A "Stages" Approach to Comparative Advantage SWP256 World Bank Staff Working Paper No. 256 May 1977 PUB ti er is prepared for staff use. The views expressed are those of the Lha PUB Ynd not necessarily those of the W\lorld Bank. LJ 8G.5 by: BelaBalassa W 57 s Hopkins University R I E,0. no.256 1MI This paper is prepared for staff use and is not for publication. The views are those of the authors and not necessarily those of the Bank. INTERNATIONAL BANK FOR RECONSTRUCTION AND DEVELOPMENT Barik Staff Working Paper No. 256 May 1977 A 'STAGES' APPROACH TO COMPARATIVE ADVANTAGE This paper has investigated the changing pattern of comparative advantage in the process of economic development. Comparative advantage has been defined in terms of relative export performance, thus neglecting the composition of imports which is greatly affected by the structure of protection. The empirical estimates show that inter-country differences in the structure of exports are in large part explained by differences in physical and human capital endowments. The results lend support to the 'stages' approach to comparative advantage, according to which the structure of exports changes with the accumulation of physical and human capital. These findings have important policy implications for the developing countries. Firstly, they warn against distorting the system of incentives in favor of products in which a country has a comparative disadvantage. Secondly, the results can be utilized to gauge the direction in which a country's comparative advantage is moving. Finally, the results permit to dispel certain misapprehensions as regards the foreign demand constraint for develop- ing country exports. Thus as, in progressing on the comparative advantage scale, one developing country replaces another in exporting particular commodities to developed countries, the problem of adjustment in the latter group of countries will not arise. This paper was prepared by Bela Balassa, Professor of Political Economy, the Johns Hopkins University in the framework of a consultant arrange- ment with the World Bank. The author is indebted to Dominique de Crayencour, Jonathan Levy and especially to Kishore Nadkarni for research assistance. The author has 'benefited from comments on an earlier version of the paper by T. N. Srinivasan and other participants of a seminar held at the World Bank. The paper will be presented at the 5th World Congress of the International Economic Association to be held in Tokyo on August 29 - September 3, 1977. A 'Stages' Approach to Comparative Advantage Bela Balassa This paper will examine intercountry differences in export structure, with a view to indicating the changing pattern of comparative advantage in the process of economic development. The investigation will be limited to exports, since the commodity pattern of imports is greatly influenced by the system of protection in the importing countries. And, as trade in natural resource Products deDends to a considerable extent on the country's resource endowmnent, we will deal with comparative advantage in manufactured goods alone. Section I of the paper will consider the relevance for the develop- ing countries of explanal:ions of international specialization based on factor proportions and technological variables. Section II will describe the product classification schemes and country characteristics used to evaluate comparative advantage. The empirica:L results on the changing pattern of comparative advantage will be presented in Section III; they will be further analysed in Section IV. Section V w:Lll indicate the policy implications of the results. I Hufbauer (1970) was the first to introduce the distinction between the neo-factor proportions and the neo-technological explanations of comparative advantage. The-former combines human capital with physical capital and relates the sum of the two to (unskilled) labor. In turn, the latter emphasizes the role of technological change, the product cycle, and economies of scale in determining the pattern of international specialization. According to Hufbauer, if technological factors "were somehow combined into a single characteristic, that characteristic might prove as powerful as Lary's single measure (value added per man) of human and physical capital in explaining trade flows" (1970, p. 196). While such a single characteristic has not been established, it has been suggested that "there appears to be a new consensus emerging concerning the power of the neo-technol- ogy theory over the neo-factor proportions theory" (Goodman-Ceyhun, 1976, p. 551). The results of several recent studies on U.S. trade tend to support this conclusion. Goodman and Ceyhun have found that "the variables describing different facets of the technology phenomena are singularly the most important variables, which suggest the importance of the neo-technology hypothesis in the explanation of international trade in manufactures" (op. cit. p. 547). Similar results have been reached by Baldwin (1971) and by Branson and Junz (1972), These authors have shown that net U.S. exports are negatively correlated with physical capital intensity. Baldwin also finds that general measures of human capital, such as the average cost of education, average years of education, and average earnings,are not statistically significant-/ in explaining U.S. trade. And, while the human capital variable is positively correlated with net exports in the Branson-Junz study, its level of statistical significance is greatly reduced once technological variables are introduced in the equation. 1/ This conclusion applies also in U.S. trade with Canada, Western Europe, the developing countries, and the rest of the world other than Japan, although statistically significant results were obtained for U.S. trade with Japan, (1971, p. 140). - 3 - Amongtechnological variables, R&D expenditure performs the best in the Branson-Junz study whereas Baldwin finds the number of engineers and scientists to be the most important explanatory factor, adding further that "probably of even more importance is the fact that a significant part of this labor group is engaged in research and development activities" (1971, p. 142). Finally, Morall concludes that "the United States' comparative advantage in skill intensive products 'must be due to mechanisms such as the product cycle model, the government subsidy of R&D explanation, the economies of scale in R&D arguments, or the dynamic shortage theory" (1972, p. 120).-/ The results obtained for the United States,however, have limited relevance for our inquiry into the changing pattern of comparative advantage for the developing countries. These countries are at the other end of the spectrum from the United States and engage in research and development to a very small degree, if at all.-/ Accordingly, we will next consider the determinants of trade between developed and developing countries. Postulating that light manufactures are relatively labor intensive and heavy manufactures capital intensive, Kojima (1970) has concluded that the factor proportions explanation is valid to trade between developed and develop- ing countries. Defining capital intensity in terms of value added per man, 1/ A dissenting voice is that of Harkness and Kyle (1975). However, the results of these authors were obtained by replacing a continuous variable (net exports or export-import ratios) with a binary variable, classifying industries into two groups according to whether exports exceed, or fall short: of, imports. This choice brings into question the validity of the results, in part because of the error possibilities involved in a binary classification and in part because large and small export-import balances are given equal weight. 2/ This conclusion also applies to Leamer's findings as to "the clear superiority of the research and development variable" (1974, p. 369) in determining export-import ratios for two-digit SITC categories in a group of twelve developed countries including the United States, who carry out much of their trade with each other. -4- taken to reflect the use of physical as well as human capital, Lary has also found that developing countries tend to export labor-intensive manu- factures (1968, ch. 4). This conclusion has been reinforced as regards U.S. imports from developing countries by Mahfuzur Rahman (1973), who defined capital in physical terms, and for German imports from developing countries by Fels (1972), who defined capital as the sum of the value of the (physical) capital stock and the discounted value of the difference between average wages and unskilled wages in particular industries, taken as a proxy for human capital. In examining trade between developed and developing countries, however, these authors have divided a continuum more-or-less arbitrarily into two segments, hence their estimates cannot be used to indicate changes in the pattern of comparative advantage in the process of economic development. Continuous variables as regards country characteristics have been used by Hufbauer in attempting to explain intercountry differences in the average values of particular product characteristics.-/ But, in his sample of 24 countries. Hufbauer has included only 9 countries which may be considered developing, and most of these belong to the semi-industrial group (1970, p. 157).-/ In turn, Hirsch (1974) has classified 18 industry groups in three categories, according to whether the correlation between export performance and value added per worker in the nonagricultural sector, estimated in an intercountry framework, was positive, zero, or negative. Hirsch has also made estimates for individual countries by regressing export-output ratios in the 1/ E.g. the average physical capital intensity of exports was related to intercountry differences in physical capital per man. 2/ On the definition used, see Section III below. - 5 - 18 industry breakdown on the skill, physical capital, scale, and natural resource characteristics cf these industries, and has grouped the results obtained for the 29 countries studied into four categories according to their per capita incomes (high income, medium high income, medium low income and low income groups). Nlo attempt has been made, however, to establish a relationship between the two sets of estimates. Considering further the low level of significance of 1:he estimated regression coefficients in the country equations, the crudeness of the fourfold country classification scheme, and 1/ the high.degree of commod:ity aggregation- , the results have remained rather impressionistic. Finally, Herman and Tinbergen (1970) and, subsequently, Herman (1975) have classified countries into eleven categories on the basis of their physical and human capita.L endowments. However, the sources cited provide no information that would permit estimating physical capital endowments and the proxy used for human capital (the cost of educating professional, technical, and related workers classified in Group 0/1 in the International Standard Classification of Occupations used by the ILO) includes personnel in liberal occupations, such as jurists, preachers, artists, and athletes while excluding production supervisors, foremen, and skilled workers that are of considerable importance in the developing countries. Also, the results derived regarding comparative advantage have not been subjected to statistical testing. 1/ An even greater level of aggregation (1-digit Standard International Trade Classification categories) is used by Banerji who distinquishes among four commodity categories and has not been successful in introducing variables directly expressing comparative advantage in the regression equations (1975, Ch. III). -6- II We have briefly reviewed recent efforts made to examine the pattern of comparative advantage,-/ with emphasis on the relevance of the results for the developing countries. It has been shown that applications of the neo- technological theory largely pertain to U.S. trade, particularly with the developed countries. In turn, in statistical investigations that have included developing countries, these countries have been considered as a group or, alternatively, they have been classified on the basis of a single criterion, such as per capita incomes or average value added per employee in the nonagricultural sector. In the latter case, intercountry regressions have been estimated by relating average product characteristics for all manufactured exports or for aggregate industry-groups to a particular country characteristic. A considerable degree of comumodity aggregation has been employed also in examining the relationship between product characteristics and the export structure of the individual countries in an interindustry framework. At the same time, no linkage has been established between the intercountry and interindustry estimates. A different approach has been followed in the present study. Thirty- six countries have been chosen for the investigation, of which 18 are developed and 18 developing. For each country, regression equations have been estimated relating their "revealed" comparative advantage in 184 product categories in the year 1972 to various product characteristics. The regression 1/ For an excellent review of earlier contributions the reader is referred to Stern (1975). - 7 - coefficients thus obtained have in turn been correlated with particular country characteristics in inter-country regressions so as to indicate the effects of these country characteristics on international specialization. The first question concerns the choice of product characteristics for the investigation. Harry Johnson has suggested to extend the concept of capital to include human capital as well as intellectual capital in the form of production knowledge, noting that, "such an extension is fully consistent with Irving Fisher's approach to the relation between capital and income" (1970, p. 17). However, as Branson observes, the aggregation of various forms of capital assu'res that they are perfect complements or perfect substitutes in production- (1973, p. 11). In the present study, we have experimented with an aggregate measure of capital as well as with separate variables for physical and human capital.i! Investment in research and development has been subsumed under the two, as this is in part embodied in physical capital (e.g. laboratories) and in part in human capital (scientists and engineers engaged in R&D). This procedure appears appropriate in an investigation of the changing pattern of comparative advantage in the process of development since, as noted above, developing countries carry out hardly any R&D, so that little is lost in combining 1/ On the complementariity of physical and human capital, see Fallon and Layard, 1975. 2/ Physical and human capital have also been separated in a recent article by Hirsch, which has come to the author's attention after this paper was completed. Hirsch makes a distinction between high-skill and low-skill industries, further separating (physical) capital and (unskilled) labor- intensive industries within each. For each group, export performance is related to incomes per head, taken as a proxy for physical and human capital (Hirsch, 1975). Thus, in contradistinction with the present study, an aggregated commodity classification scheme is used and capital endow- ment variables are not introduced in the analysis. Also, the human capital 'intensity of the different product categories is defined in terms of skill intensity, which wa.s criticized in connection with the-Herman- Tinbergen study above. -8- intellectual capital in the form of production knowledge with physical and human capital. Capital intensity may be defined in terms of flows (Lary's measure of value added per worker) or stocks (the value of the capital stock plus the discounted value of the difference between average wages and the unskilled wage, divided by the number of workers). The latter approach was used by Kenen (1965) and, recently, by Fels (1972) and by Branson (1973) The stock measure of capital irtensity (k ) is expressed in (1) (1) k.s = pi s+ hi ' Pi + i i- 1 ~~~~~~h r for industry i, where pi and hi respectively, refer to physical and human capital per man, w1 is the average wage rate, wu the wage of unskilled labor, and r the discount rate used in calculating the stock of human capital. This approach implicitly assumes that the rental price of physical capital, i.e. the risk-free rate of return and the rate of depreciation, is the same in all industries. This assumption is made explicit in expressing the flow equivalent (FE) of the stock measure of capital intensity as in (2), where rP is the (2) (FE) ki5 = pi(rp + d)+ - -W ) discount rate for physical capital and d is the rate of depreciation. In turn, the flow measure of capital intensity (k ) can be expressed as in (3) where va refers to value added per man. Now, nonwage value added - 9- f f f per m a vai = pi hi = (vai-wi)+wi = (vai-Wi) + iWiu + wil per man (vai-wi) is taken to represent physical capital intensity and wage value added per man (w human capital intensity. As far as physical capital intensity is concerned, the two measures will give the same result: in risk free equilibriumn,provided that product, capital, and labor markets are perfect and nonwage value added does not include any items other than capital remuneration. However, production is subject to risks that vary among industries and assuming risk aversion, profit rates will include a risk premium that will differ from industry to industry. Also, the situation in a particular year will not represent an equilibrium position and this fact, as well as imperfections in product, capital, and labor markets, will further contribute to interindustry variations in profits. Moreover, nonwage value added may include items other than capital's remuneration, such as advertising. Finally, while the stock measure imnputes differences between average wages and the unskilled wage to human capital, the flow measure includes the entire wage value added under this heading, thus overestimating human capital intensity by the amount of the unskilled wage. This would not give rise to problems if the unskilled wages were the same in every industry. However, unskilled wages may differ among industries due to factors such as thedisutility of work and the power of labor unions. The existence cf interindustry differences in risk,market imperfections, the inclusion of items other than capital's remuneration in non-wage value added, and the inclusion of unskilled wages in wage value added represent deficiencies - 10 - of the flow measures of capital intensity. In turn, the lack of consideration given to interindustry differences in depreciation rates and in the extent of obsolesence of existing equipment, as well as the use of historical rather than replacement values for physical capital, represent disadvantages of the stock measure. The implications of the described shortcomings of the two measures of capital intensity for the results will depend on the particular circumstances of the situation. The usefulness of the stock measure would be greatly impaired in an inflationary situation where historical and replacement values differ and the magnitude of their differences varies with the age of equipment. This is not the case in the present study since the benchmark years used for estimating capital intensity (1969 and 1970) are part of a long noninfla- tionary period. By contrast, the usefulness of the flow measure is limited by reason of the fact that profit rates show considerable variation over time and interindustry differences in profit rates cannot be fully explained by reference to risk factors.-/ These considerations tend to favor the use of the stock measure of capital intensity. Nevertheless, given the error possibilities involved, interest attaches to making estimates by the use of both measures- , which also permits us to examine the stability of the results derived under alter- native assumptions. This has been done in the present study, with emphasis given to the estimates obtained by the use of the stock measure in evaluating the results. l/ Reference is made here to U.S. data which were used in the calculations as noted below. 2/ Fels has employed both measures in correlating net German exports with capital intensity in a nineteen industry sample (1972, Table 3). In turn, Lary has used Hufbauer's data to calculate the rank correlation coefficient between country averages of value added per employee in exports and per capita incomes (1970). - 11 - For purposes oE the calculations, we have attempted to obtain data on the capital intensity of the production process for Japan, the factor intensities of which may be presumed to lie in-between the relevant magni- tudes for highly-developed and less-developed countries. However, for lack of information on physicaL capital and on unskilled wages in a sufficiently detailed breakdown, this attempt had to be abandoned and we have had to have recourse to U.S. data. The use of U.S. data in the investigation will be appropriate if factor substitution elasticities are zero or they are identical for every product category. While this assumption is not fulfilled in practice, Lary has shown variations in capital intensity to be small in U.S. - U.K., U.S. - Japan, and U.S. - India comparisons as regards his value added measure (1968, Appendix D). For lack of data, similar comparisons could not be made for the stock measure and the further investigation of this question had to be left for future research. In defining the manufacturing sector for purposes of the present investigation, we have talcen the concept used in the U.S. Standard Industrial Classification (SIC) as our point of departure. We have excluded from this category (SIC 19 to 39)) foods and beverages (SIC 20) and tobacco (SIC 21), where the high cost of transportation and the perishability of the basic material give an advantage to primary-producing countries. We have further excluded primary nonferrous metals (SIC 333) where transportation costs account for a high proporl:ion of delivered price of the basic material, and ordnance (SIC 19) where comparable trade data are not available. In turn, given the relatively low cost of transporting the raw material (Cornwall, 1972) and the prevalence of exports based on imported materials, we have retained petroleum products and wood products in the manufactured product category. - 12 - We have also retained nonmetallic mineral products bv reason of the ubiquity of the basic materials. Defining the manufacturing sector as SIC industry groups 22 to 39 less 333, the product classification scheme used in this study has been established on the basis of the 4-digit SIC categories. Particular 4-digit categories have been merged in cases when the economic characteristics of the products in question were judged to be very similar and when comparable data did not exist according to the U.N. Standard International Trade Classification, which has been used to collect trade figures. Appendix Table 1 provides information on the capital intensity of the 184 product categories chosen, using the stock as well as the flow measure of capital, and further separating physical and human capital. In turn, Appendix Table 2 shows the SIC and SITC categories corresponding to these product categories. Data on the capital stock, employment, value added, and wages used in calculating capital intensity originate from the U.S. Census of Manufactur- ing. In turn, the data for unskilled wages have been taken from the MonthZy Labor Reviews, published by the U.S. Bureau of Labor Statistics; they relate to 2-digit industries, thus involving the assumption that unskilled wages are equalized at this level. In order to reduce the effects of variations due to the business cycle and nonrecurring events, we have used simple averages of data for the two latest years (1969 and 1970) for which information was available. Finally, we have estimated the value of human capital under the stock measure by discounting differences between the average wage and the unskilled wage for the individual product categories at a rate of 10 percent.1/ 1/ This is in-between the discount rates of 9.0 and 12.7 percent used by Kenen (1965); the same discount rate was used by Fels (1972) and by Branson (1973). - 13 - As noted earliEr, the study covers altogether 36 countries. The sample is evenly divided between developed and developing countries; the countries in the first group had per capita incomes exceeding $1800 in 1972; incomes per head were be]ow $1400 in the second group. The variability of per capita incomes is 1:3 in the developed country subsample, 1:13 in the developing country subsample, and 1:56 in the entire sample. Thus, the sample, and, in particular the developing country subsample, exhibits considerable variability, which permi:s indicating the changing pattern of comparative advantage in the process of economic development. The distinctioni between developed and developing countries has been introduced in the econometric analysis through the use of a dummy variable for developed countries. At the same time, we have used continuous variables to denote country characteristics, including physical and human capital endowment. These are shown in Table 1 together with per capita incomes.- In the absence of data on the physical capital stock in the individual countries, we have used the sum of gross fixed investment over the period 1955-71, estimated in constant prices and converted into U.S. dollars at 1963 exchange rates, as a proxy for capital endowment. The data have been derived from the WorZd TabZes, 1976, published by the World Bank; they have been expressed in per capita terms. A similar procedure has been employed by Hufbauer, except that he used data for the period 1953-64 and included manufacturing investment only (1970, p. 157). The choice of a longer period in the present study reflects the fact that capital equipment is used beyond eleven years; also, 1/ On the use of per capita incomes as one of the explanatory variables, see Section IV below. - 13a - Table 1 Country Characteristics and Regression Coefficients Obtained in Estimates for Individual Countries Country Characteristics Regression Coefficients DUMMY GNPCAP GDICAP HMIND SKILLS 3s gf fsp 6sh 6fp 6fh Argentina 0 1139.65 2013.68 122.0 8.76 .32** .19* .25 -.04 .60* -1.49 Australia 1 3271.69 6675.24 183.3 10.93 .34* .78* .23* .12 .50* .09 Austria 1 2741.26 5129.79 112.9 10.08 -.31 -.93 -.33* .04 -.64* .03* Belgium 1 3701.15 5441.70 140.5 11.62 .11* .04* .24 -.11 .51** -1.30 Brazil 0 511.27 1016.00 29.3 8.57 -.69* -1.48* -.35* -.42 -.74 -.80* Canada 1 4691.51 7970.65 179.9 14.92 .75* .87* .46 .25* -.22 2.25k Colombia 0 357.08 751.59 32.3 7.41 -1.31* -2.48 -.06* -1.31 -.33 -3.82 Denmark 1 4187.67 6259.56 139.2 13.63 -.40 -.12* -.44 .05** -.15 .08 Finland 1 2877.73 6999.27 109.9 14.89 -.26 -.62 .08 -.37 -.32 -.34** France 1 3841.68 7211.24 138.8 13.46 -.07* -.08* -.06 -.00Z .11 -.50* Germany 1 4218.84 7102.15 114.3 10.71 .20 .43 -.05 .26 .05 .69* Greece 0 1407.20 2196.43 93.7 9.60 -.27* -1.05* .11* -.49** -.08* -1.90* Hong Kong 0 1048.88 1370.61 60.7 5.09 -2.30* -2.84* -1.83* -.52 -.94 -3.15* India 0 102.03 214.25 50.2 9.64 -1.10* -2.30* -.93* -.09 -.19** -4.58 Ireland 1 1840.20 2701.89 110.7 12.53 -.48** -.80* -.44 -.04** -.39 -.66 Israel 1 2416.28 4280.96 148.9 19.19 -.37* -.70* -.02** -.41 .27** -2.02 Italy 1 2176.52 3366.47 91.3 7.05 -.33** -.46* -.20* -.12 -.29* -.06** Japan 1 2740.95 4765.11 146.2 8.26 -.31* -.52* -.42 .11* -.70** .86* Korea 0 301.03 402.89 66.7 6.15 -1.67* -3.02* -.46 -1.24** -69 -3.91* Malaysia 0 408.62 494.56 34.5 9.53 -.88* -2.32* -.26 -.63* -.56 -3.65* Mexico 0 745.41 1067.02 41.1 9.22 -.91* -1.48* -.17 -.80* -.40 -2.38* Morocco 0 279.13 293.08 27.9 8.23 -1.18* -2.95* -.18* -1.06 -.23* -6.10* Netherlands 1 3466.90 5375.15 158.6 11.62 .28 .44 .22* .07 .55 -.67 Norway 1 3786.91 7806.11 107.4 13.91 .22* .01* .44* -.25 -.05 .18* Pakistan 0 104.11 197.76 33.1 4.63 -1.56* -3.11* -.93 -.63* -.51 -5.78* Philippines 0 223.50 448.72 134.2 10.98 -1.34* -2.28* -.07 -1.39* -.53* -3.03 Portugal 0 1084.26 1154.43 68.1 5.09 -.81* -2.09* -.01* -.90 -.82* -1.80* Singapore 0 1354.41 1189.84 97.6 8.00 -1.47* -2.35* -1.11 -.38* -1.03 -2.01** Spain 0 1333.76 2049.09 63.4 7:28 -.43 -.56 -.03 -.42* -.12* -.82* Sweden 1 5141.10 9452.90 129.6 20.87 .21 .15 -.16* .39* -.44 1.50 Switzerland 1 4810.02 8852.63 112.6 13.13 .04* -.10* -.44* .50* .08* -.23* Taiwan 0 481.94 629.88 103.5 7.06 -1.56 -2.61* -.68 -.87 -.98 -2.45* Turkey 0 431.16 581.22 37.5 10.32 -.42 -1.62* -.06** -.42* -.36 -2.01 U.K. 1 2765.25 4844.68 136.2 11.44 .13* .46* -.18** .34* .19 .31* U.S.A. 1 5679.47 7616.20 325.0 14.21 .84** 1.47* .24 .62 .23* 2.22 Yugoslavia 0 798.30 1162.06 110.0 13.73 -.47 -1.41 -.29 -.20 -.81 -.60 Country Characteristics: DU1IY -- 1 for developed, 0 for developing countries; GNPCAP = GNP per capita in 1972, $US; GDICAP = Cumulated gross fixed investment per capita, 1955-71, $US; HMIND = Harbison-Myers index; SKILLS = share of professional, technical and related workers in the total non- agricultural labor force. Regression Coefficients have been obtained by regressing for each country the ratio of 'revealed' comparative advantage, estimated for 184 product categories, on measures of capital intensity. The coefficient 65 has been estimated by regressing the comparative advantagg1 ratio on the stock measure of total capital intensity as in (4), while 3. and 6 I lave been obtair.ed by regressing this ratio on physical and human capital intensity, introduced simultaneously in the estimating efuaticr. as in (5), again using the stock measure of capital. Bj , . h, and 6-f are the corresponding regression coef- ficients estimated by substituting the flow measure of capital in the place of the stock measure in (4) and (5). Regression coefficients that are significant at the 5 percent level have been denoted by * and those significant at the 10 percent level by **. - 14 - we have considered all capital, and not only that used in the manufacturing sector.'/ In turn, in usLng value added per worker (1974, p. 542) as a proxy for capital endowmenit, Hirsch does not separate physical and. human capital and neglects inte:!country differences in profit rates and in unskilled wages. Hufbauer has taken the ratio of professional, technical and related workers to the labor force in manufacturing as a proxy for human capital endowment (1970, p. 158). As noted earlier, the use of this measure is objectionable, because it includes various liberal occupations while excluding production supe'rvisors, foremen, and skilled workers that are of considerable importance i;l the developing countries. A more appropriate procedure appears to be to make use of the Harbison-Myers index of human resource development.- While this index is a flow measure,-/ the use of estimates pertaining to 1965 (Harbison, Maruhnic, and Resnick, 1970, pp. 175-6) permits us to provide an indication of a country's general educational level, and thus its human capital base, in 1972, the year for which trade data have been obtained. Nevertheless, we have also experimented with the skill ratio employed by Hufbauer, utilizing the data reported in the ILO Yearbook of Labor Statistics. 1/ This choice can be rationalized on the grounds that, ex ante, capital can be allocated to manufacturing as well as to other sectors. And while adjustments would need to be made if there was complementarity between capital and natural resources in certain uses, such as mining information on the sectoral composition of investment was not available for a number of the countries under study. 2/ This index has also been used in a study of world trade flows by Gruber and Vernon (1970). 3/ It is derived as the secondary school enrollment rate plus five times the university enrollment rate-in the respective age cohorts. - 15 - III As noted in the introduction, the investigation is limited to exports since the commodity pattern of imports is greatly influenced by the system of protection. Following earlier work by the author (1975), a country's relative export performance in the individual product categories has been taken as an indication of its 'revealed' comparative advantage. For this purpose, we have calculated the ratio of a country's share in the world exports of a particular commodity to its share in the world exports of all manufactured goods. Thus, a ratio of 1.10 (.90) means that the country's share in a particular product category is 10 percent higher (lower) than its share in all manufactured exports.-/ These ratios can be considered to express a country's comparative advantage in manufactured goods that are characterized by product differentiation and are hence exported by a variety of countries. For each of the 36 countries, the ratio of 'revealed' comparative advantage, calculated for the individual product categories, has been regressed on variables representing the capital intensity of the individual product categories. Separate equations have been estimated using the stock and the flow measures of (total) capital intensity, as well as by simultaneously introducing physical and human capital under the two definitions of capital intensity. The estimating equation is shown in (4) for total capital intensity (4) log x = log a + % log k-s 1/ An alternative measure would involve relating exports to output in each country. In the absence of output figures, however, this measure could not be utilized in the present study. At any rate, it would require adjusting for country size (Balassa, 1968) while the measure used here does not require such an adjustment. - 16 - and in (5) for physical and human capital intensity for the case when the stock measure of capital -is used. The same equational forms have been used in conjunction with the flow measure of capital. The equations have been estimated in a double-logarithmic form, (5) log x = log aj + Sp log-Pi + sh log his so that the value of the $ coefficient for country j indicates the percent- age change in the country's comparative advantage ratio (xij) associated with a one percent change in capital intensity.!' A positive (negative) S coefficient thus shows that a country has a comparative advantage in capi- tal (labor) intensive products while the numerical magnitude of the S coef- ficient indicates the extent of the country's comparative advantage in capital (labor) intensive commodities.-/ The estimated S coefficients are reported in Table 1. In the regression equations utilizing the stock measure of (total). capital intensity, the S coefficient is statistically significant at the 5 percent level for 22 countries and at the 10 percent level for 26 countries. In turn, in regression equations utilizing the flow measure, the coefficient is significant at the 5 percent level for 29 countries, with no additional countries included at the 10 percent level. Note further that the a coefficients that have values near to zero have an economic interpretation even if they are not significantly 1/ Since the logarithm o:f zero is undefined, in the estimating equations, an export ratio of .01)1 has been used to represent cases when-the exports of a country in a pa:rticular product category were nil. We have also experimented with the use of a .01 ratio and have obtained practically the same results. No:r are the results materially affected if we drop the zero observations from the regressions. This and other estimates not reported in the paper are available from the author on request. 2/ Alternatively, use mar be made of non-parametric tests involving the calculation of the Spearman rank correlation coefficient between the 'revealed' comparative advantage-ratio and the individual factor intensity measures. This test has the disadvantage, however, that it cannot handle more than one explanatory variable and that it does not permit indicating the implicaticns of the intercountry results for a country's future compara- tive advantage (on the last point, see the concluding section). - 17 - different from zero; they indicate that a country is at the dividing line as far as comparative advantage in capital and labor-intensive products is concerned. The 6 coefficients estimated by using the stock and the flow measures of capital intensity are highly correlated,with a Spearman rank correlation coefficient of .956 . In turn, in estimates obtained by disaggregating capital into its physical and human capital components, a high degree of correspondence has been obtained in regard to the S coefficients pertaining to human capital intensity (Spearman rank correlation coefficient of .841) but not for physical capital intensity (Spearman rank correlation coefficient of .650). These differences are explained if we consider that human capital intensity was defined in a similar way under the stock and the flow measure of capital while this was not the case for physical capital intensity. The level of statistical significance of the coefficients, too, is lower if we disaggregate capital into its physical and human capital components. The $ coefficients are significant at the 5 (10) percent confidence level in 14 (17) cases for the physical capital intensity variable and in 13 (17) cases for the human capital intensity variable if we use a stock measure of capital. The corresponding figures are 11 (15)for the physical capital intensity variable and 21 (24) for the human capital intensity variable under the flow measure.l/ 1/ The results contrast with those obtained by Helleiner who found total (physical and human) capital intensity to have lower explanatory power than skill intensity alone. But, Helleiner's results pertain to the trade of the LDCs taken as a whole; he did not employ a stock measure of capital; and he used the average wage as a measure of skill intensity. (1976). Helleiner also used some additional variables, of which scale economies was statistically significant in trade between developing and developed countries (1976, p. 512). However, comparative advantage in products subject to scale economies is related to the size of the domestic market (Balassa, 1968) and, with developing countries having smaller markets, Helleiner's results raise problems of identification. - 18 - Next, we have tested the hypothesis that intercountry differences in the 8 coefficients can be explained by differences in country character- istics that determine the pattern of comparative advantage. This test has been carried out by regressing the 8 coefficients estimated for the indi- vidual countries on variables representing their per capita physical and human capital endowments and the level of economic development in an inter- country framework. (6) shows the estimating equation for the case when per (6) 6. = f(GDICAP., HMINDj, DUMMY) capita physical capital endowment (GDICAP) and human capital endowment 1/ (HMIND) are introduced sinultaneously in the equation/- and a 0-1 variable (DUMMY) is used to indicate whether a country belongs to the developing or developed group. In this way, we test the hypothesis that the two capital endowment variables independently affect comparative advantage. Statistically significant results have been obtained in estimat- ing equation (6) for both the physical and the human capital endowment variables, regardless of whether the dependent variable originated in country regressions utilizing the stock or the flow measure of capital intensity. In both regressions, the physical as well as the human capital endowment variables are significant: at the 1 percent confidence level, while the co- efficient of determination is .65 using the stock measure and .77 using the flow measure of capit:al intensity (equations 1.6 and 2.6 in Table 2). The level of statistical significance of the regression coefficients for the physical capital endowment variable is however much lower if it is used in conjtnction wLth the human capital intensity variable without 1/ In order to minimize problems related to heteroscedasticity, the data for the individual countries have been weighted by the inverse of the standard error of the a coefficients. While in this way greater weight is given to the 6 coefficients obtained in (4) and (5) that have a higher degree of statistical significance, broadly similar results have been obtained by using unweighted data. 2/ Regressing the rank correlation coefficients on factor endowment variables has generally confinned the reported results, although the level of statistical significance of the coefficients was somewhat lower. - 18a - TABLE 2 INTERCOUNTRY REGRESSION EOUATIONS FOR THE TOTAL CAPITAL INTENSITY MEASURE (Weighted Regressions Using Reciprocals of Standard Errors of the Dependent Variable as Weights) Dependent Equation Coefficient of Explanatory Variables Variable Number Determination GDICAP HMIND DUMMY SKILLS CONSTANT S Bj 1.1 .53 ., (6.13) -2.85(-7.12) 1.2 .60 1.32 (7.15) -3.33(-8.24) 1.3 .62 .27 (1.27) .96 (2.85) -3.32(-9.31) 1.4 .53 .97 (2.90) -.14 (-.62) -2.83(-7.02) 1.' .60 1.39 (3.96) -.03 (-.24) -3.33(-8.12) 1.6 .65 .67 (2.21) 1.17 (3.38) -.37 (-1.80) -3.3b(-9.72) 1.7 .53 .67 (1.82) .008(.32) -3.00(-4.74) 1.8 .53 .83 (1.91) -.17 (-.72) .01 (.50) -3.09(-4.36) 1.9 .62 .36 (1.02) .99 (2.81) -.007(-.32) -3.19(-5.50) 1.10 .65 .67 (1.74) 1.17 (3.27) -.37 (-1.74) .0005(.02) -3.40(-5.91) 2.1 .61 1.42 (7.30) -3.59(-8..6) 2.2 .73 2.45 (9.69) -4.20(-11.49) 2.3 .75 .39 (1.32) 1.94 (4.23) -4.21(-11.62) 2.4 .61 1.46 (2.81) -.02 (-.06) -3.59(-8.71) 2.5 .73 2.46 (5.19) -.009(-.05) -4.21(-11.30) 2.6 .77 .83 (2.07) 2.18 (4.60) -.45 (-1.56) 4.26(-11.97) 2.7 .61 1.32 (2.35) .008(.21) -3.69(-5.79) 2.8 .61 1.36 (2.03) -.05 (-.12) -.009(.23) -3.71(-5.64) 2.9 .75 .69 (1.46) 2.04 (4.28) -.03 (-.82) -3.91(-7.53) 2.1u .77 1.05 (1.97) 2.22 (4.57) -.42 (-1.41) -.02(-.53) -4.06(-7.77) Note: for explanation of symbols, see Table 1 - 19 - introducing a dummy to represent the level of economic development. This result indicates that the effects of physical capital endowments but not those of human capital endowments on international specialization depend on the level of economic development. We have also experimented with the ratio of professional, techni- cal, and related workers to the total in the place of, and together with, the Harbison-Myers index.. The skill-ratio variable (SKILLS) is not statis- tically significant at even the 10 percent level and it fails to raise the coefficient of determinal:ion. It can thus be rejected on statistical grounds. It will be reca3lled that the-level of statistical-significance of the 0 coefficients for the physical capital intensity variable in the country regressions has been generally low. Statistically poor results have been obtained also iLn regressing these coefficients on variables representing physical and human capital endowment in an intercountry frame- work as in (6). The exp:Lanatory power of the regressions is very low as is the level of statistical significance of the coefficients in cases when the physical and the human capital endowment var4ables are introduced simultaneously in the estimating equation, However, the coefficients are statistically significant: when these variables are introduced separately (Table 3). The explanatory power of the regressions is relatively high in cases when the 3 coefficients obtained in (5) in regard to human capital intensity are used as thes dependent variable. Also, both the physical and the hui:ara capital endowment variables are highly significant when inero- duced simultaneously in the equations. The level of significance is some- what lower in cases when the stock measure of capital intensity is used instead. - 19a - TABLE 3 REGRESSION EQUATIONS FOR PHYSICAL AND HUMAN CAPITAL INTENSITY MEASURES (Weighted Regressions Usirng the Reciprocals of the Standard Errors of the Dependent Variable as Weights) Dependent Equation Coefficient of Explanatory Variables Variable Number Determination GDICAP HMIND DUMMY CONSTANT Sjp 1.1 .13 .41 (2.29) -1.40 (-3.46) 1.2 .07 .19 (1.65) -1.13 (-2.98) 1.3 .14 -.08 (-.39) .52 (1.56) -1.40 (-3.42) 1.4 .16 .15 ( .49) .64 (1.81) -.21 (-1.02) -1.44 (-3.50) S sh 2.1 .58 .98 (6.86) -1.98 (-6.90) 2.2 .60 .62 (7.14) -1.75 (-6.89) 2.3 .64 .37 (2.38) .49 (1.99) -1.98 (-7.35) 2.4 .65 .49 (2.14) .56 (2.11) -.12 (-.74) -1.99 (-7.34) fp P 3.1 .26 .64 (3.43) -1.36 (-4.26) 3.2 .18 .35 (2.71) -1.13 (-3.63) 3.3 .26 -.006(-.02) .65 (1.87) -1.36 (-4.19) 3.4 .26 -.10 (-.31) .61 (1.65) .09 (.39) -1.35 (-4.10) fh 4.1 .50 3.07 (5.86) -2.76 (-6.76) 4.2 .48 1.95 (5.55) -2.47 (-6.38 4.3 .53 .91 (1.47) 1.91 (2.02) -2.78 (-6.93) 4.4 .58 2.14 (2.38) 2.44 (2.55) -1.10 (-1.83) -2.82 (-7.27) Note: For explanation of symbols, see Table 1 - 20 - IV We have further examined deviations from the relationships estimated in an intercountry context:. Upward deviations from the regression line are shown with respect to the physical capital endowment, but not with regard to the human capital. endowment, of Argentina and the United States. The results indicate that the actual capital intensity of the exports of these countries much exceeded expected va:Lues based on their physical and human capital endow- ments. The results for Argentina are explained if we consider that, during the period under study, this country represented an extreme case among the developing countries as far as distortions due to the application of protective measures are concerned. These distortions, in turn, have affected the pattern of exports and imports; in particular, with the implicit subsidy to capital goods through t:he overvaluation of the exchange rate associated with high protection, exports have been biased in a capital-intensive direction. The results for t:he United States are somewhat of a puzzle as the findings of other authors would have led us to expect that actual U.S. exports are less, rather than more, physical-capital intensive than the hypothetical exports derived from intercountry relationships. And while the solution to the puzzle may well be that the ratio of physical,to human capital intensity is even higher for the imports than for the exports of the United States, our results conflict with those of Hufbauer which show the U.S. to be below the regression line (1970, p. 169). Note, however, that Hufbauer's results pertain to an earlier year and he provides evidence that U.S. exports have become increasingly physi,al-capital intensive over time. Finally, our calculations using direct input coefficients are preferable to earlier estimates derived by the use of direct plus indirect coefficients once we admit international trade in intermediate products. - 21 - In turn, the exports of Hong Kong are less capital-intensive than expected on the basis of its physical capital endowment. It would appear that Hong Kong's export structure does not yet fully reflect the large investments in physical capital carried out during the period under consideration. Finally, deviations from the regression line are relatively small in regard to human capital endowment. Next we have estimated a matrix of Spearman rank correlation coeffici- ents for pairs of country characteristics in the 36 country sample. From Table 4 it is apparent that the extent of the correlation is the weakest in regard to the skill ratio reinforcing our conclusion as to the inappropriateness of this variable. In turn, the correlations between per capita GNP on the one hand, and per capita GDI and the Harbison-Myers index on the other, indicate the effects of investment in physical and in human capital on incomes per head. The existence of this correlation also explains that the inclusion of all three variables in the regression equation raises the standard error of the coefficients to a considerable extent.-/ Nevertheless, the fact that the level of statistical significance of the physical and human capital endowment variables much exceeds that for incomes per head may be taken as an indication of the "primacy" of the former. We have seen that the intercountry regressions provide the same general results, irrespective of whether we use a stock or a flow measure of capital intensity. This finding may be explained by the relatively high degree of correspondence in the ranking of the product categories by the two 1/ The relevant regression results with t-values in parentheses are 2 as= -3.41 - .13 GNPCAP + .77 GDICAP + 1.22 HMIND - .37 DUMMY R .65 J (8.25) (.14) (.98) (2.45) (1.77) a = -4.24 + .28 GNPCAP + .66 GDICAP + 2.07 HMIND - .45 DUMMY R2 - 77 i (11.20) (.22) (.60) (3.01) (1.53) - 21a - Table 4 Spearman Rank Correlation Coefficients for Country Characteristics in the 36 Country Sample GNPCAP GDICAP HMIND SKILLS GNPCAP 1.000 .984 .754 .674 GDICAP .984 1.000 .730 .697 HMIND .754 .730 1.000 .660 SKILLS .674 .697 .660 1.000 For explaaation of symbols, see Table 1. All coefficients are statistically significant at the 1 percent level. - 22 - measures of capital intensity that is shown by the estimated Spearman rank correlation coefficient of .782.-/ (Table 5) The rankings of the 18 two-digit industry groups, too, are rather similar under the two measures of capital intensity. Among the individual industry groups, Apparel and other textile products, Leather and leather products, and Stone, clay and glass products are relatively labor intensive while Petroleum and coal products, Chemicals,and Paper and paper products are relatively capital intensive (Table 6). At the same time, the results vary to a considerable extent within each industry group. For example, fur goods are very capital intensive although they belong to the highly labor= intensive Apparel and other textile products industry group. In turn, explo- sives are relatively labor intensive although they belong to the capital-inten- sive Chemicals industry group. Moreover, substantial differences are observed among individual product categories in terms of their factor intensity. At one extreme, we find woolen yarn and thread with (total) capital per worker of $3215, followed by earthenware food utensils ($3520), footwear ($3757), leather bags and gloves ($5483),vitreous china food utensils ($7221), costume jewellery ($8589), and games and toys ($8654), which are the most labor intensive among the 184 product categories. At the other end of spectrum, petroleum products ($191739), wood pulp ($135474), organic chemicals ($124198), synthetic rubber ($120631), carbon black ($101161), inorganic chemicals ($92762), and paper ($89089) are the most capital intensive (Appendix Table 1).- 1/ Some major exceptions are various textile fabrics, reclaimed rubber, aluminum castings, ball bearings, and railroad cars where the stock measure, and toilet articles, paints, electric housewares, electric lamps, and motor vehicles where the flow measure, shows a considerably higher degree of capital intensity than the other measure of capital intensity. 2/ The results obtained by the use of the flow measure of capital intensity are broadly comparable, although they differ in regard to particular commodities. - 22a - Table 5 Spearman Rank Correlation Coefficients for Alternative Measures of Capital Intensity (PC+HC)/L VA/L PC/L HC/L (VA-W)/L W/L (PC+HC)/L 1.OOC .782 .758 .907 .680 .835 VA/L .782 1.000 .636 .685 .951 .809 PC/L .758 .636 1.000 .488 .604 .562 HC/L .907 .685 .488 1.000 .565 .839 (VA-W)/L .680 .951 .604 .565 1.000 .631 W/L .835 .809 .562 .839 .631 1.000 For explanation of symbols see Table 1. All coefficients are statistically significant at the 1 percent level. - 22b - TA'DLE 6 &Vf AGF FACTOR IINTESITIES FOD 18 AGGREGATED PRODUCT CATEGOPILS (DOLLARSI S I r, a a ~~~~~~~~~~~~~~~af 0 fu No, o1* Po,mUCT CATEGORY wEIGHT p h, ki Pi hi ki wi 1. ?2. TEXTILE T!LL PRN:DIJCTS 75,38 9404, 17814. 27219, 3885, 5919, 9804, 422Z. 2, 23I ? pADAEL t VT7iER TEXTILE PRnOUCTS 22,98 2024. I1067. 1399l, 3583, 5165 8748, 3968, 3. ?4. LURPLE t 4+'0t POOTDLCTS 31.75 I1266. 12184. 23449. D452, 6431. 10883, 5213, a, 25. FURNIT1'PF t FIAT'JUFS 9,80 4520, 21o78, 2b198, 4431, 6669, 11100, 4505, S, 5 . D A , ,LLILU P9r0DCT5 43,30 57609, '0126, 97735, 1407. 10397, 21864, e382, 6. ?7. P1414TINCO + PULSL1S-1tG 13,54 8417, 36191. 44607, 89b2, 6941. 17903, 5323. 7. 28. CHEMICk4 * ALLIEI0 P400L,CTS 117.73 41417, 33031, 7441s, ig9BS, 9882, 29367, 6580, 8. 2C. P7TrULtUo t CLOAL P)004UC7S 2L.45 126110, St5629, 191739, 31310. 11910. 43220, 5342, 9. 3* PRV8PER . PLASTIC P'IICL'C1S 2a.11 1I188, 18579. 2876t. e400, 7784, 14190, 5922. 10. 31S. LEATb1'R + LEAIH;R 0Pc41;UCTS 7.75 5860, 1728D1 23142, 4?20. 6824, 11244. 5096, 11, 3C. STOrQL,CL 0'+ GLOSS PP4QDJCTS !3.9o 11843, 10003, 218A6t 5571 6742. 12313. 6082, 12. 33. PQT7AQY r TAL 4 *LLIEIŽ PDf0UClO q99,9j 32937, 30130, 630bb, 6774, 10385. 17159, 7373, 13, 34 FA4qICAT00 'PETAL OPOCICITS ID,99 9073, 27860, 3b933, 7330. 8715. 16045, 5927, 14. 'A ONELECTPTsTL 9AC4435r,Y6 14836 00045, 29011, 390St. 7526. 9704, 17030, 6831. 1,, 5D: LLECTQICOL PatIP'E\I * SUPPLIES 94,98 7122. 30836, 3795s, 5808, 8916. 14724, 5834, 1, '07. TANSOCAT±113r4 EiItlT 194,14 11I02. 2707, 138669, 11090, 10624, 21914. 8114, 17. 1A. INSTLJ'rF.1S + PILTED PPODtICTS 234 11i1A7. 41230, 52376, 1350, 9619, 23150. 5495, 18. 3Q. 'IIC. M4AIRIFaCTUPEI) P,OOLICTS 24.71 5667, 17761, 2342S, 6228. q93¶9 15547, 5436. ALL CtTF,,rPILS 1000,00 20518, 2eŽ78. Oe796. qbAS. 9308, 895q3, 5831, NOTE: The table shows average capital-labor ratios in a two-digit industry breakdown. The average capital-labor ratios for the individual product categories have been derived by weighting by the share of exports of the product category con- cerned in the total exports for all the 184 categories aggregated over the 36 countries. For explanation of the definitions and symbols used, see equations tl) to (3) in the text. - 23 - There is less of a correspondence in the rankings of product categories by their physical and their human capital intensity. The Spearman rank correlation coefficient between these indicators is .488 under the stock measure of capital and .631 under the flow measure. In turn, the correlation coefficient is .604 between the two measures of physical capital intensity and .839 between the two measures of human capital intensity. These differences are explained by the faci: that the flow measure of capital intensity is sensitive to inter-indusi:ry differences in profits that do not affect the stock measure whereas both measures of human capital intensity are affected by average wages in the various product categories. Among the individual product categories, organic chemicals, cellulosic man-made fibers, dyeing and tanning extracts, fertilizers, carbon black, and petroleum refining and products have relatively high physical as against human capital intensity, regardless of whether we use the stock or the flow measure of capital. The opposite result has been obtained for canvas products, radio and TV equipment, aircraft, ships and boats, and scientific instruments and control equipment. The first group includes product categories where the ratio of physical to human capital was between 1.5 and 5 using the stock measure of capital intensity and exceeded 1.2 using the flow measure. In turn, product categories in the second group had a ratio of physical to human capital of between .1 and .2 under the stock measure of capital and less than .6 under the flow measure (Appendix Table 1). Sunmmary and Conclusions This paper has investigated the changing pattern of comparative advantage in the process of economic development. Comparative advantage has - 24 - been defined in terms of relative export performance, thus neglecting the composition of imports which is greatly affected by the structure of protection. For each country, export performance has been related to the capital intensity of the individual product categories, using a stock as well as a flow measure of capital, with further distinction made between physical and human capital. Next, the intercountry differences in the regression coefficients thus obtained have been correlated with country characteristics, such as physical and human capital endowment and the level of economic devel- opment. The empirical estimates show that intercountry differences in the structure of exports are in a large part explained by differences in physical and human capital endowments. The results lend support to the 'stages' approach to comparative advantage according to which the structure of exports changes with the accumulation of physical and human capital.-/ This approach is also supported by intertemporal comparisons for Japan, which indicate that Japanese exports have become increasingly physical capital and human capital intensive over time (Heller 1976). These findings have important policy implications for the develop- ing countries. To begin with, they warn against distorting the system of incentives in favor of products in which the country has a comparative disadvantage. The large differences shown among product categories in terms of their capital intensity point to the fact that there is a substantial penalty for such distortions in the form of the misallocation of productive factors. 1/ The expression 'stages' is used here to denote changes over time that occur more-or-less continuously rather than to discrete, stepwise changes. It is thus unrelated to economic stages described by Marx, the exponents of the German historical school, and Rostow. - 25 - Possible magnLtudes of the economic cost of distortions are indicated in Table 7. This provides comparisons between production costs in the United States, assuminig that pre-tax returns and depreciation amount to 30 percent of the gross value of physical capital, and production costs in a hypothetical developing country where unskilled wages are one-third of U.S. wages-/ and the cost of capital is commensurately higher.- In the latter country, the estimated cost of capital-intensive products is 15 to 32 percent-higher, and that of labor-intensive products 38 to 52 percent lower, than in the United States, so that differences in relative costs between capital and labor-intensive products range from 1.87 to 2.76.-/ The economic cost of policy-distortions may be especially high in countries that bias the system of incentives in favor of import substitution in capital-intensive industries and against exports in labor-intensive industries. Costs will also be incurred if capital-intensive exports are artificially promoted. The results can. further be utilized to gauge the direction in which a country's comparEttive advantage is moving. For this purpose, use may be made of the regression estimates obtained as regards total capital intensity. As a first step, we substitute projected future values of a country's physical and human capital endowments in the intercountry regressions, so as to estimate the prospective values of the 0 coefficients. -/ Next, we derive the hypothetical 1/ In 1974, average wages in manufacturing in Korea were 9 percent, and in the Philippines 6 percent, of U.S. wages (ILO, Yearbook of Labor Statistics). 2/ The difference in the cost of capital has been estimated at 43.3 percent under the assumption that value added in the manufacturing sector was the the same in the two cases. It has further been assumed that the absolute difference between skilled and unskilled wages remained the same. 3/ As elsewhere in the paper, the calculations do not allow for factor substi- tution in response of intercountry differences in factor prices. 4/ In line with the stEages approach to comparative advantage, this is done on the assumption that new countries exporting manufactured goods will enter at the lower end of the spectrum. - 25a - Table 7 Hypothetical Production Costs Calculated under Alternative Assumptions (U.S. Dollars) Product Category United States Developing Country Ratio of Physical Human Unskilled Total Physical Human Unskilled Total Total Capital Capital Labor Costs Capital Capital Labor Costs Costs Capital-Intensive 1. Petroleum refining & products 37833 6563 5342 49738 54215 9405 1781 65401 1.315 2. Wood pulp 26400 4747 6382 37529 37831 6802 2127 46760 1.246 3. Organic chemicals 22635 4875 6632 34142 32436 6986 2211 41633 1.219 4. Synthetic rubber 20826 5121 6632 32579 29844 7338 2211 39393 1.209 5. Carbon black 18669 3893 6632 29194 26753 5579 2211 34543 1.183 6. Inorganic chemicals 16044 3928 6632 26604 22991 5629 2211 30831 1.159 7. Paper 14778 3983 6382 25143 21177 5707 2127 29011 1.154 Labor-Intensive 8. Games & toys 1521 359 5436 7316 2180 514 1812 4506 .616 9. Vitreous china food utensils 1608 186 6082 7876 2304 267 2027 4598 .584 10. Costume jewelry 978 533 5436 6947 1401 764 1812 3977 .572 11. Leather bags & purses 711 311 5096 6118 1019 446 1699 3164 .517 12. Earthenware food utensils 1056 0 6082 7138 1513 0 2027 3540 .496 13. Woollen yarn & thread 486 160 4228 4874 696 229 1409 2334 .479 14. Footwear 660 156 5450 6266 946 224 1817 2987 .477 All Categories 6155 2828 5831 14815 8818 4052 1944 14815 1.000 Note: U.S. production costs have been calculated by adding 30 percent of the gross value of physical capital, assumed to reflect pre-tax earnings and depreciation, to observed labor costs. In turn, for the hypothetical developing country it has been assumed that unskilled wages are one-third of U.S. wages and the cost of capital is correspond- ingly higher. The latter has been estimated to exceed U.S. costs by 43.3 percent under the assumption that value added in the entire manufacturing sector is the same in the two cases. All data are expressed per worker. - 26 - structure of exports corresponding to the estimated a cc6efficients, which are taken to reflect the country's future physical and human capital endowments. Comparing the projected export structure with the actual structure of exports, one can then indicate prospective changes in export flows. The regression estimates obtained in regard to physical and human capital intensity can also be used in the manner described above so as to indicate the relative importance of physical and human capital intensive pro- ducts in a country's future export stru'cture. Given the poor statistical results of the regressions for physical capital intensity, however, one should utilize directly t'he data reported in Appendix 1 which show the physical and the human capital intensity of individual product categories. The stages approach to comparative advantage also permits one to dispel certain misapprehensions as regards the foreign demand constraint under which developing countries are said to operate. With countries progressing on the comparative advantage scale, their exports can supplant the exports of countries that graduate to a higher level. Now, to the extent that one developing country replaces another in the imports of particular commodities by the developed countries, the problem of adjustment in the latter group of countries does not arise. Rather,the brunt of adjustment will be borne in industries where the products of newly graduating developing countries compete with the products of the developed countries. A case in point is Japan whose comparative advantage has shifted towards highly capital-intensive exports. In turn, developing countries with a relatively high human capital endowment, such as Korea and Taiwan, can take Japan's place in exporting relatively human capital-intensive products, and countries with a relatively high physical capital endowment, such as Brazil - 27 - and Mexico, can take Japan's place in exporting relatively physical capital- intensive products. Finally, countries at lower levels of development can supplant the middle-level countries in exporting unskilled labor-intensive co=modites. The prospects of economic growth through exports thus appear much brighter once we understand the character of the changing pattern of comparative advantage. Further work on the experience of individual countries over time would be necessary, however, in order to study this process in more depth. - 28 - References Balassa, B., 1965, "Trade Liberalization and Revealed Comparative Advantage", Manchester School, May, Vol. 33. , 1969, "Country Size and Trade Patterns: Comment", American Economic Review, March, Vol. 59. Baldwin, R. E., 1971, "Determinants of the Commodity Structure of U.S. Trade", American Economic Review Papers and Proceedings, March, Vol. 61. Banerji, R., 1975, Exports of Manufactures from India, Kieler Studien, Institut fur Weltwirtschaft an der Universitat Kiel, JCB Mohr (Paul Siebeck), TiUbingen. Branson, W. H., 1973, "Factor Inputs, U.S. Trade, and the Heckscher-Ohlin Model"', Seminar Paper No. 27, Institute for International Economic Studies, University of Stockholm. Branson, W. H., and H. Jumz, 1971, "Trends in U.S. Comparative Advantage", Brookings Papers on Economic Activity, Vol. 2. Cornwall, A. B., 1972, ":nfluence of the Natural Resource Factor on the Comparative Advantage of Less-Developed Countries", Inter-mountain Economic Review, Fall, Vol. 3. Fallon, P.R., and P.R.G. Layard, 1975, "Capital-Skill Complementarity, Income Distribution, and Output Accounting", Journal of PoZiticaZ Economy, April. Vol. 83. Fels, G., 1972, "The Choice of Industry Mix in the Division of Labor between Developed anc Developing Countries", WeZtwirtschaftliches Archiv, Band 108, Heft 1. Goodman, B., and R. Ceyhun, 1976, "U.S. Export Performance in Manufacturing Industries: An. Empirical Investigation", WeZtwirtschaftZiches Archiv, Band 112, Heft 3. Gruber, W. H. and R. Vernon, 1970, "The Technology Factor in a World Trade Matrix" in The Technology Factor in InternationaZ Trade, (ed., R. Vernon), National Bureau of Economic Research, Columbia University Press. Harbison, F. H., J. Maruhnic and J. R. Resnick, 1970, Quantitative AnaZyses of Modernization and DeveZopment, Industrial Relations Section, Dept. Of Economics, Princeton University, Princeton. Harkness, J. and J. F. Kyle, 1975, "Factors Influencing United States Comparative Advantage", Journal of InternationaZ Economics, May, Vol. 5. Heller, P. S., 1976, "Factor Endowment Change and Comparative Advantage", Review of Economic & Statistics, August, Vol. 58. Helleiner, G. K., 1976, "Industry Characteristics and the Competitiveness of Manufactures from Less-Developed Countries", WeZtwirtschaftZiches Archiv, Band 112, Heft 3. Herman, B., 1975, The Optimal InternationaZ Division of Labor, International Labor Office, Geneva. Herman, B. and J. Tinbergen, 1970, "Planning of International Development". Proceedings of the InternationaZ Conference on IndustriaZ Economics, Budapest, April 15-17. - 29 - Hirsch, S., 1974, "Capital or Technology? Confronting the Neo-Factor Proportions and the Neo-Technology Accounts of International Trade", WeLtwirtschaftZiches Archiv, Band 110, Heft 1. , 1975, "The Product Cycle Model of International Trade - A Multi-Country Cross Section Analysis", Oxford BuZZetin of Economics and Statistics November, Vol. 27. Hufbauer, G. C., 1970, "The Impact of National Characteristics and Technology on the Commodity Composition of Trade in Manufactured Goods" in The TechnoZogy Factor in InternationaZ Trade, (ed., R. Vernon), National Bureau of Economic Research, Columbia University Press. Johnson, H. G., 1970, "The State of Theory in Relation to Empirical Analysis" in The TechnoZogy Factor in InternationaZ Trade, op. cit. Kenen, P. B., 1965, "Nature, Capital and Trade", JournaZ of PoZiticaZ Economy, October, Vol. 73. Kojima, K., 1970, "Structure of Comparative Advantage in Industrial Countries: A Verification of the Factor-Proportions Theorem", Hitotsubashi Journal of Economics, June, Vol. 11. Krueger, A. 0., 1974, Foreign Trade Regimes and Economic DeveZopment: Turkey, National Bureau of Economic Research, Columbia University Press. Lary, H. B., 1968, Imports of Manufactures from Less-DeveZoped Countries, National Bureau of Economic Research, Columbia University Press. , 1970, "Comments on The Technology Factor in a World Trade Matrix" in The TechnoZogy Factor in InternationaZ Trade, op. cit. Leamer, E., 1974, "The Commodity Composition of International Trade in Manufactures: An Empirical Analysis" Oxford Economic Papers, November, Vol. 26. Mahfuzur Rahman, A. H. M., 1973, Exports of Manufactures from DeveZoping Countries, Centre for Development Planning, Rotterdam University Press. Morall, J. F., 1972, Human Capital, Technology.& the RoZe of the U.S. in InternationaZ Trade, University of Florida Press, Gainesville. Schydlowsky, D. M., (forthcoming), "Argentina" in DeveZopment Strategies in Semi-IndustriaZ Countries by B. Balassa and Associates. Stern, R. M., 1975, "Testing Trade Theories", in InternationaZ Trade and Finance, (ed., P.B. Kenen), Cambridge University Press. - 30 - APPENDIX TABLE I SECTORAL CHARACTERISTICS FOR 184 PROUUCT CATEGORIES (IN DOLLARS) SECToR PRODUCT CATEGORY Pi hs ki P hi kf P/hT p /h 1. COTTON FA4PICS(GREY) 11770, 15906. 27676, 2871. 5819, 8690. ,740 .493 2. SYNTHETIC FABRICS 11690, 22165, 33855. 3734, 6446, 10180. 527 .579 3. WOOLLEN FaRPICS 9490, 20229. 29719. 3792, 6248. 10040, .469 .607 4. NARROW FAPRICS 6820, 15092. d1912. 3778. 5742. 9520. ,452 .658 5. HOSIERY + KNIT FABRICS 6930. 10406. 17336. 4501, 5269. 9770. .666 .854 6, KNIT OUTEPWEAR 4410. 12826. 17236. 2609, 5511. 8120. ,344 .473 7, KNIT UNDEoWEAR 4330. 6864. 11194, 2083, 4917. 7000. .631 ,424 8. COTTON FARPICS(FINISHED) 13170, 23804 36974, 3349, 6611. 9960. .553 ,507 9, WOVEN CARPETS + RUGS 11550. 23188 34738. 4995, 6545, 11540 498 ,763 10. NONWOVEN CARPETS t RUGS 10400. 24937, 35337, 9569. 6721. 16290. .417 1,424 11. YARN + THpEAD, EXCEPT WOOL 11710, 11396. d3106. 3652, 5368. 9020. 1.028 8680 12. WOOLLEN YaRN + THREAD 1620. 1595. 3215. 799, 1221. 2020, 1,016 8654 13. FELT GOODS 13280. 38588, 51868. 6985. 8085. 15070. ,344 ,864 14, LACE + EMPJIOJDERY 4530, 28545, 33075. 4853, 6787, 11640, .159 ,715 15. TEXTILE PADDINGS 9460, 24145, 33605. 6336. 6644. 12980. ,392 .954 16. COMBED FIpERS + PROCESSED TEXTILE WASTE 9240. 14905. 24145. 4080, 5720, 9800. .620 .713 17. NONRUBBERTZED COAtED FABRICS 12370. 44198, 56568, 7024. 8646, 15670, .280 812 18. COROAGE + TwINE 9100, 193277 28427, 4880, 6160, 11040, .471 ,792 19. TExTILE GoODS NES 11640. 29568. 41208, 7247, 7183. 14430. .394 1.009 20. HENS AND pOYS OUTER APPAREL 1680, 9450, 11130. 3181. 4869. 8050. ,178 ,653 21. NONKNIT UiDERwEAR i920, 7320, 9240, 3042. 4648, 7690. .262 .655 72. TIES,COPSFTS + GLOVES 1990, 15577, 17567. 4324. 5476, 9800, .128 ,790 23. WOMENS ANr CHILORENS CLOTHING 2070. 12453, i4523. 3383, 5167, 8550, .166 .655 24. HATS 4 CAPS 2050. 11217. 13267. 3405. 5045, 8450. .183 675 25, FUR GOODS 5320, 46677. S1997, 11051. 8589. 19640, .114 1.287 26. LEATHER CLOTHING 1860, 7783. 9643. 2739, 5211. 7950, .239 .526 27. CURTAINS + DRAPERIES 3220. 12276, 15496, 3605. 5145. 8750. .262 ,701 28, TEXTILE PaGS * SACKS 5810. 11581, 17391, 3592. 5078, 8670, .502 .707 29, CANVAS PRnDUCTS 2410. 19684. d2094. 3306. 5684, 9190. .122 .562 30. PRESERVEO WOOD 11820. b618. 20438. 5618. 6072, 11690, 1.372 ,925 31. SAWMILL PoODUCTS 10940, 10586, 21526. 4058. 6272. 10330, 1.033 ,647 32. PREFABRICATED WOOD 5990, 22740. e8730. 4596. 7484. 12080, .263 .614 33, VENEER + PLYWOQD 11850, 21362. 33212. 4370. 7350. 11720. .555 ,594 34. WOODEN BOyES + CRATES 5570. 3358. 8928. 4131. 5549. 9680. 1.659 ,744 35. COOPERAGE PRODUCTS 7860. 15868. 23748, 3935. 6805, 10740. .497 578 36. WOOD PRODiuCTS NES .9450. 8952. 18402. 4925, 6105, 11030, 1.056 .807 37. FURNITURE + FIxTURES 4520. 21678. - 6198. 4431, 6669, 11100. .209 ,664 38. WOOD PULP 88000. 47474. 135474. 16028, 11132. 27160. 1,854 1.440 39, PAPER, EXcEPT FOR CONSTRUCTION 49260, 39829. 69089. 8651. 10369, 19020. 1,237 ,834 40. PAPERBOARD 42830, 35581. 78411, 13657, 9943. 23600. 1.204 1,373 - 31- APPENDIX TABLE I (CONTINUEDI SECTnR PRODUCT CATEGORY hs ks P h k i at, STATIONERy 8460, 20604, d90e64, 6941. 7869. 14810. .411 ,882 42, PAPER BAGS + CONTAINERS 11950. 16255, d8205. 5366, 8004. 13370. .735 ,670 43, PAPER P9OnUCTS NES 16340. 22431. 38771. 13t00, 8620. 22220. .728 1.578 449 BUILDING PhPER+ PAPER PRODUCTS 30400. 27341. 57741. 8656. 9114. 17770. 1,112 ,950 a5S NEWSPAPEQS + PERIODICALS 8990. 34989. 43474, 7946, 8199, 16440. .243 ,935 496 BOOKS 7820. 38557. 46377. 12964. 8856. 21820, .203 1.464 47. HISCELLANEOUS PUBLISHING 8780, 36563. 45343, 5313. 8647, 13960, *240 614 48, ENGINEERING + PRINTING 8870. 33713. 42583. 7047. 9513. 16560. 263 .741 49. INORGANTC CHEMICALS 53480. 39282, 92762. 19675. 10556. 30230. 1.361 1,864 50. ORGANIC CHEMICALS 75450. 48748, 124198, 25581, 11509. 37090. 1.548 2,223 51. PLASTIC MATERIALS 4 PRODUCTS 17790, 14789. 32579. 8559, 7911. 16470. 1.203 1,082 52, SYNTHETIC RUBBER 69420. 51211. 120631. 26251, 11759. 38010. 1.356 2,233 53. CELLULtlSIr MANMADE FIBERS 34300, 8002. 42302. 8746, 7434, 16180. 9.287 1.176 54, SYNTHETIC FIRERS 40820. 28863. f9683. 13757. 9523. 23280. 1,414 1,445 55. BIOLOGICAL + MEDICINAL PRODUCTS 17670. 42256. 59926. 28068, 10862. 38930. ,418 2.584 56. SOAP + CLEANSERS 18830. 30622, 49452. 34207. 9693. 43900. .615 3,529 57. TOILET PREPARATIONS 8770. 17967, 26737. 39617, 8433. 48050. ,488 4,698 58, PAINTS 11650. 26604, 38259. 12644. 9296. 21940. .438 1.360 59. DYEING + TANNING EXTRACTS 27530. 9239. 56769, 12141, 7559. 19700, 2,980 1.606 60. FERTILSERS .39710. 10748, 50458. 11543, 7707, 19250. 3.694 1.498 61. MISC., AGPICULTURAL CHEMICALS 30560. 24698, S5258, 29967, 9103. 39070. 1,237 3,292 62, ADHESIVES + GELATIN 18500. 34856, 53356, 19986. 10124, 25110. 531 1,480 63. EXPLOSIVES 6760, 26423. 33183, 1887. 9273. IlO1. .256 ,204 6t. PRINTING tNK 10810, 29033, 39843, 8726., 9534. 18260. .372 .915 65. CARPON BLaCK 62230. 38931. 101161. 30489. 10521. 41010. 1.598 2.898 66. MISC. CHEmICAL PREPARATIUNS 17820. 26071. 43891. 17251. 9239, 26490. ,684 1.867 67. PETROLEUM REFINING + PRODUCTS 126110. 65629. 191739. 31310. 11910. 93220. 1.922 2.629 68. ASBESTOS t ASPHALT PROOuCTS 19810, 43088, 62898. 11119. 9651. 20770. .460 1.152 69, TIRES + TUBES 23050. 45124, 68174. 11792. 10958, 22750. .511 1.076 70, FOOTWEAR 2200. 1557. 3757. 3360. 5611. 8970. 1.413 .599 71, RECLAIMED RUBBER 21620. 22534. 44154. 4059. 8691. 12750. ,959 ,467 72, MISC; RUBqFR PRODUCTS 10280, 20621, 30901. 5413, 8507, 13920. ,499 .636 73, LEATHEP 7470. 23815. 31285, 4840, 7480. 12320, .314 ,647 74, INDUSTRIAL LEATHER BELTING 5130, 25564, 30694, 6324, 7656. 13980, .201 ,826 75. LEATHER UPPERS 3250, 8041, - 11291. 4144. 5896, 10040, ,404 ,703 76, LEATHER BAGS + PURSES 2370, 3113. 5483. 3509, 5401, 8910. ,761 ,650 77. MISC. LEATHER GOODS 2890, 7623, 10513. 2567, 5863, 8430, ,379 .438 78, FLAT GLASS 35560, 49700, 85260, 12076. 11054, 23130, ,715 1,092 79. GLASS CONTAINERS 9980, 7366, 17346, 5710, 6260. 11970, 1,355 .912 80. CEMENT + CONCRETE 41110, 31110, 72220. 11447, 9193, 20640, 1,321 1,245 81. BRICK + STRUCTURAL CLAY TILES 14700. 8708. 23408. 4252, 6948, 11200, 1,688 ,612 82. REFRACTORyES 22060. 30907, 52967, 8609, 9171, 17780, ,714 .939 83, VITREOUS PLUMBING FIXTURES 11810, 24748, 36558, 7498, 8562, 16060, ,477 .876 84, VITREOUS CHINA FOOD UTENSILS 5360. 1861, 7221, 3390, 6260, 9650, 2,880 .541 85, EARTHENiOARE FOOD UTENSILS 3520, 0, 3520, 764, 5606, 6370, R .136 86, PORCELAIN PRODUCTS 9620, 11122. eO742, 3743, 7197. 10940, ,865 .520 87, CONCRETE + BRICK PRODUCTS 11980, 20101, 32081, 6702, 8088, 14790, .596 ,829 88. LIME 46700, 28313, 75013, 8969, 8911, 17880. 1,649 1.006 - 32 - APPENnIx TA4LE I (CONTINUED) SECTOR PRODUCT CATEGORY pi h k p/hs pf f 89. GYPSUM PPnDUCTS u9070, 33163, 62233, 12184, 9396, 21580. 1.480 1,297 90, CUT STONE PRODUCTS 7100. 7783. 14883. 3072, 6858, 9930, .912 .448 91. ABRASIVE DPODUCTS 15580. 29249. 44829. 8467, 9013, 17480. .533 .939 Q2. ASBESTOS pRODUCTS 13320, 23891. 37211. 6199, 8 71 4 6ab70, .558 ,732 93. MINEPAL WonOL 21330. 32757. So087. 11098, 9362, 20a40. .651 1,185 Q4, MISC. NONMETALLIC MINERAL PRODUClS 12330. 22278. 34608. 4827, 8313. 13140. .553 581 95. STEEL + STFEL P'OoUCTS 37850. 31089, 68939, 6729. 10631, 17360. 1.217 ,633 96, IRON FOUNr)RIES 11460. 23994, 35454. 4569, 9511, 14080. ,478 ,480 97, STEEL FOUNIDRIES 11440. 14716, 26156, 4112, 9208, 13320, .777 0447 s8, WROUSHT CnPPER 23400. 30342, 53742. 8544. 10036, 18580. .771 .851 99, WROUGHT ALUMINIUM 32930. 30085. 63015, 5647. 10013, 15660. 1.095 564 100, NONFENROUS METALS NES 19760. 25371. Q5131. 9154, 954b. 18700. .779 ,959 101. ALuMINILM CASTINGS + ST4mPjNGS 13160. 37694, 50854 4a745, 10025, 14770, ,349 .473 102. BRASS,RaRnZE + COPPEP CASTINGS 10260, 21309. 31569. 4502, 913e, 13640. .481 .493 103. IRON + STEEL FORGINGS 13070, 34100. 47170, 4765, 11145, 15910. ,383 ,a28 104. PRIMARY METAL PRODUCTS NES 15840, 18649. 34489, b554, 9476, 16030. ,849 ,e92 105. METAL CONTAINERS 18720. 39734. S8454, 10848, 9902, 20750. 471 1,095 166. CUTLERY 10100, 204'73. 0573. 136941 7976. 21670, ,493 1,717 107. A4NO * EDE TOOLS 8650, 29390. 38040. 7959, 8871. 16830, ,294 ,897 108 HANMOSAVS ; SA'iHLADES 7780, 18343, 26123, b609, 7761, 14370. .424 .852 109. HARODARE NES 10200, 28076. s8276, 6925, 8735. 15660. ,363 ,793 110, SANITARY + PLUMBING FIXTURES 10560. 25764. 36324, 6391. 8509, 14900. .410 ,751 111. NONELECTRIC mEATING EQUIPMENT 7070. 24427, 31497, 7057, 8373, 15430, .289 .843 112, STRUCTURAL METAL PRnDUCTS 6930, 27283. 44213. 5264, 8666, 13920, 254 ,608 113. PLATEWOPR * BOILERS 9010, 34228. 43238. 7293. 9347, 16640, .263 .780 114, BOLTS + NtjTS 11800, 36358, 48158, 6527, 9563, 16090. ,325 ,b83 115. SAFES * VaULTS 6110, 38511, 44621. 13432, s778, 23210. .159 1.374 116. FABRICATEn METAL PRODUCTS NES 10120, 26116, 36236. 6257, 8543, 14800. .388 ,732 17. STEAM ENGTNES * TURBINES 134e0, 45537, 58997, 7183. 11257. 18440. .296 ,638 118. INTERNAL COMPUSTION ENG1NES 15990. 24173. 40163, 8019. 10411, 18430, .661 .770 119. FARM MACHTNERY 10230, 20823, 31053, 6930, 8810, 15740, *491 ,787 120. CONSTPUCT?ON + DRILLING MACHINERr 10600. 35814, 46414, 6483, 10287, 16770, .296 ,630 t21. CONVEYING + CARRYING EQUIPMENT 6610, 30535, s7145, 7443, 9757, 17200. ,216 ,763 122. INDUSTRIAL TRUCKS * TRACTORS 7110, 26440. 33550, 7309, 9351. 1666O, ,269 ,782 123, MACHINE TnOLS 11620, 41454. 53074. 5639, 10851, 16490. .280 .520 124, mETAL * WnODOORKING AACHINERY 9320, 23959, 53279. 6747, 9103, 15850. .389 741 125, FOOD PRODUCTS MACHtNERY \7570. 29204. 36774. 6928, 9b22, 16550, ,259 .720 126, TEXTILE * LAUNDRY MACHINERY 9440, 21364, 30804, 7610, 8550. 16160, .442 ,e90 127, PAPER MAK5NG MACHINE9v 10960, 31110. 42070, 5966, 9814, 15780. .352 ,608 128. SPECIAL INDUSTRY MACMINES NES 8540. 3094j, 39481, 6868, 9802, 16670, 276 t701 129. AIR CCmPpFSSORS 4 pUMPS 8880, 27219, 36099, 7770, 9430, 17200, .326 ,824 130, BALL + ROLLER BEARINGS 15800, 30400, 46200, 4714, 9746, 14460, ,520 ,484 131, INDUSTRIAL FURNACES + OVENS 276o, 8020, 10780, 2566, 4004, 6570, 344 ,641 132, GENERAL INOUSTRIAL MACWINERY NES 7910, 24793, 32703, 6738, 9182, 15920, ,319 .734 133, TYPEWRITERS 10250, 25301. 35551, 12402, 9238, 21640, .405 1,342 134, COMPUTERS 7400, 41578. 48978. 9217, 10863, 20080, .176 ,849 135, CALCULATTNG + ACCOUNTING MACHINES 7560, 20586. d8146, 8275, 8765, 17040, .367 ,944 136, SCALES * RALANCES 4870, 23451, d8321, 6893, 9047, 15940, ,208 ,762 - 33- APPENDIX TABLE I CCONIINwIEo) SECTnR PRODUCT CATEGORY s h Pi f f Ps f f 137, OFFICE MArHINERY NES 9210. 26824. 3603.4 9625. 9385. 19010. .303 1.026 138. AuTOMATIC MERCHANDISING MACHINES 7190. 7716. 14906, 6431, 7479, 13910, .932 860 139. REFPIGERATION MACHIqEqf 8320. 20541, Z88b1, 8307. 8753, 17060. u05 ,949 1uO, NONELECTPyCAL mACHINERY NES 7950. 22605. 30555, 5102. 8968. 14070. .352 569 141. ELECTRIC mEASURING INSTRUMENTS 570, 30040. S571o. 6232, 8838. 15070. .189 ,705 102. TRANSFORMFRS,MOTORS + GENERATORS 8500. 28378. 36878, 5660, 8670, 14330, .300 .653 143, CAR9ON * r.RAPHITE PRODUCTS 24760. 29715. t4475, 10636, 8804. 19440. .833 1.208 1u0, HOUSEHOLO COCOKING E'QUIPMENT 6690, 22527. 29217. 6044, 80H6. 14130, .297 ,748 14$. HOUSEHOLD REFRIGERATORS + FREEZERS 8870. 33611. 42481. 8603, 9197. 17800, .264 ,935 lab, ELECTRICAI HOUSEWARES + FANS 6240, 10702. 16942. 9004. 6906. 15910. .583 1,304 107. SEWING MACHINES 10810. 38991, 49801. 6884, 9736. 16620. 277 ,707 1l8. ELECTRc IAmPS 9830. 13094. 22924, 10608. 7142. 17750. .751 10485 1u9. LIGHTING FIxTURES 6650. 19563, 26213, 7006. 7794, 14800, ,340 ,899 150. PAOIO + Tv EQUIPmENT 5450. 42988. 46438, 4951, 10129. 15080. ,127 ,489 151. PHONOGRAPHIC RECORDS 6250, 9939. 16189, 8232, 6828. 15060, ,629 1.206 1F2, TELEPHONE + TELEGRAPH APPARATUS 9360, 37508. 46868, 6941, 9579. 16520, 250 ,725 153. ELECTRONIC COMPONENTS + ACCESSORIES 7910, 22078. Z9988, 3909. 8041. 11950. ,358 ,486 154, STORAGE BATTERIES 9940, 34723, 44b663 7780. 9310. 17090, .286 ,836 155, PRIMARY BATTERIES 7750, 10431, 22181. 12803. 7277, 20080. 537 1,759 156. XRAY APPAPATU1 + TUhES 5860, 42854. 48714. 7132. 10118, 17250. .137 .705 157, AUTOMOTIVE ELECTRICAL EQUIPMENT 8440, 34409, 42849, 7764, 9276, 17040, ,245 ,837 158, MOTOR VEHTCLES + NOOIES 13210. 25855, S9065. 13616, 11064, 24680, ,511 1.231 159, TRAILERS 3720, 6420, 10140. 4832. 7338. 12170, .579 .659 leO, AIRCRAFT 6820, 41083 48303, 5339. 12121, 17460. .164 440 161, AIRCRAFT FNGINES + EQUIPMENT 8830, 35132, 43962. 4268. 11482, 15750, ,251 ,372 182. SHIPS + B8ATS 5730, 32125, 37855, 1538, 9392, 10930, ,178 .164 163. LOCOMOTIVFS + PARTS 11780. 50318. 62098, 18106, 11204, 29310, 234 1,616 164. RAILROAD rARS 9660, 36258. 45918. 1771, 9799, 11570, .266 ,181 165, MOTORCYCLFS,BICYCLES + PARTS 5940, 15325, 21265, 4381, 7709, 12090, ,388 .568 166, SCIENTIFIr INSTRumENTS + CONTROL EQUIP, 6040, 35157. 41197. 5394. 9016. 14410. ,172 ,598 167. OPTICAL INSTRUMENTS 6600. 434067, 50067, 5965, 9845, 15810. ,152 ,606 168, MEDICAL APPLIANCES + EOUIPMENT 6470, 25155, 31625. 8762, 8008, 16770, ,257 1.094 169, OPTHALMIC GOODS 6020, 15288, 21308. 6438, 7022. 13460, ,394 917 170, PHOTOGRAPHIC EQUIPMENT + SUPPLIES 17720, 57378, 75098, 21716, 11234, 32950, 309 1,933 171, WATCHES + CLOCKS 3840, 17270, 21110, b386, 7224, 13610, ,222 ,884 172, JEWELRY + SILVERWARE 4680. 24708, Z9388, 6119, 7911, 14030, .189 ,773 173, LAPIDARY wORK 6370, 29030, 35400, 8507, 8343. 16850, ,219 1,020 170. MUSICAL INSTRUMENTS + PARTS 5250. 17739. 22989, 4287, 7213. 11500 .,296 .594 175, GAMES * TnYS 5069, 3585, 8654, 3922, 13408, 17330, 1,414 ,293 176. CHILnRENS VEHICLES 6860, 10903, 17763. 4550, 6526, 11080. ,629 ,698 177, MISC. SPORTING GOODS 5180. 8676. 13856. 5205, 6305, 11510. ,597 ,826 178, WRITING INSTRUmENTS + MATERIALS 630, 1S678, 22308. 6327, 7003, 13330. 423 ,904 179, COSTUME JFWELRY 32bO. 5329, 8589w 4988, 5972, 10960, .612 ,835 180. 8UTTONS 5120, 11302, 1b422, 4470, 6570, 11040, 453 .680 181, NEEDLES,PTNS + FASTENERS 6560, 11756, 18316. 6765, 6615, 13380. .558 1,023 182, ROOmS + pRUSHES 5800, 5540, 11340, 5336, 5994. 11330, 1,047 ,890 183. HARD FLOOR COVERINGS 28460, 38547, 67007, 15684, 9296, 24980, .738 1,687 184, MISCELLANFOUS MANUFACTURES NES 4920, 8842, 13762, 4484, 6316, 10800, .556 .710 NOTEt ON THE CORRESPONDENCE OF THE PRODUCT CATEGORIES WITH THE US STANDARD INDUSTRIAL CLASSIFICATION AND THE UN STANDARD INTERNATIONAL TRADE CLASSIFICATION, SEE APPENDIX TABLE 2 FOR AN EXPLANATION OF THE DEFINITIONS AND SYMBOLS USED, SEE EQUATIONS (1) AND (3) IN THE TEXT - 34 - Appendix Table 2 Correspondence of the Sector Classification Scheme with the Standard Industrial Classifidation and the Standard International Trade Classification Sector SIC SITC 001 2211 652.1 002 2221, 2262 653 less 653.2, 653.3, 653.4, 653.9 003 2231 653.2 004 2241 655.5, 655.9 005 2251, 2252, 2256, 2259 841.4 less 841.43, .44 006 2253 841.44 007 2254 841.43 008 2261 652.2 009 2271 657.5 010 2272, 2279 6-57.6, 657.8 011 2281, 2282, 2284 651 less: 651.2,.5, .8, .9 012 2283 651.2 013 2291 655.1 014 2292, 2395, 2396, 2397 654.0 015 2293 655.8 016 2294, 2297 262,6, .7, .8, .9; 263.4, 266.23 017 2295 611.2, 655.4 018 2298 655.6 019 2299 651.5, 651.9, 653.3, 653.4, 653.9 020 2311, 2321, 2327, 2328, 841.11 2329 021 2322, 2341 841. less: 841.11, 841.12 022 2323, 2342, 2381, 2389 841.2 023 2331, 2335, 2337, 2339 841.12 2361, 2363, 2369 024 2351, 2352c 655.7, 841.5 025 2371 842 026 2386, 3151 841.3 027 2391, 2392 656.6, 656.9, 657.7 028 2393 656.1 029 2394 656.2 030 2411, 2491 242 - 35 - Sector SIC SITC 031 2421, 2426, 2429 243, 631.8 032 2431, 2433 632.4 033 2432 631.1, 631.2; 631.4 less 631.42 034 2441, 2442, 2443 632.1 035 2445 632.2 036 2499 244, 631.42, 632.7, 632.8, 633 037 25 821, 895.1 038 2611 251 039 2621 641 less; 641.5, 641.6 040 2631, 2641 641.5 041 2642, 2645, 2649, 2761, 642.2, 642.3 2782 042 2643, 2651, 2652, 2653, 642.1 2654 043 2646, 2647, 2655 642.9 044 2661 641.6 045 2711, 2721 892.2 046 2731, 2732 892.1, 892.3 047 2741, 2751, 2752, 2771 892.4, 892.9 048 2753, 3555 718.2 049 2812, 2813, 2816, 2819 513 less 513.27, 514, 515, 533.1, 561.1 050 2815, 2818 321.8, 512, 521, 531, 532.3, 551.2 051 2821, 3079 581, 893 052 2822 231.2 053 2823 266.3 054 2824 266.21, 266.22 055 2831, 2833, 2834 541 less 541.9 056 2841, 2842, 2843 554 057 2844 553 058 2851 533.3 059 2861 241.2, 532 less: 532.3, 599.6 060 2871, 2872 561 less 561.1 061 2879 599.2 062 2891 599.5 063 2892 571.1, 571.2 064 2893 533.2 065 2895 513.27 - 36 - Sector SIC SITC 066 2899 551.1, 571.3, 599.7 067 2911, 2992, 2999 331 less.: 331.01; 332 068 2951, 2952 661.8 069 3011 629.1 070 3021, 3141, 31.42 851 071 '3031 231.3 072 3069 621, 629 less: 629.1, 841.6 073 3111 611 less 611.2 074 3121 612.1 075 3131 612.3 076 3161, 3171, 31.72 831 077 3199 612.2, 612.9 078 3211 644.3, 664.4, 664.5 079 3221, 3229, 3231 651.8, 664 less: 664.3, 664.4, 664.5; 665 080 3241, 3273 661.2 081 3251, 3253, 3259 662.4 082 3255, 3297 662.3, 663.7 083 3261 812.2 084 3262 666.4 085 3263 666.5 086 3264, 3269 663.9, 666.6, 723.2 087 3271, 3272 663.6 088 3274 661.1 089 3275 273.2 090 3281 661.3 091 3291 663.1, 663.2, 697.9 092 3292, 3293 663.8, 719.94 093 3296 663.5 094 3299 663.4 095 3312, 3313,' 3315, 3316 693.2, 693.3, 694.1, 698.3, 698.6, 719.93, 3317, 3566, :3481, 67 less: 671.3, 678.1, 678.5, 679 3493 096 3321, 3322, 3494, 3497 678.1, 678.5, 679.1, 719.92 097 3323 679.2 098 3351 682 less: 682.1 099 3352 684 less: 684.1 100 3356, 3357 681, 683.2, 685.2, 686.2, 687.2, 688, 689 less: 689.31, 693.1, 723.1 - 37 - Sector SIC SITC 101 3361, 3461 697.2. 102 3362, 3369, 3392 698.9 103 3391 679.3, 698.4 104 3399 671.3 105 3411, 3491, 3496 692.2 106 3421 696 107 3423 695.1; 695.22, 695.23 108 3425 695.21 109 3429 698.1 110 3431, 3432 812.3 111 3433 719.13, 812.1 112 3441, 3442, 3444, 691, 693.4 3446, 3449 113 3443 692.1, 692.3, 711,1, 711.2, 711.7 114 3452 694.2 115 3492 698.2 116 3499 719.66, 729.91 117 3511 711.3, 711.6, 711.8 118 3519, 3714 711.5 119 3522 712, 719.64 120 3531, 3532, 3533, 695.24, 695.25, 695.26, 718.4, 718.51, 719 91, 3544, 3545 719.54 121 3534, 3535, 3536 719.31 122 3537 719.32 123 3541, 3542 715.1. 124 3548, 3553 715.22, 715.23, 729.6, 719.52, 719.53 125 3551 718.3, 719.62 126 3552, 3582, 3633 717.1 less: 717.14, 725.02 127 3554 718.1 128 3559 715.21, 717.1 less all but 717.14, 717.2, 718.5 less: 718.51, 719.19, 719.51, 719.61, 719.8 129 3561, 3564, 3586 719.21, 719.22 130 3562 719.7 131 3567, 3623 719.14, 729.92 132 3569 719.11, 719.23 133 3572 714.1 134 3573 714.3 135 3574 -714.2 - 38 - Sector SIC SITC 136 3576 719.63 137 3579 714.91 138 3581 719.65 139 3585 719.12, 719.15 140 3599 719.99 141 3611 729.5, 729.99 142 3612, 3621 722.1 143 3624 729.96 144 3631 697.1 145 3632, 3639 719.4, 725.01 146 3634, 3635 725 less: 725.01, 725.02 147 3636 717.3 148 3641 729.2, 729.42 149 3642 729.94, 812.4 150 3651, 3662 724 less: 724.91, 729.7, 729.93, 891.1 151 3652 891.2 152 3661 724.91 153 3671, 3672, 3673, 722.2, 729.3, 729.95, 729.98 3674, 3679 154 3691 729.12 155 3692 729.11 -15.6 3693 726 15;i 3694 729.41 158 3711, 3712, 3713 732 less: 732.9 159 3715, 3791, 3799 733.3 160 3721 734 less: 734.92 161 3722, 3723, 3729 711.4, 734.92 162 3731, 3732 735 163 3741 731.1, 731.2, 731.3 164 3742 731 less: 731.1, 731.2, 731.3 165 3751 732.9, 733.1 166 3811, 3821, 3.822 861.8, 861.9 less: 861.92, 861.94 167 3831 861.1, 861.3 168 3841, 3842, 3843 541.9, 733.4, 861.7, 899.6 169 3851 861.2 170 3861 861.4, 861.5, 861.6, 862 171 3871 864 - 39 - Sector SIC SITC 172 3911, 3912, 3914 897.1 173 3913 667 174 3931 891 less: 891.1, 891.2 175 3941, 3942 894.2 176 3943 894.1 177 3949 894.4 178 3951, 3952, 3953, 895 less 895.1 3955 179 3961 897.2 180 3963 899.5 181 3964 698.5 182 3991 899.2 183 3996 657.4 184 3999 613, 861.92, 861.94, 894.5, 899 less: 899.2, 899.5, 899.6 PUB HG3881.5 .W57 W67 no.256 Balassa, Bela A. A "stages" approach to comparative advantage / PUB HG3881.5 .W57 W67 no.256 Balassa, Bela A. A "stages" approach to comparative advantage /