DISCUSSION PAPER DRD116 Intra-Industry Specialization in a Multi-Country and Multi-Industry Framework Bela Balassa and Luc Bauwens December 1984 Development Research Department Economics and Research Staff World Bank The World Bank does not accept responsibility for the views expressed herein which are those of the author(s) and should not be-attributed to the World Bank or to its affiliated organizations. The findAgs , interpretations, and conclusions are.the results of research supported by the Bank; they do not necessarily.represent official policy of the Bank. The designations employed, the presentatio-:l of material. and any maps used in this document are solely for the convenience of the reader and·do not imply the expression of any opinion whatsoever. on the part of the World Bank or its affiliates concerning the legal status of any country, territory, city, area, or of its authorities, or concerning the delimitations .ofits boundaries ,or national affiliation· .. .' ..e Intra-Industry Specialization in a Multi-Country and Multi-Industry Framework Bela Balassa and Luc Bauwens . The paper tests alternetive hypotheses as to the factors determining the extent of intra-industry tr, non-ferrous metals (SIC 333), as well as several 4-digit categori;s covering textile waste, preserved wood, saw m~ll products, prefabricated wood, veneer and pLywood, wood pulp. dreing and tanning extracts, fertilizers, adhesives and gelatin, carbon black. petrol.eum refining and products, asbestos and .sphalt products, cement andco~crete. lime, gypsum products, cut stone products, and lapid~ary work. It al;;o excludes ordnance (SIC 19), for which comparable trade data are not a~ailable. - 4 - particular products have been judged to be very similar. 11 The use of an economically meaningful classification scheme is of importance, so as to identify 'genuine' as compared to spurious intra-industry trade, which latter is an artifact of the classification scheme employed. The individual industry categories have further been matched against the 3- and 4-digitcategories of the United Nations Standard International Trade Classification (SITC). !I Section I of the paper describes the methodology utilized. Sections II and III, respectively, review th~ intercountry and the interindustry determinants of intra-industry specialization. Section IV provides the empidcal results obtained for bilateral trade among all. t.b.e., cQuntt'·Les._... covi:.:red, among deve Loped count ries, among deve loping couO"t:""~."Ii," ....ell as between developed and developing countries. Finally, Section V compares the results obtained for the various groups of countries. 11 The principal criteria have been high substitution elasticities "in ~roduction and in consumption. !I Among other empirical studies of intra-industry trade, Havrylyshyn and Civan (1983) and Pagoulatos and Sorensen (1915) used 102 3-digit SITC categories; Loertscher and Wolter selected 59 such categories because of a lack,of sufficient reliable export data for others (l980, p. 28Sn); Caves chose 84 3-digit SITe categories which could be matched with 4-digit SIC categories (1981, p. 206); Toh utilized 112 4-digit SIC categories for which comparable trade data could be "derived from aggregating comparable and not to; ~any SITC numbers in order to keep the extent of statistical aggregation bias to the minimu~' (1982, p. l88); Lundberg (1982) made calcaulations for the 77 manufacturing sectors of the International Standard Industrial Classification; Berlstrand (1983) used 3 digit categories wfthin SITC class 7, and Clair. Gaussens, and Phan (1984) utilized S-digit categ~ries in SITC classes 5 and 1. None of these authors attempted to replac; the statistic~l categories by more appropriate industry categories or to exclude natural- resource products. In turn, several of them introduced variables to evaluate the implications for the index of intra-industry trade of the heterogeneity of the statistical categories that is not necessary if an economically meaningful system of classific(tion is used. . - 5 - 1 The index of intra-industry trade, IIT'k' · has been defined as in J 1 e e (1), where X'k' and stand for the adjusted exports and imports of , J 1 M'k' J 1 industry i 1n trade between countries j and k. The formula makes adjustment for imbalance in total trade between countries j and k, when Xjk and Mjk represent the total exports and imports of country j in trade with country k.!! The index takes values from 0 to 1 as the extent of intra- industry trade increases. ~" ··d,'~ X· k , ",M'k','i .... ;;;' e e X'k' - M'k'1 J = 1 ; X'k · r,Ck J (1) IIT'k' : .1 J 1 - - J 1 e e X'k'1 + M'k'1 J J e X'k + M'k X'k' +M'k' where = X J' J and e = u J J X'k' u J jki 2X .ljki 1 njki jk 2Mjk In the regression equations explaining intercountry and interin~ustry differences in the extent of intra-industry trade, IIT'k' has been used as J 1 the dependent variable. In turn, the explanat6ry variables include the country characteristi~s and the industry characteristics described in Sections II and III, respectively. Various consider~tions are relevant to the choice of the functional form utilized in the estintation. '.-.. To begin ,....ith. a Linear or lo~linear - l! While Aquino (978) made adjustment for the imbalance in trade in ;anufactured goods, the present study follows Balassa (l979) in adjusting for the imbalance in total trade, so as to allow for inter-industry specialization between pri~ary and manufactured gooas that is of particular importance in trade between developed and developing countries. - 6 - equation may give estimated values that lie outside the 0 to 1 range. While a logistic function does not have this shortcoming, itslogit transformation 11 cannot handle values of a and 1. At the same time, although values of 1 (representing complete intra-industry specialization) do not occur in the sample, values of 0 (representing complete inter-industry specialization) are of importance. In trade among all the countries concerned, there are potentially 106,856 observations. !I IIT'k' is, however, not defined in 41 percent of J 1 the cases, because Xjki = ~jki = 0 ; i.e. no trade takes place in a particular industry cate.gory between two particular- ce-un'Ct'res. Amon~t..b.ii.-·-· remaining 62,770 observat ions, 51 percent are equal ta aT''" ~d~e· 'eltfi"e'r- X'k' or M. k, is zero; i.e. there is complete inter-industry specialization. J 1 J 1 Given the importance of the zero Observations, the choice has been made for the nonlinear least squares estimation of the logistic function that can handle such observations. We have thus estimated (2), where Z'k' is the J 1 vector of the explanatory variables, a 1S the vector of the regression coefficients, and E'k' is a random disturbance ter~. Estimation has J 1 1 ( 2) IIT .k' = ---";;;'--.--- + 0:,. , J 1 JItl l+exp -8 Zjki t been done by decomposing 8 Z'k.as shown in (3), where C refers te country J 1. ·f ... -- characteristics and [ to industry characteristics. · - It 11 1n (IIT'k,/I-IIT'k') = b'Z'k" + u'k' J 1 J 1 J 1 J 1 2/ There are 38 countries trading with 37 countries in 152 commodity categories, but we eliminate one-half of the observations since IITjki = IITkj i . - 7 - , (J) B Z ok' = J 1 It is apparent that while none of the individual terms in (J) includes both the cou~try and the industry dimensions of the variation of the dependent variable IIT'k' , they are both incorporated in the entire J1 function. This means that the effects of country characteristics on the index of incra-inaustry specialization are assumed to be invariant across industries and the effects of industry characteristics on the index of intra-industry special izat ion are assumed to be invariant acro$S'·e~M~:f'.,~~t1!g~:l,t: < II ~I In examining trade in differentiated products, Linder advanced the proposition that "the more similar the demand structures of two countries, the more intensive, potentially, is the trade between these two countries" (1961, p. 94). He further argued that while "a whole array of forces influences the demand structure of a ~ountry ··· the level of average income is the most important single factor and that it has. in fact, a dominating influence on the structure of demand [so that] similarity of average income levels could be 11 In order to estimate industry-specific effects of country characteristics and country-specific effects of industry characteristics on the extent of intra-industry trade, one would ideally estimate (3) with industry-specific, c:leff i ci en t s bi C and country-specific coefficients bJ'kI. illStead of the . coefficients be and bI · However, this would involve the e~imation of several thousands of coefficients. Such an estimation would be excessively costly and might be infeasible in practice, even if the sample size is in principle sufficiently large. 21 This section 1S based on Balassa, 1984a. - 8 - used as an index of similarity of demand structures" (Ibid.). The converse of this proposition is that "per capita incomes differences are a potential obstacle to trade ···· When per capita income differences reach a certain magnitude, trade can only take place in certain qualitatively homogeneous products" (Ibid., p. 98). In utilizing a model where intra-industry trade occurs in differentiated manufactured goods produced under economies of scale, Helpman subsequently provided proof of the proposition that, in the case when the home country has a lowe~ (or equal) capital-labor ratio than the foreign country and factor prices are equalized, "ifwereallocate·the.~.Ld.'s.!abor and capital stock in a way which increases the foreign.co~ntt'y's capital-labor ratio and reduces the home country's capital-labor t'atio without disturbing commodity prices and factor rewards, then the share of intra-industt'y tracie ···witt decline" 0981, p. 325). Now, "since the higher the capital- labor ratio the higher is income per capita (in a cross country comparison), this raises the hypothesis that a country's share df bilateraL intra-industry tt'ade is negatively correlated with the absolute difference in bilateral incomes per capit.a" (Ibid., p. 337). Helpman also provided pr~of of the proposition that, in two countries that have the same capital-labor ratio, "a redistribution of resources which .... preserves ea::h country's initial capital-labor ratio increases the 'lolume of trade if it reduces the inequality in country size, and it reduces the volume ... of tt'ade if it i~reases the ipequality in country size. The volum~of trade - is largest when both countries are of equal sizelt (Ibid. p. 327). On the assumptions made, the entit'e increase in trade takes the form of intra- industry trade. Correspondingly, one may hypothesize that the extent of - 9 - intra··indust1.7 trade between any two countries will be negatively correlated with differences in ~heir size. The twO propositions were combined by Dixit and Norman who showed that "if the two countries are of similar size, and have no clear comparative advantage across industries, then we will see the predominant pattern of trade as one of intra-industry trade" (1980, p. 288). Comparative advantage is .defined in terms of di fferences in factor endowments,· for which per capi ta income differences may again be used as a proxy. Linder further suggested that "the higher the per capita lncome, the higher will be the degree of quality characterizing the demand structure as a products tend to be differentiated, the extent of intra-industry trade between any two countries is expected to be greater, the highe~ is their (average) per capita income. Finally, Lancaster showed that, owing to economies of scale, the equilibrium number of differentiated manufactured products will be the greater, the larger is the size of the market !I (1980, p. 158). Correspondingly, it may be hypothesized that the extent of intra-industry trade between any two-countries will be positively correlated with 'h-:!ir (average) size. We have considered various hypotheses linking the level of" per capita lncomes and country size, as well as intercountry ~ifferences thereof, to the · 11 As Lancaster notes, this result will not obtain if economies of scale are derived from a homogeneous production function of constant degree. - 10 - extent of intra-industry trade. According to these hypotheses, the extent of intra-industry trade ig expected to be positively correlated with the average per c"'!;..i.ta income and the average size of the two countries and negatively correlated with intercountry differences in per capita incomes and in country size. While the hypotheses have originally been formulated in a two-country model, in the present case they will be tested in a multi-country model. At the same time, it should be recognized that empirical testing has not permitted intrOducing some of the restricc.ive'ct'S'sumptt-Ofts ma!1.e.;bY. the authors in developing their hypotheses. In testing the stated hypotheses, per capita income has been represented by GNP per head and country size by GNP. 11 But, rather than taking absolute vatue~ of intercountry differences in per capita incomes and size, use has been made 0; a relative inequality measure that takes values between 0 and 1. This measure 1S superior to utilizing the absolute values of the differences. w~:ch latter are affected by the magnitudes of the particular country characteristics in the different countries. The relative inequality measure is shown in (4), (i) INEQ = 1 + ({w) In ('01) + (l-w) In (1-w)]/ln2 ... .,.. .... - ·.. 1/ While the domestic consumption of manufactured goods would have been a ~ore appropriate measure of the size of domestic market for these products, the necessary date are not available for some countries and are subject to considerable error in regard to others. At the same time, from available iQformation it appears that the consumption oE manufactured goods and: the gr6ss national product are highly correlated. - 11 - where w refers to the ~atio ofa particular country characteristic in country j to the sum of this characteristic in country j and partner country k. The next question concerns the introduction of transportation costs. In models of intra-industry trade, such as that of Krugman (1980), transportation costs will reduce the volume of such trade. However, the literature does not p~ovideus with a presumption that intra-industry trade will thereby be affected relatively more (or less) than inter-industry trade. Such a presumpt~o!l may be established if information flows are introduced. There is no need to provide"tr'rt''b''rmation on the characteristics of standardized (non-differentiated) products, such as copper metal, steel ingots, and caustic soda, which have uniform specifications across the world and hence their trade is determined largely by r~lative costs, giving rise to inter-industry specialization. However, there is need for information on the characteristics of differentiated products, such as machinery, transport equipment. and consumer goods, which are subject to intra-industry trade. It can be assumed that the availability of information decreases, and its cost i~creases, with distance. Correspondingly, it may be hypothesized that the extent of intra-industry trade between any two countries will be negatively. correlated with the distance bet~en them. ... Distance has been measured in terms of miles between the centQ:'s of geographica~ gravity for · each pair of countries. ·- The existence of common borders will also contribute to information flows. Furthermore, as Grube!' and Lloyd suggested, in countries sharing a co~~on border, intra-industry trade may occur "in products which are functionally homogeneous but differentiated by location" (1976, p. 5). Thus, - 12 - it may be hypothesized that ~he extent of intra-industry trade will be greater between countries that share a common border than between countries which do not have common borders. At the same time, the separate introduction of distance and border variables permits testing the hypothesis that common borders have economic significance for intra-industry trade beyond that of distance. In :he econometric investigation, the existence of common borders has been represented by a dummy variable, which takes the value of 1 when the two countries share a common border and is 0 othen/ise. In a model incorporating speci.fie capi.t~l 'antli::Onstant re~'J-AS to scale, Falvey found that the volume of intra-indust~!f"",~t'aee-'~iJ.,.l.:/ ...a.r:(. inversely with the level of tariffs and of trade restrictions in general (1981, p. 50S). Sut, again, the question is if tariffs will affect intra- indUStry trade relatively more than inter-industry trade. The senior author suggested that such would be the case in the event of trade liberalization in general and economic integration in particular. This is because adjustments to reducti~ns in ttade barriers would occur largely throu~h rationalizing operations and changing the product composi~ion of ~ndividual industries, with national product differentiation contributinl to intra-industry trade (Salassa, 1977). The same author showed that trade liberal izati,n (1977) and economic: 'f integration in the European (1966a, 191.5) and the Latin American (1979) area ... .... .. - were in fact accompanied by.increases in the extent of intra-industry trade ~ among the countries in question. In the present investigation, the hypotheses will be test~d that the extent of intra-industry trade between any two countries 1S negatively correlated with the average level of their trade - 13 - restri:tions and positively correlated with participation in integration schemes. Estimates of tadfE levels are not available for a number of countries and the tariff equivalent of quantitative import restrictions 1S not known with any confidence for others. Correspondingly, an indicator of trade orientation has been used to represent the extent of trade restrictions. Trade orientation has been defined in terms of deviations of actual from hypo~hetical values of per capita exports. Hypothetical vdlues have been derived from a regression equation that, i'1 addition to the per capita income and popuLation variables utilized in early worJiI'. ~y Chenel:"Y H960J. includes variables representing the availabil ity'of' mineraL resources and propinqui ty . to markets. JJ Mineral resource availability has been represented by the ratio of mineral exports (Xm) to the gross national product while propinquity has been defined as the weighted average of the inverse of distance between country j and partner country k.{D., ), the weights being the gross national product of . Ji<. the partner countries (y ) The results are reported in k. · .. 'f .. ..- ... · · 1/ ALthough the paper deals with trade in manufact~red goods, the extent of ~rade orientation in yegard to all products in the appropriate variable. This is be~ause protection is a relative concept and trade barriers on primary product affect trade in manufactured goo~d.·as '<...reI 1. ' , ." ,,~~~~-~. ~'"IAssociation are positively correlated, with all the coefficients being highly significant statistically in the relevant equations. Also, the language variables have the expected positive sign whenever they are statistically significant, which is the case in most instances. In turn, the extent of intra-industry trade is expected to be positively correlated with product differentiation, represented by the Hufbauer measure of produet differentiation, marketing costs, and the variability of profit rates, and neg~tively c8rrelated with product standardization, represented by economies of scale and industrial concentration. All ~he regression coefficien~s have the expected sign and are. ;. generally significant statistically, the exceptions being the standard -.. · deviation of profit rates in-the case of trade among developed and among ·- 11 This conclusion does not apply to the size variables as far as trade among ~eveloping countries is concerned, however. - 28 - developing countries; the economies of scale variable in the case of trade among developing countries and bet~een developed and developing countries; and the product differentiation v~riable in trade bet~een developed and developing countries. The variables associated ~ith foreign investment (foreign direct investment and foreign affiliates) have a negative sign and are statistically significant~ except for trade among developed countries. [n turn, the offshore procurement variable has the expected positive sign and it is highly statistically significant in all three cases. 11 And, ~hile the average tariff variable does not reach minimum lev&ls- Oct Soi.g.nifi.oaI1c-e in any of the cases considered, the tariff dispersion variable is statistically significant except for trade among developing countries. The estimates presented in this paper combine the intercountry and the interindustrty determinants of the extent of intra-industry trad~. The explanatory po~er of the regression equation for intra-industry trade among all the countries under consideration is fairly high, not~ithstanding the relatively small number of estimated parameters compared to the number of observations. At the same time, the coefficient of determination is the highest for trade among the developed countries tha.t have a relatively more homogeneous economic structure and for ~hich intra-industry trade represents 1 an importarit propOrtion of total trade. [n turn, the heterogeneity of the .... .. sample and the relatively large proportion of zero observations ayears to · have r.::duced the explanatory po~er of the regressions for intra-i;dlJstry trade among developing, as ~ell as bet~een developed and developing, countries. 11 It ~illbe reca~led that the iast-ment oned three variables are not ;elevant for trade among developing count.r es. - 29 - References Aquino, Anconio (1978), "Incra-Induscry Trade and Incer-Industry Specialization as Concurrent Sources of Internacional Trade in Manufactures," Weltwirtschaftliches Archiv. 114(2), 175-96. Balassa, Bela (1966a), "Tariff Reductions and Trade in Manufactures Among Industrial Countries," American Economic Review, 56(3), 466-73. Balassa, Bela (1966b), "American Direct Investmencs in the Corrunon :-Iarket," Banca Nazionale del Lavoro Quarterly Review, June, 121-46. (1967), Trade Liberalization among Induscrial Countries: Objectives and Alternatives, New York, ~cGraw Hill for the Council on Foreign Relations, . Chapter 5. (1975), European Economic Integration, Amsterdam, North-Holland, Chapter 2. (1977), "Effects of Commercial E'olicy on Intemati·onal. Trade~ the Locat ion0f Produc t ion, and Fac tor Movement so'!: i-:l.~Be.,::,,," i 1:'ei'rtTn'-~Per:"Ove Hesselborn and Per Magnus Wijkman, eds. The International Allocation of Economic Activity, London, Macmillan, 1977, pp. 230-58. (1979), "Intra-Industry Trade and the Integracion of Developing Countries in the World Economy," in Herbert Giersch, ed., On the Economics of Intra-Industry Trade, Tubingen, J.C.B. Mohr (Paul Siebeck), 245-70. (1982), "Structural Adjustment Policies in Developing Countries," World Development, January 1982, pp. 23-38. (1934a), "Intra-Industry Trade among Exporters of :-Ianufactured Goods," paper presenced at the Symposium on Intra-Industry Trade in Brussels in June 1984; to be published in the Proceedings of the Conference. World Bank Development Research Cencer Discussion Paper No. 87. (1984b), "The Determinants of Intra'-Industry Specialization "in United States Trade," Washington, D.C., World Bank, Development Research Department, Discussion Paper No. 86, Washington, D.C. June 1984. Bergstrand, J.H. (1983), "Measuf-ement and Determinants of Intra-Industry International Trade," in P. K. ~. Tharakan, ed., Intra-Industry Trade, Amsterdam, North Holland, 201-~ · · Brander,James A. (1981), "Intr~-Industry Tra.de in Idencical Commodities," Journal of International Economics, 11(1), February, 1-14. Caves, Richard E. (1981), "Intra-Industry Trade and Market Structure 1n the Industrial Countries," Oxford Economic Papers, 33(2), July, 203-23. :-lichael E. Porter, and A. Michael Spence, with John T. Scott (1980), Competition in the Open Economy: A Model Applied to Canada, Cambridge, Mas~., Harvard University Press. - 30 - Chener-y, Hollis B. (1960), "Patter-ns of industdaL Gr-owth," American Economic Review, 50(4), September, 624-54. Cl.air-, Cl.aude, Olivier Gaussens, and Due-Loi Phan (1984) "Le corr.mer-ce inter-national intr-a-branche et ses determinants d'apr-es le schema de concur-rence monopalistique: une verification empir-ique," Revue Economique, 35 (2), ~ar-ch, 347-78. Dreze, Jacques (1960), "Quelques reflex ions sereines sur l.'adaptation de l.'industrie bel.ge au ~ar-che Commun," Comptes r-endus des Travaux de La Societe Royal.e d'Economie Pol.itique de BeLgique, No. 275, December. Dixit, Avinash K. and Victor- Norman (1980), Theor-v of Inter-nationaL Tr-ade. Digswel.l. Pl.ace, Wel.wyn. James Nisbet & Co. and Ca~br-idge Univer-sity Press. Eastman, H. C. and S. Stykolt (1967), The Tar-iff and Competition In Canada, Toronto, ~acmillan. Fal.vey, Rodney F. (1981), "Commer-c.iaL.l'qlicy and Intr-a-Industry Trade., Jour-nal of Internat ional. Economics, 11 (4), November;' 4~1OI,.~2·."·· " Gr-ay, H. Peter and John P. ~artin (1980), "The ~eaning and ~easurement of Product Differ-entiation In International. Trade," Weltwir-tschaftliches Ar-chiv, 116(2), 322-29. Gr-ubel, Her-ber-t G. and P. J. Lloyd (1975), Intr-a-Industry Trade, London, ~acmillan. Ha'lryLyshyn, Oli and E·· Civan (1983), "Intra-Industr-y Trade and the Stage of Development: A Regr-ession Analysis of Industrial and Developing Countries," in P.K.~. Thar-akan ed., Intr-a-Industry Tr-ade, Amsterdam, North Holland, 111- 140. Helpman, Elhanan (1981), "Inter-national Tr-ade in the Presence of Pr-oduct Differentiation, Economies of Scale and ~onopolistic Competition: A . " Chamberl in-Heckscher--Ohl in Approadt," Journal of Inter-nat ional Economics, 11(3), August, 305-40. Hufbauer, G. C (970), "The Impact, of National Character-is tics and T~chnology on the Commodity Composition of Tr-ade in ~anufactur-ed Goods," in RaYliond Vernon, ed., The Technology Factor in Inter-national Tr-ade, New Yor-k,·National Bureau of Economi~ Research, 145-231. -- · · · Kr-ugman, Paul R. (-1979), "Scale Economies, Pr-oduct Differentiation, .!hd the Patter-n of Tr-ade," Amer-ican Economic Review, 70(5), December, 950-59. Lancaster, Kelvin (1980), "Intra-Industr-y Trade under Per-fect ~onopolistic Competition," Journal of Inter-national Economics, 10(2), ~ay, 151-75. Linder-, Burenstam Staffan (1961), An Essay on Tr-ade and Tr-ansportation, New ; Yor-k, John Wiley. j :} - 31 - Loertscher, Rudolf and Frank WoLter (1980), "Determinants of I~tra-Industry Trade: Among Countries and Across Industries," Welt'.rirtschaftiiches Archiv. 116( 2), 280-92. Lundberg, Lars (1982), "Intra-Industry Trade: The Case of Sweden," Weltwirtschaftliches Archiv, 118(2), 303-16~ Pagou1atos, Emilio and Robert Sorensen (1975), "Two-Way International Trade: An Econometric Analysis," Weltwirtschaftliches Archiv, 111(3), 454-65. Toh, Kiertisak (1982), "A Cross-Section Analysis of Intra-Industry Trade in U.S. Manufacturing Industries," Weltwirtschaftliches Archiv, 118 (2), 282-301 · ·