DISCUSSION PAPER Report No q; DRD87 I l Intra-1ndustry Among Exporters of Manufactured Goods Bela Balassa Hay 1984 Development Research Department Economics and Research Staff World Bank The views presented here are those of the author, and they should not be as reflecting those of the World Bank Abstract This paper examines the determinants of intra-industry specialization in bilateral trade among countries exporting manufactured goods. It is shown that the extent of intra-industry trade between any two countries increases with their average income level, represented by per capita incomes, and with their average size, represented by GNP, and it decreases with differences in their income level and their size. It further appears that the extent of intra-industry trade is positively correlated with the trade orientation of the countries concerned and with the existence of a common border and it is negatively correlated with the distance between them. INTRA-INDUSTRY M~ONG EXPORTERS OF MANUFACTURED GOODS Bela Balassa * May 1984 * Professor of Political Economy at the Johns Hopkins University and Consultant to the World Bank. This paper was prepared in the framework of the World Bank's research project~ "Changes in Comparative Advantage in Manufactured Goods" (RPO 672-41). The author is greatly indebted to Marcus Noland for reviewing alternative explanatory variables and suggesting possible specifications and to Luc Bauwens for undertaking the arduous task of estimation and offer~ng valuable comments on the previous draft of the paper. Thanks are also due to Linda Pacheco for data collection and to Jerzy Rozanski for generating the trade data. However, the author alone is responsible for the opinions expressed in the paper that should not be interpreted tc represen~ the views of the World Bank. INTRA-INDUSTRY TRADE AMONG EXPORTERS OF l1Al the volume of intra-industry trade will decline (1980, pp. 953-54). Thus, one may hypothesize that the extent of such trade between any two countries will be negatively correlated with transportation costs. 4 In a model incorporating specific capital and constant returns to scale, Falvey found that the volume of intra-industry trade will vary inversely with the level of tariffs (1981, p. 50~). In analogy to transportation costs, this conclusion can be extended to the case of intra-industry trade in products subject to economies of scale. One may then put forward the hypothesis that the extent of intra-industry trade between any two countries will be negatively correlated with the average level of their trade restrictions. Furthermore, Grube! and Lloyd suggested that, in countries sharing a common border, intra-industry trade may occur "in products which are functionally homo- geneous but differentiated by location" (1976, p. 5). Correspondingly, it can be hypothesized that the extent of intra-industry trade will be higher between countries that share a common border than between countries which do not have common borders. The described general country characteristics are expected to influence the extent of intra-industry trade in all the countries under consideration. There are some further factors that may affect intra-industry trade between particular countries which have some common features. These features include participation in integration arrangements; common language, and former colonial ties. This author suggested that economic integration will promote intra-industry trade (1966) and examined the relative importance of such trade in European (1975) and in Latin American (1979) integration projects. In the following, it will be hypothesized that economic integration increases the extent of intra- industry trade among the participating countries~ Familiarity with each.other's products may also contribute to intra- industry trade between particular countries. Common language breeds familiarity aa does the existence of past colonial ties. It may, then, be hypothesized that the existence of a common language and former colonial ties will increase the extent of intra-industry trade between any two countries. 5 II As noteJ above, the investi6ation covers 38 exporters of manufactured goods that, in their large majority, are characterized by product differentiation and are subject to economies of scale.l Calculations have further been made for intra-industry trade among 18 developing countries, among 20 developing countries, and between developed and developing countries. As defined for purposes of the study, developed countries cover the income range between $6758 and $2254 in 1973 at the exchange rate prevalent in that year. In order of their per capita GNP, the countries in question are Switzerland, United States, Sweden, Denmark, Germany, Australia, Canada, Norway, France, Belgium, Netherlands, Japan, Finland, Austria, United Kingdom, Israel, Italy, and Ireland. Using the same principle of ordering, with per capita GNP ranging from $2031 to $96, the developing countries included in the study are Spain, Singapore, GrP.ece, Argentina, Hong Kong, Portugal, Yugoslavia, Mexico, Brazil, Taiwan, Malaysia, Tunisia, Korea, Morocco, Turkey, Egypt, Thailand 1 Phillipines, India, and Pakistan. Manufactured goods have been defined by reference to the United States Standard Industrial Classification. Excluding natural resource products, alto- gether 167 industry categories have been established by merging 4-digit cate- · gories in cases when the economic characteristics of the products in question were judged to be very similar. This has reduced the possibility of the hetero- geneity of product categories that gives rise to spurious intra-industry trade. The index of intra-industry trade for any pair of countries (IITjk) has been derived as in (1) where xjki and Mjki refer to the adjusted exports and lExcluding the 19 cases when no trade occurs between pairs of countries, there are thus altogether 684 observations. In every case, data reported by the country with the higher per capita income has been utilized in the investiga- tion. 6 imports of commodity i in trade between countries j and k. The form~la makes adjustment for imbalance in total trade between any pair of countries, when Xjk and Mjk' respectively, stand for the total exports and imports of .country j in trade with country k.l The index takes values from 0 to 1 as the extent of intra-industry trade incr.eas~s. (1) IIT • 1 - jk e xjk + Mjk e xjk + Mjk where X • X 2xjk and Mjki • M jki jki jki 2Mjk A linear or loglinear regression equation may give estimated values that lie outside the 0 to 1 rangeo A logistic function does not have this abort- coming, but its legit transformation cannot handle values of 0 or 1.2 Since a value of zero provides relevant information, representing the extreme case of inter-industry specialization, nonlinear least squares estimation of the lA consistent adjustment procedure was first proposed by Aquino (1978). However, while Aquino adjusted for the imbalance of trade in manufactured goods, in the present investigation adjustment has been made for the imbalance in total trade, so as to allow for inter-industry specialization between primary and manufactured goods (Balassa, 1979). 2The logistic function takes the form IITjk ~ 1/(1 + exp - e'Xjk), and its logit transformation is tn(IITjk/1 - IITjk) e'xjkt where Xjk is the vector of ID the explanatory variables. 7 ~ logistic function has been used in the present investigation.l With income levels expressed in terms of per capita GNP (Y/P), country size has been represented by the gross national product (Y); averages of the two variables (AY/P and AY) have further been calculated for every pair of countries. At th~ same time, rather than taking absolute differences between per capita GNP and ~lP to represent inter-country differences in income levels and country size, an inequality measure has been introduced, with INEQY/P denoting inequality in per capita GNP and INEQY representing inequality in GNP. The measure, shown in equation (2) for INEQY, is one of relative inequality that takes values between 0 and 1. (2) INEQY = 1 + [w ~n(w) + (1 - w)in(l - w)]/in 2, yj where w = y + y , j k In the absence of data on average transportation costs between countries, use has been made of a distance variable, with distance (D) measured in terms of miles between the centers of geographical gravity of any two countries. In turn, the existence of a common border has been introduced in the form of a dummy variable. Estimates of tariff levels are not available for a number of countries and the tariff equivalent of quantitative import restrictions is 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 lThere are 99 cases when IITjk is 0 because either Xjki or Mjki is equal to zero; in no case is IITjk equal to 1. 8 has been defined in terms of deviations of actual from hypothetical values of capita exports. Hypothetical values have been derived from a regression equation that, in addition to the per capita income and population variables utilized in early work by Chenery (1960), includes variables representing the availability of mineral resources and propinquity to markets. Mineral resource availability has been represented by the ratio of mineral exports cxm) 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, the weights being the gross national product of the partner countries L(YkDjk)/IYk. The results are reported in equation (3), with t- k k values shown in parenthesis. As is apparent, the equation has a high explana- tory power and all the regression coefficients are significant at the 1 percent level, using a one-tail test. X (3) iog _1 =- 0.1864 + 0.9212 iog(Yj/Pj)- 0.3541 iog] pj (0.38) (15.02) (6.83) t Yk/Djk -2 + 0.0251 ~j/Yj + 0.0598 l R = 0.9404 (2.91) (2.06) 1y k For any pair of countries, their combined trade orientation (ATO) has been introduced in the estimating equations to test the hypothesis that the extent of intJ:a-industry trade is positively correlated with trade orientation. In tur.n, dummy variables have been included to represent participation in integration arrangements, such as the European Common Market (EEC), the European Free Trade Association (EFTA), and the Latin American Free Trade Association (LAFTA). Dummy variables have also been used for common languages, such as English, Spanish, French, German, Portuguese, and Scandinavian. Finally, dummy variables have been introduced to denote the fo~~er colonial ties of United Kingdom (UKCOL) and France (FRCOL). 9 III The empirical estimates for the.entire group of 38 countries are reported in Table 1. Equation (4) shows the results of estimation including the common character~.~tics of the countries cited in Section II, tog~ther with dummy variables for economic integration, common language, and colonial ties. In equation (5) those variables of equation (4) have been retained that were sta- tistically significant at least at the 10 percent level. Finally, equation (6) includes common country characteristics onlyo The results support the hypotheses put forward in Section I as far as the common charac~e~l~tics of the countries in question are concerned. Thus, the regression coefficients of the average per capita income, income inequality, average country size, inequality in country size, trade 04ientation, distance, and border variables all have the expected sign and are statistically signifi- cant at the 1 percent level in equations (4) to (6). Among the economic integration variables, the EFTA and the LAFTA dummies have the expected sign and are statistically significant at the 1 percent level. In turn, the EEC dummy has the expected sign, but is not significant even at the 10 percent level. While this may seem surprising in view of the emphasis given to intra-industry specialization in the European Common Market from the mid- sixties (Balassa, 1966) onwards, the results appear to indicate that general country characteristics, including the border variables, largely explain the extent of intra-industry trade among the Common Market countries. The English and Spanish language variables have the expected sign and are significant at the 1 perc~nt level whereas the other language dummies have very low t-values. The latter conclusion also applies to the French colonial ties variable; in turn, the variable for English colonial ties has the expected sign and it is statistically significant at the 5 percent level. The results for French language and colonial ties are not affected if only one or the other of the variables is included in the estimating 10 Table 1 · Estimates of Intra Industry Trade for Court tries Exporting Manufactu_red Products: Developed and D.eveloping Countries Combined. (regression coefficients, with t-values in parenthesis) Equation (4) Equation (5) Equation (6) Constant 3.140 (11.21) 3.216 (11.78) 3.151 (11.36) .tn AY/P 0.662 (8.62) 0.670 (8.82) 0.698 (9.52) INEQ Y/P -1.115 (6.62) -1.106 (6.60) -1.214 (6.92) .tn AY 0.400 (11.84) 0.407 (12.40) 0.368 (11.46) It;EQY -o.S20 (7 .31) -0.847 (7.76) -0.818 (7 .23) ATO 0.483 (10.54) 0.492 (10.93) 0.494 (10.98) tn D -0.525 (19.43) -o.532 (20.24) -0.531 (20.04) 130RDER 0.331 (3 .69) 0.381 (4. 89) 0.400 {4.94) EEC 0.154 (1.28) EFTA 0.368 (4.37) 0.371 (4.89) LAFTA 2.047 (6.74) 2.050 (6.79) ENGLISH 0.233 (2. 26) 0.218 (2 .12) SPANISH 1.400 (5 .14) 1.413 (5 .21) FRENCH 0.081 (0.25) GERMAN 0.107 (0.56) PORT. 0 .. 024 (0.02) SCAND. 0.068 (0.46) PRCOL -o.141 (0.21) UKCOL 0.457 (2.00) 0.473 (2 .07) -2 R 0.8721 0.8772 0.8573 A (J 0.0707 0 .. 0705 0.0741 N 684 684 684 NOTE: For the definition of the variables and explanation of methodology, see text. 11 equation.l Again the results may seem surprising in view of the existence of close trade ties between France and its former colonies and, to lesser extent, among the J~tter. It would appear, however, that trade among the countries con- cerned largely involves interindustry specialization. Table 2 provides the results obtained for intra-industry trade among deve- loped countries. Equation (7) includes all the relevant variables; it omits the dummies for LAFTA, Spanish and Portuguese languages, and colonial ties. Equation (8) also excludes variables whose level of statistical significance did not reach 10 percent while equation (9) incorporates only the common country characteristics. With the exception of intercountry differences in income levels, the results support the hypotheses confirmed by the estimates for the entire group of countries. However, apart from the EEC dummy, the statistical significance of the coefficients is lower than in the previous case, and two variables (AY'/P and BORDER) are significant at only the 5 percent and one variable (English) at the 10 percent level. The lack of statistical significance of the variable representing inter- country differences in per capita incomes may be explained by the fact that, with income differences being much smaller among developed countries than in the entire group of countries, the demand structure of the countries concerned is also more similar. Correspondingly, one may not expect large variations to occur in the extent of intra-industry trade as a function of income differences. In turn, the lower level of significance of the English language variable may be attributed to the fact that several of the English-speaking countries are 1In this connection it may be added that the negative sign for the French colonial ties variable may be explained by intercorrelation between this variable and the variable for French languagee .. 12 Table 2 Estimates of Intra Industry Trade for Countries Exporting Manufactured Products: ---- Trade among Developed Countries. (regression coefficients, with t-values in parenthesis) Equation (7) Equation (8) Equation (9) Constant 2.299 (2.82) 2.243 (2.87) 2.497 (3 .10) tn AY/P 0.596 (2.32) 0.660 (2. 69) 0.680 (2~66) INEQ Y/P -o.775 (0.77) -0.548 (0. 53) tn AY 0.298 (3. 99) 0.299 (4.07) 0.277 (3.82) INEQY -o.579 (2. 84) -0.608 (3. 05) -0.677 (3.34) ATO 0.792 (3.20) 0.765 (3. 26) 0.738 (2.98) in D -0.445 (8.95) -0.452 (9. 28) -0.478 (9.64) BORDER 0.265 (2 .11) 0.343 (3 .14) 0.401 (3.59) EEC 0.252 (1.65) 0.210 (1.46) EFTA 0.353 (3.13) 0.404 (3. 99) ENGLISH 0.271 (1.74) 0.244 (1. 59) FRENCH 0.037 (0 ,, 09) GERMAN 0.226 (0. 95) SCAND. 0.218 (1.13) -2 R 0.9467 0.9458 0.9390 .... a 0.0846 0.0841 0.0886 N 153 153 153 NOTE: See Table 1. 13 in the developing country sample. Finally, the EEC dummy is (barely) signifi- cant at the 10 percent level in equation {7), indicating the relevance of this variable once developing countries are omitted from the calculations. The level of significance of the EEC dummy variable (slightly) falls below the benchmark in equation (8) that excludes the variables which were not significant at the 10 percQnt level aG in equation (7). The same sort befalls the English language variable, but again by a small margirt. The results may be explained by intercorrelation with the omitted variable~. By contrast, the EFTA dummy remains statistically significant at the 1 per- cent level. Also, the level of significance of the average per capita income and the border variables improves to 1 percent as the variables that were not significant at the 10 percent level are eliminated from the estimating equation. Nor are these results affected if only the variables representing the common country characteristics are retained as in equation (9). The EEC, EFTA, the German and Scandanavian language, and the colonial ties variables have been omitted from equation (10), estimated for intra-industry trade among developing countries and reported in Table 3. The per capita income in-equality variable is again not significant statistically, reflecting con- siderations similar to those adduced for the developed country group. Apart from this variable, the results again support the hypotheses confirmed for the entire group of 38 countries, although the inequality in country size and the border variables are significant at only the 5 percent level. The LAFTA and the Spanish and English language dummies are statistically significant at the 1 percent level, representing an improvement in the case of the latter variable over the previous results. In turn, if these variables are omitted as in equation (12), the level of significance of the inequality in country size and the border variables improves to 1 percent. Finally, all the variables representing the common characteristics of the countries concerned are statistically significant at the. 1 percent level in equation (13) that provides estimates for intra-industry trade between pairs of 14 Table 3 Estimates of Intra Industry Trade for Exporting Manufacture~ Products: ~tries Trade among Developing Countries: (regression coefficients, with t-values in parenthesis) Equation {10) Equation (11) Equation (12.) Constant 5.121 (3. 92) 4.782 (3.76) 6.738 (5 .80) 1n AY/P 1.117 (4.02) 1.291 (5.11) 1.349 (4.53) INEQ Y/P -0.564 (0. 98) -0.283 (0.50) in AY 0.872 (3 .57) 0.80'1 (3 .43) 1.356 (7.56) INEQY -1.592 (2 .14) -1.522 (2.11) -3.344 (5. 73) ATO 0.875 (7.88) 0.869 (8.26) 1.031 (8.65) in D -0.609 (5. 79) -0.611 (6.03) -0.531 (5.04) BORDER 0.880 (2.48) 0.821 (2.38) 1.118 (3 .84) LAFTA 2.187 (5. 56) 2.274 (5.80) ENGLISH 0.776 (3. 33) 0.641 (3 .18) SPANISH 1.556 (3. 90) 1.610 (4.02) FRENCH 0.404 (0 .• 05) PORT. 0.620 (0.55) -2 0.7402 R 0.7385 0.6671 .... c 0.0501 0.0498 0.0558 N 175 175 175 NOTE: See Table 1 15 developed and developing countries (Table 4). Among the dummy variables for economic integration, common language, and colonial ties, the EFTA dummy per- taining to trade between the developed member countries and Portugal is signifi- cant at the 10 percent level while the French language and the British colonial ties variables reach this level of significanct! under some specifications but not under others.l It is apparent that, apart from the incomet inequality variable that loses its statistical significance in cases where the intercountry variati~n of incomes is more limited, the variables representing the common characteristics of the countries concerned are highly significant statistically, regardless of the choice of the country sample or the specific:ations of the estimating equations. Several of the variables pertaining to economic integration, common language, and colonial ties are also statistically significant but, with the exception of the developing country group, they add little to the explanatory power of the equations. Also, the coefficient of determination remains affected practically unaffected if variables whose level of significance does not reach lO percent are omitted from the regression equations. The coefficient of determination is the highest (0.95) in the equation el~laining intra-industry trade among developed countries~ This result is of particular interest in view of the fact that, as first suggested by Linder, the extent of intra-industry specialization is much greater in trade among developed cou1ntries than in trade among developing countries or between developed and lwith the French language and French colonial ties variables being highly cor1~elated, the forrner los·es its statistical significance if the latter is ommi.tteci from the estimating equation. C<)rrespondingly, both have been dropped from equation (14). 16 Ta'ble 4 Estimates of Intra Industry Trade !.2.!_ Countries Exporting Manufactu£ed Products: Trade between Developed and Developi·n.g Countr:Les. (regression coefficients, with t-values in pare1rtthesis) Equation (13) Equation (14) l~quation {15) Constant . 2.749 {5 .59) 2.693 (5. 61) ;z. 591 (5 .45) .tn AY/P 0.741 (3. 63) 0.765 (3.81) 0.768 (3 .86) INEQ Y/P -1.373 (5 .48) -1.322 (5.33) -·1.383 (5 .. 58) tn AY 0.408 (6.34) 0.403 (6.28) 0.378 (6 .03) INEQY -0.635 (2. 79) -0.622 (2.77) .,..0.561 (2 •.50) ATO 0.333 (4.38) 0.348 (4. 70) 0.35lJ (4.~~1) R.n D -o.483 (9.38) -0.482 (9.54) -·0.476 (9. 65) BORDER 0.789 (4 .08) 0.785 (4.06) 0.787 (4.()6) EFTA 0.354 (1. 71) 0.343 (1.66) ENGLISH 0.165 (0 .81) FRENCH 0.982 (2 .. 17) FRCOL -o.010 (1.37) UKCOL 0.541 1\:1.50) 0.675 (2.08) -2 R 0.6993 0.6966 0~6920 • a 0.0713 0.0713 0.0717 N 356 356 356 17 dev,eloping countries •. 1 At: · t:he same time, it is noteworthy that the explanatory pewter of the regression eq,uation is high, with coefficients of determination of 0.7«• and 0.70, in the latt1er two cases also; it is 0.87 in the equation per- taioing to the entire group of countries. IV Various hypotheses put forward in the theoretical lite1~ature as regards the: dete1t1ninants of intra-industry specialization have been test'ed in this paper for tlt"ade~ among countries exporti111g manujfa<~~tured goods. This has been done in r.:tgard to trade among all the 'c:ountrjles', meeting certain crit~eria, as well as for trade among developed countrle1s, among developing countries, and between deve- loped and developing countries. The results show that the extent of intra-industry trade. between any two · countrief;J increases with th.eit' average inc,ome level, rE!presented by per capita GNl:•, anc.t w.f. th their averag1e si.ze, represeli\ted by GNP, and it decreases with dif- fex"ences in their income level and in thei1~ size. Howe~ver, dj~ffe1~ences in income levels are not significantly related to the extent of intra-industry spe- cialization in trade among dev·eloped and a:mong developing countries, where incc,me, differences are much smaller than betw·een developed and developing countries. It further appears that the extent of intra-industry trade is positively corr.,e!lated w.ith the trade orientation of the countries concerned and with the existence of a common border and it is negatively correlated with distance bet- ween them • Taken together, these common country characteristics explain much of 'the variation in the e~tent of intra-industry specialization, with the coef- fic:ient of determination ranging from 0. 95 in trade among developed countries lThe average index of intra-industry trade is 0.300 for trade among deve- loped countries, 0.039 for trade among developing countries, 0.081 for trade between developed and developi.ng countries, and 0.120 for trade among all countries studied. 18 to 0.67 for trade among developing countries. Introducing variables for economic integration, common language, and colo- nial ties increases the explanatory power of the regression equation relati V4!ly little, the exception being the equation for trade among developing countrieu. At the same time, several of these variables are highly significant statist!·· cally. Thus, it is apparent that participation in the European Free Trade Association and in the Latin American Free Trade Associa,tion increases the extent of intra-industry trade. The regression coefficient for EEC membership is also positive, but its level of statistical significance rarely reac~es 10 percent~ Finally, English and Spanish as common languages and English coloni•tl ties appear t:o contribute to intra-industry specialization .. While the statistical significance of the estimates f'or the various count:ry groups is generally high, the regression coefficients of several C)f the variables, vary among the groups. The statistical significance oi~ these dif- ferences haa been tested in regard to the three subgroups by util1:.zing the results obtained in equations (9), (12), and (15) that c.ontain cotillllon country characteristics, so as to ensure comparability in the re:sults. As shown in equations (16), (17), and (18), the dif:ferences :in the regression coefficients are not significant statistically in rega.rd to the average per capita GNP, income inequality, and distance variables at the 10 per- cent level, and in regard to the border variables at the 5 perce1nt level. in any of the three comparisons (Table 5). Nor are there statistically significant differences at the 5 percent level in regard to any of the variables if the estimates for intra-industry trade between developed and developing countries 19 Table 5 :Differences betwt~en Regression Coefficients · Estimated in~he Three Subgroups(a) (with t-values in parenth~esis) Trade atllong develop- Trade bet-w·ee11 Trade between ing coun1tr!es and developed and developed and trade ~am.~ong developed developing C()Untries, developing count- eountrjletJ and trade amo1ng de- ries. and trade veloped countries among developing countries Equation (16) Equation (17) Equation (18) Cousta1nt 4.241 (2.58) 0.094 (0.12) -4.148 (2.62) .f.rt AY/P 0.669 (1 •.53) 0~088 (0.31) -0.581 (1.34) I'NEQ Y/P 0 .. 264 (0.24) -0 .. 836 (0. 95) -1 .. 100 (1.42) tn AY. 1.079 (4.50) 0.101 (1.1.7) -0.978 {4.07) INEQY -2.667 (3.4!)) 0 .. 116 (0.42) 2.783 (3 .. 53) ATO 0~293 (l.l5i) -0.380 (l.. 77) -0.673 (3 .94) D -o.052 (0.37') 0.002 (0.03) 0.055 (0 .. 38) BORDER 0.717 (1.85) 0.386 (1.79) -0.331 (0.78) joint test(b) 31.38 11.07 35.05 NOTES: (a)The d.:f.ffere·nces are derived from pairwise comparisons between equations (9)~ (12), and (15). (b)The statistic for the joint test of equality is distributed as a ehi-square random variable, with eight degrees of freedom. 20 and among develop~d countries are compared. However, differences in the regression coefficients of the average country size and the size inequality variables are significant at the 1 percent level in the other two comparisons. The same conclusion applies to the trade orientation variable, when the estima- tes for the trade between developed and developing countries and among developing countries are compared. In addition to testing the statistical significance of differences between the individual regression coefficients, combined tests for these coefficients have also been carried out. As it is apparent from the results of Table 5, the hypothesis that the regression coefficients, taken jointly, are equal is rejected in the comparisons of the equation pertaining to intra-industry trade among developing countries with the other two equations; it is accepted in the comparison of the estimates for trade among developed countries and between developed and developing countries. This paper has shown that the determinants of intra-industry specialization are similar in trade among developed countries, among developing countries, bet- ween developed and developing countries, as well as in trade among all the countries concerned. At the same time, it appears that coefficient values differ in regard to intra-industry trade of the developed and of the developing countries. This matter would require further study so as to examine the reasons for the observed differences. 21 Appendix Comparisons with Estimates by Loertscher and Wolter Luc Bauwens* The only comparable estimates of the determinants of intra-industry spe- cialization in bilateral trade are those by Loertscher and Wolter who made calculations for OE'CD countries other than Australia and New Zealand.1 Table 6 reports the Loertscher-Wolter r.esults, together with estimates for the same group of countries derived by using the data of this study. In order to ensure comparability, the variables used in the pre~ent study have been redefined to correspond to those of Loertscher and Wolter. This has involved using absolute differences of per capita GNP (DY/P) and GNP (DY) in the place of the inequality measures, omitting the trade oriehtation variable, and combining the dummy variables for economic integration (EI) and common languaes (CL). However, a cultural group (CG) dummy variable has not been included because of its intercorrelation with the language variable. In interpreting the estimates, it should be noted that the index of intra- industry trade used by Loertscher and Wolter (Pjk) is scaled to vary between 0 and 100. Furthermore, in applying logit analysis with weighted least squares, Loertscher and Wolter weighted the independent but not the dependent variables by [Pjk(100- Pjk)] 1fl (1980, p. 288). This is incorrect and in the estimates reported in equation (20) the dependent variable has also been appropriately *The author is Researcher at the World Bank. 1The authors used data for 59 out of 102 three-digit SI.TC categories to compute the index of intra-industy trade, utilizing equation (1) reported in the text, except that they made adjustment for imbalance in bilateral trade in manufactured goods rather than total trade (cf. p. 6 above). 22 weighted. At the same time, as noted below, the coefficient values and their level of significance are affected to a considerable extent if the incorrect weighting procedure is used. Table 6 further reports the results obtained by nonlinear least squares estimation of the logistic function used elsewhere in this investigation. The results show that the average size variable, the size differential variable, as well as the distance variable are statistically significant at the 1 percent level in all the equ~tions. The per capita income differential variable is also significant at the 1 percent level in the present study. In turn, the average per capita income variable is not significant in the Loertscher-Wolter study while it is significant at the 1 percent level in the estimates of this study. Among the remaining variables, the economic integration dummy is statistically significant at the 1 percent level in the estimates derived by weighted least squares in both studies whereas the common language dummy is significant at the 1 percent level in the nonlinear least squares, and at the 5 ]percent level in the weighted least squares, estimates of this study and not at all in the Loertscher-Wolter study.1 1Applying the incorrect weighting procedure used by Loertacher and Wolter to the data of the present study, the average per capita income, average size and distance variables are significant at the 1 percent level, the differential size variable barely attains the 10 percent level of significance whj,le the other va-r.iables are far from reaching this level. Furthermore, the absolute values of the coefficients are affected to a considerable extent, although all retain the expected sign. Finally, the standard error of the implied residuals is 0.146, while the adjusted coefficient of determination drops to 0.462. The latter value is nonetheless much higher than in the Loertscher-Wolter results (0.147). 23 Table 6 Estimates of Intra-Industr Trade amon OECD Countries (regression coefficients, with t-va ues in parenthesis Loertscher-Woltera This studl weighted least weighted least nonlinear least squares squares squares Equation (19) Equation (20) Equation (21) Constant -78.300 (u.a.) 1.855 (3.75) 2.(.. 19 (5.03) R.n AY/P -0 •.210 (0.25) 0.877 (6.06) 0.954 (6.60) R.n TJY./P -o. i~36 (8.92) -0.075 (1. 89) -0.063 (1.66) R.n AY 0.619 (7. 85) -0.580 (6.47) 0.617 (7.25) R.n DY -0.2:51 (2.67) -0.246 (4.34) -0.243 (4.73) R.n D -o .4.os (5.59) -0.438 (9.28) -0.527 (11.05) Border 0.094 (0.03) 0.259 (1. 93) 0.195 (1.61) EI 0.535 (2.12) 0.385 (2.36) 0.418 (2. 77) CL 0.166 (0.09) 0.286 (2 .13) 0.342 (2 .80) CG 0.363 (0.99) n.a. n.a. -2 R 0.147 0.798 0.927 e n.a. o.o89h 0.088 N 187 190 190 Note: (a)All the regression coefficients should be multiplied by 10-2. This is explained by the scaling of the index of intra-industry trade· and the lack of weighting of the depend~nt variable. (b)standard deviation of the differences between the observed values of IITjk and the estimated values computed by using the estimates of equation (20). · 24 References Aquino, Antonio (1978)» "Intra-Industry Trade and Inter-Industry Specialization as Concurrent Sources of International Trade in Manufactures," Weltwirtschaftliches Archiv 114(2), 175-96. Balassa, Bela (1966), "Tariff Reductions and Trade in Manufactures Among Industrial Countries," American Economic Review 56(3), 466-73. ·~~~--~(1975), European Economic I~tegration, Amsterdam, North-Holland, Chapter 2. "Intra-Industry Trade and the Integration of Developing ----:::---:---:--(1979), Countries in the World E•!onomy," in Herbert Giersch, ed., On the Economics of Intra-Industry Trade, Tubingen, J.C.B. Mohr (Paul Siebeck), 245-70. Chenery, Hollis B. (1960), "Patterns of Industrial Growth," American Economic Review 50(4), September, 624-54. Dixit, Avinash K. and Victor Norman (1980), Theory of International Trade, Digswell Place, Welwyn, James Nisbet & Co. and Cambridge University Press. Falvey, Rodney F. (1981), "Commercial Policy and Intra-Industry Trade," Journal of International Economics, 11(4), November, 495-512. Grubel, Herbert G. and P. J. Lloyd (1975), Intra-Industry Trade, London, Macmillan. Helpman, Elhanan (1981), "International Trade in the Presence of Product Differentiation, Economies of Scale and Monopolistic Competition: A Chamberlin-Heckscher-Ohlin Approach," Journal of International Economics, 11(3), August, 305-40. Krugman, Paul R. (1980), "Scale Economies, Product Differentiation, and the Pattern of Trade," American Economic Review, 70(5), December, 950-59. Lancaster, Kelvin (1980), "Intra-Industry Trade under Perfect Monopolistic Competition," Journal of International Economics, 10(2), May, 151-75. Linder, Staff an Burenstam (1961), An Essa't.. on Trade and Transformation, New York, John Wiley & Sons. Loertscher, Rudolf and Frank Wolter (1980), .. Determinants of Intra-Industry Trade: Among Countries and Across Indust-ries," Weltwirtschaftliches Archiv, 116(2), 280-92.