 2.23 4.65 90 0.50 Wood manufactures (632) 1,472 16,689 0.43 -0.10 80 0.50 Machines for special industries (718) 10,763 120,617 0.93 0.04 113 0.50 Structures and parts (691) 1,783 19,834 0.72 1.05 105 0.49 Prepared meat (011-013) 21,934 237,912 0.17 0.23 102 0.49 Motor road vehicles (732) 206,996 2,112,750 1.25 3.17 122 0.47 Plywood and veneers (631) 3,707 35,630 0.20 0.16 57 0.46 Lace and ribbons (654) 1,386 13,157 3.56 2.29 68 0.46 Special textile products (655) 4,945 46,919 0.88 0.59 69 0.46 Prepared sugar (061-062) 11,456 102,655 0.30 0.07 140 0.44 Prepared dairy (022-024) 23,495 204,019 4.31 18.17 180 0.43 Metal manufactures, not specified elsewhere (698) 5,984 51,430 0.90 0.87 84 0.43 Electric power machinery (722) 14,278 121,717 1.18 -0.33 88 0.43 Materials of rubber (621) 3,636 30,780 3.13 3.13 100 0.43 Glassware (665) 5,381 45,017 2.21 2.88 107 0.42 Nails, nuts, and bolts (694) 3,021 24,782 2.86 0.81 88 0.42 Preserved fruit (053) 4,486 36,053 0.05 0.08 116 0.41 Domestic electrical equipment (725) 12,568 97,322 2.19 3.76 92 0.41 Base metal household equipment (697) 5,592 40,452 2.72 -0.84 79 0.39 Nonfur clothing (841) 19,342 138,805 0.63 0.71 64 0.39 Metal tanks and boxes (692) 3,960 28,099 2.97 -0.09 82 0.39 Copper (682) 3,001 21,161 0.54 0.05 120 0.38 Total 417,212 4,344,657 1.49 3.06 118 0.48 Note: Mercosur countries are Argentina, Brazil, Paraguay, and Uruguay. a. Numbers in parentheses are the SITC codes. b. The regional orientation index (R^ for Mercosur exports of product/ is defined as R^ = [(x^l Xlr) I (x^l Xlo)\ x 100, where xn and xo/ represent the value of exports of/ in Mercosur's intratrade and to third countries, respectively. Similarly, Xlr and XK reflect the total value of member countries' exports within and outside the arrangement. An index value above unity indicates a greater tendency to export the good to regional markets. c. The factor intensity index for industry / (L) is defined as L = [(V / N^ / (Vt I N,)] x 100, where V and V, represent value added in industry / and in all manufacturing in the United States, respectively, while N and N, represent the number of workers in the industry and in all manufacturing activity, respectively. The higher the index, the higher the capital intensity of the production process. Industries with an index value of 100 would have a labor/capital intensity that is average for all manufacturing activity. For information on how the index is derived, see Lary (1968) or World Bank (1992). Source: Derived from United Nations Comtrade statistics. Table 5. Mercosur Exports with the Largest Change in Regional Orientation toward Mercosur Markets, 1988-94 Revealed Exports comparative advantage Commodity 1988 1994 1988 1994 Change, 1988-94 index*, 1994 Nonalcoholic beverages (111) 349 26,238 2.35 48.47 46.12 0.05 Lead (685) 642 219 3.03 25.42 22.39 0.00 Prepared dairy (022-024) 23,495 204,019 4.31 22.49 18.17 0.13 Nonwheat meal or flour (047) 4 954 0.05 17.26 17.21 0.04 Perfumes and cosmetics (553) 4,766 86,282 5.22 13.37 8.16 0.14 Wheat meal or flour (046) 65 35,051 0.22 5.67 5.44 1.08 Cork manufactures (633) 18 721 1.18 6.30 5.13 0.05 Preserved vegetables (055) 23,404 48,745 17.66 22.61 4.95 0.13 Articles of paper (642) 15,763 72,249 2.16 7.10 4.93 0.20 Nonmotor road vehicles (733) 3,118 35,854 2.23 6.88 4.65 0.13 Alcoholic beverages (112) 4,137 81,671 1.87 6.48 4.61 0.19 Agricultural machinery (712) 39,608 121,294 2.08 5.88 3.81 0.45 Domestic electrical equipment (725) 12,568 97,322 2.19 5.94 3.76 0.23 Road motor vehicles (732) 206,996 2,112,750 1.25 4.42 3.17 0.45 Materials of rubber (621) 3,636 30,780 3.13 6.26 3.13 0.32 Glassware (665) 5,381 45,017 2.21 5.09 2.88 0.38 Syntheticfibers(266) 13,381 21,170 6.28 9.14 2.87 0.11 Rice, glazed or polished (042.2) 22,583 148,079 9.28 11.65 2.37 1.03 Lace and ribbons (654) 1,386 13,157 3.56 5.86 2.29 0.22 Food preparations not specified elsewhere (099) 7,727 45,412 2.10 4.35 2.25 0.28 Structures and parts (691) 1,783 19,834 0.72 1.77 1.05 0.39 Tobacco manufactures (122) 2,032 112,681 0.95 1.86 0.91 0.39 Textile yarn and thread (651) 26,523 118,120 0.85 1.73 0.88 0.90 Metal manufactures not specified elsewhere (698) 5,984 51,430 0.90 1.77 0.87 0.33 Nails, nuts, and bolts (694) 3,021 24,782 2.86 3.66 0.81 0.28 Nonelectric power machinery (711) 25,140 290,687 0.40 1.20 0.79 0.45 Nonfur clothing (841) 19,342 138,805 0.63 1.34 0.71 0.27 Plumbing and lighting equipment (812) 3,819 14,363 2.23 2.93 0.70 0.13 Electrical distributing machinery (723) 6,821 35,775 1.55 2.22 0.67 0.26 Glass (664) 4,851 25,079 1.27 1.93 0.65 0.36 Total 488,345 4,058,540 2.83 5.97 3.14 0.31 a. Numbers in parentheses are the SITC codes. Mercosur countries are Argentina, Brazil, Paraguay, and Uruguay. b. The regional orientation index (R^) for Mercosur exports of product / is defined as R^ = [(x^ I Xn) I (x^ I Xlo)] x 100, where x^ and x . represent the value of exports of/ in Mercosur's intratrade and to third countries, respectively. Similarly, Xa and Xto reflect the total value of member countries exports within and outside the arrangement. An index value above unity indicates a greater tendency to export the good to regional markets. c. The revealed comparative advantage index (C() is defined as C y = [(xoj/ Xw) / {x* / X*w)] x 100, where x* and X ^ represent world exports of product/and total world exports exclusive of the intratrade of the regional trade arrangement member countries, respectively. Values above unity indicate that the region has a revealed comparative advantage in the product. Source: Computed from United Nations Comtrade records. 16 THE WORLD BANK ECONOMIC REVIEW, VOL 12, NO. 1 uct groups that experienced the greatest reorientation of trade toward the region (as measured by the regional orientation index) over 1988-94. In addition, it shows the modified 1994 RCA index (based only on trade performance in third markets) for each item. The two measures can be compared to determine whether the goods that assumed steadily increased importance in intratrade were among those that Mercosur was able to export competitively to third countries. In other words, the comparison shows whether Mercosur countries showed any evidence of export strength in these goods in independent markets where they were not protected by discriminatory trade arrangements.1 The results reflected in table 5 are discomforting. For the 30 groups with the largest regional shift, only two (SITC 046—wheat meal or flour—and SITC 042.2— glazed rice) had RCA indexes slightly above unity, while the index averaged only 0.27 for the other items. These results strongly suggest that Mercosur members experienced a strong comparative disadvantage for these goods in markets that did not incorporate discriminatory trade measures against outsiders. The re- vealed comparative advantage index averaged only 0.07 for the top five prod- ucts listed in the table. Mercosur countries do receive OECD preferences for some of the products in table 5 under the Generalized System of Preference (GSP) schemes that were adopted in the early 1970s. These GSP preferences have since been eroded by tariff cuts in the multilateral trade negotiations and now average about 1 to 2 percentage points for products that are eligible for such treatment. In contrast, section III shows that the preferences that Mercosur countries extend to one another are far higher than the G SP margins that developing countries have ar- gued are vital to their ability to compete in international markets. The state- ments that developing countries have made on the GSP strongly suggest that dis- criminatory tariffs set at Mercosur's levels have the potential to displace a significant amount of exports from third countries. Analysis of the underlying 1988-94 changes in Mercosur's RCA indexes for the products in table 5 reveals another disturbing trend. The average index value actually declined from 0.42 to 0.31 over the six-year period. Furthermore, the reductions were widely distributed within the group, with 21 of the 30 (70 per- cent) items recording lower RCA values in 1994 than in the earlier period. In short, the evidence suggests that Mercosur became less, rather than more, inter- nationally competitive in products where trade was reorienting most rapidly toward the region. What factors are responsible for this surprising reorientation of trade? Evi- dence suggests that Mercosur's own trade barriers were the cause. The analysis in section III shows that goods, such as those listed in table 5, generally are 1. The regional orientation and R CA indexes both depend on the shares of each good in Mercosur's total exports. To the extent that these are measured with error, the two will be negatively correlated. There is no reason to be particularly worried by this, however, because trade shares are fairly robust at this level. Besides, nearly all our analysis relates changes in regional orientation to the starting level of revealed comparative advantage. Yeats 17 protected by higher-than-average discriminatory trade measures. Local produc- ers would have a strong incentive to seek the higher prices available on sales to Mercosur markets. Given the option of selling locally at higher prices, producers would have a strong incentive to divert exports from more competitive foreign markets to less competitive regional markets. As a result, RCA indexes would decline for the products that are growing fastest in intratrade. Figure 1 provides a graphical view of the major changes that were occurring within the product composition of Mercosur's intratrade. The upper half of the figure shows aggregate RCA indexes for the 15 product groups (defined here at the two-digit SITC level) that accounted for almost all (92 percent) of the 1988- 94 change in Mercosur's intratrade. The width of each product's bar is propor- Figure 1. Regional Orientation and Revealed Comparative Advantage Indexes for Mercosur's Intratrade, 1994 Revealed comparative advantage Index 100 0.75 050 " II 0 25 II o g Z 5 Product group Regional orientation 2.00 1.00 Produa group Note-. The products in the figure account for 92 percent of the change in Mercosur's intratrade in 1988-94. The width of each bar is proportional to the share of the produa in the 1988-94 increase in Mercosur's intratrade. Products are defined at the two-digit level of the srrc Revision 1. See text definitions of the indexes. Source. Author's calculations. Table 6. Mercosur's Regional Orientation Index and Revealed Comparative Advantage Index by Product Decile, 1988-94 Regional Change in Revealed Percentage of industries Mercosur intratrade orientation regional comparative with a revealed comparative (thousands of dollars) index11 orientation index, advantage index? advantage index greater Product dealer 1988 1994 1988 1994 1988-94 1988 1994 than unity, 1994 First 127,939 810,619 3.84 13.47 9.63 0.43 0.24 7.6 Second 325,573 3,033,899 1.63 4.26 2.64 0.64 0.48 7.6 Third 212,759 1,027,964 1.39 1.86 0.47 0.88 0.54 15.4 Fourth 302,175 1,303,419 1.50 1.67 0.17 0.96 0.99 46.1 Fifth 108,996 597,859 l.ll 1.16 0.04 2.15 2.14 38.5 Sixth 167,733 840,060 1.33 1.12 -0.21 2.65 3.68 38.5 Seventh 287,832 595,436 1.15 0.78 -0.37 3.04 3.60 69.2 Eighth 222,842 469,499 1.73 0.96 -0.77 2.05 2.60 46.1 Ninth 373,557 516,184 3.48 1.33 -2.14 0.86 2.06 46.1 Tenth 226,930 514,461 31.82 10.17 -21.65 0.24 1.99 23.1 Total 2,356,437 9,709,400 a. Product deciles are ranges that include 10 percent of all observations after they have been ranked in either ascending or descending order. The deciles are based on 128 processed product groups. b. The regional orientation index (Rf) for Mercosur exports of product / is defined as R( = [(xn I Xn) I {x^ I Xa)] x 100, where xn and x , represent the value of exports of / in Mercosur's intratrade and to third countries, respectively. Similarly, X^ and X^ reflect the total value of member countries exports within and outside the arrangement. An index value above unity indicates a greater tendency to export the good to regional markets. c. The revealed comparative advantage index (C) is defined as C, = [(x^ IXK) I (x ^ / X *J] x 100, where * ^ and X*m represent world exports of product / and total world exports exclusive of the intratrade of the regional trade agreement member countries, respectively. Values above unity indicate that the region has a revealed comparative advantage in the product. Source: Computed from United Nations Comtrade records. Yeats 19 tional to its share of the total increase. Three groups—transport equipment, nonelectrical machinery, and electrical machinery—accounted for more than half of the total 1988-94 increase in Mercosur's intratrade, even though the RCA indexes for these products were very low (in the 0.25 to 0.40 range). To help interpret these data, the lower half of the figure plots the matched regional ori- entation index for each group. For these 15 product groups, a consistent pattern of (very) low RCAs and high regional orientation is generally observed. In only one product group (iron and steel), which accounted for only about 3 percent of the increase in intratrade, did Mercosur achieve a revealed comparative advan- tage index above unity. In short, figure 1 provides little evidence that Mercosur's intratrade was evolving along lines consistent with current comparative advan- tage. Rather the products recording the largest shift toward the region were those for which Mercosur had not demonstrated an ability to export competi- tively elsewhere. While the discussion of table 5 focuses on extreme (positive) changes in the regional orientation index, table 6 provides a composite view of the complete profile of intratrade changes occurring within Mercosur. Table 6 presents the regional orientation indexes for 1988 and 1994 for the 128 processed products, ranked in descending order based on changes in this index and grouped by decile. The table shows the average RCA and regional orientation indexes by decile and compares the 1988-94 change in the regional orientation index with Mercosur's average RCA indexes. The results presented in table 6 reveal a pattern that differs from that ex- pected on the basis of the comparative advantage measure. For the first decile products, those that registered the largest shift toward the region (an average regional orientation index increase of 9.63), the average 1994 Mercosur RCA index was only 0.24. Similarly, for the second decile products, the regional ori- entation index change took a value of 2.64, in spite of the fact that the average RCA index for these goods was under 0.50, and actually declined from its earlier 1988 level. The average RCA index declined between 1988 and 1994 for the top three decile product groups, indicating that Mercosur became less, not more, internationally competitive for goods experiencing the greatest shift toward intratrade. Because the goods in these three deciles enjoy well-above-average levels of protection against third countries (see section HI), domestic producers would have a strong incentive to divert trade to local markets (causing the RCA indexes to decline) in order to profit from the higher prices. Mercosur's regional orientation indexes showed a fairly consistent tendency to move counter to the revealed comparative advantage measure over the decile ranges. For example, in the sixth through tenth decile products, where the 1994 RCA index ranged between 1.99 and 3.68, exports to third countries grew far faster than those to the region. The fairly high index values in the fifth through tenth deciles were in part due to the fact that the distribution of index values is right-skewed. That is, the index is bound by zero but can range (in theory) to infinity. One or two products in a decile group with high RCA index values may 20 THE WORLD BANK ECONOMIC REVIEW, VOL. 12, NO. 1 have influenced the results for these decile groups. The key point that follows from this complete profile of trade changes is that the regional orientation of exports is growing most rapidly for products where there is little evidence that Mercosur has a current comparative advantage. m. MERCOSUR'S TRADE BARRIERS AND RECENT TRADE CHANGES What caused the observed changes in the pattern of trade to diverge so widely from the pattern expected on the basis of efficiency conditions and current com- parative advantage? It seems highly unlikely that they were due to changes in natural factors such as freight costs. Transport costs can be reduced by adopting new shipping technologies or by making major improvements in port and han- dling facilities, but these are unlikely to occur in a relatively short time period such as that covered by this study. The discriminatory nature of Mercosur's own trade policies might explain the changes in the pattern of trade. The Mercosur countries had been liberaliz- ing imports on a most-favored-nation basis for several years when, in 1991, they introduced a widespread set of preferential tariff cuts. United Nations Comtrade records indicate that in 1991 intrablock trade accelerated sharply. If preferences were highest for the most dynamic products in Mercosur's intratrade (products that were shifting most rapidly toward the region), this would suggest that Mercosur's trade barriers affected the reorientation of exports. Evidence relat- ing to this point could come from an analysis of the margins of preference that Mercosur's trade barriers provide to member countries. Are these margins high enough to account for the increases in intratrade that occurred during the 1991-94 period, when tariff preferences on all but a few products were being implemented? Several sources of statistics on Mercosur's tariffs and nontariff barriers are available for analyses of these points. A cooperative project between UNCTAD and the World Bank, named SMART—Software for Market Analysis and Re- strictions on Trade—compiled statistics on many OECD and developing coun- tries' pre-Uruguay Round trade barriers (see UNCTAD and World Bank 1989 for a description of the SMART database and operating system). Because the records include both Brazil's and Uruguay's 1988-89 tariffs (along with data on Brazil's nontariff measures), they provide partial details on Mercosur's trade barriers at very fine levels of detail. These two countries account for more than 60 percent of Mercosur's total imports. Thus the SMART records provide a useful profile of the structure of external protection. However, the Mercosur countries (particularly Brazil) subsequently implemented major unilateral re- ductions in most-favored-nation tariffs, making the earlier statistics an unreli- able guide to current levels of protection. For this reason, the analysis uses post-Uruguay Round tariff data directly from the World Trade Organization's Integrated Data Base (iDB). Known exceptions and departures from the re- ported World Trade Organization (WTO) statistics (for example, tariffs and Yeats 21 nontariff restrictions on automobiles) were incorporated into the data. These statistics, like those in SMART, are recorded at the level of the national tariff line, so the two sources of information are comparable. The SMART and IDB records, together, afford reasonable insight into changes in the pattern of pro- tection in the early 1990s. As far as intrablock preferences are concerned, Article 5 of the Treaty of Asuncion, which created Mercosur, required the implementation of progressive, linear, and automatic tariff reductions starting in 1991 with a view to arriving at zero tariffs for the entire tariff area by 31 December 1994. Although there was some slippage in this objective, the World Trade Organization (1996: 39) re- ports "that by the end of 1994 intraregional trade between Brazil and its Mercosur partners has been duty-free since 1 January 1995 except for 29 tariff line items" (out of 9,107).2 The same report quotes Brazilian authorities as stating that duty-free trade covered close to 95 percent of intraregional (Mercosur) trade in 1994. These observations indicate a high correlation between the reported ex- ternal tariffs (tariffs facing countries that are not members of Mercosur) and the degree of preferences offered on intra-Mercosur trade. Moreover, if in fact the bulk of the preferences was not implemented until late 1994, the trade reorien- tation would have arisen from partial preferences; full implementation of the preferences in 1995 and 1996 would have had an even larger impact. This analysis computes the average levels of current "applied" tariffs facing non-Mercosur countries and takes them as a measure of intrablock preferences. The analysis also attempts to identify discriminatory trade barriers by using two measures of the potential importance of nontariff barriers (NTBs). The first, an NTB trade coverage ratio, shows the share of all imports (measured in current values) subject to NTBs. The second, often referred to as a frequency index, shows the share of all tariff line products covered by one or more nontariff restrictions. Although these measures have some limitations (see Laird and Yeats 1990 for a discussion), they show the extent to which nontariff restrictions reinforce the effects of the block preferences. The analysis uses data from 1988, and NTBs have been subject to significant liberalization since then. As such, the NTB data are most useful for indicating the extent of distortions that were in place in the 1980s but probably are less reliable in the present context. However, again ap- pealing to the normal persistence in patterns of protection and noting that the Treaty of Asuncion provides for the removal of NTBs on intrablock trade, the data very likely identify the elements of discrimination. Given the tariff data, however, the thesis that discrimination exists does not depend on these NTB measures. 2. Behar (1995) reports that the first preferential reduction in Mercosur's tariffs took place in June 1991. A 1994 background study for a World Bank country economic memorandum gave the schedule for reducing tariffs on intratrade. This schedule set a minimum margin of preference of 47 percent and augmented the margins of preference that already exceeded this minimum (due to previous arrangements) by an additional 7 percent. A 7 percent cut every six months would follow until the zero tariff objective was fulfilled. As noted, in the case of Brazil, this objective was achieved for all but 29 tariff line items. Table 7. Tariffs and Nontariff Barriers on Mercosur's Imports by Product Decile, 1988-94 Nontariff barrier Changes in Change in Mercosur Average external ratio, 1988 the regional Product intratrade, 1988-94 tariff (percent) Frequency Trade coverage orientation index, decile" (thousands of dollars) 1988 1996b index0 index6 1988-94' First 682,680 54.4 18.1 41.9 83.5 9.63 Second 2,708,326 46.9 16.7 24.5f 36.6f 2.63 Third 815,205 56.1 17.5 42.1 51.8 0.47 Fourth 1,001,244 46.0 9.6 17.7 28.7 0.17 Fifth 488,863 46.0 10.3 20.4 8.3 0.05 Sixth 672,327 43.6 16.7 17.1 4.3 -0.21 Seventh 307,604 38.4 10.9 18.5 27.5 -0.37 Eighth 246,657 39.1 11.7 27.1 31.9 -0.77 Ninth 142,627 38.4 9.0 4.0 5.4 -2.15 Tenth 287,531 40.7 10.9 8.2 28.2 -21.65 a. The 1996 tariff statistics are drawn from the WTO's Integrated Data Base and are the average duties actually applied by Argentina, Brazil, and Uruguay to imports from non-Mercosur sources. They are based on the lower of the following two rates: the legally bound tariff or the current most-favored-nation applied rate. A description of the methodology used in computing these averages can be found in Finger, Ingco, and Reincke (1996). b. See table 6. The deciles are based on 128 processed product groups. c. The frequency index (F) for importing country / shows the percentage of tariff lines covered by some preselected group of nontariff measures and is defined by F y = (ZDNi + N() x 100, where N; is tariff line item i, D ; is a dummy variable that takes a value of unity if one or more NTBS is applied to the item and zero otherwise, and N, is the total number of lines in the product group. The summation is made over all countries exporting to importing country /'. d. The trade coverage index (C.) is defined as C = [(ED^,.^ x V ^ ) /XVl(_J x 100, where Vll_m is the value of imports in tariff line item i in year t-n, and Dn is a dummy that takes a value of unity if an NTB is applied to the item and zero otherwise. If n and m are zero, the index is based on current trade values; otherwise base year weights are used. e. The regional orientation index (R^ for Mercosur exports of product / is defined as R( = [{x^ I XJ I (x^ I Xro)] x 100, where xn and x^ represent the value of exports of / in Mercosur's intratrade and to third countries, respectively. Similarly, X^ and Xu reflect the total value of member countries exports within and outside the arrangement. f. Includes Brazil's recent restrictions on automobile imports and domestic automobile assembly regulations. Source: Computed from SMART and the World Trade Organization's Integrated Data Base. Yeats 23 The frequency index (F;-) for importing country / shows the percentage of tariff lines covered by some preselected group of nontariff measures. Similarly, the trade coverage index (C7) is defined as (5) C, = [(ZD,,_ x V ^ ) / IV,.^] x 100 where V, ,_„ is the value of imports in tariff line item / in year t-n and Dit is a dummy that takes a value of unity if an NTB is applied to the item and zero otherwise. If n and m are zero, the index is based on current trade values; other- wise base year weights are used. The UNCTAD records contain information about the following types of nontariff restrictions that are included in the frequency and coverage ratio tabulations: tariff quotas, antidumping duties, restrictive import authorizations, total pro- hibitions, suspended import authorization licenses, antidumping investigations, state monopolies, differential health and safety regulations, and differential prohibitions based on noncommercial considerations. The records may under- state the current importance of some restrictions like antidumping actions be- cause they do not incorporate data on the surge in new cases that occurred in 1992-93. Table 7 uses the product deciles from table 6 to summarize several statistics relating to NTBs. The data strongly suggest that tariff preferences, and the pro- tection they provide for intratrade, were a major factor behind the recent pat- tern of trade changes. For example in 1996, Mercosur tariffs on the most dynamic products—the first decile products from table 6—averaged 18 percent, about 5 percentage points higher than the average duty on all goods in the 10 decile groups. Items in the second and third decile groups had average discrimi- natory tariffs of about 17 percent. Those in the fourth through tenth decile groups typically declined to about 10 to 12 percent. These numbers exceed by far the average tariff margin provided by the GSP (about 1 to 2 percentage points). Many developing countries strongly sought to preserve the GSP tariff margins in the Uruguay Round negotiations on the basis that they strongly enhanced the com- petitive position of their exports. The Mercosur tariffs also exceed the average margin within EFTA and the European Union (where the average is below 3 per- centage points for industrial goods). Moreover, even including agricultural goods in the Mercosur common external tariff for the year 2000, the average margin of preference would be about 10 percentage points. In 1995 the average external tariff for the second decile products in table 7 would have been about 22 percent, because the Brazilian tariff on autos was raised to 70 percent on some imports. Recent actions within the World Trade Organization accent the importance that member countries attach to the dis- criminatory trade barriers against third countries in this sector. For example, a recent article in the Journal of Commerce (Zarocostas 1996) quotes a WTO re- port that the auto industry is the most highly assisted manufacturing activity in the country, with effective protection estimated at more than 250 percent. The article also notes that a strict tariff quota had been set on auto imports and that 24 THE WORLD BANK ECONOMIC REVIEW, VOL. 12, NO. 1 the United States, Japan, the European Union, and the Republic of Korea had all filed complaints with the WTO alleging discrimination in Brazil's auto trade re- gime. The fact that these new protectionist measures had to be introduced, in spite of the major expansion of domestic production, suggests that any scale economies accompanying the regional arrangement are proving very elusive. Table 7 shows that NTBs were initially structured along lines that reinforced the trade-distorting effects of the agreement's preferential tariffs. According to the UNCTAD data, nontariff restrictions were applied to about 21 percent of all tariff line items, which is almost half the corresponding ratio (41.9 percent) for Mercosur's first decile products in 1988. Similarly, the trade coverage ratio for products falling in the first three deciles averaged almost 60 percent (about three times the ratio for all imports) and reached 84 percent for the first decile group. IV. POLICY IMPLICATIONS This article identified dramatic changes in the Mercosur countries' trade pat- terns over the period 1988-94, particularly the last few years. It also argued that these changes were substantially due to the changes in trade policy and probably mainly those introduced under Mercosur transitional arrangements. Statements of this kind depend on comparing actual data with a counterfactual. This article presented the implicit counterfactual that, but for these policy developments, Mercosur's trade patterns would not have changed much and the shift toward apparently uncompetitive capital-intensive intrablock trade would not have oc- curred. Thus, if the Mercosur countries had achieved an equivalent degree of liberalization on a nondiscriminatory basis, they would have maintained a more efficient import structure, paying less or obtaining better goods, and they would have purchased more from their trading partners outside the block. The analysis compared trade patterns in 1994 with those in 1988 (not with some free-trade counterfactual), assuming that they were no more distorted in 1994 than in 1988. In fact, it is likely that trade in 1994 was less distorted geographically because, in the earlier period, trade barriers were applied on a most-favored- nation basis with the exception of a few sectoral arrangements. In the relatively short-run period covered by the analysis, it is unlikely that comparative advantage changed significantly (the RCA analysis supports this point). It seems likely that a most-favored-nation liberalization would not have resulted in major changes in the share of imports. Given the size of the trade effects identified, the analysis finds compelling evidence that preferences and the application of industrial policies within regional trade arrangements can be distortionary. There may be other explanations for the changes in trade pattern, but none seems nearly so direct, or so likely, as the role of trade policies and discriminatory trade barriers. A separate, albeit related, question based on a second counterfactual is whether Mercosur countries' trade was more distorted in 1994 than in 1988. Unfortu- nately, analyses of trade data alone cannot answer this question completely. Yeats 25 Although the trade pattern is more distorted, it is possible that the efficiency of the tradeoffs between domestic and foreign goods in general (roughly speaking, the level of trade) increased sufficiently to offset this. Comparing 1994 with 1988, the Mercosur trade arrangements most likely both created and diverted trade. Even trade-diverting customs unions can improve economic welfare if they lead to declines in producer and consumer prices. Thus, while Mercosur trade policies are distortionary relative to what could have been achieved, they may well have been positive relative to the starting point for member countries. This discussion raises an important semantic point: what is Mercosur? Broadly speaking, the sample period for most of the analysis (1988-94) witnessed two classes of policy change: the general liberalization of trade barriers facing all partners, which was stronger over 1988-91 but not restricted to that period, and the preferential intrablock liberalizations, which occurred mainly after the Treaty of Asuncion in 1991. Lack of data on the precise path and structure of tariffs make it possible to consider only the combined effects of these two liber- alizations, loosely referred to as "the effects of Mercosur" rather than "the ef- fects of the trade policies of the Mercosur countries." Thus the estimate of Mercosur effects includes, strictly speaking incorrectly, both the effects of the general trade barrier liberalizations over 1988-91 (and any continuing tariff reductions after 1991) that produced an impressive surge in global imports and the effects of the Argentinean-Brazilian bilateral sectoral liberalizations over 1988-91. Including the effects of the general trade barrier liberalizations leads to an understatement of the discriminatory effects of Mercosur proper; includ- ing the effects of the bilateral liberalizations leads to an overstatement. It should also be noted that the restrictive auto regime is strictly a national rather than a Mercosur tariff regime. Some commentators on this article have argued that Mercosur's effect has been to constrain the restrictiveness of this regime. The alternative approach of limiting the analysis of trade changes to the pe- riod since the Treaty of Asuncion was not pursued here precisely because it was unclear how much discrimination was built into 1991 and how much general liberalization after 1991 was strictly due to the Mercosur agreement. Indeed, given that Mercosur has so little institutional structure outside its member gov- ernments, it did not even seem meaningful to try to disaggregate the changes in trade policy into Mercosur and national components. An example of a national component would be Brazil's recent unilateral increase in tariffs on automotive products and the imposition of quotas on these goods. The findings of this study appear to constitute convincing evidence that re- gional preferences can affect trading patterns strongly and detrimentally for both member and nonmember countries. The changing trade patterns suggest that Mercosur was not internationally competitive in sectors where intratrade grew most rapidly. Domestic producers reoriented exports to local markets, presum- ably in order to charge the higher prices associated with the most restrictive trade barriers. This reduced the potential exports of third countries to Mercosur and under many circumstances may have reduced their welfare relative to an 26 THE WORLD BANK ECONOMIC REVIEW, VOL. 12, NO. 1 equivalent nondiscriminatory trade liberalization. Winters (1996) discusses the effects of regionalism on third countries. To the extent that nonmembers' exports are restricted, discriminatory trade barriers would deny access to higher-quality and lower-priced goods for con- sumers in Mercosur's internal markets. For example, Chudnovsky, Lopez, and Porta (1996) study Mercosur's trade restrictions and regulations governing do- mestic production and imports of automotive products. They conclude that the quality of produced (within Mercosur) vehicles continues to be much lower and prices much higher than in other producing countries. The authors also note that any "technological externalities" associated with the automotive regula- tions have been quite limited. Toulan and Guillen (1997) downplay the proposi- tion that regional integration arrangements are a useful vehicle for improving a country's international competitiveness. Specifically, they argue that Some claim that trading blocks can serve as a testing ground for eventual global integration as they allow firms to gradually develop international- ization skills. In many ways this argument is similar to the infant industry protectionist argument, in which barriers are used to protect domestic industries until they develop the skills necessary to compete internationally. Unfortunately, such policies have a fairly poor record in Latin America. The same potential fate could lie in store for firms operating under the protection of Mercosur, in which their level of competitiveness is confined to the demands and pressures of the Mercosur market, rather than the global one. While it is still too early to tell whether firms are in fact viewing Mercosur as a launching pad, interviews with managers do not reveal that they are in fact doing this. (Toulan and Guillen 1996: 11.) If Mercosur is consistent with WTO's rules (Article 24) for the formation of customs unions—the working group examining this has yet to report—the re- sults in this article might provide a useful input into a review of those rules. The Uruguay Round Understanding on Article 24 calls for regular reviews of re- gional trade arrangements and proposes that trade creation and trade diversion be analyzed for new arrangements. The empirical evidence examined in this study accents the importance of these functions. Given the recent proliferation of re- gional trade arrangements, this study's findings highlight the need for further empirical research on the domestic and international effects of these arrange- ments in order to assess the pros and cons of regionalism. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Anderson, Kym. 1983. "Prospects for Trade Growth among Pacific Basin Countries." Developing Economies 21 (December):376-85. Yeats 27 Behar, Jaime. 1995. "Measuring the Effects of Economic Integration for the Southern Cone Countries: Industry Simulations of Trade Liberalization." Developing Econo- mies 33(March):3-31. Brown, A. J. 1947. Applied Economics. London: George Allen and Unwin. Chudnovsky, Daniel, Andres Lopez, and Fernando Porta. 1996. "Intra-Industry Trade and Regional Integration: The Case of the Auto Industry in Argentina." In M. I. Lord, ed., Latin American Trade in Manufactures: A Handbook of Regional and Country Case Studies. Ann Arbor: University of Michigan Press. Donges, Juergen. 1987. "Comment." In Oli Havrylyshyn, ed., Exports of Develop- ing Countries: How Direction Affects Performance. Washington, D.C.: World Bank. Drysdale, Peter, and Ross Garnaut. 1982. "Trade Intensities and the Analysis of Bilat- eral Trade Flows in a Many-Country World: A Survey." Hitotsubashi Journal of Economics 2(2):62-84. Finger, J. Michael, Melinda Ingco, and Ulrich Reincke. 1996. The Uruguay Round: Statistics on Tariff Concessions Given and Received. Washington, D.C.: World Bank. Havrylyshyn, Oli. 1987. "Evidence of Differences between South-South and South-North Trade." In Oli Havrylyshyn, ed., Exports of Developing Countries: How Direction Affects Performance. Washington, D.C.: World Bank. Havrylyshyn, Oli, and Martin Wolf. 1987. "What Have We Learned about South-South Trade?" In Oli Havrylyshyn, ed., Exports of Developing Countries: How Direction Affects Performance. Washington, D.C.: World Bank. Kojima, Kiyoshi. 1964. "The Pattern of Trade among Advanced Countries." Hitotsubashi Journal of Economics 5(1): 16-36. Laird, Sam, and Alexander Yeats. 1990. Quantitative Methods for Trade Barrier Analy- sis. London: Macmillan Press. Lary, Hal. 1968. Imports of Manufactures from Developing Countries. New York: Na- tional Bureau of Economic Research. Sapir, Andre. 1992. "Regional Integration in Europe." Economic Journal 102(415, November):1491-1506. Toulan, Omar, and Mauro Guillen. 1997. "Beneath the Surface: The Impact of Radical Economic Reforms on the Outward Orientation of Argentine and Mendozan Firms, 1989-95." Journal of Latin American Studies 29(May):395- 418. Tuong, Ho Dae, and Alexander Yeats. 1980. "On Factor Proportions as a Guide to the Future Composition of Developing Countries' Exports." Journal of Development Economics 7(fall):276-97. UNCTAD (United Nations Conference on Trade and Development) and World Bank. 1989. A User's Manual for SMART. Washington, D.C.: World Bank. Winters, L. A. 1996. "Regionalism and the Rest of the World: Theory and Estimates of the Effects of European Integration." International Economics Department, World Bank, Washington, D.C. Processed. World Bank. 1992. Global Economic Prospects and the Developing Countries 1992. Washington, D.C. WTO (World Trade Organization). 1996. Trade Policy Review—Brazil. Geneva. 28 THE WORLD BANK ECONOMIC REVIEW, VOL. 12, NO. 1 Yeats, Alexander. 1989. "Shifting Patterns of Comparative Advantage: Manufactured Exports of Developing Countries." Policy, Planning, and Research Working Paper 165. International Economics Department, World Bank, Washington, D.C. Pro- cessed. Zarocostas, John. 1996. "wro Criticizes Brazil for Straying from Reform." Journal of Commerce (November 1).