.Lan...r 1 . .. t. . .. . . lauiiuww . . .. . ..... . . . . . . . a a 4 4 . v- o . , .. . .. . C n firmwm .7 .1 . . . . - .. 3.76% ...~ 54.. u . ... .., . .VMN-umhl ._ I, n H t 4 u. 4 o . . u. . U . . . 4 503?.” u. u - Eds-1t...” lww‘w. .5 I O g ' (mg . . . . km.........”.._....nn..... JAY . .97.. . . .. . . '_ t. o. 4 ...$. .unmwlwofi.‘ - ; . . .. i. p . ..u. a . .. .4. . ... . .x . u .4. .m. ......,. an. ...... 9 o .. . 1 .. . . . .42. Idlt‘r‘fi’} .- 3413 f . . . ‘ “1.“! x. i'v . . 3.8.» 5...}... . g.U.t..U.s.flu...rsz - It'll . 1 , 4.. v?.l.msll..wuw Dian} .. .szm. .;.r. .9- . t V O. ‘1. C a r - ., . . .. . . 1mm»- "lawma. n... _ .- w. n. H... a...» .. , 4)! k”) L 1.01.... 1.. v - “3.1.34.4451'1.” ML . 1 .4 1 a A I . 51440.! c . .. 1 $23.. r e . {.4 . - . . . .65.. ...ru~..uk..r.l....i.l€. . , ,..Uuuwmnwdvyb,hdn.lx¥.uu_ . .53 $ u<.M.r.vv n. nmiu 3f £5»).an RE... . ..rn.b..r......rmow....:. .. v. 31$. .41..» It'sll..\.(t 43' 4 .4. | d . 5d" 1.!- lie. 7.3.! v. n 4 u 1. . if. ., . t. . filinggprl: ...:;.nk| In J . 4. n . a», “3%. 5%. . . 53:... 111. i}.-. .. .o .- .... dun-In”! ”0“.“ij (40.32.4541 I. . ”a! . IQ\¢¢4~!”~01 nvv Shun-Or (3%“: 15004.3. . 1 .«r‘l a. . ‘O‘Iuht‘ .1 I11.» . . . . I! ‘?.v1 ( v t . A . . . Hum. .,.-......... .. 1.7.3.... . as» . . , . ...II I. .P. . . it”sovuvltltn..o htlll A..- :4 O! o “OTAIWIA 4.1» - ll‘ 4 y n x .v I\ icill ‘hN'I-IIO' r} L U. 0 1"‘ .Nn T . .|I\l|!lv 3 .t . .. p- :1 .3 : .. . :3 -. Yum: .. i} x. -2 .h .. (fl... vamimuno - 5mm..~fimd.u...fiadnflwmrhnmnu$u: .me UP..- Jeruwxuh w...... . www.mr..n.w.u...znl.v«.1 unnmavfnukvml .(fllwlqwflu .. 1 .n . a. I .14. - l: I n I . 4.9904 if." v I tlu ! 0 AD! I . tr! 4t NICO-I. or. 1.1). .. 1 Gr 1 a .. It lllll. .! r I . 4.. .. v M»... .i. .Ll. . 144.93“? w. . ..J...M0u.7cb..:..n...1..urn... I 7... an... tunl-..’ 2:1 v H. a. I 0'44... . o . .c .. .fik‘ ,. .71. 141.721. T only . a .a u. 8 s: n .t - « n- I ll. 1?. . .ulflaoiu fluka v 51.”: 43L: 44:4 avnvlu .. . 4 III .4: .II a r o . . .71» .u .100. . . v! '0.-. 3:... 91. ... ll! III 3 o .0 X c. £3534... 53.14; .«lihkhrttu. . 1.! 11"! Iv - ’11: 9‘43] 1. IIAOOI 7!.113}. .95 u .2" . . r. a m .-o.U.nI..xv.....n1.fll 4 4‘54!de Huh-Fl ..o~.n.ul 3qu .4” . .lln «'toq-OHVOOFv14l~¢’\~4~¢V‘.It .1 bl-.'l\.v..l‘l I: ”‘1': 1. . If ell l‘oatg zv:'.1l|a.l|r.lv .Lfl. 91.1.1090‘. . I I‘M '0 - o .‘I in! Italy! ‘suikn .' .04 Hr . in r « . . Il.|\’ ‘g .\C ‘ofiiiwl . 94.! 444444 n. .. . In II.I It‘fluw nu . v 1 t! ‘ . . r . II II I .. lulllw'" .IIWL..!-JL»O .v. .Yi l‘c-VT I. .lll’l CY‘OU'I .0 44.7 ... aaA.lOl_.AIPO. t ~ .4 All,‘ -Dl}..-v1|‘xv 4..- "U. hz'filx ' .. IQ u .1 I i. I 1’ I ' '4'... . t ‘- '0‘ '0‘ [V I“ .i .[Il‘fllto‘ QDII 4 .turi v. n»!1 I'lf \' 4‘ ‘ IHII in I (”13.01. , . -34...» t, 8 \..!v 1...! n l (I- . .7!“ ‘3': : iii”! 0 t . . . .y . . . . , a. .0 . n? .U 1..---. r I | .lulllu 'll'l ’ 4 .l-£t I I it ‘ n 4.: ..4 - J 4 V lib‘ \ I 0 ‘i .luiutk .mdenKJMv -u minimal ,Hulv : rhUUtht1WINLH !u..y!u.1.w..i.l-..lh 4...... ”m 1.! V5 :1 i . t 1! 0 rn inialnlcldwuvzv.‘lil.‘llfli(x,} .l u ..I! v - ”In?” 1 -lls..nl. L . ., . .q I 3.-. . I: .l [0-14143I’I‘Iivtt... 1. . . . . . . 00.!a- _ . 0.? ’ror‘lr.lvl-.n 11‘. . , . , .. . . ~ . IIPOI.‘: v.0 ll,<flu¢..|..1n ”'0'“. . v 4. v 3 14' II _ nu" '4 c4- lkva..t"Lv||oQ.|Xl‘ n0 loului l 1 i 3935. W I: I 1|'\14 .. fun .ldt .I.l . . np'ru .. 1. 7.1.2.! it... -12.... . . - . ‘1 vl‘lflnfiflak... ... || 1).“: “Ll‘b‘ I. t 4332!“? r vl I_I .II.. -I) t1!!! .1! ._.0.. '0. .. V I It |rI $. in: in! .I. !«vl.o lit- . ‘0 J IV! -II ”UVIrIIIAJ .. .. {M11 - V w! | I, . 13-15134; This is to certify that the thesis entitled AN EMPIRICAL EXAMINATION OF SEVERAL THEORIES OF THE COMMODITY COMPOSITION OF TRADE presented by Robert A. Brusca has been accepted towards fulfillment of the requirements for Ph.D. Economics degree in /"7m’Z'/€£/éu Kym/ye Major professor Date November 16, 1977 AN EMPIRICAL EXAMINATION OF SEVERAL THEORIES OF THE COMMODITY COMPOSITION OF TRADE By Robert A. Brusca A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Economics 1977 ABSTRACT AN EMPIRICAL EXAMINATION OF SEVERAL THEORIES OF THE COMMODITY COMPOSITION OF TRADE By Robert A. Brusca Since Leontief produced his famous paradox, international trade theorists have reexamined the assumptions underlying the Heckscher-Ohlin (H-O) model. Their efforts have produced new theories and caveats. Currently, several models provide explanations for the commodity composition of trade. This study empirically evaluates the H-0 and several post-H-O models: human skills, scale economies, the technological gap, and preference similarity. The total factor requirements, calculated from the U. S. l970 input-output (I—O) table, are used to examine the l975 trade patterns of nineteen developed and developing countries. U. S. coefficients are applied to other countries' trade flows. The study focuses on trade in manufactures, but shows the effect of adding natural resource intensive products to the trade flows. The human skills and scale economy theories are evaluated using 1-0 and multiple regression analysis. In both cases the I-0 classifications (l2l sectors) constrain the level of industrial detail. The input-output tests find support for each theory, while Robert A. Brusca regression analysis rejects the scale theory based on economies internal to the plant. Although the regression results generally support the conventional human skills theory, a three factor approach including unskilled labor, skilled labor, and capital better explains the trade patterns of the developed countries not at the extremes of the endowment rankings. The three factor model is buttressed by the relative factor intensities revealed by a three factor I-O model. The technological gap model is not itself tested, but the human skills tests provide some insights by identifying workers who supply services required in high technology industries. Multiple regression analysis reveals the United States as the only country in the world which derives an advantage from workers providing highly technical services. The "revealed comparative advantage" (RCA) approach developed by Balassa reinforces this finding by iden- tifying commodities in which the most highly developed countries have an advantage. The RCA rankings are related to the skill inten- sities of the products. Current trends toward protectionism are better understood by assessing changes in the RCA pattern over time and due to different standards of comparison in the same time period. These methods also produce casual empirical support for Linder's trade model. In a separate test, Linder's preference similarity hypothesis is supported by the trade patterns of the most developed countries in a sample of twenty-six. The test involves a dependent variable with a truncated distribution and uses a quadratic specification. The difficulty in testing Linder's hypothesis is discussed thoroughly. Robert A. Brusca The test reveals some favorable statistical evidence, while other supportive but insignificant evidence gains credibility due to the unconstrained nature of the quadratic form. The study also addresses several general issues. The total and immediate factor requirements are shown to be highly similar across industries. When each is used to explain the export per- formance of nineteen countries' trade, similar regression results are produced. Thus the immediate coefficients (which correspond to the value added produced in an industry) may be used to test trade theories without loss of correspondence to the correct total factor requirements. If this finding is not sensitive to aggregation, it implies that greater industrial detail and, therefore, more refined tests are possible. To my wife, Candice ii ACKNOWLEDGMENTS I wish to express my gratitude to Professor Mordechai Kreinin, Chairman of my dissertation committee, for his encourage- ment and helpful suggestions. I would like to thank Professor Lawrence H. Officer for his useful comments and Professor Anthony Y. C. Koo for providing the input-output table used in this study. I am indebted to all those at the Bureau of Business and Economic Research at Michigan State University, especially Acting Director Professor David I. Verway. I would like to thank Robert Nevius for his very capable computer programming and Nancy Heath for her diligent and accurate typing of this dissertation. Finally, I wish to express my indebtedness to my wife, Candice. Without her encouragement and support this task could not have been completed. TABLE OF CONTENTS Page LIST OF TABLES . . . . . . . . . . . . . . . . vii LIST OF FIGURES . . . . . . . . . . . . . . . ix Chapter I. INTRODUCTION . . . . . . . . . . . . . . 1 REFERENCES . . . . . . . . . . . . . . 6 II. THE HUMAN SKILLS THEORY OF INTERNATIONAL TRADE . . . 7 2.1 Introduction . . A . . . . . . 7 2. 2 The Theory and Its Assumptions . . . . . 9 2. 3 Previous Empirical Tests of the Theory . . . l4 2. 4 Conclusions . . . . . . . . . . . . 19 REFERENCES . . . . . . . . . . . . . . 21 III. THE HUMAN SKILLS THEORY OF INTERNATIONAL TRADE. EMPIRICAL EVIDENCE . . . . . . 24 3.l Introduction . . . . . . . . . . . 24 3.2 Methodology . . . . . . . 25 3.3 Application of the Skill Indexes . . . . 34 3.4 Empirical Evidence for the United States . . 37 3.5 Revealed Factor Intensity: The Export and Import Patterns of Nineteen Countries . . . 42 3.6 Skill Endowments and Revealed Skill Intensity . 48 3.7 A Three Factor Revealed Approach to the Assessment of Trade Patterns: Evidence for Ten Countries . . . . . . . . . 5l 3.8 Conclusion . . . . . . . . . . . . 56 REFERENCES . . . . . . . . . . . . . . 58 IV. THE SCALE ECONOMY THEORY . . . . . . . . . . 60 4.1 The Theory . . . . . . . . . . . . 60 4.2 Internal Economies: Previous Empirical Tests . . . . . . . . . . . . . 62 4.3 Conclusion . . . . . . . . . . . . 66 REFERENCES . . . . . . . . . . . . . . 67 iv Chapter V. VI. VII. VIII. IX. AN EMPIRICAL TEST OF THE SCALE ECONOMY THEORY . Introduction Methodology. . Empirical Test of the Scale Economy Theory. Nineteen Countries . . Empirical Examination of the Scale Economy Theory: The United States Conclusion . . . . REFERENCES. 01 01 010101 0 o o o 0" b CAIN-J THE LINDER MODEL 6.1 The Theory 6.2 Previous Empirical Tests 6. 3 Some Empirical Considerations . REFERENCES . AN EMPIRICAL EXAMINATION 0F LINDER' S PREFERENCE SIMILARITY HYPOTHESIS. . 7.1 Introduction 7.2 Methodology . . 7.3 Empirical Evidence 7.4 Conclusions . REFERENCES . REVEALED COMPARATIVE ADVANTAGE. A POLICY . PERSPECTIVE . . . . . . . 8.1 Introduction 8.2 Methodology. . 8.3 Revealed Comparative Advantage for Five Countries and the EEC- 6 . . 8.4 Comparative Advantage: Changing Patterns . 8.5 Conclusion . . REFERENCES. A MULTIPLE REGRESSION ANALYSIS OF THE HUMAN SKILLS AND HECKSCHER-OHLIN THEORIES: SOME IMPLICATIONS . 9.1 Introduction 9.2 Methodology. . . 9.3 The Human Skills and Heckscher-Ohlin Theories of International Trade: An Empirical Analysis . . 9. 4 Total and Immediate Factor Requirements 9.5 Conclusion REFERENCES . Page 69 69 69 72 75 79 80 95 V 95 106 119 120 121 121 122 126 135 141 142 143 143 144 150 160 163 164 Chapter Page X. SUMMARY AND CONCLUSIONS . . . . . . . . . . 165 APPENDICES . . . . . . . . . . . . . . . . . 174 A. 1970 Input-Output Sectors . . . . . . . . . 175 B. Regressions: Nineteen Countries . . . . . . . 183 vi Table 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 5.1 5.2 7.1 7.2 LIST OF TABLES U. S. Labor Requirements by Skill Classes, Per Million Dollars of Exports and Competitive Import Replacements . . . . . . . . . Distribution of U. S. Labor Requirements by Skill Classes per Million Dollars of Exports and Compe- titive Import Replacements . . . . Skill Ratio of U. S. Imports, Exports, and Imports/ Exports for Various Skilled/Unskilled Ratios and for Selected Groupings of Industries . . Skill Indexes for Exports and Imports of Nineteen Countries' Trade in Nineteen Countries' Trade in Manufactures . Correlations Between Skill Indexes for Exports and Imports for Nineteen Countries . . Skill Index Ratios of Tim/Yix Derived from Nineteen Countries' Trade in Manufactures . . . Skill Endowment Rankings and Skill Intensity Rankings (Yix) of the Exports Seventeen Countries 1975. . . Revealed Relative Factor Intensity with Three Factors: Capital, Skilled Labor, and Unskilled Labor--Manufactures Trade Scale Economy Relationships to National Character- istics . . . . . . . . . Multiple Regression Test of the Scale Economy Theory . . . . . . . . . . CosX1M Values Between National Export and Import Vectorg, Three Digit SITC Basis . . . Regressions for Constant Export Vectors CosX M. —b + b G. + b2 Gj 1975 World Manufacturing Trade j .0. I j vii Page 38 4O 41 43 44 47 49 54 73 78 107 110 Table 7.3 7.4 7.5 8.1 8.2 9.1 9.2 9.3 9.4 10.1 A.1 B.1 Expected and Estimated Values of Per Capita GDP for the Maximum of the Linder Similarity Function, COSXiMj = bo + b161 + b265-.1975 World Manufactur- es Trade for 20 CountriesJ . . . . . . . . Regressions for Constant Import Vectors CosX1M1 = a0 + a1e1 + a262.1975 World Manufacturing Trade . . . . . . . . . . . Expected and Estimated Values of Per Capita GDP for the Maximum of the Linder Similarity Function, CosX1M = a0 + a1G1 + azG1. 1975 World Manufactures Trade for 21 Countries . . . . . . . . . . Revealed Comparative Advantage . Revealed Comparative Advantage: Changing Patterns Average Relative Importance of Skill Classes Across Sectors; U. S. Labor Coefficients . The Sensitivity of the Theoretical Tests of the Human Skills and Heckscher-Ohlin Theories to the Use of Immediate Versus Total Factor Requirements and to the Selection of a Dependent Variable: An Inter- national Comparison for 1975 Manufactures Tradeé-A Summary of Results . . . . . . . . . Measured Relative Factor Abundance 1975 . Correlations Between Total and Immediate Requirements Skill Indexes . . . . . . . Summary Table of Test Results: 1975 Manufactures Trade . . . . . . . . . . . . . Input Output Sectors of the 1970 U. S. Input-Output Table . . . . . . . Multiple Regression Analysis of Nineteen Countries' Trade Patterns: 1975 Manufactures Trade with the World, Using Total and Immediate Factor Requirements Measured from U. S. Coefficients . . . viii Page 112 116 118 127 136 150 152 156 152 T70 176 184 Figure 4.1 6.1 6.2 7.1 LIST OF FIGURES Hypothetical Distribution of Firms in Two Countries . The Tent Shaped Similarity Function for Fixed Country i and Different Countries j The Tent Shaped Similarity Function fbr Fixed Country j and Different Countries i . The Similarity Function ix Page 61 88 89 102 CHAPTER I INTRODUCTION The fundamental purpose of the pure theory of international trade is to explain the commodity composition of trade. Currently, several hypotheses offer complementary or competing views. In this study we test these theories in i561ation, in direct opposition, and simultaneously where appropriate. The Heckscher-Ohlin, human skills, and scale economy theories comprise the group of supply models which are tested. Linder's demand driven model is also tested. Although the technological gap theory is not tested, examination of the human skills model provides some insights by focusing on technically oriented laborers. We address these issues by first considering the supply models. In Chapters 11 through V the models are presented, the literature is reviewed, and new evidence is offered. The following two chapters concern the opposing predictions of Linder and the supply models. In Chapter VIII we take a policy approach, centering on issues ignored in the previous chapters. Chapter IX provides a simultaneous test of the relevant theories. The Heckscher-Ohlin theory centers on two factors of pro- duction: labor and capital. Countries which are relatively abundant in capital are predicted to have a comparative advantage in 1 commodities whose production requires a relatively intensive use of capital. When labor is relatively abundant, countries derive an advantage in commodities which are relatively labor intensive. A related three-factor model is tested empirically in Chapter III; the two-factor H-O model is examined in Chapter IX. The human skills theory of international trade is set in the same "factor proportions" framework as the H-0 theory. There are three factors of production: skilled labor, unskilled labor, and capital. But, the relative availability of skilled and unskilled labor is the sole determinant of trade flows. Capital, it is argued, does not influence trade patterns due to its relative international mobility. Multinational corporations play a primary role, since a corporate empire transcends international boundaries allowing capital to move freely. A large and growing proportion of international trade involves transactions within the multinational corporation (7). Thus the human skills theory concentrates on the relative availability of labor of differing qualities. ‘Countries which are relatively abundant in skilled labor derive an advantage in products which use skilled labor relatively intensively. Chapter III tests this theory in isolation, using input-output analysis. The chapter also explores the possibility that the relative availability of all three factors, skilled labor, unskilled labor, and capital, determines trade patterns. The scale economy theory is tested in Chapter V. It asserts that comparative advantage is determined by relative plant size when economies of scale are internal to the plant. The hypothesis is examined under the assumption that (1) scale economy benefits are passed on from sector to sector to the export market, and (2) only final stage scale economy benefits confer an advantage to exporters. The Heckscher-Ohlin (H-0) and post H-O supply models identify the pertinent production characteristics of commodities. By relating these to countries' national endowments, the theories single out commodities in which nations will have a comparative advantage. Thus, supply theories of international trade assert that greater differences in national endowments create greater opportunities for gains from trade. In this sense, the preference similarity model of Burenstam Linder is different. Focusing on international variations in con- sumer preferences, Linder sets out to show that the greatest oppor- tunities to reap gains from trade lie between nations with highly similar demand structures. This prediction is in direct conflict with the supply driven models. The opposing viewpoints are tested in Chapter VII. In Chapter VIII a different approach is pursued. Among a selected sample of commodities, the products are identified in which the eleven most highly industrialized countries in the world have their greatest advantage. The comparative advantage ranking of commodities is given for five countries plus the EEC (original six) as a unit. Labor force characteristics and technological aspects of high and low ranking products are related to each country's "revealed comparative advantage." Changes in the conmodity rankings over time and under differing assumptions are used to help explain the current trend toward protectionism and the increased reliance on orderly marketing agreements. Additional insights into Linder's preference similarity hypothesis are obtained. The various themes explored in the earlier Chapters are brought together in Chapter IX, where the relevant theories are tested simultaneously using multiple regression techniques. The trade patterns of nineteen developed and developing countries are examined. As most tests of international trade theories are con- ducted using immediate factor requirements (those factors employed in the final stage of fabrication), we conduct an investigation as to the similarity between these and the theoretically correct total factor requirements (which include the factor requirements of the inputs and inputs into the inputs, and so forth). The purpose of this empirical analysis is to test certain logical implications of competing and complementary theories; there- fore, we must have confidence that the empirical results are objec- tive and meaningful. In several previous empirical studies (1,3,4,10) the critical test results1 have been found to be sensitive to the choice of a dependent variable. If two dependent variables are reasonably good measures of comparative advantage, but produce widely different critical test results, it is very difficult to judge exactly what we learn from the divergent findings. To see whether this problem plagues this study, the regression analysis in 1The signs and significance 0f the independent variables in regression equations. Chapter IX is conducted using two different, but reasonable, depend- ent variables to measure comparative advantage. The following chapters are intended to illuminate a variety of issues in international trade theory. The analysis, basically, is cross-sectional, although the revealed comparative advantage approach uses comparative statistics. The year 1975 was chosen for two reasons: first, it is the most recent year for which an inter- national data set is available; second, by 1975, fluctuating exchange rates had been in existence long enough to have settled at their equilibrium level, thereby allowing commodity trade flows to adjust. The supporting data on commodity characteristics also are very recent. The human skills occupational classes are from the 1970 Census of Population, the scale elasticity parameters are estimated from data in the 1972 Census of Manufactures, and the input-output table is of 1970 vintage. Therefore, this study embodies the most current infor- mation generally available. This is important because most previous studies were conducted in the early to mid-19605. Since then, tariff barriers have declined, exchange rates have begun to float, Japan's development has increased enormously, and multinational corporations have exerted considerable influence on international trade flows. New tests with more recent data are warranted. 10. REFERENCES Baldwin, R. E. "Determinants of the Commodity Structure of U. S. Trade." The American Economic Review (March 1971): 126- 46. Bhagwati, L. "The Pure Theory of International Trade: A Survey." Surveys of Economic Theopy, Growth and Development 2 (New York: 1965): 173-5. Branson, William H., and Junz, Helen B. "Trends in U. S. Trade and Comparative Advantage." Brookings Papers on Economic Activity 2 (1971): 285-346. . "U. 5. Comparative Advantage: Some Further Results." Brooking Papers on Economic Activity_3 (1971): 754-9. Hufbauer, G. C. "Factor Endowments, National Size, and Changing Technology: Their Impact on the Commodity Composition of Trade in Manufactured Goods." In The Technologpractor in International Trade. Edited by R. Vernon. New York: Columbia University Press, 1970. Katrak, H. "Human Skills R and D and Scale Economies in the Exports of the United Kingdom and the United States." Oxford Economic Papers (November 1973): 337-60. Kreinin, M. E. International Economics: A Policy Approach. New York: Harcourt Brace Jovanovich, Inc., 1975. Leamer, E. E. "The Commodity Composition of International Trade in Manufacturers: An Empirical Analysis." Oxford Economic Papers (1972): 350-74. Leontief, W. "Domestic Production and Foreign Trade, The American Capital Position Re-examined." Economia Internazionale (February 1954): 9-45. Weiser, L., and Jay, Keith. "Determinants of the Commodity Structure of U. S. Trade: Comment." American Economic Review (June 1972): 459-66. CHAPTER II THE HUMAN SKILLS THEORY OF INTERNATIONAL TRADE 2.1 Introduction The human skills theory is based on the proposition that the relative availability of skilled and unskilled labor is the funda— mental determinant of international trade patterns. Although capital is a factor of production, it is relatively more mobile internation- ally than labor, and therefore, less likely to determine trade patterns. Skilled workers are relatively more difficult for a nation to acquire than capital. Labor migration does not occur in suffi- cient magnitude to alter initial endowments importantly. An existing stock of unskilled labor may be transformed into skilled labor, but this requires an intensive national training program. Physical capital, on the other hand, can be purchased directly at prevailing market prices; the profit motive will attract financial capital. These considerations are relevant, especially to a world economy whose nation's have engaged in trade for a considerable period of time. The dynamic environment has provided each country with the opportunity to supplement an initial endowment of a relatively mobile factor of production, such as capital. But, since labor is immobile, if the skill intensity rankings of commodities across countries are similar, relative skill endowments will determine trade flows. The preceding discussion conveys the essence of the skills approach. In this chapter the theory is formalized. Empirical evi- dence establishing the validity of the necessary assumptions is presented. Then, the contrasting assumptions of the human skills and human capital theories are set out, and it is shown that the former approach is more consistent with existing empirical evidence. The chapter concludes with a survey of the human skills literature. Most tests of the skills theory are conducted using immediate or direct skill coefficients. Since it is the main point of this chapter that the theory is not properly tested by these coefficients, let us establish the terminology which will be employed in this and in all subsequent chapters. Input-output tables distinguish between three types of factor input requirements: immediate, direct, and total. Immediate characteristics are qualities of the product itself; direct characteristics are qualities of the product plus its first stage inputs; total characteristics include the direct char- acteristics plus those embodied in the inputs of all other stages of production (the inputs into the inputs etc.). The importance of these distinctions is established in the following section.1 1Those who are familiar with Keesing's work (10,11,12,13) are aware that he claims to use direct coefficients. Immediate coeffi- cients seem to be more consistent with his arguments (10, p. 288). I will interpret Keesing's statements to mean that the direct factor requirements were used. Although this may be incorrect, it will serve to make a point. If this is a misinterpretation, it is not important, since both direct and immediate coefficients are inap- propriate. * 2.2 The Theory and Its Assumptions According to the human skills approach, production functions are linear and homogeneous in the first degree. Labor services from workers of specified skill classes comprise the factor inputs; all factors are perfectly divisible. Define the amount of labor services of each type, t, necessary to produce one dollar's worth of output in industry j as Stj' Let the structural relationship of industries in the economy be defined by the matrix A, where aij e A (i = l, 2, . . . m; j = 1, 2, . . . n). Each aij represents the amount of industry i's output required to produce one dollar's worth of output in industry j. The total factor requirements of industry j are obtained by combining the skill vectors st for each industry j with 2 the Leontief inverse of the matrix, A. Thus, the total requirements for each type of labor t in industry j are: 2Keesing has pr0posed that the direct factor requirements (obtained by replacing r.. in equation 2-1 with a1-) can be used, as in an open economy inputgjcan be obtained through rade (10, p. 288). However, the theory which takes for granted that which it purports to explain is not logically sound. Furthermore, direct factor requirements include only the factor requirements which are specific to the final stage of fabrication and the first stage material inputs. Ignoring the inputs into the inputs, etc., implies that the total factor content of a product is not adequately measured regardless of the location of the supplier of that input. In any event, use of direct coefficients in no way implies that the inputs actually were produced domestically. Therefore, the use of direct coefficients is not defensible. By comparison, the total factor requirements measure the factor service content of all inputs and inputs into the inputs etc. Thus they measure the total factor content of a given product. 10 r.. (2-1) . n; i = 1, 2, . . . m) II —J u N u where: (t rij e [I-A]-], the Leontief inverse matrix. The total labor content of a nation's eXports may be obtained by computing,3 111 m S = 2 Z S . .. . - ( 1),, MIM 11:11] xJ (2 2) for t = 1, 2, . . . n; j = l, 2, . . . m, where Xj is the value of a nation's exports of commodity j. U. S. technical coefficients (rij's) and skill vectors (Sij's) are used throughout. The theory explains trade in manufactures, ignoring products which are highly dependent on natural resources. The immobility of natural resource inputs prevents the product which intensively uses them from being produced where factor prices would be most advantageous, except by chance. The application of U. S. skill coefficients to foreign coun- tries is a procedure which must be justified. Two general questions are pertinent: (1) What are the theoretical implications? (2) Is the procedure empirically valid? Taking up the theoretical issue first, it is possible, although unlikely, that every country produces each commodity with 3The factor content of imports is obtained by replacing X. with M. in equation 2-2, where M. is the value of a nation's import; of commodity j. J 11 exactly the same skill mix as the United States. In this case, the U. S. coefficients measure the skill content of exports and imports perfectly, causing no distortions. However, if capital and unskilled labor can be substituted for one another, but not for skilled labor, U. S. coefficients mismeasure the unskilled labor content of trade.4 The amount of skilled labor embodied in a given trade flow is accu— rately measured, but the amount of unskilled labor may be mismeas- ured. (In fact, it will be too low if the United States substitutes its relatively abundant capital for its relatively scarce unskilled labor.) If high and low skilled labor are easily substituted for one another, "skill intensity reversals" can occur. These are a response to divergent factor price ratios which cause a relatively skill-intensive product in one country to be unskilled intensive in another. These reversals disturb the skill-intensity orderings of commodities between countries. When this happens, it is no longer possible to assert that a relatively skilled labor abundant country will export skill-intensive commodities. Since those same commod- ities may be produced by a relatively labor-intensive process, a country which is relatively abundant in unskilled labor may enjoy a conflicting advantage. Thus, the admission of substitution possi- bilities destroys the theoretically deterministic nature of the theory. 4Berndt and Christensen recently have estimated an aggregate production function for the U. S. and found that capital and skilled labor are complements while capital and unskilled labor are substi- tutes. Skilled and unskilled labor also are substitutes, but capital and unskilled labor are more easily substituted for each other (3). 12 Nonetheless, the model is useful so long as substitution does not alter the essential relationships across industries. If, for example, substitution between capital and unskilled labor occurs to the same extent across all industries, their relative skill-intensity rankings are not affected, and the model accurately measures the relative skill intensity of a given trade flow. Whether or not reversals constitute a serious problem is an empirical question. The existing evidence to be discussed in the next paragraph, supports the nonreversability assumption. Rank correlations between the 1958 average wages paid in thirteen industry groups across twenty-three nations produced 182 out of 253 positive and significant (1% level) Spearman's correlation coefficients (8, p. 174). If wage rates are a good proxy for skill intensity across industries, these results imply that the interna- tional skill intensity ordering of commodities is rather uniform. Similarly constructed rank correlations on an average earnings basis between the United States and seven other countries produced even larger positive rank correlation coefficients, all of which were significant at the 1% level (8, p. 174). However, more direct evi-' dence is available. Keesing (13) has used analysis of variance to 5 compare directly the immediate industrial requirements of scientists, engineers, and technicians (R & D) and also white-collar workers 5The relevance of these findings based upon the immediate requirements is established following a presentation of the empirical results. 13 across the manufacturing industries of seventeen countries.6 Consid- ering a subset of nine developed countries, 83% of the total variance of the R & D coefficients is attributable to differences between industries. Less than 3% is associated with differences between countries. The industry effect is very significant, while the coun— try effect is not statistically significant. A second test, involving the white-collar labor coefficients, revealed that 79% of the total variance of the coefficients is explained by industry effects; 11% is explained by country effects. Once again the industry effects are highly significant, while the country effects are significant only at the 10% level. When eight smaller and poorer countries are included in the sample, the country effect for white-collar workers is not quite significant at the 5% level.7 These results are based on immediate coefficients, and it has been argued here that only the total coefficients are theoreti- cally correct. Yet, Keesing's results are both relevant and impor- tant. The total coefficients are derived from the immediate coef- ficients [Stj = f (Sti)’ equation 2-1]. Since the relationship is an aggregation of a series of linear combinations, the test results presented above apply rather straightforwardly to the total require- ments coefficients. 6Based on data from a 1966 study by Horowitz, Zymelman, and Herrnstadt (7). 7Keesing fails to report the importance of this effect and the industry effect. The smaller and poorer countries which were added to the sample are Finland, Norway, Ireland, Israel, New Zealand, Yugoslavia, Argentina, and Chile. 14 These findings make an additional contribution. International trade studies attempt to explain a flow of commodities by using industry characteristics. However, the industry product mix varies across countries. Close correspondence between countries' industry level skill characteristics implies that differences in the product mix of industries do not affect empirical results importantly. Alternatively, this finding may be evidence that the actual industry product mix does not vary importantly between countries. If this is the reason for the international similarity skill coefficients, we have confidence that the U. S. technical coefficients (rij's) are accurately measuring industrial interrelationships. 2.3 Previous Empirical Tests of the Theory Attempts to measure the relative importance of labor hetero- geneity in determining trade flows can be classified into two divi- sions: human capital and human skills. The former approach begins from the proposition that labor essentially is homogeneous. From that beginning, empirical studies set out to measure the extent to which an industry's labor force embodies human capital over and above a specified base level. Generally, this is measured as the excess of the industry wage over a selected base wage (4,5,14,15). Alternative approaches estimate the amount of embodied capital directly from the cost of education (2,6), or from the income flows accruing to laborers (25). These empirical studies generally are confined to an analysis of U. S. trade patterns, although several have inspected other individual countries (4,25). One common 15 application is to try to resolve the Leontief paradox (or its equiva- lent for other countries) under the assumption that physical and human capital can be aggregated. Using input-out (I-O) analysis, Bharadwaj and Bhagwati found that when human capital estimates were added to India's tangible capital stock, the relative capital—labor ratio of India's exports increased (4, p. 139). A re-examination of U. S. 1947 trade patterns reveals that the Leontief paradox can be reversed by using wage differentials capitalized at 9.0 percent in combination with the physical capital stock (15, p. 457). Baldwin's I-O study (2) showed that a one million dollar bundle of 1962 U. S. exports embodied the services of more highly educated laborers than a comparable import bundle. However, the aggregation of net physical plus human capital did not resolve the Leontief paradox until natural resource intensive industries were excluded from the trade flow. By considering human capital as a third factor of production, West German exports were found to embody that input most intensively and simple labor services least inten- sively. Thus it was concluded that West Germany is most abundant in human capital, then physical capital, and least abundant in simple labor (25, p. 160). The classification of human as distinct from physical capital is fundamental to the skills approach. Various occupational cate- gorizations designate laborers with different skills. By identifying skilled and unskilled classes, industries can be ranked by their relative skill intensity. Obviously, this is not completely 16 unrelated to the human capital approach. Wage rates across indus- tries are influenced certainly by the occupational mix of the indus- try. Thus, a paper by Waehrer provides an empirical bridge between the two approaches (27). She finds that an occupational skill 7 explains a great deal of the variation in wages across indus- index tries (R2 = .74). Furthermore, the skill index explains each indus- try's trade balance as a percentage of industry shipments better than its yearly wage (16, p. 196). The occupational index is a fundamental tool of the human skills approach which measures the skill intensity of an industry. Although several specific indexes have been employed (10,12), the common objective is to devise a measure of the ratio of skilled to unskilled workers. The index is used either as an independent vari- able in a regression equation across industries (9, 27) or to reveal the factor intensity of an aggregate trade flow (10,12). Existing evidence seems to favor the skills approach. In a recent study, blue-collar and white-collar workers were found to be distinct inputs which cannot be aggregated (3). Separately, human capital and the physical capital/labor ratio have been found to influence U. S. export performance in different directions. Branson and Junz found the United States derived an advantage from human capital intensity and a disadvantage from physical capital intensity across industries (5). This result undermines studies which combine 7Waehrer's skill index and occupational groupings may be found in (27, p. 29). 17 physical and human capital (2,4,15). We shall, therefore, confine our attention to the empirical studies of the human skills theory. Most tests of the theory are based upon the use of direct or immediate skill coefficients, but there is one exception. Using total factor requirements, Baldwin (2) found that in 1962 the United States was a net exporter of the services of professional and tech- nical workers, craftsmen and foremen, clerical workers, and all types of farm labor. His regression analysis revealed the United States derived a significant advantage in industries which used scientists and engineers, craftsmen and foremen, and farmers and farm laborers relatively intensively. U.S.-Japan bilateral trade showed the U. S. advantage to be associated with the intensive use of scientists and engineers, and farm workers in an industry. The U. S. disadvantage was found to lie in industries which intensively used laborers and service workers. In trade with the Western European countries, the U. S. enjoyed a significant advantage in industries which required large proportions of scientists and engineers and farm laborers. Typically, skill indexes are used to test the theory. Their most common application is in conjunction with input—output analysis. Keesing has performed this type of test based on direct factor requirements. The method, described at the beginning of this chapter, requires the computation of the amount of services from laborers of each class embodied in a given export and import flow. Indexes are constructed to measure the relative skill intensity of each country's exports and imports using U. S. labor coefficients. The following skill classes have been used: 18 1. Scientists and engineers II. Other professional and technical workers III. Managers IV. Machinists, electricians, and tool and diemakers V. Other skilled manual workers VI. Clerical, sales, and service workers VII. Semiskilled and unskilled workers From these classifications, several index are formulated: A = (I + II + III + IV + V)/VII B = (I + II + III)/VII C = (IV + V)/VII D = [2(1 + II) + IVJ/VII The index chosen does not seem to be important. The rankings of nine countries according to indexes A, B, and C computed from 1951 export flows of manufactured goods are very similar, as are the import rank- ings by thos indexes (10). For these nine leading industrialized countries, Keesing has found that the export rankings are approxi- mately the inverse of the import rankings. Using index A, 20 out of 36 possible pairings revealed a rank ordering of countries such that a country which ranks above another always has the greater skill con- tent in the bilateral exchange of exports. From this Keesing con- cludes that labor skill availabilities influence trade patterns. However, he has made no attempt to measure actual factor endowments. In a second study the commodity coverage and the set of sample coun- tries was expanded (to include developing countries), and index 0 was applied to 1962 manufactured trade flows; similar test results were obtained. Although no "perfect" export and import ordering emerged, the Spearman's correlation coefficient between the export 19 and import rankings of thirteen countries was .878 (12). Using all fourteen countries, a rank correlation between their export indexes and the corresponding country per capita income ranking was .93. However, this still does not connect skill-intensity rankings to relative skill endowments. The results are interesting, but do not consitute tests of any theory, particularly not of the factor pro- portions framework which Keesing claims underlies the results (10, p. 5). Separately, this same skill index, computed from the imme- , diate requirements, has been used successfully as an independent variable in an equation explaining U. S./U. K. exports. The United States was assumed to be skill abundant relative to the United Kingdom (9). 2.4 Conclusions There is ample evidence that various measures of hetero- geneous labor inputs explain trade patterns. The skills approach has produced evidence that labor skills influence international trade patterns, but the underlying causal factor has not been inspected. It is entirely possible that Keesing means for us to infer from the skill intensity rankings the factor endowment rankings which "must" underlie his test results. But this is not a test. Unless these two rankings are found to be highly similar across countries, there is no support for a factor proportions theory based upon labor skill availability. Furthermore, we have no idea as to the distortion of 8Hong Kong was omitted from this calculation without explana- tion. Hong Kong's export index is the lowest in the sample, but its import index is not reported. 20 the skill indexes computed from the direct compared to the total requirements. Use of direct requirements is not defensible. They are no easier nor harder to use than total requirements, and they are theoretically inferior; thus, there appears to be no rationale for their use. The case for immediate requirements is different. These are free of the effects of imported inputs, they are theoretically incorrect, but they are easier to use and are capable of achieving far greater industrial detail than an input-output table allows.) Therefore, the relationship between the immediate and total require- ments is of interest. Nonetheless, a proper test of the human skills theory must use the total factor requirements and relate the result- ing evidence to cross-national skill endowment rankings. REFERENCES Arrow, K.; Chenery, H.; Minhas, B.; and Solow, R. "Capital- Labor Substitution and Economic Efficiency." Review of Economics and Statistics (August 1961): 225-50. Baldwin, R. "Determinants (If the Commodity Structure of U. S. Trade." American Economic Review (March 1971): 126-46. Berndt, E., and Christensen, LI “Testing for the Existence of a Consistent Aggregate Index of Labor Inputs." American Economic Review (June 1974): 391-404. Bharadwaj, R., and Bhagwati, J. "Human Capital and the Pattern ~ of Foreigh Trade: The Indian Case." Indian Economic Review (October 1967): 117-42. Branson, W., and Junz, H. "Trends in U. S. Trade and Comparative Advantage." Brooking's Papers on Economic Activity 2 (1971): 285-346. Fareed, A. ”Formal Schooling and the Human-Capital Intensity of American Foreign Trade: A Cost Appraoch." The Economic Journal (June 1972): 629-40. Horowitz, M.; Zymelman, M.; and Herrnstadt, I. Manpower Require- ments for Planning: An International Comparison Approach Vol. II, StatisticaT‘Tables. Boston: Northeastern ‘ University, Department of Economics, 1967. Hufbauer, G. "The Commodity Composition of Trade in Manufactured Goods." In The Technology Factor in International Trade. Edited by R. Vernon. New York: NBER, Columbia University Press, 1970. Katrak, H. "Human Skills, R. and D. and Scale Economies in the Exports of the United Kinguom and the United States.' Oxford Economic Papers (November 1973): 337-60. 10. Keesing, 0. "Labor Skills and International Trade: Evaluating Many Trade Flows with a Single Measuring Device." Review of Economics and Statistics (August 1965): 287- 94. 21 12 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 22 Keesing, 0. "Labor Skills and Comparative Advantage." American Economic Review (May 1966): 249-58. . "Labor Skills and the Structure of Trade in Manu- factures." In The Open Economy. Edited by Kenen and Lawrence. New York} C61umbia University Press, 1968. . "Different Countries' Labor Skill Coefficients and the Skill Intensity of International Trade." Journal of International Economics (January 1971): 443-52. Kenen, P. and Yudin, E. "Skills Human Capital and U. S. Foreign Trade." International Economics Workshop, Columbia University, New York, 1965. . "Nature, Capital, and Trade." Journal of Political Economy (October 1965): 437-60. "Skills, Human Capitol, and Comparative Advantage." Conference on Human Resources (NBER 1968). Kravis, I. “Wages and Foreign Trade." Review of Economics and Statistics (February 1956): 14-30. Kreinin, M. "Comparative Labor Effectiveness and the Leontief Scarce Factor Paradox." American Economic Review (March 1965): 131-40. Leontief, W. "Domestic Production and Foreign Trade, The American Capital Position Re-examined." Economia Internazionale (February 1954): 9-45. . "Factor Proportions and the Structure of American lrade: Further Theoretical and Empirical Analysis." Review of Economics and Statistics (November 1956): 387-407. . "International Factor Cost and Factor Use." American Economic Review (June 1964): 335-45. Minhas, B. An International Comparison of Factor Cost and Factor Use. Amsterdam: North-Holland Publishing, 1963. Philpot, G. "Labor Quality, Returns to Scale and the Elasticity of Factor Substitution." Review of Economics and Sta- tistics (May 1970): 194-9. Robinson, R. "Factor Proportions and Comparative Advantage, Part I." In Readings in International Economics. Edited by R. Caves and H. Johnson. Irwin, 1968. 25. 26. 27. 28. 29. 23 Roskamp, K., and McMeekin G. "Factor Proportions, Human Capital and Foreign Trade: The Case of West Germany Reconsidered." Quarterly Journal of Economics (February 1968): 7152-60. Valavanis-Vail, S. "Leontief's Scarce Factor Paradox." Journal of Political Economy. (December 1954): 523-28. Waehrer, H. "Wage Rates, Labor Skills, and United States Foreign Trade." In The Open Economy. Edited by P. Kenen and K. Lawrence. New York: 1968. Weiser, L., and Jay, K. "Determinants of the Commodity Structure of U. S. Trade: Comment." American Economic Review (June 1972): 459-66. Yeung, P., and Tsang, H. "Generalized Production Function and Factor Intensity Crossovers: An Empirical Analysis." Economic Record (September 1972): 387-99. CHAPTER III THE HUMAN SKILLS THEORY OF INTERNATIONAL TRADE: EMPIRICAL EVIDENCE 3.1. Introduction In this chapter we assess empirically the human skills theory for a group of nineteen countries. They represent a cross section of both developed and developing nations. We address two major questions concerning the theory: 1. 00 human skills influence the commodity composition of international trade? 2. Can the influence of human skills be related directly to a nation's relative skill endowment? Input-output analysis1 is utilized to provide answers to these ques- tions and to reveal the relative factor intensity of ten countries' trade when physical capital is introduced to the analysis. The structure of the input-output table imposes the funda- mental limit to disaggregation. A more detailed input-output (I-O) table would be of little value as the present level of disaggregation 1The input-output table was compiled by the Bureau of Econ- omic Analysis, U. S. Department of Commerce. It represents the input-output relations of 121 sectors of the U. S. economy for 1970. The compositional skills data are from U. S. Census of Population 1970, Occupation by Industry. The Annual Survey of Manufacturing 1971 provides the labor-output ratios necessary to perform the analysis. This data set represents the most current available statistics to assess the theory. 24 25 nearly exhausts the most detailed census classification by occu- pation and industry (12) at least for traded commodities.2 Throughout, this analysis assumes U. S. skill requirements to characterize the production processes in all countries. The assumption is supported by statistical evidence offered by Keesing (8); its implications have been addressed in the previous chapter. 3.2 Methodology A neo-factor proportions test of the human skills theory requires computation of the total skill requirements of each industry. The following procedure describes how to transform the immediate skill requirements into total requirements on an indus- try level basis. Define the direct requirement input-output table of the economy as A. 2For several industries the census classifications are not detailed enough. The primary and secondary ferrous and non- ferrous metal industries (I-O sectors 49-57) suffer most from this deficiency. For these sectors some averaging of the compo- sitional skills data occur. No skill data for the sector "space vehicles and guided missles" were available so the aircraft industry's coefficients were used as proxies for the immediate coefficients. Nonetheless, overlapping data are not very common and census to I-O concordances are considered quite acceptable. r. 611, 612, . . ., a1"- a a . . . a __n1’ n2’ ’ nn Here each aij represents X.. .11. X. .1 where: Xj is the total gross output of industry j X.. is the output of industry i absorbed by ‘3 industry j, each expressed in value terms using producer's prices.3 The total requirement input-output table of the economy is the Leontief inverse, _ F— [I-A] 1 = r11: r12: ° ° .. rin r , r , . . ., r L__n1 n2 nn “L where: r.. is the total requirement of industry i's output needed to produce one dollar's worth of industry j's output. ~ . The vectors, Sj, representing the immediate skill require- ments of each industry j per one million dollars of value of shipments are defined by, 3Prices reveived by producers. 27 Sj = [51. 52. s3. s4. 55. $6. s7]j. $1 = [5111 (3-1) 4 of this chapter which correspond The occupational groupings to the subscripts, t, are I. Engineers and Scientists II. Other professional, technical, and managerial III. Clerical and sales IV. Craftsmen and foremen V. Operatives VI. Nonfarm laborers and service workers VII. Farm Laborers Each Stj represents the number of man years from the tth skill cate- gory required to produce one million dollars of output of industry j. For example, 511 is the number of man years of service from scientists and engineers required in industry one to produce one million dollars' worth of industry one's output. ZSj is the number of man years of all types of labor service required by industry j to produce one million dollars' worth of output. The immediate skill coefficients are combined with the total requirement matrix to yield the total skill requirements for each industry. Define the total skill requirements of industry j for labor of type t as gtj, then _ n s . = Z s r (3-2) tJ 1:1 ti ij where: t = l, 2, . . . 7, j = l, 2, . . . 121. 4A detailed description of the occupational groupings may be found in (12). 28 The E S tj's are interpreted similarly to the stj's except that the tj's incorporate the labor services from all stages of production required to produce one million dollars worth of output. For example, 511 is the total number of man years of service from scientists and engineers required to produce one million dollars' worth of industry one's output. The representative one million dollar export and import bundles are obtained by computing the percentage weight of each commodity in the actual export or import bundle. Define Xj as the proportional weight of commodity j in the export bundle. We also may interpret Xj as the value of exports of commodity j expressed in millions of dollars. The import bundle Mj is computed and defined similarly.5 5In order to calculate the amount of labor services embodied in each nation's exports and imports, we must first adjust the 1975 trade data to the 1970 price level in which the a11's and r1j's of the input-output table are expressed. Then an adjustment is made to transform the value of the import and export vectors into produc- er's prices (14). This adjustment removes the transportation and wholesale trade mark-ups which are included in the valuation of the international trade flows. This is done because wholesale and retail trade are treated as separate sectors in the I-0 table. Transporta- tion is also a separate I-O sector, therefore, transport costs incurred in "moving" a product from one sector to the next must also be excluded. These adjustments avoid double counting. 29 The total labor services of each type, t, required to produce the one million dollar export bundle is estimated by,6’7 (St)x = #[( Z Stl rijlxj 1 (3-3) similarly for imports, (spm = 2 [( 3 st1r11m11 <3-4) J-11=-1 Using the total skill requirements from equation (3-3) and (3-4) five measures of skill intensity are computed. As these indexes test the human skills theory, the following empirical results must not be interpreted as a search for a skill index which performs best. Instead, we should view each as testing a different aspect of the theory. The indexes are defined below. Generally, we do not expect the results of the tests to vary greatly as the measure of skill 6For the computations of the total skill requirements by type i = l, 2, . . ., 121. Therefore, the contribution of all sectors-- excluding government--is accounted for. Since our main concern is with manufactures, natural resource intensive products were excluded from the first stage of the analysis. Manufactures are then defined as I-O sectors 18, 19, 22-92, excluding 42. These sector numbers refer to the values taken by j in the computation of the skill indexes in equation 3-5' - 3-9' for manufactures. In a separate calculation each nation's skill indexes were computed for all trade- ables, except oil. For these calculations j = 1,2,. ,8,10, , 41, 43, . . . 92. Finally, oil was added, then j = 1,2, I, 92. 7Due to the sharp increase of the price of oil and oil's increased prominence in the import vector of most of the countries studied here, it has been omitted from the analysis. Here "oil" refers to crude petroleum and natural gas (I-O sector 9) and refined petroleum (I-O sector 42). 30 intensity is changed. However, Y3 represents less of a skilled to unskilled disparity than the other measures and may be expected to not perform as well. The greatest disparity between skilled and unskilled labor is represented by index Y1- However, this neglects a large portion of each industry's labor force. Therefore, expect index Y2 to give the most objective test of the skills hypothesis. It combines a A relatively large skill disparity with consideration to the bulk of the labor force in each industry. The seventh skill category, farm labor, although clearly unskilled, was omitted from the general indexes because it is the only class which is specific to an industry. Therefore, the size of E7 is an industry's skill vector is totally determined by that indus- try's requirement of agricultural goods. To isolate this effect, 75 alone includes s, in the denominator. Skill index Y4 focuses attention on scientists and engineers. Keesing found this occupational classification to be the major determinant of trade flows when the direct requirements coefficients are used (6). The index represents the percentage of the total labor services consumed by an industry which are supplied by workers of type 1. This measure is the total requirements counterpart of Keesing's variable. The skill indexes are calculated for exports and imports. The ratio of a given index for imports versus exports (M/X) is the usual factor proportions test statistic. The export indexes are: 31 YTX = (ST)X/(S6)X (3'5) (sl)x + (S2)x YZX (S5)x + (S6)x (3-6) (3-7) —-___‘.’_(....._ - Y4X ' . (3 8) = (Sl)x + ($2)x Y5x 7 ( Es t=5 (3-9) t)x The formulas representing the indexes for imports are obtained by replacing the subscripts, x1 in equations (3-5)-(3-9) with m. The usual factor proportions test statistic for the ratio of skill embodied in imports versus exports is defined by, er = Ylm/le (3'10) YZY' = Yzm/sz (3'11) (3-12) Y3r = Y3m/Y3x 32 Y4r = Y4m/Y4x (3-13) (3-14) Y5r Y5m/Y5x The formulas by which the export indexes [equations (3-5)- (3-9)] are estimated are derived by substituting for (St) from equation (3-3) into equations (3-5)-(3-9) for the appropriate values of t. The resulting estimators are: 1'1 1 [gfls1irijlxj] A O = J _ u le (3 5) n 1[(151s61r1 J.ijT 11M: 11M: 5 n n 2 .§ [( § .§ Sti'ij)xj] 1'] t‘] 1'] (3'6.) "(<6 n ) 1 ' Z Z Z s .r.. X. j=1 t=5 i=1 t' '3 J Y2x = I 4 n l J ( X Z s r.. X. T t=l i=1 t1 '3 3 6 n 1[(t:5 ifistir15)x3] = 3 (3-7'> 11M: 11M: 3 1'1 [( z S11r1j)xj] i=1 . 6 n (3'8 ) l[(t§1 iEl Stirlj)xj] I {—1. 11M311M3 d j 33 n 2 n A 1: [( E .E Stjrij)xj] Y5x n [ 7 n ) ] ' Z Z 2 s .r.. X. j=l t=5 i=1 13 '3 3 Similarly, the estimators of the skill index for imports are derived by substituting for (St)m from equation (3-4) into the import equa- tions analogouS'UDequations (3-5)-(3-9). The resulting estimators defined as Ylm, Y2m’ Y3m’ Y4m’ y5m, are identical to equations (3-5')-(3-9') except the Xj is replaced with Mj. The ratio of the factor (service) content of a country's imports versus its exports is estimated by equations (3-10')-(3-14'). A A A Y". = Ylm/le (340') A A A er = Y211/sz (3—11') Y3r = Y3m/Y3x (3-12.) Y4r = Y4m/Y4x (3-13.) Y5r = Y5m/Y5x (3-14') This concludes the technical description of the formulation of the skill indexes. Although they are computed deterministically, the indexes derived from the computational form have been written with "hats" to stress that the underlying data from the Census 34 Bureau is estimated. However, for expositional convenience the "hats" and the terminology "estimate" will no longer be used. 3.3 Application Of the Skill Indexes The human skills theory suggests two empirical tests. The first requires knowledge of a country's relative skill endowment. Once the endowment ranking is known, the theory predicts the relative skill intensity of imports compared to exports (the value of Yir)' For countries with the greatest relative skill endowment, we expect Yir < 1, which implies that the skill intensity of exports exceeds the skill intensity of imports. For those countries in which skilled labor is relatively scarce, Yir > 1 is predicted. Although it is difficult to formulate expectations for countries whose relative skill endowments are at neither extreme, we can expect the skill endowment rankings to agree roughly with the inverse of the country rankings obtained by first ranking each country according to the value of Yir' This test statistic divides out the effect of imported inputs which become embodied in a nation's exports. The double count- ing of the factor services embodied in imports which are re-exported changes the value of Yir (compared to the case in which this effect is absent), but does not cause it to be greater or less than unity. Since this is our main interest, the critical nature of the text statistic is not affected. However, when Yix and Yim are considered 35 separately, the double counting may be of concern as it is not divided out.8 The second test requires separate consideration of the skill intensity of a country's exports and imports. These intensities are measured by Yix and Yim’ respectively. Countries are ranked separately, according to the skill intensity of their exports and imports. If skills (or lack of them) are the motivating force behind these trade patterns, we expect to find a negative and significant rank correlation between countries ranked by Yim versus Yix' Corroboration of the theory by this test implies that labor skills influence trade patterns. This, of course, is one of the theoretical predictions. But unless the skill intensity rankings of aggregate trade flows can be related to endowment rankings, the theory does not receive support. Correlations between the endowment rankings and either Yix or Yim establish this link. However, both of these measures contain the bias imparted by imported inputs. The bias of Yim will grow larger as the proportion of national imports demanded as inputs for subsequent exports increases. The export skill intensity index becomes more biased as exports contain a larger proportion of imported inputs. However, if the total and immediate skill coefficients are highly similar across industries, Yix is relatively less biased than Yim’g 8This problem is addressed below; an empirical perspective is provided in section 3.5. 9The similarity between the total and immediate factor requirements is established in Chapter IX. 36 To clarify this point, let us assume that the human skills theory is correct. We would expect a relatively skill abundant coun- try to export relatively skill intensive products. The imported inputs for these exports also would be relatively skill intensive. However, if skill availabilities influence trade patterns, imported inputs would be relatively less skill intensive than the value added by the final stage of fabrication (measured by the immediate coeffi- cients) of the export product. Since these imports are pulled by input demand, their skill intensity may exceed that of imports demanded for final consumption or investment purposes. This would cause Yix to be smaller than its "true" value (with imported inputs excluded), while Yim would be "too large." (When a country has a relative abundance of unskilled labor, the direction of these biases would be reversed.) Similarity between the total and inmediate skill coefficients would imply that Yix is not biased much while Yim would become more biased as input demand pulls a larger proportion of total imports. Thus, Yix would measure skill intensity across countries better than Yim’ if the human skills theory is correct. Differing domestic tariff structures and national demand patterns imply that export skill indexes are superior to import skill indexes. National import composition is affected greatly by these differences. Export composition, on the other hand, is influenced more uniformly across countries as these distortions are faced by all exporters. Therefore, skill indexes measured from export patterns will be used to relate skill intensity to relative national skill abundance. 37 The skill endowment rankings which provide the standard of comparison are compiled from international data on the occupational mix of each country's labor force. The ILO (10) provides this infor- mation. The relative skill endowment of each country is measured by calculating the ratio of skilled to unskilled workers in each coun- try. Skilled laborers are assumed to be professional, technical, administrative, and clerical workers. Service, farm, and production workers comprise the unskilled portion of the labor force. 3.4 Empirical Evidence for the United States Tables 3.1 and 3.2 provide detailed summaries of the labor content of U. S. exports and imports in 1975 by occupational group- ings. Table 3.1 clearly shows that the United States is a net exporter of each type of skilled labor service while it net imports each type of unskilled labor service, for a balanced one million dollar bundle of manufacturing exports and imports. Since the United States is the most skill abundant country in the world (Table 3.7), these findings represent strong support for the human skills theory. Table 3.1 refers to manufactures trade alone. When oil and other natural resource intensive products are entered in the export and import vectors, the United States no longer is revealed to net export each type of skilled labor services; nor does it net import each type of unskilled labor services. Yet it is precisely because of their natural resource intensiveness that these products were excluded. This point may be clarified by using oil as an example. "Crude petroleum and natural gas" products and "petroleum products" 38 TABLE 3.l.--U. S. Labor Requirements by Skill Classes, Per Million Dollars of Exports and Competitive Import Replacements (Total Requirements in man years, 1975 Manufactures Trade with the World) Manufactures Exports Imports Minfigpgmggrts I. Scientists and engineers 2.80 2.08 .72 11. Other professional, tech- nical and managerial 9.13 8.45 .68 III. Clerical and sales 12.33 12.01 .32 IV. Craftsmen and foremen 12.07 11.64 .43 V. Operatives 21.64 24.36 -2.72 VI. Laborers (non-farm) and service 5.29 5.54 - .25 VII. Farmers and farm laborers __;21_ _1499_ ;_;92_‘ TOTAL 64.17 65.08 - .91 SOURCE: Commodity Trade Statistics (Magnetic Tapes) and (12). 39 are skilled intensive relative to most other products. The United States and other countries, purchase oil from the countries which product it; but, is this because oil is a relatively skill intensive product? In fact, if the oil producing countries are at all skill abundant, it is largely due to the cooperation of multinational oil companies who have provided the skilled labor required by oil pro- duction. But we are trying to test a theory which implies that the direction of causation is the reverse of this. It is the relative immobility of the natural resource that makes the flow of factor services required to produce it irrelevant with respect to the theory. Nonetheless, the effect of adding oil and other natural resource products is revealed in Tables 3.2 and 3.3. With oil included in the million dollar bundle, the United States imports relatively more skilled labor services than when we consider only manufactures trade. This occurs because the United States is a net importer of oil, which shows up as a skill intensive commodity. However, y3 alone reveals imports as skill intensive relative to exports. This index incorporates the smallest skill to unskilled dispersion. When other natural resources are added, U. S. imports are measured as relatively skilled intensive by Y5 also. The only index which includes farm labor (unskilled) is YS' However, 1975 U. S. trade is affected by large shipments of wheat to the U.S.S.R. In addition, U. S. imports of agricultural goods were down by 10% in 1975 compared to 1974. These factors tend to make U. S. net agricultural exports larger than "normal." Since agricul- ture is very unskilled intensive, these factors depress the skill 40 TABLE 3.2.--Distribution of U. S. Labor Requirements by Skill Classes per Million Dollars of Exports and Competitive Import Replacements (Total Requirements in percentages 1975 Trade with the World) . Import/Export Skill Classes Exports Imports Ratio 1. Scientists and engineers All industries 3.38 3.31 .980 Excluding oil 3.37 2.79 .828 Manufactures 4.36 3.20 .733 11. Other professional, tech- nical and managerial All industries 12.31 13.74 1.116 Excluding oil 12.28 12.50 1.018 Manufactures 14.22 12.99 .913 III. Clerical and Sales All industries 17.21 19.60 1.139 Excluding oil 17.15 17.57 1.025 Manufactures 19.22 18.46 .961 IV. Craftsmen and foremen All industries 15.42 15.93 1.033 .Excluding oil 15.42 16.13 1.046 Manufactures 18.81 17.89 .951 V. Operatives All industries 27.04 31.00 1.147 Excluding oil 27.07 33.11 1.223 Manufactures 33.73 37.43 1.110 VI. Laborers (nonfarm) and service All industries 7.82 8.91 1.139 Excluding oil 7.81 8.73 1.118 Manufactures 8.25 8.51 1.031 VII. Farmers and farm laborers All industries 16.81 7.51 .447 Excluding oil 16.90 9.17 .543 Manufactures 1.42 1.53 1.080 SOURCE: See Table 3.1. NOTE: The tabled values give the percentage distribution of the labor force, embodied in exports and imports, by occuaptional class for each of three balanced bundles of exports and imports, i.e., all industries, all industries-excluding oil, and manufactures. TABLE 3.3.--Skill Ratio of U. S. Imports, Exports, and Imports/ Exports for Various Skilled/Unskilled Ratios and for Selected Groupings of Industries. 1975 Trade with the World) (Total requirements Exports Y1 Y2 Y3 Y4 Y5 All Industries .4320 .4502 .4517 .0338 .3037 Excluding oil .4315 .4485 .4507 .0337 .3022 Manufacturing .5289 .4427 .5973 .0436 .4282 Imports Y1 Y2 Y3 Y4 Y5 All Industries .3717 .4272 .4920 .0331 .3595 Excluding oil .3196 .3653 .4581 .0279 .2997 Manufacturing .3761 .3523 .5168 .0320 .3410 Imports/Exports Y] 72 Y3 Y4 Y5 All Industries .8604 .9489 .0892 .9800 .1838 Excluding oil .7406 .8145 .016 .8279 .9868 Manufacturing .7111 .7958 .8652 .7334 .7964 SOURCE: Table 3.2 Skill Ratios: Y1 = I/VI, Y2 = (I + II)/(V + VI) (I + II + III + IV)/(V + VI) Y3: Y4 = I/(II + III + IV + v + VI + VII) Y5 = (I + Ill/(v + v1 + v11) NOTE: For skill class definitions, see Table 3.2. 42 ratio of U. S. exports. Despite the anomalies created by large oil imports and large agricultural exports, the U. S. is shown to export skill intensive commodities by three of the five import/export test statistics when all industries are included. Still, the results for manufactures alone are considered the best test of the theory. 3.5 Revealed Factor Intensity: The Export and Import Patterns of Nineteen Countries In this section we assess the proposition that the skill con- tent of goods influences international trade. The approach used here is based on the method which Keesing first employed (4). The pro- cedure does not require knowledge of the relative skill endowment of each country. In the next section we shall relate these findings to national skill endowments. As U. S. skill coefficients are used throughout, we assume that skill intensity reversals do not occur. The assumption becomes less reasonable as the country to which it is applied becomes less developed. Since we are considering export and import indexes separately, we have the problem of bias discussed in the previous section.10 Table 3.4 presents the skill indexes for the manufacturing exports and imports of the nineteen countries which comprise the sample. The correlations between the exports and import skill inten- sity rankings of nineteen countries are listed in Table 3.5. This table presents correlation coefficients for manufactures trade along 10This bias should be relatively large for Hong Kong, Taiwan and Singapore. Hong Kong's import skill intensity ranking has been perverse in two previous studies (5,4). 43 NNN. CNN. NNN. CNN. NNC. CNC. NCC. NNN. NNN. NCN. CNNNNNNC NNC. CNN. NCN. NNN. NNC. NNC. NCC. CNN. NNC. CCN. NNCCN NNN. NNN. CNN. CNN. NNC. NNC. NCC. NNN. CNC. NNN. NCNCC NNN. NCN. NNC. NCN. NNC. CNC. NNN. CNN. NNN. NCN. CCCN CCCN CNN< CNN. NNN. NNN. NCC. NNC. NNC. NCN. NNC. CNC. CNN. CCNNNNN zaz CNN. CNN. CCN. NNC. CNC. NNC. NCN. NCN. NNC. CNN. NNNNCNNCC Ncwmoo NNN. CCN. NNN. CNN. NNC. NNC. NCN. CNN. NCC. CCN. NN>NNNCCCN NNC. NNN. NNN. NNN. CNC. NNC. NNC. NNN. NNC. NNN. CNNCN mnogsm Nonpo NCN. CNN. NCN. NNC. NNC. CNC. NNN. NNN. NNN. NNN. CCNNCLN NCN. NNN. CNN. NNN. CNC. CNC. CNN. NNN. NCN. NCC. NNNECNC NCN. NNN. CNN. NNN. NNC. CNC. NNN. NNN. NNN. NNN. CNCCCCCNCC-ECNCNCN CNN. NNN. NCN. NCC. NNC. CNC. NCN. CNN. CCC. CNN. NNNNN NNN. NNN. CCN. NCN. CNC. CN9 NNN. NNN. CNN. NCC. CCCNLN NCN. NNN. NNN. NNN. NNC. NNC. NNN. NCC. NC9 NNC. NCCNNNNCNCZ CNN. NNN. CNC. NNN. CNC. NNC. CNN. CNN. CNN. CNC. NCNCLCC NNN: NNN. NCN. CNN. NNN. NNC. NNC. CNN. CNC. NNN CNC. CCCCCNC Caywmp CNN NNN. NNN. CNC. NNC. CNC. NNC. NNN. CCC. NNN. NNC. CNCNC CNN” NNN” NNN. NNN. NNC. NNC. NNN. NNN. NNC. NNN. NCNCNC NCN NNC NNN. NCN. NNC. CCC. NNN. NCC. NNN. CNN. NNNNNN CCNNCC upmaogmca Em» xm> Em» xm» Ev» xC> Em» xm» EN» xp> Nguczoo NCNNCC NNN CNN: CCNNN NNCN .NNCCENNNCCNN NNNCNV NCNCNCCNCCCZ cN muNNN .NmNNNcaou :mmNmCNz No NNNCCEN CCC NNNome NCN mewocN NNNxm--. C. m NquN 44 TABLE 3.5.--Correlati0ns Between Skill Indexes for Exports and Imports for Nineteen Countries (Total Requirements, 1975 Trade with the World) Type of Correlation Y1 72 Y3 Y4 Coefficient Manufactures Spearman -.556 -.607 -.391 -.574 (.007) (.003) (.049) (.006) Kendall -.404 -.462 -.298 -.439 (.008) (.003) (.038) (.005) Excluding Oil Intensive Industries Spearman -.463 -.498 -.054 -.l93 (.023) (.015) (.413) (.215) Kendall -.322 -.404 -.088 -.l46 (.028) (.008) (.300) (.191) All Industries Spearman -.361 -.244 +.056 +.OO9 (.065) (.158) (.410) (.486) Kendall -.263 -.l7O -.006 -.0526 (.058) (.156) (.487) (.377) NOTE: The correlations are between the skill intensity rankings of the exports versus the imports of the nineteen sample countries in Table 3.4. The export and import bundles were balanced with respect to each of the commodity groupings named above. 45 with, total trade, and total trade excluding oil. Since the indexes underlying the correlations are ordered high to low, we expect nega- tive and significant correlations. Considering trade in manufactures alone, we find that both the Kendall and Spearman's coefficients are negative and significant, indicating that countries which export relatively skill intensive products import unskilled intensive prod- ucts. Therefore, the skill content of goods influences trade patterns. If Hong Kong is omitted from the nineteen country sample, we find the relevant Spearman's correlations are: py1 = -.821, p2 = -.891, py3 = -.637, py4 = -.851. All are easily significant at the 1% level and all are substantially greater (in absolute value) than when Hong Kong is included.11 11For Hong Kong, the export and import data concorded to the input-output sectors confirm the hypothesis that Hong Kong's imports are strongly related to its input demand for the products in which Hong Kong has an export advantage. Hong Kong is a large net exporter of apparel and hosiery and knit goods. It substantially net imports textile products of a more primary nature which are also rather unskilled intensive, but are important inputs for Hong Kong's textile exports. Although synthetic fibres are not relatively unskilled intensive, this product group is another example of Hong Kong's export advantage determining its import demand. In order to see if these results could be generalized, and to assess the magnitude of the bias of the y m's, a crude index of the bias was constructed for the nineteen sample countries. Imports as a percentage of GDP pro- vides the crude measure of bias. Although this does not directly relate the export demand for imported inputs to imports, the rankings by this measure, across countries, ought to be more or less correct. When this is done, we find Yim is most biased for: (1) Hong Kong, (2) Ireland, (3) Belgium-Luxembourg, (4) the Netherlands, (5) Korea. For Hong Kong, the bias is exceptionally large as the value of its imports is nearly identical to its GDP. The least bias was found for: (l) U. S., (2) India, (3) Japan, (4) Australia, (5) West Germany. The association of bias with geographic country size is not surprising. The lack of association of bias with stage of development is notable. 45 When to our million dollar bundle we add natural resource intensive products--excepting oil--the proposition that the skill intensity of goods influences trade receives less support. Although the direction of the correlation is as we expect for all skill indexes, 73 does not meet the usual standards for significance. The lack of importance of scientists and engineers, in determining the trade flows of natural resource intensive products (excluding oil), is indicated by the decline in significance of p74 when these products are included in the million dollar bundles. When oil intensive products are included, the significance of all the correlation coefficients declines. Several Spearman's coeffi- cients (for 73 and Y4) appear with perverse positive signs although they are not significant. Obviously, the is due to the relative importance of oil in the import bundles of the developed and rela- tively skill abundant countries which are predicted to export, not import, skill intensive commodities such as oil. We have explained how the inclusion of natural resource intensive products can subvert the analysis, and shown that their admission has that effect. We shall, therefore, proceed to focus on manufactures and ignore all natural resource intensive products, including oil. Considering trade in manufactures again, Table 3.6 presents the usual factor proportions test statistic (Yir = Yim/Yix) for our sample of nineteen countries. In view of our previous discussion, it is satisfying to note that Hong Kong is a net importer of unskilled labor services. It is worth repeating that the values in Table 3.6 are biased in terms of their value, but are accurate with respect to 47 TABLE 3.6.--Skill Index Ratios of Yim/Yix Derived from Nineteen Countries' Trade in Manufactures (Total Requirements. 1975 Trade with the World) Country er Y2r Y3r Y4r Y5r Ungrouped United States .711 .796 .865 .733 .796 Canada 1.299 1.014 .998 1.079 1.020 Japan .629 .896 .780 .799 .849 9:29 United Kingdom .805 .901 .913 .868 .891 West Germany .850 .839 .833 .843 .824 Netherlands .859 .871 .905 .874 .871 France .983 1.109 .994 1.001 1.011 Italy 1.109 1.220 1.107 1.227 1.201 Belguim-Luxembourg 1.108 1.015 .994 1.062 1.019 Denmark .962 .9991 .960 1.020 .993 Ireland .946 1.010 1.052 1.025 1.454 Other Europe Spain 1.413 1.323 1.118 1.444 1.313 Yugoslavia 1.317 1.234 1.114 1.281 1.237 Oceana Australia 1.419 1.154 1.119 1.304 1.221 New Zealand 1.876 1.410 1.407 1.751 1.537 Asia Hong Kong 1.289 1.535 1.357 1.564 1.492 Korea 1.487 1.681 1.482 1.743 1.673 India 1.756 1.856 1.605 2.041 1.941 Pakistan 1.924 1.856 1.759 2.270 1.987 1This ratio, if rounded to three decimal places, is equal to unity. 48 whether they are greater or less than one. In Table 3.6 Yir < 1 designates countries with relatively skill intensive exports, while Yir > 1 implies relative unskilled labor intensity. The United States, Japan, West Germany, the Netherlands, and the United Kingdom are shown to export skill intensive commodities by every ratio. This corresponds to our casual intuitive knowledge of their relative skill abundance. Denmark, Canada, France, and Belgium-Luxembourg exhibit a skill intensive pattern according to at least one ratio. The problem of bias notwithstanding. it is encouraging to find that Pakistan, India, and Korea are indicated as having the least skill intensive exports of all nineteen countries in our sample by nearly every ratio of skill indexes. 3.6 Skill Endowments and Revealed Skill Intensity In this section we will perform the most critical skills test. Although we already possess ample evidence that skills influence international trade patterns, unless these findings can be linked to countries' skill endowment rankings, the previous evidence is little more than an interesting statistical finding. The skill endowment index is constructed from national occu- pational groupings provided by the ILO (10). These data are highly inclusive of service professionals (such as ministers) in the most skilled occupational class. Thus, inferring endowment rankings from the percentage of workers in this class alone is a tenuous procedure, although it has been done (4). Instead, Table 3.7 employs a more inclusive index which ought not be as sensitive to ILO classification 49 TABLE 3.7.--Ski11 Endowment Rankings and Skill Intensity Rankings (Yix) of the Exports Seventeen Cbuntries 1975 figfifime“ Country1 “(11:32?" 1’1 x Y2x Y3 x Y4x Y5x 1 United States .814 l 1 Canada .804 11 Netherlands .617 3 3 6 2 3 United Kingdom .598 2 2 5 Germany .555 5 Japan .522 4 _ 4 7 Belgium- Luxembourg .505 10 10 10 10 9 8 Denmark .444 6 7 3 7 7 9 France .443 7 8 8 6 8 10 Ireland .342 8 9 l3 9 15 11 Italy .274 9 12 12 11 11 12 Hong Kong .216 15 16 16 15 16 13 Spain .209 12 11 9 13 10 14 Yugoslavia .200 13 13 11 12 12 15 Korea .128 14‘ 14 14- 14 13 16 India .075 16 15 15 l6 14 17 Pakistan .059 17 l7 17 17 17 3353233"e3d33¥£il“‘°" 223?? 223:? 2:39:32); 223?; and revealed rank SOURCE: ILO (10), and Table 3.4. 1New Zealand and Australia are omitted. If included,Australia would rank fourth and New Zealand seventh in the expanded sample. Australia's Y1 rank is consistently about 12 while New Zealand's is about 16. Nhen inlcuded in the rank correlations, all are positive and significant at the 1% level, but range in value from around .65 to 75. 2SKILL ENDONMENT INDEX--(Professiona1,technical, administrative and clericaIWService, farm, and production workers) 50 problems. Others may be calculated, but due to the classification problem, they are less reliable in their correspondence to the actual (but unknown) skill endowment ranking of countries. The endowment index used in Table 3.7 considers professional, technical, adminis- trative and clerical workers as skilled. Service, farm, and produc- tion workers are treated as unskilled. With the shortcomings of the data in mind, the resulting rankings are considered to be as accurate as the data allow. The theory predicts that we will find a positive correlation between the endowment and skill intensity rankings.)2 The appro- priate rankings and corresponding correlation coefficients appear in Table 3.7. The correlations are very high, positive, and significant. These results, which strongly favor the theory, are somewhat less striking when the two omitted countries, Australia and New Zealand, are included. These countries, although clearly developed, are not highly industrialized. However, the ILO data are suspect for several reasons. First, Australia has, over the least decade, made a con- certed effort to attract skilled labor by offering to skilled pro- fessionals, free round trip transportation if they remained in the country for a specified number of years. Yet, Hufbauer (4) estimated Australia's skill endowment ranking as third among these same coun- tries in the early 1960's. (New Zealand ranked sixth at that time.) The relatively high ranking does not appear consistent with Australian policy for the ensuring period. For this reason, the Australian 12The Yix measures of skill intensity are used as they are less biased than the Yim indexes. 51 ranking is suspected of being incorrect. Lacking an alternative ranking criterion, both countries are omitted from the test. In terms of individual countries, the skill indexes for Canadian exports alone are persistently out of line with the endow- h in the endowment ment ranking. No country falling lower than 10t ranking is ever revealed to be a net exporter of skilled labor serv- ices (Table 3.6). Also, every country in the top ten is shown to export relatively skilled labor intensive commodities according to at least one skill ratio. Although the endowment rankings roughly approximate the ranking of countries according to the size of Yir (Table 3.6), Canada's trade patterns are generally more akin to those' of a less skill endowed country. Japan trades as if it were more highly endowed with skill. 3.7 A Three Factor Revealed Approach to the Assessment of Trade Patterns: Evidence For Ten Countries Thus far, we have conducted out tests as if only two factors of production exist: skilled and unskilled labor. However, it has been suggested that a three factor, factor proportions model best explains trade patterns. Recently Branson and Junz (2) have offered this hypothesis after finding a significantly negative coefficient for the capital/labor ratio in their multiple regression analysis of 52 U. S. trade. Since a similar finding is made in Chapter IX, we shall construct a three factor model.13 The total requirements for capital were computed by using the gross book value of capital augmented by working capital in the form of materials and work in progress. Finished goods inventories are treated as if held for the industry about to consume them and dis- tributed to the consuming industry using the input-output table (direct requirements). This yields the immediate capital stock for all industries. The capital stock is converted to a flow by dividing it by the value of shipments for industry j, call this kj. Using the same input-output notation as earlier, " n (3 15) jE](jE]kJ . rlj)xj _ Ktx where, K is the total requirement of capital (measured ih thousands of dollars) to produce one million dollars worth of exports. 13Although both the multiple regression and input-output tests begin with some underlying data set, there is little statisti- cal evidence which would lead us to believe that each test--of the same theory--wou1d produce the same results. The multiple regression analysis is based solely on capital/labor ratios. The input-output analysis, although employing these same ratios sums the flow of capital services by means of the capital/output ratio weighted by the percent of exports (or imports), and divides it by a similarly derived measure for labor services (of each type). Since the multiple regression analysis completely ignores capital/output and labor/output ratios, corresponding results between the two tests is by no means assurred. We will compare the results of this chapter to those of Chapter IX in the latter chapter. The choice of countries for this test was determined by the findings in Chapter IX. Countries for which the capital/labor ratio was significant in the multiple regres- sions are included here, as are those countries for which the skills variables did not "work" as well as expected. 53 Replacing X with M in equation (3-15) yields the comparable import requirement of capital, Ktm' To obtain capital/skilled labor and capital/unskilled labor ratios the Ktx and Ktm values are divided by the respective £(S and 2(St)m values [equations (3-3) and (3-4)] t)x which are summed over the relevant t's. For this analysis occupa- tional classes I, II, III, and IV are designated at skilled V, VI, and VII are unskilled. U. S. capital and labor coefficients are used to measure the relative skill intensity of manufactures trade. The results of the three factor calculations are presented in Table 3.8. The U. S. is revealed to have relatively more skill embodied in its exports than its imports. Relative to imports, U. S. exports are also more skilled than capital intensive, but more capital intensive then unskilled. Thus, the factor intensity ordering revealed by the U. S. trade flow is, skilled labor > physical capital > unskilled labor. Canada is shown to derive its greatest advantage from capital intensive industries; its greatest disadvan- tage is in labor intensive industries. It is well known that Canada trades most intensively with the U. S. Transportation costs are certainly one reason for this. However, if we accept the three factor model and the relative factor abundance which these calcula- tions imply for both the U. S. and Canada, we find another, if some- what unconventional, explanation for U. S.-Canadaian trade. The U. S. derives its greatest advantage from the same factor in which Canada is most scarce, skilled labor. Therefore, relatively skill intensive exports from the U. S. are readily absorbed by Canada, since Canada needs to import the services of relatively scarce skilled 54 TABLE 3.8.--Revea1ed Relative Factor Intensity with Three Factors: Capital, Skilled Labor, and Unskilled Labor--Manu- factures Trade (Total Requirements, 1975 Trade with the World) x m m/x x m m/x United States Netherlands sk/un 1.31 1.11 .848 1.22 1.09 .886 k/sk 19.74 21.13 1.070 23.10 21.30 .921 k/un 25.76 23.38 .908 28.27 23.11 .818 k/l 11.18 11.108 .994 12.71 11.08 .872 United States: sk > k > un Netherlands: k > sk > un New Zealand Japan sk/un .86 1.25 1.452 1.29 1.00 .780 k/sk 26.02 21.58 .829 21.90 21.48 .981 k/un 22.44 27.02 1.204 28.21 21.56 .761 k/l 12.05 12.00 .996 12.33 10.76 .823 New Zealand: un > k > sk Japan: k > sk > un Australia Korea sk/un 1.04 1.18 1.137 .77 1.18 1.54 k/sk 25.27 19.75 .781 19.95 22.83 1.15 k/un 26.32 23.33 .888 15.30 27.02 1.77 k/l 12.89 10.72 .832 8.66 12.37 1.43 Australia: k > un > sk Korea: un > sk > k TABLE 3.8.--Continued 55 x m m/x x m m/x Yugoslavia Canada sk/un 1.053 1.23 1.167 1.19 1.20 1.011 k/sk 22.32 23.52 1.053 22.78 18.94 .831 k/un 23.51 28.91 1.230 27.04 22.73 .840 k/l 11.45 12.97 1.133 12.37 10.33 .836 Yugoslavia: un > sk > k Canada: k > un > sk Denmark Belgium-Luxembourg sk/un 1.22 1.17 .961 1.11 1.11 .992 k/sk 18.96 21.66 1.413 23.39 20.92 .894 k/un 23.11 25.38 ' 1.098 26.05 23.12 .887 k/l 10.41 11.69 1.122 12.32 10.98 .891 Denmark: sk > un > k Belgium-Lux.: k > sk > un NOTE: sk = skilled laborers, classes I, II, III, and IV (man years) un = unskilled laborers, classes V, VI, VII (man years) k = physical capital (thousands of dollars) _I 11 total labor (man years) 56 laborers. The relative abundance of capital provides Canada with an advantage in a factor which is neither most abundant nor scarce in the U. S. The U. S. may be said to be indifferent to the absorbtion of this factor. Given that neither country is relatively abundant in unskilled labor, it is most likely that their bilateral trade patterns will be governed by the factors in which each has the great- est relative advantage. The relative unimportance of transport costs will tend to make small advantages relatively more important in determining their bilateral trade. The Netherlands exhibits a strong skill pattern in its trade as well as its endowment ranking (Table 3.7), but its trade flows are found to be even more intensive in capital than skilled labor. The trade patterns of Australia, Japan, and Belgium-Luxembourg are capi- tal intensive relative to both skilled and unskilled labor. Denmark, Korea, and Yugoslavia exhibit a factor content of trade which implies that they lack capital relative to the other factors of production. 3.8 Conclusion The two factor skills theory is very consistent in its corre- spondence to several theoretical predictions; the skill content of exports with respect to imports, the skill index for exports with respect to the relative endowment rankings of nations, and the national endowment rankings with respect to the ratio of the skill- import/skill-export indexes. With the exception of Canada, all of the anomalies can be accounted for. As predicted, when natural 57 resources are included, the skills theory does not explain the commod- ity composition of trade as well. The relative endowment rankings which are revealed by the three factor model and accepting the three factor version of the theory, produces some interesting results. The relative factor endowment rankings for the U. S. are exactly as expected. When coupled with the revealed Canadian endowment position, we find a very unconventional explanation for U. S.-Canadian trade. The three factor approach is useful in explaining the Leontief paradox. We infer that physical capital is neither relatively scarce nor relatively abundant from U. S. manufacturing trade. In a two factor sense, the relative capital content of U. S. trade is stable. A 1962 study of U. S. manufacturing trade, showed the capital/labor ratio for imports divided by that for exports to equal .99 (1). For 1975, that ratio is unchanged (Table 3.8). The form of this three factor model is different from any other that has been used in the past. Usually human capital is estimated, and combined with the physical capital stock. The approach used here is considered superior for several reasons. First, it has greater value in use. Occupational groupings can be made very detailed. Since they are rather objective and uniform, we can link comparative advantage to specific and identifiable characteristics in the economy. Second, the human capital tests could be performed from the occupational data by assigning to each class its average wage rate. 10. 11. 12. REFERENCES Baldwin, R. "Determinants of the Commodity Structure of U. 5. Trade." American Economic Review (June 1974): 122-46. Branson, W. and Jung, H. "Trends in U. S. Trade and Compara- tive Advantage." Brookings Papers on Economic Activity 2 (1971): 285-346. Carter, A. Structural Change in the American Economy. Cambridge, Massachusetts: Harvard University Press, 1970. Hufbauer, G. "The Commodity Composition of Trade in Manufactured Goods." In The Technology Factor in International Trade. Edited by R. Vernon. New York: NBER, Columbia University Press, 1970. Keesing, 0. "Labor Skills and International Trade: Evaluating Many Trade Flows with a Single Measuring Device." Review of Economics and Statistics (August 1965): 287-94. "Labor Skills and Com arative Advantage." American “Economic Review (May 1966 : 249-58. "Labor Skills and the Structure of Trade in Manu- facturers." In The Open Economy. Edited by Keven and Lawrence. New York: Columbia University Press, 1968. . "Different Countries Labor Skill Coefficients and the Skill Intensity of International Trade.“ Journal of International Economics (January 1971): 443-52. Miernyke, W. The Elements of Inppt-Output Analysis. New York: Random House, 1965. International Labor Office, Yearbook of Labor Statistics. Geneva, 1975. Internal Revenue Service. Corporate Income Tax Returns, 1970. U. 5., Bureau of the Census. Census of the Population by Industry by Occupation, 1970. 58 59 13. U. S., Bureau of the Census. Annual Survey of Manufacturing, 1970. 14. U. S., Department of Commerce. Survey_pf Current Business. November, 1974. 15. U. S., Department of Commerce. Wholesale Prices and Price Indexes, 1975. CHAPTER IV THE SCALE ECONOMY THEORY 4.1 The Theory Of the several possible versions of the scale economy hypothesis, we are concerned with scale economies internal to the plant. When scale economies are present, large plant size confers a comparative cost advantage to producers. In the pre-trade stage, home market size is a factor if plants as large as the most efficient size elsewhere cannot be supported. If scale economies do not persist across all sizes of plants, domestic producers in small countries may not be able to satisfy a given level of domestic demand with plants of optimum scale. As home demand grows, firms are faced with the choice of building new establishments or adding capactiy to existing optimally sized plants. Thus, the size of the home market may be important in the pre-trade stage. But when the economy is opened and trade is allowed, the potential market is expanded. However, except for products that are highly standardized, it is unlikely that producers in small countries will be able to depend on the foreign market (6). Several recent studies provide empirical support that home market size is important when internal economies exist. The average employment size of manufacturing establishments across industries is strongly correlated with indicators of market size (15,16). This 60 61 highlights the obstacles small nations face in achieving industrial efficiency when scale economies are important. Suppose, for a given industry, the distribution of plant takes the same shape across countries. Figure 4.1 illustrates the effect of market size on these distributions. Country 8, having a relatively larger market, also has a larger average plant size. Exporters gain their advantage due to their size, given the level of scale economies in this industry. Thus Country 8, having a larger proportion of its firms reaping scale benefits, has an advantage compared to A. Suppose we denote So as the minimum scale necessary for a firm to absorb transport costs, penetrate tariff barriers and compete in the foreign market. Then, being larger than Country A, Country 8 has absolutely more plants in a given size class, even when the relative frequency of plants is the same in each country, thus B has relatively more plants which reap the advantage. This effect also operates for B's less efficient firms; however, they are partially protected by existing tariff and quota barriers. % of plants S Size Figure 4.1.--Hypothetica1 Distribution of Firms in Two Countries. 62 When scale economies are internal, the assumption of perfect competition must be dropped in order to explain the survival of sub— optimal plants (4, p. 134). Caves, Khalilzadeh-Shirazi and Porter have proposed that a price umbrella is maintained by dominant sellers. If this assumption is correct, we expect to find producers, who export and service the home market, earn relatively larger profits. Empirical evidence consistent with this hypothesis has been found. United Kingdom exports as a percentage of industry output are both positively and significantly associated with the profit rate in the industry (4, p. 137). 4.2 Internal Economies: Previous Empirical Tests The scale economy theory has been tested by measuring "scale" as the proportion of an industry's employees working in establish- ments with 250 or more employees (1). Using net exports as the dependent variable in regressions estimated across industries, this variable failed to emerge as a significant determinant of the commodity composition of U. S. trade. The coefficient of the scale variable was negative for U. S. trade with the world, "others," Western Europe and Japan; significantly negative for the last two. The scale hypothesis was weakly confirmed by U. S. trade patterns with Canada and the LDC's; neither coefficient was significant. Input-output analysis was employed to calculate the relative plant size embodied in U. S. exports and import replacements. This tech- nique revealed U. S. exports as relatively more scale intensive then 63 imports. Excluding agricultural commodities decreased the relative importance of scale embodied in exports compared to imports. These test results indicate either, that scale economies are not determinants of U. S. trade patterns, or that size alone is not a sufficient proxy for scale economies. A simple measure of internal scale economies in an industry has been proposed by Hufbauer (6). The extent of scale economies internal to the plant are measured by alpha in the following equation: v. = kn? (4-1) where, V. represents the ratio between value added per man for the employment size class i and the average value added per man for all establishments in the four digit industry. n represents the average number of workers employed per establishment in size class i; k is a constant a represents the scale elasticity parameter.1 Therefore, a = .05 implies that a doubling of plant size increases output per worker by 5 percent. The scale elasticity parameters. estimated by Hufbauer (6) have been employed frequently to test the scale economy theory (2,3,6,8,17). Using scale elasticity parameters, the scale account has been tested in isolation by relating the scale embodied in a nation's manufactured exports to the size of national manufacturing output (6). Ilhe potential biases inherent in this measure are fully discussed in Chapter V. 64 The scale content of exports is estimated by 66.x. where Xj is the J J J proportional weight of commodity j in a nation's manufactured export bundle; aj is the scale elasticity parameter for industry j. These calculations were performed at the three digit SITC commodity level for 102 SITC's (classifications 5, 6, 7, and 8). For a sample of 24 nations, the rankings between the scale intensity of exports and national manufacturing output were positively, but insignificantly correlated.2 However, rank correlations between the scale intensity of exports and per capita gross domestic product produced a positive and significant rank correlation. This indicates that scale economy benefits are associated with industrial sophistication, but provides no support for the scale economy account. Branson and Junz used the scale elasticity parameter in regressions estimated across 101 three digit SITC manufacturing industries (2). Human capital, physical capital, and a measure of technological intensity were also employed as independent variables. The coefficient of the scale elasticity parameter was positive and significant thereby explaining 1964 and 1967 U. S. net exports. In a subsequent study, Branson scaled the dependent variable, using X/(X + M) across industries. When this is done the coefficient of the scale elasticity parameter is no longer significant although it is always positive (3). This scale economy measure achieved better results when the U. S. share of developed countries' exports was used as the dependent 2Viewing national market size as a proxy for average national plant size, this serves as a test of the scale economy account. 65 variable and regressions were estimated across industries (17). The positive and significant (1 percent level) coefficient of the scale measure indicates that scale economies were a determinant the commodity composition of U. S. trade in 1960 and 1967. Using the scale elasticity parameter in a different context, Katrak has argued that whenever a b of a b aj (NT/NT) > (NJ/Ni) (4‘2) country a's exports of commodity i will be relatively greater than country b's (8, p. 342). In the equation, Ni is the number of employees in the ith industry; ai is the scale elasticity parameter of the ith industry; the superscripts represent the country. Rank correlations between 1962 U. S./U.|<.exports to the world and the relative scale effect produced correlation coefficientscyf.59 and .76 for seventeen and fourteen manufacturing industries respectively. Both results are significant at the 5 percent level. The relative scale variable, (N3/N?)aj, was also employed in multiple regression analysis. It significantly explained U. S./U. K. exports. The relative scale variable performed significantly irre- spective of the industry groupings, the functional form of the equa— tion, and the year of observation (1962,1964,1966). The scale elas- ticity parameter, and relative industry size were entered separately in the regressions in conjunction with the same other independent variables. The relative scale effect, embodying both relative size and scale was found to perform better than either size or scale alone. 66 This is strong empirical evidence, however, the theoretical basis (that the entire output of a nation's industry is produced in a single plant) is questionable. Yet, the finding that industry size and the average employment size of the industry's plants are highly correlated alows N3/N? to be interpreted as a proxy for rela- tive average plant size, thus imparting stronger economic signifi- cance to Katrak's findings. 4.3 Conclusion The most general conclusion based upon empirical evidence is that size or relative size (industry or plant) is not a sufficient criterion by which to measure scale economies. A measure of the scale intensity of industries is essential. If the scale elasticity parameter is to be used, it ought to be in conjunction with a measure- ment of relative plant size. For tests performed in the aggregate, (such as Hufbauer's) market size may serve as a proxy for plant size due to the empirical relationship between the two measures (6). Nonetheless, it seems desirable to explicitly incorporate relative plant size by following the procedure established by Katrak. This test form creates the best direct linkage between the theory and the empirical test. 10. 11. REFERENCES Baldwin, R. "Determinants of the Commodity Structure of U. S. Trade." American Economic Review (March 1971): 126-46. Branson, W., and Junz, Helen. "Trends in U. S. Trade and Com- parative Advantage." Brookings Papers on Economic Activ- jty_(1971): 285-346. . "U. 5. Comparative Advantage: Some Further Results." Brookings Papers on Economic Activity (1971): 754-9. Caves,FL; Khalilzadeh-Shirazi, J.; and Porter, M. "Scale Econ- omics in Statistical Analysis of Market Power." Review of Economics and Statistics (May 1975): 133-140. Hufbauer, G. C. Synthetic Materials and the Theory of Inter- national Trade. Cambridge, Mass.: Howard Printing Press, 1966. ' "Factor Endowments, National Size, and Changing Tech- nology: Their Impact on the Commodity Composition of Trade in Manufactured Goods." In The Technology Factor in International Trade. Edited by R. Vernon. New York: Columbia University Press, 1970. Jones, R. "Variable Returns to Scale in General Equilibrium Theory." International Economic Review (October 1968): 261-72. Katrak, H. "Human Skills R and D and Scale Economies in the Exports of the United Kingdom and the United States." Oxford Economic ngers (November 1973): 337-60. Keesing, S. 8. "Population and Industrial Development: Some - Evidence from Trade Patterns." American Economic Review (June 1968): 448-455. Kemp, M. C. The Pure Theory of International Trade and Invest- ment. New Jersey: Prentice Hall, 1969. Matthews, R. C. O. "Reciprocal Demand and Increasing Returns." Review of Economic Studies (February 1950): 149-58. 67 12. 13. 14. 15. 16. 17. 68 Meade, J. E. A Geometry of International Trade. London: George Allen & Unwin Ltd, 1952. Chapter V. Ohlin, B. Interregional and International Trade. Rev. Edition. Cambridge, 1967. Owen, N. "Scale Economics in the EEC." European Economic Review (July 1976): 144-63. Pryor, F. L. "The Size of Production Establishments in Manu- facturing." Economic Journal (June 1972): 657-66. Scherer, F. M. "The Determinants of Industrial Plant Sizes in six Nations." Review of Economics and Statistics (May 1973): 135-45. Weiser, L., and Jay, Keith. "Determinants of the Commodity Structure of U. S. Trade: Comment." American Economic Review (June 1972): 459-66. CHAPTER V AN EMPIRICAL TEST OF THE SCALE ECONOMY THEORY 5.1 Introduction In this chapter two tests of the scale economy theory are performed. The first test uses the concept of scale elasticity para- meters introduced by Hufbauer (3) in conjunction with input-putput analysis to assess the trade patterns of nineteen countries. The second employs multiple regression analysis to inspect the scale economy hypothesis for U. S. trade. The analysis is limited to trade in manufactures. 5.2 Methodology, Scale elasticity parameters are utilized to measure the extent of scale economies in each industry. The data are from the recently completed 1972 Census of Manufactures. This census reports the relevant data by the employment size class of establishments. The value added and employment statistics are arranged in employment size classes for establishments ranging in size from one to four employees up to 2,500 (plus) employees. The four digit Standard Industrial Classification (SIC) constitutes the level of disaggrega- tion generally available. Over 300 manufacturing industries were utilized in the analysis. The scale elasticity parameter, a is defined by the following equation: 69 70 v. = kn? (5-1) where: Vi is the value added per worker in the ith class size, "i is the average employment size of establishments in the it class size, k is the constant. The regression equation which was estimated is, .. = . + . .. + .. - lnv1J 1nkJ aJlnn1J en (5 2) where: eij is the error term. This equation was estimated across establishment class sizes, 1, for each SIC industry, j. Use of the scale elasticity parameters implies that increases in value added per worker due to increased plant size are passed on in the form of lower prices. However, other factors, unaccounted for in (5-2), affect output per worker; therefore, the estimates of the scale elasticity parameter may pick up the effects of these omitted variables. Possible sources of bias are:1 1. Heterogeneous product mix. Within a given four digit industry different plants may produce different products. If rela- tively skill intensive or capital intensive products are associated with large p1ants,& is biased upward. If the association is with smaller plants, 6 is biased downward. 2. VaryingyFactor Proportions. Among plants producing the same product, different qualities of labor or different mixes of 1Presented in Hufbauer (3). 71 capital to labor may be systematically associated with plant size. If skilled labor intensiveness and capital intensiveness are asso- ciated with large plants, 3 is biased upwards. If the association is with smaller plants, the bias is downward. 3. Technology. If larger plants tend to be newer plants, A a will reflect the effects of improved technology and overstate the measured scale effect. 4. Market Power. To the extent that market power may affect the analysis, it will impart an upward bias to 6,as market power is derived from size. The estimated values of on. are concorded to the input-output classification and weighted by the employment size of each industry in the 1-0 sector in order to get one scale measure (aw) to repre- sent the sector. However, according to the input-output relations specified by the table, each industry absorbs a portion of the output of other industries to utilize as inputs}2 To take account of this, it was assumed that any scale economy benefits are passed on to the consuming industry. Therefore, the awj's (j = input-output sector) are weighted by the elements of the total requirements matrix, rij’ thus: O'tj = §(“wi ' rij)/§rij (5-3) 2The use of an input-output table to test a scale economy theory implies a basic contradiction. Here, the input-output table is viewed as a tool which measures the interrelationships of indus- tries at a point in time. Curvilinear isoquants are assumed to exist; the input-output table identifies a point on each isoquant for each industry. 72 measures the total scale economy benefits enjoyed by industry j. Con- sidering a balanced export and import bundle, the tendency of nations to have an advantage or disadvantage in scale intensive products is measured by: (5-4) where, x. is the value of exports of commodity j in a million dollar export bundle m. is the value of imports of commodity j in a million dollar import bundle ' ' Thus, RS measures the scale content of imports relative to exports. 5.3 Empirical Test of the Scale Economy Theory: Nineteen Countries Lacking plant size data across the sample countries, it will be assumed that average plant size is larger and therefore, scale economy benefits are greater, the larger is the domestic market.3 National manufacturing employment is the best index of market size by which to test the theory. Value added in manufacturing may also be used; however, it provides an inferior test as it contains the scale effect which the theory explains. Table 5.1 presents the scale con- tent index and several relevant national characteristics. The corre- lation between manufacturing employment and this index is positive 3Pryor (6) has found that the average employment sizes of manufacturing establishments are positively correlated with indica- tors of market size. 73 op mmm._ mp me m N_¢.¢ 0 com. ccmpmmN 3mz m mNN.¢F Pp eNN.— m mmm.m m— Nmm.N m_~mcum:< mcmwuo NF Nmo.m op mmm.p up NmP.P mp mom.~ mw>mpmom=> p— cmm.mp m mmo.N ep mNm.P N Fem. :mmam maOcsm cmzpo N— mmN._ mp Npm mp mN—.N up NNm._ acmFmgH mF NNN.m mp oNe w 0No.m mp mmm.p meEcmo o— mop.¢~ m_ mm_._ o om¢.m m mom. mesonswx34 nszwmpmm o emu.¢m N NNm.m NP mou.N m mmc. megH a Nqo.nw m N¢N.m w Nmo.m v eNm. mucus; m Noe.mp ¢~ omo.~ N mo_.m op cpN._ mucmpcmsumz N omm.mmp m FNo.m m mm_.o N owe. Xamscmw ummz m omm.Nm a mom.n —e mmm.m m Nam. soumcwx umywmp UNN m www.mop N mam.op o_ Nm_.¢ _ mne. :mamw . . M “We...“ a. ma. gamma oom.No _ NNN.m . . . P e P um no; :2 Aooo.ooov» meme» eae Aooov Aooov a xcmm mcwgapumwacwz xcmm acmezo_qsu xcmm new m N cw umuu< w3~m> N mcwczuummacmz N muwamu can xcmm a Acpcaou mowum_cmpumgmsu Pmcowpmz — Awnmch mmczuommacmz mumpv muwumwgmuumcmnu ~mcowpmz op mawgmcowwmpmm agocoum mpmum--.P.m N4mcou Nucmcgau __< "Neoz xoonsmm> Pmuwpmwuwpm .z .2 "mumzom Apo.v Apo.v Am_.v mocmuwwwcmwm NNN.+ www.+ mmm.+ mm can mowpmwcwpumgmcu :mepwm cowumescou scam m.cmscmmam mp omm NF NNe mp ¢m_ NF mmm.¢ :mumwxwm up Pmp.e m mom.¢ mp oN. NP N—m.p mwccH mp omm.N NP um_._ Np eom NP ome.e mmgog mp mum my «Fe m_ mom.~ op mo_.N mcox ace: owm< Aooo.ooovw meme» :65 Aooov Aooov m Nana mcwcauummzcmz xcmm “cosmopaEN xcmm mow m N cw cmnu< m:_m> N mcwcapumuzcmz N mu_amo can xcmm m ngczou muwumwcmpumgmsu pmcowumz emacwpcou--._.m ugm m>wpmmmc mumcwswpm op cmngo cw mco an cmucwsmsm we: cmumEmLma m_mum comm .cowmmmgmmc mnu cw vmcmucm can: .cowmmmgmmc any cw new: wpnmwcm> Naocoum mpmum Pmauoo as“ nonwgommu mpnmwcm> ucmucmamu on“ on uxm: cowumuo: asbp ._e>ep Nm age we peauaeaem_me .mmmmcpcmgma cw cabana muwumwpmum u och .mucow Loam. map No mcowpgoaoca mm uwcmucm ago manmwcm> Ppwxm mg» .msmppoc Lo mccmmaosu cw uwgammms mucoaxm um: .m .2 m? Azixv mmmwcumzccw mmocom mwpmcums< use .cmamw .mcmcmu m:_a mmmgpcsoo Amy umm asp mo mucoaxw us» an vmvm>wu magoaxm .m .2 mm Axw\xv ”Amim cowpmzcmv mpnmp o-H as» »n umpgmwmz mgammws «FmUm w>wumpmg mg» mw m .Aeim cowumzcmv mgmumsmcmn zuwuwummpm mpmum umusmwmz 01H mg» mm ac mgmmemgmn zpwuwummpm mqum vmusmwmz acme izopgso asp .azc m? 35 mews: mpamwgm> ucmvcwawv ms“ :0 pawLUmpzm use an umHmUPvcw m? co_mmmcmmg msu cw vow: mpnmwgm> m—mum one .vmummu mam wpamwgm> aeocoum mpmum mcp we mwgammms ngm>mm "whoz ANN.V Aaa.v Lea.v AmN.. Amm.v .Aem.mv Aao.v ea e_a.m .N. ma._- NN.mm- ou.m~ mm.ap NN._- N_.mm mm. - Axw\xv o_-m Aam.v Amm.v ANa.v Aas.v Amm.v «Amm.mv Aao.v m kNN.N Pm. No.N- m_.am- _m.a_ mN.mP mN.P- N6.Nm am. Axw\xv m-m ANa.v AON.PV ANa._v Aoa.v eamo.mv eAON.Nv AN_.FV ea eON.m am. N_o. - maF. - ~_P. amo. - mac. - emu. _mo. 52-xv m-m ANo.v Amm.v Ama.v AMN.V Ao~.V efiam.mv Apm.v d eou.~ om. mm. - _m.mm- om.mF Na.ap aa._- mm.nm m_.N- a Axw\xv N-m mup>cmm mFQmem> Loam; o» . mgmmcwmcu u a F Asocoum meuamo mcmcmmmg mgmmumm mm>wumcmno .ucm accumcou wwnmwmm> N apaam to eaeam a statue“ memwpeaaam a a an Ammo, upto: 6:6 gap; teeth mmmcapummzcmz .m .zv Ngomgh Naocoum mpmum asp No amok :owmmmgmmm «Fawupzzii.N.m m4m a2 and b1 > b2, although a1 and a2 are statistically indis- tinguishable. For equation 6-1 R2 = .235, for equation 6-2 R2 = .503. These results are fully consistent with neither orthodoxy nor Linder although they are partially consistent with both. Hufbauer concludes that each theory (orthodox and Linder) can be assigned a sphere of influence. He states, "judging solely from the cosine exercise, Linder . . . works best in accounting for 90 trade withjp_the rich country zone. By the same token, orthodoxy . . does better at explaining the commodity composition of manu- factures withjp_the poor country zone. As for trade between zones, the cosine results agree with Linder if the zones are close together, and with orthodoxy when the zones are widely separated" (5, p. 205). If by "cosine exercise" Hufbauer is referring to the regres- sion analysis he receives no support for these statements. He has estimated only two regressions across the sample countries. The regressions capture the effects of trading "upstream" and “downstream“ regardless of the level of per capita GDP in a particular country. Thus, one cannot reach different conclusions for rich versus poor countries based on the regression results. His tests imply that for a given country i (at any level of per capita GDP) the export composition of poorer countries becomes more similar to the import composition of a given country i as their per capita GDP's approach that of country i. This is as Linder predicts. However, as country i's import composition is compared to the export composition of richer countries, the two vectors continue to become more similar, but at a decreasing rate (b2 < b1). A similar finding is made for any given country's export pattern (compared to other countries' import patterns) except a2 < a1 is not a statement for which there is statistically significant support. There is no way to know for certain how Hufbauer reached the conclusions which he did. However, simply viewing the matrix of CoinMj values (4, pp. 224-26) leads one to somewhat similar, 91 although admittedly tentative, conclusions. For example, the values of CoinMj are higher when countries i and j are "rich" than when they are "poor." This would be consistent with the hypothesis that Linder's theory explains trade within the rich country zone. Further speculation is possible but not warranted because no other general- izations seem as clear. If Hufbauer's conclusions (whatever their basis) are correct, then his test procedure is wrong. As it was pointed out earlier in this chapter, Linder's hypothesis of trade intensity does not receive support across all countries. However, trade intensity does seem to increase with demand similarity for countries which are rela- tively rich.2 When Hufbauer reaches a similar conclusion for preference similarity, he effectively destroys his own empirical analysis which inspected the aggregated effect across a pooled set of countries. Yet Hufbauer's conclusions are attractive. The gains from trade according to Linder's theory would seem to be greatest for relatively rich countries. The potential to increase the menu of choices available to consumers, or to eliminate monopoly returns to technological advances, are relatively more important considerations when there is a greater amount of discre- tionary income. Furthermore, Linder's theory, as he applies it to poorer countries, is purely static. If the demonstration effect operates or if a relatively poor country is trying to develop, its 2See Sailors (7) and Fortune (2). Both articles are briefly reviewed above. 92 import pattern would be more similar to that of a richer country. In order to properly address these issues, countries must be tested individually, not collectively. 6.3 Some Empirical Considerations The basic testing procedure of Linder's hypothesis is similar regardless of whether it is a test of preference similarity or trade intensity. In each case the researcher establishes the object country's level of per capita income (or GDP) as the critical point at which something should occur (maximum trade intensity or peak similarity). Although this is what Linder predicts and therefore it is a reasonable test of the theory, it is entirely possible that no particular change occurs at that specified point. Furthermore, given the very loose definition of representative demand, the choice of a specific point as the center of the range of representative demand is too confining. The tests performed by Fortune (2) and Sailors (7) have an additional undesirable property. Each of these studies employs the absolute value of the difference in per capita income as an inde- pendent variable. This prohibits a separate consideration of up- tream and downstream trade patterns. Although Hufbauer used aggregated data, he found Linder's hypothesis of preference simi- larity to be corroborated for downstream trade, but not for upstream trade. Fortune and Sailors' method does not even allow inspection of this possibility. Furthermore, as countries become richer, there are fewer upstream observations. Similarly, as countries become 93 poorer there are fewer downstream observations. Therefore, Splitting the sample at a critical level of per capita income, to separately study the differences of upstream and downstream trade (if there are differences) can only be done for countries in the middle of the sample. Even for these, the number of degrees of freedom, roughly, is cut in half. Thus, if these effects are to be separately consid-. ered, a new type of test must be devised. REFERENCES Bhagwati, J. "The Pure Theory of International Trade: A Survey." In Surveys of Economic Theory, Growth, and Development. New York: 1965, 2, pp. 173—5. Fortune, J. "Some Determinants of Trade in Finished Manu- factures." Swedish Journal of Economics (September 1971): 311-317. Hoftyzer, J. "Empirical Verification of Linder's Trade Thesis: Comment." Southern Economic Journal (April 1975): 694-698. Hufbauer, G. C. "Factor Endowments, National Size, and Changing Technology: Their Impact on the Commodity Composition of Trade in Manufactured Goods.“ In The Technology Factor in International Trade." Edited by R. Vernon. New York: Columbia University Press, 1970. Linder, S. B. An Essay on Trade and Transformation. Stockholm, 1961. Linneman, H. An Econometric Study of International Trade Flows. Amsterdam, 1966. Sailors, J.; Qureshi, U.: and Cross, E. "Empirical Verification of Linder's Trade Thesis." Southern Economic Journal (October 1973): 262-268. "Empirical Verification of Linder's Trade Thesis: Reply." Southern Economic Journal (April 1975): 698- 700. 94 CHAPTER VII AN EMPIRICAL EXAMINATION OF LINDER'S PREFERENCE SIMILARITY HYPOTHESIS 7.1 Introduction According to the Linder hypothesis nations which have the smallest per capita income differences will tend to exchange products which are highly similar. Between two countries as per capita income differences increase both the volume of trade and its similarity will diminish. A nation's exports are, therefore, similar to its imports and to the imports of countries with similar per capita incomes. In this chapter we will test the preference similarity aspect of Linder's hypothesis. 7.2 Methodology The 1975 exports and imports of twenty-six countries and their respective per capita gross domestic products (GDP) constitute the data set. The GOP data are from the U. N. (l) and are almost entirely of 1974 vintage. Where 1974 GDP statistics were not avail- able, the most current year1 was used by adjusting it to the 1974 dollars by means of U.N. G.D.P. deflators for developing and developed 1This is not a major problem. India's GDP is the least current (1972). 95 96 economies (l). The data used are for 102 separate manufacturing industries; SITC's 5, 6, 7, and 8, at the three digit level for trade with the world. A country's exports and imports may be expressed as vectors in which each element of the vector is the percentage value of the particular SITC in the given vector of manufactures trade. Defining exports and imports in this way, country i's export vector Xi is comprised of elements xi", where n denotes commodity n. Similarly, country i's import composition is represented by Mi where "in is the percentage of that country's total manufacturing imports of commodity n. The cosine of vectors Xi and Mj provide an index of the similarity of two nation's trade.2 2X4 0 M. n in jn /_' 2 2 §(Xin) ' §(M jn) CoinM. = (7-1) When CoinMj equals one, the two vectors are identical. When the cosine equals zero, they are completely dissimilar. The similarity functions for which theory suggests estimation are, CoinMj = a0 + 316i + azdg + ui (7-2) for j = k; 1 = l, 2, . . .. 26. ”i is the error term. 2This measure has been used by Hufbauer (3) and Linneman (5). The following discussion is based on these sources plus R.G.D. Allen (1). 97 _ 2 CoinMj - bo + b16j + bsz + “j (7-3) for j = 1, 2, . . ., 26; i = k where Gi and Gj are respectively the per capita GDP's of the ith and jth countries, uj is the error term. The correspondence between these functions and Hufbauer's tent functions (3) is unmistakable, although there are important differences. First, the similarity functions can be estimated for each individual country; no aggregation is required. Second, the functions are not constrained to reach a critical point at a specific level of per capita GDP. The critical point is estimated where the function best fits the data. Linder predicts that the critical point will be a maximum. Considering equation 7-2, this implies that a nation with the same level of per capita GDP as nation j (Gj) will have an export pattern which is most similar to j's import pattern. Turning to equation 7-3, Linder predicts that country i's export pattern will be most similar to the import pattern of a country experiencing the same level of per capita GDP as i, thus 7-3 should attain a maximum at Gi' However, equations 7-2 and 7-3 cannot be estimated directly. Since the dependent variable is defined over the range from zero to one, this constraint must be included in the specification. The following logistic model incorporates this restriction, CoinMJ. = l 2 (7-4) a + a G. + a G. + u. 1 + e o l 1 2 l l 98 where the variables are defined the same as in equation (7-2). This function can be estimated in the following form, 1n(————1———-1)=a +aG +aG2+u (7-5) CoinMj o l i 2 i i This form restricts CoinMj to the range between zero and one. Given this restriction the dependent variable in the equation is defined over the interval, (- ”7 m). Therefore, the necessary econometric assumptions are satisfied. However, it is no longer clear that this Specification lends itself to a test of Linder's hypothesis. In order to constitute a proper test, equation 7-4 must be capable of attaining an interior global maximum over the range of possible values of G. Allowing the exponent of "e" to have a quadriatic term admits the possibility of a maximum, a minimum, or an inflection point. The point at which the critical value occurs is easily located by taking the first derivative of the function, setting it equal to zero and solving for G. This is done below omitting the subscripts i, j. a + aIG + a G2 + u dCosXM o 2 _ -(a1 + 2a2G) e de - a + alG + asz + u 2 (7'6) (1+e ) However, the denominator is positive, and since "e" to any power is positive, 99 dCosXM = dG 0 when -(a1 + 2a2G) = 0 Therefore, the critical value of the function is reached where Next, we must evaluate whether this is a maximum, minimum, or point of inflection. This requires taking the second derivative and sub- stituting 'al for G. In order to simplify this procedure define, 2a2 f(G) = -(a1 + ZazG) 2 a + a1G + azG + u 2 h(G) = (l + e I 2 a0 + a]G + azG + u 9(6) = e Using this notation,equation (7-6) can be expressed as, dCosXM = f G G dG h G Then, collecting common terms, the second derivative of equation (7-4) is 100 2 d CoinM. = f’(G q“; + [f(G)]2 2 h(G) ' dG 2 ao+aIG+azG +u E-hLG) + 2(1+e [h(G)]2 31(6)] (7-7) where, f’(G) = -2a2, the other terms are defined above. In equa- a +alG+aGZ+u tion (7-7) h(G), [f(G)]2 and (1 + e ° 2 ) are each greater than zero; f’(G) can be positive or negative. The expres- sion (7-7) must be evaluated by substituting -a]/2a2 for G in order to determine the nature of the critical point. And, since this is not a simple quadratic function, we must inspect the possibility that an estimated maximum (or minimum) is not global. Therefore, this particular specification (equation 7-4) does not easily lend itself to a test of Linder's hypothesis. Incorporating the restriction that o < CoinMj < 1, does not produce a function with desirable properties; therefore, let us look at the nature of the problem when the restriction is ignored. The entire cosine distribution is "piled up" between zero and one, and E(CoinMj) is not restricted to values within that interval. From an operational point of view, this is not much of a problem unless many observations lie near the extremes of the interval Specified above. Table 7.1 presents the values of CoinMj. Only 6% of the total observations are found to fall within the two 10% tails [i.e. P(.l < CosXM < .9) = .94]. Although there are relatively 101 few extreme observations, each may be relatively important. Thus, the relative scarcity of extreme observations is a necessary, but not a sufficient condition to allow direct estimation of equations (7-2) and (7-3) using OLS. Since separate regressions are to be run for each country, we are interested in the concentration of extreme observations on a country level basis. Considering equation (7-2), we find that the following countries have the greatest concentration of data points in the two 10% tails of the cosine distribution; France (39%), West Germany (31%), Pakistan (27%), Hong Kong (23%), the United Kingdom (15%), and Canada (8%). Korea, Italy, and Israel have one observa- tion (4%) in the tails; the remaining countries have none. The cosine values which apply to equation (7-3) are more concentrated in the tails of the distribution for Canada (19%) than any other country. The next highest concentration is for Sweden, Australia, Finland, and New Zealand (12%). For the remaining countries, we find that, ten have two observations in the tails (8%), four countries have one (4%), and seven countries have none. These casual observations imply that the estimates of equa- tion (7-2) are the least reliable for France, West Germany, Pakistan, Hong Kong, and the United Kingdom. For the remaining countries the problem does not seem serious. The estimates of equation (7-3) are generally less affected as only Canada has a rather large proportion of observations in the tails of the cosine distribution. Although objective skepticism is warranted, there is no evidence that the 102 other regressions will suffer due to the truncated distribution of CosXM. These conclusions allow us to retain the quadratic specifi- cation which Linder's theory addresses. The quadratic form also is suited to test the Linder versus orthodoxy controversy. Linder predicts that each function will attain a maximum where orthodoxy predicts that it will attain a minimum. Yet the test itself does not assure that either theory will receive support even if all the coefficients are significant. The test is "independent" of either theory, but suitable for evalua- ting both. To clarify this point, refer to Figure 7.1. The figure CoinMj A B Gc D. (D per capita GDP Figure 7.l.--The Similarity Function. depicts a hypothetical similarity function as it might be estimated by equation 7.2.3 Thus it describes how the similarity of country j's imports and other countries' exports changes as those other 3The interpretation for equation 7-3 is analagous, except that the object country's export vector is compared to the import vectors of other countries. 103 countries are richer or poorer than country j. In order to interpret the test results we must know the per capita GDP level in country j and find the level of per capita GDP at which the similarity function peaks (as it does in the example). Suppose the level of per capita GDPikn~country j (Gj) is equal to GB. Then the figure represents the case in which Linder is strictly supported. Had the similarity function attained a minimum at GB (where Gj = G3) the figure would have shown support for orthodoxy. Orthodox trade theory predicts that a country's import vector will be most similar to the export vectors of countries which are most dissimilar in terms of per capita GDP. However, this test does not restrict the similarity function to attain its critical value at (or near) Gj' But, even when Gj is not close to GB, the results may be in favor of either orthodoxy or , Linder. Suppose, in Figure 7.1, country j is the poorest country in the world, with per capita GDP equal to GA and further, that for no country in world does per capita GDP exceed GB. Then the figure depicts an orthodox result. The "downturn" in the quadratic is mean- ingless as no countries have levels of per capita GDP greater than GB. Next, assume that per capita GDP varies across countries from G to G . The critical value of the function may be attained either A C within this range or not. When it falls outside of the observable range of per capita GDP, the sample from which the estimate came only provides information on the slope and convexity of the similarity 104 function; the critical value is not meaningful since there are no countries which are that rich (or poor). When the object country's per capita GDP is not close to GB (and GB lies within the observable range) neither orthodoxy nor Linder receives strict support. The Specification of equations 7-2 and 7-3 differs substan- tially from previous tests of the Linder hypothesis. The equations do not impose symmetry with respect to a partciular level of per capita GDP. The test procedures of Fortune and Sailors (see Chapter VI) impose symmetry on upstream and downstream trade patterns. Thus, their tests of Linder's trade intensity hypothesis produce a single estimate of upstream and downstream trade patterns. If Linder (or orthodoxy) is correct their procedure is perfectly valid. However, if for upstream trade, trade intensity reacts differently to per capita GDP differences as compared to downstream trade, the imposi- tion of symmetry aggregates dissimilar effects together. In fact, Hufbauer's regressions which explain preference similarity imply this result, but his data are aggregated across countries. The quadratic form proposed here does not have these shortcomings. The test is considered superior because: it allows separate consideration of upstream and downstream trade; separate consideration of individual countries; and does not require the researcher to split the sample at a specified point (see Chapter XI). Although symmetry is still imposed on the relationship between similarity and per capita GDP, the regression, not the researcher, determines where this occurs. Next, we must develOp a test statistic which can be used to evaluate the theory. As we have established above, we are 105 interested in two results: (1) whether the similarity function attains a maximum or a minimum, and (2) the level of per capita GDP at which the critical value occurs in comparison with a country's own level of per capita GDP. The point at which the critical value occurs is found by taking the first derivative of the similarity function, setting it equal to zero and solving for G. For equation 7-2, the estimated critical value of the function occurs at: > I D) > _..| C) I 11 (7-8) 5‘ N 01) For equation 7-2 the estimated critical value occurs at, 0 U) ._o G = (7-9) )(I N 0') In equation 7-2 the object country is country j. Therefore, we are intersted in comparing Gfi with Gj; define the test statistic as, Gfi-G. Rn ='_TGTTJ'° 100 (7-10) J In equation 7-3 the object country is country i, define the test statistic as, (2)-(hiri- 100 (7-11) 106 Each of these test statistics measures the percentage differ- ence between the estimated point at which the critical value occurs (Gfi, 6;) and the predicted point (Gj, Gi)' If Rm (RR) is close to zero and the function attains a maximum, Linder receives support; if the function attains a minimum orthodoxy is supported. If Rm and RR are not close to zero, or within a reasonable range, neither theory is supported. However, in this case a country's placement in the sample of countries ranked by per capita GDP may provide some clues to the proper interpretation. 7.3 Empirical Evidence Table 7.1 presents the cosine coefficients between each nation's manufacturing export and import vectors. For convenience the countries are ordered by per capita GDP. For each column, the country listed across the top of the table has its export vector held constant while "comparing" it to the import vectors of other countries. For the rows, the import vector of the country listed at the left has been held constant while export vectors of the countries named at the head of the table vary along the row. The values of per capita GDP were obtained by converting local currency units to dollars, using 1975 mid-point exchange rates. This procedure has been found to undervalue the Per capita income levels of less developed countries (2), making per capita income differences between developed and developing countries appear to be greater than they actually are. If Linder's theory is correct, and if upstream and downstream trade patterns are symmetric, this 107 MNN—. ovop. wmpc. Nopo. Nch. oNoN. N_NN. .Nom. oove. Novv. «NNN. opmm. oon. sou:— moNn. opom. woos. NNco. ovFN. MNoo. Noon. .ooo. ooNN. pomo. mcvo. o_NN. moeo. ecumFxma NNPN. opoN. euro. Npom. Nomo. ooov. ooov. oNpN. _oNo. pmoo. NcoN. mNNo. oNNo. uocox mmoN. nsmN. ammo. Noou. NmFN. coco. pme. moms. oNom. Nvoo. some. .mos. mooo. a_>mpm -o a» omNm. mowN. oF—m. omns. oNNm. coco. oocN. NNNo. comm. oooo. ommo. No_o. omoo. fiamaucoa mwoN. NooN. Paco. ammo. o_No. o_oN. oPoP. mpov. omvc. ovov. .m.N. Noam. omNo. acox mcox quNN. NomN. omen. opmN. oooo. «poo. oNoN. one“. ocNo. Noon. omNm. moan. NNoN. :_eao opoo. oomc. moow. m_om. coNN. oooo. wooN. pooo. moNo. cook. Nooo. oooo. oooo. teapot— om_m. mooN. no_m. OMNo. mmom. q_oo. ommN. oooo. oomo. N_No. omoo. ono. Nomo. x_ou_ omom. oNov. eNmN. OMFN. ooNN. mono. oNoN. pooo. NooN.. «moN. oooo. muon. mNmo. Eonoc_x weave: mooN. o_NN. memo. mono. oooN. NNNc. omen. NNoo. osmo. ompo. o_oo. .ooo. moon. poocm_ oNom. o_oN. p_co. NNqN. oNoo. Nome. oooN. oopo. ooco. mooo. omNm. ooNo. NmNo. comma eNom. ooNN. NNNN. mmoo. ooNo. omoc. oNN_. ompv. —moo. oomN. mNoN. Nmoo. nmmm. mwcum3< oNNN. FNNN. Nona. ommN. NNoN. memo. NNNN. Noco. osem. ova. acoo. oon. ammo. new—moN 3mz coon. oopm. momm. oNoN. moms. ooNo. NPNv. ONoo. ammo. ooNo. em_o. evoo. ammo. ocu_c_m moon. ocom. «New. poem. _Nmo. moNo. eNmm. oops. oopo. m_oo. omNo. NmNo. pooo. mucocg moum. opoc. mm—m. MNom. om_o. meow. oooN. Nooo. NNmo. mon. oooo. pmpo. o_mo. mu:m_gwauwz morn. Nmom. MPMm. NNoo. Nooo. ONPo. pomp. NNoo. pvoo. ono. QFPm. mums. mmoo. mxzo 1.5 _ Tom .mom. NNom. mNmm. amps. oNoN. _oNc. comm. NNNo. ooNa. Noam. Nmon. Nooo. mo_o. um_oeum:< _NMN. m—ce. mmoo. mooo. ooNo. e_¢q. —oNo. opqo. meoo. mNco. coco. NmmN. ooom. xezeoz nopo. comm. «mum. Nmoo. moon. opNo. oooo. Nous. ooNo. omNN. coco. ammo. aomc. Newscmo moon. Nch. onco. NMNN. mooo. moov. oocN. NFNo. nooN. ooNN. oooo. vaN. omoc. Acostwo puma omoN. oNoN. oooo. come. Nmmo. omvo. moo_. mem. Npmo. "NNN. oo—o. «NNN. oooN. mmumuo o¢u_:: onmp. coop. NNmm. moee. NNoN. oooc. ooNp. opom. m_om. evvm. oon. coon. oFNN. «cozmu eNpm. Mmcm. .Nmo. oNNN. m_mN. ¢_qo. oon. moon. mama. ammo. ocmo. omow. oooo. cmomzo oNpm. mmmm. snow. NNoN. omNo. memo. noo_. ocoo. mono. —ooN. oooo. moNN. coco. ucmp uLmNumzo : Ac m a uc~_nwm van—c.u oucmcu moan—cmnumz -anoflom a_—ocum:< aazcoz Newscmo mfimww aww_ww moncmu :wuozo ucmFLaNuwxm mopguczoo ucwugoasm momma op~o u_m_o munch .mcooum> uco;s~ new ucoaxo pocowuuz coozuuo mwapa> mm_cucaou mcvucoaxo III 9.381.: 39: 108 .mcocoaoo upon man” as» :. macoaxm upco: on you» no toga musoaxu omega uazu ccwe=mma no uuc~.cuaxu mmc_ m.—occm_ soc» cmuasmumo no: «can ucocxc cameraEmc on» up~m Lo—au—uLoq a c. mucoaxw m.PmoLm~ can: .cw>mxo: .ANV :. vouconmc ago: has» .zczoco omcop «no: .z.c ecu soc; opoapwa>oca mo: poocmc Lo» mono ucoaxo ommp Aummch cu umwscwgc acc aumamu con an .z.c .muvumcuaum Amoach umaoccczv p umcwcco mo'gucsou "who: munch xu_cosEoc "ocmcco o_~—. chc. mooc. cccm. om_N. cmNc. Nch. Nm_m. poom. .ch. «cow. oomm. omoc. o_c=~ och. chF. emv—. Ncco. chN. ocvc. _coo. mNov. coco. ocNm. copp. Ncoo. mcoN. ccumpxac mooN. Nocc. NcmN. Nmoo. och. coco. como. NNNe. cho. cNoo. cmep. co_o. coco. owcox oNVN. pccp. voo—. «moo. NooN. onc. ooNo. oocv. macs. ocNm. Ncpp. mcoo. ammo. mwma_m . 03> .omN. chfi. __cN. como. cNNm. mom_. coco. oomo. eooc. Pocc. oqNN. _Nmm. pocc. .omaucoa mNoo. c_ov. cv_¢. oNcm. Ncco. Nch. N_om. cmNo. .ccv. Ncco. och. omcm. N_No. ccox ccoz NoNN. voNc. cvo_. pmcc. ocNN. occp. .v_o. cho. ocNm. chc. Ncmp. oeoo. N_cm. :_aco ocmm. coNN. occv. _N_o. oNNo. cmNm. o_No. oo_m. Nocc. mccm. omop. oNoo. comm. cam—mc_ Nc_m. hoop. ommN. cmco. oocm. coop. memo. _Noo. ocmc. coco. mmmp. ccvm. ocNm. m_uu~ Nch. coop. mmcm. oooo. ommo. ompm. eNco. oc—o. comm. _mcc. comm. ocNo. ~oom. Eovmcou cmo_: comm. ococ. cxm—. ova. cNmm. eNoc. oomo. oo_m. pm_o. Nmmm. poom. .oco. cNNo. —mocm_ omcv. opmp. NNoe. coco. ocNm. .eom. coco. .vmm. Nooo. comm. ccvm. eNme. cNoo. cacao mccv. oNNN. ocmv. mono. moNo. ooom. ovco. —mco. ocNm. ooNN. Pom_. mcmm. ommo. o_cuma< ocNm. Ncop. occp. Nmoo. mopm. ammo. NNNo. comm. comm. occc. Nc—_. moms. Pmom. ccopmmN 3oz ocNm. cch. NcoN. cmoo. c_om. ocN—. _mmm. .ooo. moNc. VNcc. NvN—. oc_c. ech. ucap:_m opmm. cpop. .com. NNoo. cva. cNoN. Nc_m. c_oo. _ch. cmmc. Nom_. cNmm. comm. mucmcu cope. __cp. empo. cmoo. Nooo. .qce. NNco. oNNo. mm_m. Ppoc. oNcp. covm. cc_o. mace—Lozuwz «mom. comp. c_Nm. ooom. cNNo. mmoN. coco. NcNo. oooc. NNFc. «pom. cpmm. cmuo. oxac . usaw .mc Nomm. Noc_. cocN. ocom. NN_v. .mvN. _oco. _coo. cNoc. mc_c. oNv_. pon. ocmm. om_mcom:< pNQN. mch. Nccm. mNcc. comm. mmmN. owes. acme. cho. coco. eocp. cmpc. coco. magnoz onm. Nccfi. cficm. mmvm. che. eNcN. oNNN. momo. FPmo. chc. o_op. NNec. chc. guacamc cpce. ccoN. .ovo. c—No. nooo. mcmo. coco. omen. cNNc. mNNm. cch. Nooo. comm. memFCmc you: cocm. mmNc. ccmN. Npoe. ocNm. o—oN. FcNo. oomm. coco. comm. Noop. cmcm. ccoo. mucouo cmu_cc co_N. m_oc. mmNp. Nccm. NmN—. mocc. moNo. co_m. mmNN. cNNc. eooc. oNoN. come. ocmcoc omom. moo”. NNNe. oomo. oNNv. cccm. _cmm. Ncmo. Nccm. omcc. cNep. cNNm. cmmc. :mumzo mono. omnp. ooNv. cc_o. momo. Nome. ocoo. ccco. ofioc. cmoc. oppm. memo. cmom. vea— -La~uw3o ccox socccmx o_u:— :mumvxau amLox ~m>o_moc:> pmccucoa one: :mocm cam—wc~ Amoco cwumcc ppmmcmu cocoa accum=< mumguczou ccmugoceo mm_gu::oo ccmococxo w “Psflh Eula-TI". VI. r..-l girth”: ..1.“ III vwccmucounu.—.m mcmou goo—N Nmo No «o o.NF No co c.mm om m« o.c« co om N.m_ cm Np N.NNN N« o o.—N— muchaPo o.ooo c. N c.NNN «o «o N.oN _« .m N.Nm «« . Nm o.«« N« cN o.oop «o No N.NN mm__uxou a: one; o.omo m« N« o.oo Nm cm m.oN cN cN ..pcp o« «« c.cN o« co c._c~ aN NN ..o«N mu_caow upwuxmu po_uoao omo m« N« o.oo oN ow N.«_— mm c« o.o« No no m.—N «« «N c.N«N co —« c.N—N Nuoumoceumo womp.m__ap «mo o« No c.m« c N c.NcN «o No «.cN cm om N.om oo mo c.o« mm c« o.ocp muwcooo uwuogucmo co.o.moo cN —N m.« m« No c.cm No co c.c No No o.o « o c.oom mP o_ «.oo— mu_cnmo cmfiooz N.moo co co o.N— «N NN o.oc— Np NN «.cN_ «N am c.cm — N ..Noo c« No m.mo mo—Nuxmu cm>ooz cacao c.moo «N m N.oFN mm cN _.—c o« Nm c.Nc oo co m.«N co co o.oo No co «.oc mu_cnom couuoc Noo N« —o c.mo cN c— N.oc— No co N.oN N N m.NNmN c« c« c.mc_ mN oN c.~«_ mcoo_uu_omgocxmcopco> o._oo mo Nm c.oN c« m« N.m« N« mm N.—m cN oN c.c No on N.NN. co m o.oo. vacuum—n .ccox couuou «..oo _o «« N.No N« N_ c.mm— co No N.m —N .N c.c «o a «.coN oo o« N.mc_ conunwpoca .ccom couoou m..mo NN NN c.N .o Nm o.NN oo mo m.oN oo .o c.c c« « o.oom oo Nm m.oN. scum .oo: N._mo op cN N.o«p c« oo N.oN « « c.NNm om Nm —.Nm mN om «._mp o« _o N.om muuaooca cocoa N«o om NN ..mo o« o« «.c« N — o.Nco « « o._oo «o «o o.«« No NN N.Nm ucoongwaoa o Luann _«o oN —« c.oo o NF o.o«r —N .N c. o. cN N.cm op m« c.ccp o_ N« c.Nc_ among o mwc_u cognac ".mNo. «« No o.m« cN cm o.oo pp p. o.oo— mo «o N.o N — m.«No mo co N.oo muuauoca cognac coguc c.Nno N— pm c.cm mm .o o.pm N Np N.mo_ «m .o «.NN o— cm o.oNp — — o.m«N muuacocc cognac _No No co N.N« oo oo m.o .N _N m.NN_ N« m« ..cN m pp c.«oN N. c. o.No~ mc_xm can m—o «o No _.mm co No o.op o« No o.o« oN cm m.N« Pm mm o.oNP « m N.«NN .coe .meuomc N_o oN NN o.N_— N N m.m«« om «m c._c o« _« N.NN «m cN o.cmp co «o «.oN concomc __o mp a. m.oN, «« mo o.oN Nm m« N.NN om N« «.oN o. oN m.o«p c. Nm m.oN. muuzuocc .cuwewcc moo N« c« c.cN NN oN N.oc pm cN _.mc— oo oo «.mN c« No m.«c cm a. m.o«_ mu_uma~a Foo o m_ N.cm_ «o co o.o m o m.ch —« No N.NN c— o. «.NNN NF m« N.Ncp mwSNmONon NNo o N _.mo_ o— N_ N.Ncp mo «o ..N. N c m.opN co co o.mm mm «o P.oc .mcs .mco~w__ucmu poo NN «m N.cc .o No —.c c« mo «.«« .o co o.NP N m_ o.omp m o_ «.oo. mupuoEmou o Newsaccma «.moo c« «N «.cpp mo co —.NN No cN «.N «o oo o.m mm c. o.NNN c« .N m.o«— menwcon o —_o _o_ucommo _oo mN oN N.Nc No mo m.«N oN om N.mc c« c« ..oN N— «— m.mm_ op cm o.«m. _auwuamuoecaca .«o o« c« N.No c« —« «.c« «N «N «.N—— oN oN N.c« NN c. m.oNN «N o— N.m«— mozm_cco> o muc_am mmo No co c.om F« «m o.«o oo oo c.c— NN NN c.c Nm op c.cN_ om mp c.No. emu upcmmco u_umoucmo _mo Np N. ..mmp cm c« N.oo «m o« m.co «— op m.NN— om mo o.oo NN N« «.cN— m—ou_saco u_c«ccoc~ o.«.m~o NN cN N.—NN cN oN N.oN co .o m.«N mm «o m.oN Nm c« m.oN. _« mm «.o_. upou_so;u umcacco Npo 22. :9. :9. :5: «s. :9. 82. (2. “Se :53. $2. :2 :5: cs”. .5“ Jam :9 ”S”. mccwxcom noucu mocwxcom xmuc~ mmcwxcmm xouc~ occvxcom xmcc_ mmcpxcox xwucu mmcwxcma xmu:~ UNmo mououm vow—cc coaac couozm avocac Eocccmx nou_c: o-ooo ocauco>u< o>wuuuoneou ao~au>u¢--.—.o u4muo .muuayso you monou.o:. . . scat. to. o» sc.g «so moc.x=oc ..< .o.aou as» :. assuage xovc. ..uuoan~c 5.0;» so venue as. mac—acne on» .«-c co.umaao ca oc.vcouua voua.:u.au m. xocc. ocN .mc mm «o m.c« co mo m.c. . N o..N« o. c. ..Nm N. .m m.Nm. o oN o.N«. mczu.cL:N N.o «m om N.NN oo oo ..N. o N o.o«N NN .m N.o« .N .N o..o. NN NN o.c«. o:..omg o cc.oea.m N.m co co N... m m «.coN c. m. m.No. m. N. m.oc. .o .N o.oN Nm oo N.«o muaoo o mc.zo omN . . ..on NN NN c.. mo mo c.c Nm om «.Nm N« No m.oN mo NN o.om , u.ecuc.< «mN cN cm m.Nm om «« o..« «. o. c.c«. c« mo N.c. «. NN c.N«. N mN c.om. mo.uxu.c mmN o« .o o.«m «N mm m.oo c c. o.oo. . . c.cc«m m« cm c.cN N o c.Nm. $63.5 cc... mmwac N.o.N.NmN om m« N.co N o o.c.N oN cm o.oc o N c._cN oo oo o.om «« No N.om um..aosou:< o ...NmN .N mN c.«.. c. 8 o.oo. a. 8 N.oN. c. .. o.oN. MN «m ..Nm. «o .o m.oN «gatorwwoouéflcoc ..NmN c. o. m.om. a. «N N.cm No co «.NN m. c. o..c. oo cN o.NN c. mm o.NN. mm.o_:m> Noz..om .mN N N. m.om. oN om N.No m. N. N.Nm. .m o« m.oN mN o« c.Nc. cN o« c.Nc. «Louocucmc u.cuoo.o ..NNN Nm mm «..N .m om «..o mN cm N.cm Nm o« o.cN o N o.omm «m «« ..oc. Neoc.zooe .uu.cuum.m me.c c.NN c« oo c.«« NN NN m.mc «« «o m.m« «o No o.N. mm N« «.c.. m «. N.co. Acmc.;uos m..rme ..N.N .m mm N.NN mm cm ..No cm mm N.cN cN «m m.om N« «o ..oc c. N. N.«o. Ng~:.nuoe c:.xcox .uumz o.N N « m.mc. N. mN o.oo NN Nm c..c c. o. m.m.. a. mm o.Nm. No cN N..o mcmc.;uae mo.Noc «.N m o o.oN. N. «. «.oN. NN NN c. o m. «.No. N. cm o.Nm. c« oo o.oo account» o.N.N cm Nm «.No No cN c.« o. o. o..«. c c. N.No. m m c.Nc« o« No m.mc Ngwc.guoe .acau.:o.cc< c.N.N N. c m.co. «m Nm c.No cm «« o..N N. «. N.m«. cN mN ..co. oo oo N.NN Neoc.goae ccwuocwcmc buxom ..N - mo N.c« - .m «.«N - cc «.cN - oo c.c - c. m.ooN - N c.NN. occaocz .:.N N.Nco - c. N.oN. - mo N.cm - oo N..m - mN N.NN - N« o.«c. - NN c..«. azcnocx .u:.N N.owo - c« o.oo - No N.«N - .« N.cN - o o.m.« - .o ..om - mN ..o«. unmaocz .cmmc N.oco o. o ..c«. N« o« o.om N. «. N.co. No co c.c. .o oo m.oN mm .m N.NN. uncaocz .e:c.e:.< N.«co « .. m.om. mm o« o.om cN om ..cm m o ..«.m m. Nm c.«m. .N mo N.No “smoocz ..mxo.z N.moo oo mo N.N« co NN m.mc c« o «.oo. No NN c.cN No NN o.oo. co mm c.«.. .gcaocz .meaoc N.Nco om N« c..o m c. o.NN. m« No «.«« NN oN c.co No co o.cN oN «m ...N. mc:.uu.u o mwa.q .mocaN oNo oN cN m.Nm o. o. N.o.. N« co «.o« cN .N m.oN mo .o o.oo «. «N o.o«. .o.cm.me co.uuccumcou max..mm oNo co mo m.mN .N m. N..m. m o m.ooN No No c.c mo oo N.N« Nm cN c.o«. mn.cum o maoo: oNo No No c.o. c. o «.cNN NN mN ..c.. m« cm o._m co NN ..mN m« om c.o.. mumozm amo.a.c.m.mmcm>.£. «No «o oo c.c. .. « N.NNN Nm NN N.«c. co c« m.om No co m.mo N« om «.c.. memo .mmuo o co». mNo oo co o.om . . m.NNo « m c.c.m N. c m.NcN co No N.om .o oo N.mN cac. c.N .No 22 .3“ .42. 36.. is. .5“ 25.. .5“ .5: 35.. .3: £8. :5”. .5”. .5”. :5: :9. .6”. moc.x:ux none. moc.xcc¢ xmnc. mcc.xcom xmcc. mcc.xccx xovc. mc:.x:om xmcc. mc:.xcm¢ xmcc. uN.m mouoao vuu.cc :maoc cwomxo nuacoc soccc.x cmo_:c o-uom . .Iil‘ Y‘all.|".lil 11.-.-.N.," IIII‘ N-n .ULIhJ IND-l.l'llll.l. I umacmucOUiu.—.o mcmdh 129 employed in each industry. Among the remaining top 10 industries, only nickel, tractors, and railway vehicles are not relatively tech- nological. No other particular skilled labor class is highly asso- ciated with the products in which the U. S. has its greatest advantage. However, the U. S. advantage is generally concentrated in products which are skill intensive according to index y] used in Chapter III. Either method of computing RCA identifies the same products as those in which the U. S. has a relative disadvantage. Wool yarn, woolen fabrics, pottery, other woven fabrics, footwear, and blankets are among the lowest ranked products. All of these are relatively intensive in their use of unskilled labor. Ships and boats, univers- als, plates, and sheets, iron and steel bars, and hoops and strips also fall near the bottom of the RCA rankings. Ships and boats utilize skilled blue collar labor more intensively than any other manufacturing industry. The steel industry products are the fourth most intensive in their use of skilled blue collar labor.4 There is little evidence that the EEC as a unit derives an advantage in technologically oriented products. For the most part the RCAI rankings are dominated by products which use Operatives relatively intensively such as materials of rubber, footwear, manu- factures of leather, travel goods and handbags, blankets, and bleached cotton yarn. When the RCAw ranks are computed, travel goods and handbags, blankets, and bleached cotton yarn fall to rankings 4These rankings are from the input-output table sectors used in Chapter III. 130 between 30 and 50. Also,severa1 more technologically oriented products appear near the top of the rankings; textile machinery and metal working machinery. Bicycles and furniture move into the top ten from rankings in the mid 20's. At the other end of the spectrum, the RCAI rankings show the EEC to have a relative disadvantage in aircraft, paper and paper board, office machinery, musical instruments, other rubber articles, other woven fabrics, tractors, pig iron, leather, fur clothing, made up textiles and bodies chasis and frames. When RCAw is com- puted, these rankings are affected, but not in terms of the skill requirements of the products. Two prominent features characterize EEC external export patterns: (1) the lack of importance of technology, except as a determinant of disadvantage, and (2) a concentrated advantage in products which use unskilled and semi-skilled labor relatively intensively, implying a conflict with the export patterns of develop- ing countries. The products in which the United Kingdom has its greatest advantage are not very similar in terms of their skill content. The RCAI rankings show the U. K. to have an advantage in agricultural machinery and other electrical machinery; both products are tech- nologically oriented. Other high ranking products are: other rubber products, other woven fabrics, wool yarn, woolen fabrics, floor cover- ings, and unbleached cotton yarn. All of these use unskilled labor relatively intensively. Pottery and wrought tin, both of which 131 require inputs of skilled blue collar labor intensively, also rank among the U. K.'s top ten products. When these rankings are compared to the RCAw ranks, two very unskilled intensive products (unbleached cotton yarn and wool yarn) drop around 50th. They are replaced by perfume and cosmetics and explosives; the latter product is rela- tively technological.. Similar products are identified as those in which the U. K. has a disadvantage by either RCA index. Fertilizers and synthetic fabrics are the most technologically intensive products. Other low ranking products use skilled blue collar labor rather intensively: universals, plates, and sheets, ships and boats, railway vehicles, tubes and pipes, pig iron, autos, hoops and strips, railway con- struction materials, and iron and steel bars. Travel goods and hand- bags, paper and paper board, and cotton fabric intensively require unskilled labor and rank near the bottom of the RCA scale.a Canada's comparative advantage is derived from an abundance of natural resources. Paper and paperboard, wrought lead, wrought nickel, and fur clothing have high rankings. Automobiles, buses and trucks, and bodies chassis and frames rank high due to the U. S. - Canadian auto agreement whch took effect just prior to the base period of this study. Canada also has an advantage in several tech- nologically oriented products: synthetic yarn, fertilizer, and agricultural machinery. Canada's revealed comparative advantage with respect to the sample countries is very similar to its advantage with respect to the world. 132 The Canadian disadvantage is highly concentrated in textiles. Unbleached and bleached cotton yarn, wool yarn, woolen fabrics, blankets, tulle, lace, and embroidery dominate the lower RCA rankings. Canada also has a disadvantage in several natural resource intensive products: wrought tin, glass, and aluminum. The U. S. has an advantage in the last two. Sweden's comparative advantage is greatest in furniture, paper and paperboard, articles of paper and explosives. Pig iron, ships and boats, hoops and strips, wrought copper, plumbing and heating, apparatus, and buses and trucks are skilled blue collar labor intensive products in which Sweden also has an advantage. High on the comparative advantage scale are other rubber articles, and clothing; both are unskilled labor intensive. When RCAw is used to assess Sweden's comparative advantage, clothing and wrought copper drop far back in the rankings. Little else substantively changes. Sweden's disadvantage is revealed to be about the same by either RCA index. Textile products predominate in the lower rank- ings: woolen fabrics, unbleached cotton yarn, synthetic yarn, wool yarn, and synthetic fabrics. Several specific capital goods also appear: tractors, aircraft, and railroad vehicles. The remaining items are an assortment of non-durable consumer goods and inter- mediate-inputs: perfume and essential oil, musical instruments, jewelry, tires and tubes, fertilizer, synthetic organic dyes, and pottery. Among these items only railway vehicles and tractors require a relatively intensive amount of skilled blue collar labor. The textile products are relatively intensive in unskilled labor. 133 5 is concentrated in steel and in products Japan's advantage which are relatively intensive in their use of steel as an input. Pig iron, iron and steel bars, universals, plates, and sheets, tubes and pipes, and hoops and strips are the top ranking iron and steel products. Japan's advantage in steel is complementary to its advan- tage in automobiles, ship and boat building, and tractors. All of the above named products are produced with a relatively large propor- 6 Japan also has an advantage in tin of skilled blue collar labor. leather, synthetic fabrics, unbleached cotton yarn, musical instru- ments, pottery, and tires and tubes. When the Japanese advantage is assessed relative to the world, unbleached cotton yarn drops substan- tially in rank. Also, scientific medical and optical equipment rises five rankings to 13th. The former product is unskilled intensive; the latter is relatively technological. Japan has a general disadvantage in consumer non-durable goods: fur clothing, fur skins (an input), perfume and cosmetics, perfume and essential oil, footwear, glassware, and jewelry. For the most part, these products embody low Skill labor. Several very 5Japan's advantage in domestic electronic equipment evidently is hidden by the fact that the "other electrical machinery" grouping is the most aggregated commodity classification employed. The input- output sectors break-out radios and TV'S. There, Japan's advantage is clearly shown. 6Judging from U. S. skill coefficients (a tentative judgment given the state of the U. S. steel industry) most of the products' names above are produced in input-output sectors which rank among the top five manufacturing sectors in terms of their intensive use of skilled blue collar labor (Chapter III). Only automobiles and tractors (20th) rank lower than 5th. 134 technologically oriented products also appear among the lowest rank- ings: aircraft, medical and pharmaceutical products, and explosives. Recently Jorgenson and Nishimizu completed a study in which they concluded that the level of technology in the Japanese economy as a whole reached parity with the U. S. four years ago (2). In this study Japan's technological edge is assessed by three different methods: (1) the RCA method of this chapter, (2) input-output analysis using skill classes, and (3) multiple regression analysis. Common to all of these methods is the use of U. S. coefficients. Although this is reasonable, it may be inaccurate for Japan's most important product, steel (15.5% of Japan's total exports). Nonethe- less, in this section it has been shown that there are several other highly technological products in which Japan has a marked disad- vantage: aircraft and drugs. Also, the multiple regression analysis in Chapter IX fails to produce a significant relationship between the percentage of scientists and engineers in an industry and several measures of Japanese export performance. Finally, according to the relative capital endowment rankings, relative skill endowment rank- ings (both Table 9.3), and the revealed skill intensity rankings (Table 3.4), Japan is not superior to the U. S. in its general abun- dance or use of technology, capital, or skilled labor. Instead, Japan's advantage has been built by specialization and focusing its strength in several key sectors (see Chapter IX). This has certain consequences for Japan's future growth that will be addressed later in this chapter. 135 8.4 Comparative Advantage: Changing Patterns Since Balassa first introduced the concept of revealed com- parative advantage, there has been a substantial shift in the RCA rankings among the developed countries. Balassa's study covered the period from 1953 to 1962. Table 8.2 shows the changes in the rank- ings for several products and product groups between the 1953 to 1962 period and the 1967-1975 period. Japan shows the most marked shift among the sample countries. Japan has moved from a dominant position in footwear, textiles, and clothing to dominant positions in steel and automobiles, while increasing its strength in ship and bOat building. The Japanese position in office machinery has also improved substantially. Thus, over a twenty year period Japan has transformed its pattern of exports from that of a developing country to one more characteristic of a highly developed country. Over the same period the U. S. has retained its disadvantage in textiles, clothing, footwear, and ship and boat building. The U. S. disadvantage in automobiles has importantly diminished and U. S. strength in office machinery has increased. I have no direct evidence as to why the U. S. disadvantage in automobiles has been lessened. However, Linder's theory appears to provide the relevant explanation. European and Japanese auto producers were familiar with the technology of compact car production and design; through product differentiation, they penetrated the U. S. market-- '136 TABLE 8.2.--Revealed Comparative Advantage: Changing Patterns gagging: EEC-6 Hglgggm ‘ Canada Sweden Japan 325:3: TextilesI RCA162 34.3 41.3 48.2 52.2 17.8 52.4 RCAI75 37.3 26.1 47.9 46.6 32.5 43.6 RCAW75 47.2 37.3 49.2 49.1 37.8 48.9 Clothing RCA162 6 66 49 28 15 59 RCAI75 12 17 24 9 48 50 RCA”5 52 45 39 33 53 56 Footwear RCA162 64 46 61 1 69 RCAI75 2 55 33 26 66 65 RCAW75 8 55 38 35 66 63 Automobiles RCAI62 1 24 53 7 54 60 RCA”5 52 ' 66 7 38 6 43 Ships and Boats RCA162 41 45 57 3 8 68 RCAI75 56 71 17 13 3 69 Steel Products2 RCA162 18.5 51.8 26.8 13.7 33.7 53.8 RCAI75 35.7 65.7 33.3 26.8 8.0 55.2 Office Machinery RCAI62 55 59 17 13 43 14 RCAI75 70 33 16 32 23 4 SOURCE: Balassa (1). NOTES: RCA16 indicates that the rankings in that row are from Balassa's 1953-62 period ans use the index RCA , RCA175 indicates that the RCAI75 index was used for the period 1967-75. RC 75 indicates that the RCAw index was used for the period 1967-75. 1Includes 651.2, 651.3, 651.4, 651.6, 652, 653.2, 653, 655, 656, 658, 653.10, 654, 655, 656.6, 656.0, and 657--the average rank is reported. 2Includes 671, 673, 674, 475, 676, and 678--the average rank is reported. 137 satisfying a previously neglected demand. In addition, higher oil prices have made this market segment relatively more important. Only in recent years have the U. S. producers offered cars which are essentially similar to the imported models. Therefore, it is the increased responsiveness of U. S. producers to satisfy domestic tastes which has tended to decrease the U. S. disadvantage. The change in the United Kingdom's comparative advantage pattern is not consistent with respect to the factor intensities of the products whose rankings have changed. The U. K. has developed an advantage in textiles, improving its average comparative advantage ranking from 41.3 to 26.1 between the 1962 and 1975 periods. However, when the 1975 ranking is computed relative to the world standard, the U. K. ranking slips back to 37.3. The same is true of the U. K. position in clothing, and to a lesser degree, footwear. Thus, the U. K. has developed an advantage among products which are better pro- duced by developing nations. The U. K. position in automobiles, ship and boat building, and steel has worsened; but the disadvantage in office machinery has been neutralized, as office machinery moves from 59th to the middle of the RCA scale (33). The pattern for the EEC-6 is highly similar to this. Canada has cultivated an advantage in clothing, and somewhat improved its position in footwear. Its greatest advance has been in automobiles and ship and boat building. Sweden has lost part of its advantage in ship and boat building, and all of it in automobiles. The Swedish position in office machinery and steel products has also 138 declined substantially. Relative to the sample countries- Sweden has gained in clothing, footwear, and slightly in textiles. This change in the revealed comparative advantage pattern indicates that Sweden has lost her advantage in several important skilled blue collar labor intensive products, and one very technological product. Meanwhile, the Swedish advantage relative to the sample countries has shifted into products which are relatively intensive in their use of unskilled labor. Thus, considering labor as a heterogeneous input clarifies the dislocation in the domestic labor market. The current conflict between the EEC, U. K., Sweden, and the developing nations is revealed by the differences between the RCAI75 RCAW75 rankings for clothing and textiles (Table 8.2). Here there are two factors to be considered. The RCAI rankings improved between the two periods partly because of Japan's withdrawal from these non- durable consuming goods industries. However, the 1975 RCA rankings which are relative to the world standard fall when compared to the sample country's standard partly due to the granting of preferences to developing nations. Between 1973 and 1975 the EEC-9 accounted for 72% of the world growth in textile imports (6). This has occurred despite the signing of a multi-fiber agreement in 1973, aimed at limiting imports. The European Community is seeking to freeze imports from Hong Kong, Taiwan, and South Korea, but allow export growth for other less developed countries. The developed nations have a true advantage-- relative to the world--in only a few textile products. Rising unem- ployment and stagnant demand for textiles have tended to make short 139 run protectionist solutions overly attractive. The entire picture is clouded further by multinational corporations which are located on both sides of the existing and proposed barriers. Steel products have also been subjected to trade restrictions. The essence of the steel problem is a lack of aggregate demand for a product with relatively high fixed costs. This creates incentives for dumping. The U. S. steelmakers have accused the Japanese of dumping and receiving government subsidies. The Japanese claim that their advantage is due to superior technology and wage costs that are 30% below U. S. levels. Japan sends 20% of its steel exports to the U. S. (and more counting the steel embodied in automobiles) com- pared to 4% to the EEC (7). The Japanese steel industry is heavily dependent on exports which comprise 36% of its output. This is 50% of the total amount of steel traded internationally (excluding internal EEC shipments) (8). The device which has been used to control this potentially dangerous situation is the orderly marketing agreement. Japan and the EEC reached an agreement under which Japanese steel exports to the Community could be limited. Subsequently, the U. 5. steel pro- ducers claimed that this agreement deflected more steel to the U. S. market. The U. S., failing to reach an orderly marketing agreement with the EEC and Sweden, imposed import quotas on their shipments of specialty steels. However, the U. S. and Japan were able to reach an orderly marketing agreement in speciality steels (9). 140 The willingness of the Japanese to enter into these agree- ments is a direct result of their reliance on steel exports. Japan can ill afford a highly restrictive unilaterally imposed barrier against these exports. It is reasonable to assume that a mutually agreed upon limit will be less restrictive than one which is uni- laterally imposed. Furthermore, as it is an agreement, it may be open to renegotiation as circumstances change. The failure of the U. S. and EEC to reach an agreement on steel is due to a basic ideological conflict. The EEC is far more committed to free trade than the U. S. Although the Community is not opposed to the use of the orderly marketing agreement, it is seen as a device of last resort. At the time that the U. S. attempted to negotiate the agreement, the European Community felt that their problems in steel were as great as in those of the U. S., but U. S. economic growth was progressing faster. Given these circumstances, the Community felt the U. S. should not request unwarranted protec- tion (3). Nonetheless, the EEC has entered into other agreements to protect its markets. The EEC and Japan have negotiated quota agree- ments concerning imports of steel, cars, ball-bearings, and ships. Similar agreements with other countries may be forthcoming (5). The increased prominence of orderly marketing agreements has prompted a response from GATT. A GATT study estimates that new restrictions now apply to 3 to 5% of world trade flows (4). The products most commonly restricted are textiles, clothing, shoes, steel, ships, and household electrical appliances. Although these 141 agreements are allowed under current GATT rules, their increased usage tends to subvert the basic GATT goal of free trade. Further- more, although these are agreements, they are agreements reached between parties with differing bargaining power. The agreements are most commonly struck between developed and developing countries or between Japan and other developed countries. At best this approach constitutes a second best solution to current world problems. The developed countries seek these agreements, not to improve their wel- fare, but in response to labor union and industry pressures. Consumer lobby groups are too weak to effect a balanced viewpoint on the issue of protection. 8.5 Conclusion The tendency toward protectionism is world wide. The most fundamental characteristic of protectionism is the unwillingness of countries to reallocate resources from traditional industries where they no longer have an advantage into industries where they have an advantage. This solution is not simple to implement, given the slow upturn of the world economy. Furthermore, multinational corporations located in both developed and developing countries, charges of dump- ing, and subsidiation, cloud true assessments of comparative advan- tage. It is probable that these issues will not be resolved until the recession is clearly gone and demand recovers; thus making alternatives to protectionism politically more desirable. REFERENCES Balassa, 8. "Trade Liberalisation and 'Revealed' Comparative Advantage." Manchester School (May 1965): 99-123. Business Week, September 26, 1977, p. 55. European Community, August/September, 1976, no. 196, p. 43. New York Times, September 23, 1977, p. A-l. The Economist, July 9, 1977, p. 51. The Economist, July 30, 1977, p. 75. The Economist, August 6, 1977, p. 75. The Economist, June 28, 1975, p. 91. kooowmm-bwm Wall Street Journal, June 8, 1976. 142 CHAPTER IX A MULTIPLE REGRESSION ANALYSIS OF THE HUMAN SKILLS AND HECKSCHER-OHLIN THEORIES: SOME IMPLICATIONS 9.1 Introduction In this chapter several major themes are brought together. As such it draws heavily on the preceding chapters, especially Chapter III (Human Skills). The major function of this chapter is to provide an empirical assessment of the human skills and Heckscher- Ohlin theories using multiple regression analysis to isolate the effects of the individual theories. These theories are tested by employing the relevant total requirements variables. As most of the current econometric evidence which pertains to these theories is based upon the use of the incorrect immediate and direct requirements, we shall inspect the relationship between skill indexes, compositional skill variables, and capital/labor ratios when both the total and the immediate requirements are utilized. If there is a close positive relationship between variables based upon the total requirements and their immediate counterparts, we may conclude that our current stock of information is left more or less intact. HoWever, to increase the certainty of this conclusion, both the immediate and then the total requirements will be employed as independent variables using the same dependent variable. 143 144 Having already completed an input-output evaluation of the hunan skills theory, we have an opportunity to make a comparison between the inferences drawn from that technique compared to those drawn from the use of multiple regression analysis. From a strict theoretical standpoint, there should be no differences. However, each of these empirical methods has strengths and weaknesses. The critical distinction between I-0 and regression testing is that the former procedure assesses aggregated characteristics across countries, while the latter assesses characteristics aggregated at the industry level across industries. By taking these into account and balancing one set of results against another, we can obtain insights that would be unavailable had we simply chosen one mode of analysis. First, the methodology is set forth. Next, the multiple regression results which provide the proper test of the theory are presented. After comparing these results to those of similar regressions which use the immediate requirements coefficients as explanatory variables, the similarity between the immediate and total requirements is assessed. 9.2 Methodology Here comparative advantage is measured by two different dependent variables. The choice of dependent variable is important because it is this variable which the theory tested purports to explain. If the variable is a poor measure of comparative advantage, then the test of the theory is not valid. However, for any reasonable measure of comparative advantage the same inferences 145 should be able to be drawn. If this cannot be done, the problem is reduced to a rejection of the theory versus a rejection of the dependent variable as a valid measure to be explained.1 For each country, the dependent variable is defined as its exports minus imports (net exports) and also its exports as a share of the exports of all the countries in the sample (export share). From a theoretical standpoint, net exports is the preper variable by which to measure comparative advantage for a factor pro- portions test. The net exports variable subtracts out the imports and focuses on the net flow of goods. Clearly, a factor proportions account is meant to address exports and imports. Although X/M may also qualify on these grounds,it does not give weight to each indus- try in accordance with its impact on the allocation of domestic resources. However, when the effects of commercial policy are con- sidered, the inclusion of imports creates a distortion. Tariffs, quotas, nontariff barriers and especially the rise in the promi- nence of orderly market agreements, distort the trade flows across industries for each country in a fashion specific to the commercial policy of each. If commercial policy is geared for protection, use of the net export variable imparts a bias against the theory so tested. However, because of the year under study (1975), tariff barriers should not affect this analysis as much as they 1Although the latter choice involves circuitous reasoning, it is a rather common conclusion. If a dependent variable constitutes a particularly bad measure of comparative advantage, it should not even be used. However, if there are some problems with a given variable, but we have expectations as to the net inpact of those problems, there may be a rationale for employing that variable. 146 have affected previous studies due to the lower tariff levels. Orderly market agreements will affect the analysis; however, they are mostly confined to a few particular commodities.2 The export share variable does not net out imports, although it is less affected by differences in commercial policy. Since it measures exports to the world, the individual country dif- ferences in commercial policy are not as great a factor since all countries face more or less the same barriers. Therefore, the export share variable may be a superior measure against which to test the theories, but both dependent variables should lead to the same gen- eral conclusions. The skill categories employed here are those which are common to the literature (1,2). Although the availability of data allows a far more detailed breakdown of the labor force by occupation, the introduction of too many skill classes increases the probability of spurious correlations. The selection of seven skill categories is considered to embody the optimal trade-off between requirements for detail versus economic distinctiveness among the skill classifica- tions. The critical question is: Which skill classes are important determinants of trade patterns? The answer to this question deter- mines the functional specification of the empirical test. The skill classes which are most important must be included in the regressions; otherwise, specification errors are introduced. The omission of a 2See Chapter VIII. 147 relevant explanatory variable introduces a bias and precludes inspec- tion of the excluded variable. These considerations are important because the skill variables sum to unity across industries. Thus, the perfect multicollinearity between the skill variables and the con- stant term in the regression precludes inclusion of all the skill classes or requires a constrained regression. The most objective way to deal with these problems is to use a constrained regression. When the constant term in the regression equation is surpressed, the problem of perfect multicollinearity is resolved. The regression is perfectly objective as it is in no way dependent on the researcher's choice of which of the skill variables is most important. By including all skill classes in the regression, information on each nation's comparative advantage among the most skilled variables and among the least skilled variables also can be obtained. The independent variables entered in the multiple regressions measure three main economic characteristics: skilled labor intensity, unskilled labor intensity, and capital intensity. Four variables measure different aspects of skilled labor intensity; three variables measure unskilled labor intensity. The following analysis sets forth the special aspects of each of these variables within its major group. High Skill Labor Classes. Expect each of these classes of labor to be positive determinants of comparative advantage for relatively skilled labor abundant countries; negative determinants for relatively unskilled labor abundant countries. 148 1. Scientists and engineers: This class contains the most skilled of the skilled laborers. In addition to being skilled, these are the laborers most important to research and development activities. 2. Other professional technical and managerial: This class contains the most heterogeneous mix of skilled white collar workers. 3. Clerical and sales: This is the least skilled of the skilled white collar labor classes. However, as the services which workers of this type provide are demanded most intensively by workers with yet more skill, the class can be viewed as a general proxy for skilled white collar labor. 4. Craftsmen and foremen: These are the most highly skilled laborers in the blue collar work force. Workers of this type provide the bulk of the skilled labor services which are most closely related to production activity. Unskilled Labor Classes. 5. Operatives: This unskilled labor classification contains the largest proportion of unskilled labor.3 Although these workers are not the least skilled of the unskilled group, they possess skills which are very easily acquired. The factor endowment theorem is based upon the proposition that in a relatively labor abundant coun- try, it is the large supply of unskilled labor which makes labor rela- tively cheap. Therefore, the relative size of the "operatives" classi- fication across industries makes it the most important classification 3Generally speaking, it contains the largest proportion of any type of labor across industries, although for a number of indus- tries the proportion of craftsmen and foremen is the largest (see Table 9.1); nearly double the proportional average of any other labor classification. 149 with respect to the role of unskilled labor in the context of a factor proportions model. 6. Nonfarm laborers and service: this too is one of the most unskilled of the unskilled labor classes, but it is relatively less important than the operatives class. As workers in this class— ification hold no special skills, their sole affect on the trade patterns is through the relative wage effect which has only one third of the impact of the operatives classification. Unskilled Labor--Special Considerations 7. Farm laborers: Workers in this class are at least as unskilled as workers in the previous unskilled class. However, no manufacturing sector demands this variety of labor service in the 4 immediate sense. Consequently, the variable performs more as a proxy for the extent to which agricultural strength is "passed on" to the manufacturing sectors.5 If national advantages in agriculture are passed along, this variable ought to indicate an advantage for Canada, the United States, and Australia.6 4For this reason the farm labor variable is not defined when immediate skill coefficients are used. 5Note that the processed food sectors are not included in the multiple regression analysis. 6Agricultural goods and therefore farm labor is a most inten- sive input in the textile sectors, but least intensive in the capital goods sectors. DevelOping countries have a very large volume of their exports concentrated in textile products.‘ For these countries the farm labor variable will probably have a positive coefficient. However, in these cases the variable does not necessarily imply national agricultural strength. Instead, it merely indicates that the exports of the country are relatively intensive in agricultural inputs. In fact, the agricultural inputs may not even be provided 150 Capital/Labor Ratio 7. Capital/labor ratio: Relatively capital abundant coun- tries derive an advantage in the production of relatively capital intensive commodities; relatively labor abundant countries derive a disadvantage in those commodities. Table 9.1 summarizes the average relative importance of each of the skill classes across industries.7 TABLE 9.1.--Average Relative Importance of Skill Classes Across Sectors; U. S. Labor Coefficients Skill Class* I II III IV V VI VII Percent of Total Labor Force 3.4 14.3 19.1 18.1 31.6 9.4 4.1 SOURCE: 1970 U. 5. Census of Population. *Roman numerals correspond to the arabic numerals above which designate the skill classes. 9.3 The Human Skills and Heckscher-Ohlin Theories of International Trade: An Empirical Analysis In this section the comparative advantage of nineteen coun- tries is assessed using the total requirements characteristics domestically. Developed skill abundant countries have a disadvantage in textile products due to the high unskilled labor content of these products. Therefore, if despite this conceptual bias the farm labor variable is a positive source of comparative advantage for the particular countries named above, this is even stronger evidence that national agricultural strength is passed on to Unemanufacturing sectors. Due to the complementarity between farm labor and this particular natural resource characteristic, the asymmetrical inter- pretation is thought to be advisable. 7In Chapter II the non-occurrence of factor intensity reversals was established. 151 estimated from United States data. Table 9.2 summarizes the results 8 of fifty-seven regressions. The following regressions were estimated, X-M = a‘s1 + a2s2 + a3s3 + a4s4 + ass5 + ass6 + a7s7 + a8k/1 + e1 (9-1) X/ZX - a‘s1 + azs2 + a3s3 + a4s4 + a555 + a656 + a7S7 + a8k/1 + ei (9-2) where ei is the error term and for each country: X-M is net exports (thousands of dollars) X/ZX is each country's share of exports among sample countries 5. is the proportion of laborers of class i required by each industry computed from the total require- ments coefficients. The regressions are estimated across the manufacturing sectors for each country. The relevant regressions for each country are summarized in Table 9.2 as (X-M)-T and (X/ZX)-T where the 1 indicates that the independent variables used are measured by the total require- ments.9 Table 9.3 presents the measured relative factor endow- ment position of each country. For the sake of completeness, 8The full results which underlie this table are presented in the appendix. The summary table only provides the sign of the coef- ficient and its level of significance for coefficients which are significant at the .20 (20%) level. Thus, the table sifts out insignificant results to reveal more clearly the most important trends. The significance levels are rounded to two decimal places. 9For the moment the regression results produced by using the immediate requirements will be ignored. 152 «N. .00.. + .00.. + .-Mx0.x 0N. .00.. + .0... - .- 00.. 00. .00.. + .00.. - .00.. - .-.0-.. 0000.000000 0.. .00.. + .00.. 0 .0... 0 .00.. - ...00.0. 0.. .00.. - ..0.. + .00.. + .00.. + .00.. - ...x0\.. 0.. .0. . - ..0.. + .-.0-.. menacmc 000: .N. .00.. + ..... + .00.. + .-.N.... 00. ..0.. + .-.xu... .N. .N... - .8. + .0... + .0... + 7.5-... 00000.0 000.00 0.00. 0.. .00 . + .0... - .-..0\.. 0.. .00.. - .00 . + ..0.. - .0... + .-.x.... 0.. ..0.. + .-.z-.. 00006 0N. ..0.. + .00.. + .-.00... 0N. .00.. + .00.. + .-.x0.xw 0.. ...z.. . ouocou N0. ..0.. - .00 . + .0... + .00. + .00.. - M00.w + .-.00... .0. .00.. - .00.. - .00. + ..0.. - 00. + ...x.... 00. .0... - .00.. + ..0.. - .00.. + .-.0-.. 0000.0 mm..0c taco-.95 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 50 Logo. o. 0.00000. ou.>.wo 000 003000. 00.00 .0.000000: 00000.00o ..ouum. a .0..00u 0C0. oco 000L000. 00>.uocmac can use can .00.:0uaN new .0 wcmN N .o 0.000 0.0200. 020.00: 00500.009 .00..0.u .oco.00o.000 0000c 000.000.00 . 000 ..0 .0 0 0. ... .. . 0.00.00. ucmocwcmc co. 000.cmanc .00o..0ccm.c. :4 III (F .Itlrh ..IllM-i AI-l .! Ill-III 00.00..0> 0000000000. .0. 0.o>0. mocau...cm.o can 0cc.o 000.0...00c 00.000... .0 00000030 $6.00.... 09.30.2000: oNc. ”0.00..o> acoucmamc a .o co..uo.0o 00. o. 000 0.0020L.000¢ .0000. .000. 0:000> 000.0000. .0 000 00. 00 00.0000. 0..00-00000.o00 000 0....0 00000 000 .0 0.00. .00..00000. 00. .0 0......0000 00.--.~.0 0.00. 153 .0. .00.“ + ..0.. + .0... + .-.xu... N0. .00. + .-.xu... 0.. .0... + .-.z-xw 0.>0.00 a» NN. .00.W + ..0.. + 0..w + .-.x... 00. .00. - .N... - N0. + .00.. - .-.x... 00. .00.. - .00.. - .00.. + .0... - .-.0-.. c.00m omocao .mmmm N0. .00.. + .00.. - .00.. + .00.. - .-.xu... N0. .00.. - .00 . + .00.. - .00.. + .00.. - .-..0\.. 00. .00.. - .0... - .00.. + ... . - .-.0-.. 000.00. 0.. ..0.w - M00. + .N... + .-.x0.0. 0.. .00. - 0.. + .-.x0.x .N. .00.. - .-.0-. acmecwc .0. .00.. 0 .00.. + .-.00... 0N. .00.. + .-..0\.. ... .0... 0 .0... + .0... - .-.0-.. mgcooewng -E:.m.wm N0. .0... - .0... - .00.. + . .0... 0 00.w - ...x0\.. 00. .0... - .00.. + .0. - .-..0\.. 00. .00.. - .00.. + .-.0-.. 0.0.. 0.. .00.. + .N... + .00. - .-.00... 0N. .00.. + . .00. - .-.N.... 00. .0... - .-.0-.. 00000. .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 0000. 0» 0000000. 00.>000 000 000000. 00.00 .0.000000: 0.00:.cco .000MMN 20x .0».009 5000 0:0 0.00000. 00>..0Laao 0:0 0:0 0:0 .00.:000N 000 .0 o N N .0 0.000 000000. 500.002 00500.000 .00..0.o .000.000.000 .0000 000.000.00 000 ..> .0 0 0. ... .. . 0.00.00. ucmucmamc 00.00.00> 0000000000. 00. 0.000. 00000...0c.o 000 000.0 000.0...00c N "MIMI EIIHHFIIYHM-lhhuflh u fikflfiggJ fill." aahfl.dflaflu "I“II I uw::.ucou-i.~.a m.m(p 1554' .... 000000000000 000.0000. 0.000 00 ... 0000000.:000 .0000 0.00 500. 00000003 0003 0.000 000 .0 0000 000 00 00500 00.00.00) 0000000000. 000 0000003 0000000 0.00.00> 000000000 000 000.0 . 00 . 00.0000.000 00. .0.0E00 000 0. 00.000000 000000.: 000 .0 0000000 000 .0 00.0000000 0 00 0000000 0.0000000 0000 0. 0.00.00, 000000000 000 0000 00000.00. .xu\xv .000..00 .0 .0 .0 000000000 0. 0000000x0 00000x0 000 0. 0.00.00> 000000000 000 0000 00000.00. .:-xvr ..000000 0000 00.. .0000. .0 00000 ..0 00 0>.00.00 .00.000 00. 0. 0.00. 0000.\.00.00u 00. .0.00.00> 000000000 000 .0 0005 000 00. 00000000 0 000 0. .Emm. 00000000 0. 00.0: N0 .0 00.0) 00. .n- ..z-.. 0.>0.0000. .xcm..2-xv 00000. .0.0 ..:-xv 0000000x00-00.0.0m N 00. ..z-xv 000000 .0 00.000000 000 00.: .0>0. am 000 00 0000...:0.0 000 000.00000 ..< .0000000. 0000 0. 00.00 00 000 00000000000 .0>0. ...00 000 0000 000. 000 000. 00.0.0000 00000.0000000. 000 .0 0000000 0.0.00 000 0000000 00000. 000 000.0000000 ..0 .0.000 000 .0 0000 000 00 00000000 000.0...000 000 .0 .0>0. 00000...00.0 000 0>.0 00000000000 0. 00000.. 00. .0.000 000 0. 0000.00. 000 .0>0. 00w 000 00 0000...:0.0 0000.0...000 00000 0.00 .0000.00. 00.00.00> ..0 0>00 000.0000000 ..0 ”0.02 00.00. "mumaom 00. .00.. + ..0.. - .0... - ...x0\.. .0. .00.. + .00.. - ..0.. + .-.x... 00. .00.. + .00.. - .00.. + .-.0-. 0000.000 00. .00.W + ....0 - .0... - ...x.... 00. ..0.. + .0... - .00. + .-.00000 00. .00.. + .0... - ..0.. + .-.z-.. 0.00. 00. .00.. - .00.. + .0... - ....0 - .-...\00 .0. .00.. - .00.w + .0... - .-.x0\.0 00. .0... + .00. + ..0.. - .00.. - .-.z-.. 000ox 00. .0... - .00.. - .00.. + .00.. - .0... - .-.00\x. .0. .00.. - .00.. + .00.. - .0... - ...x.... 00. _ ..0.. - .00.. + .00.. - . .00.. - .-.0-.. 0000 000: 0.0< 0.. .00.. + .-.00000 00. .00.. + .00.. + .0... - .-.xu... 00. .00.. + .-.0-.. van ..mmN 302 0.. .00.. + .-.x0\.. .0. .00.. + ..0.. + .00.. - .00.. + 0-..... 00. .00.. + .00.. + ....0 - .-.0-.. . a. ~00uma< 000000 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 .000 00.0 50 00000 00 00000000 00.>00m 000 0050000 00.00 .0.000000z 00000.0cu .00WMM. a .00.000 5000 000 00000000 00>.00000o 000 000 000 .00.:000. 000 .0 h N .0 0.000 009000. 500.002 00200.000 .0u.00.u .000.000.o00 0000c 000.000.00 000 ..> .> 0 0. ... .. . 0.00.00. . 000000000 00.00.00> 0000000000. 00. 0.0>00 00000...00.n 000 000.0 000.0...0ou ”“0! fl5r|r I: 1 '1.E'\“FII .. |. 1LrlrlL .d . l|.llll Ill. Ila" I...‘.I1nl|.|l uiu Ei' .||'I||I|b.: .n 00:0.0:09uu.~.0 w0m x ucmFoQN 3oz so A :3 A x <+ <+ m- A HH> x aepaeuwsn ocomuo x A xm A c: a+ NF m- o o > p+ uw>o~moma> x <- mp —+ o o N+ cwaam omega“ cacao x m _+ o ~> H _+ ucmFocH x A c: A xm <- <. m m+ 4 ~+ xcmEcmo xm A c: A x <+ ofi m+ 4 4 > m- oesonme=4 issmw_mm x __ ~+ p- »_mpn x m _.+ .— _.+ muse”. es A xm A x <+ <+ N e+ a ~> .> o meea_toeeoz x m F+ 4 >H o Acostaw “mm: x m _+ H ~+ seamewx ooywmp gum :3 A xm A x e _+ >H ~+ cocoa xm A c: A x <+ m N- 4 4 e- anmcmu c: A x A xm <- _ m- 4 A H~> .> H o mmuuum umu_:: vmmaoemc: m e Ppwxmca ppwxm acme»wwwmmwmwmw—mm Azixv xcrx am i mmEm x~>1ivc . i 2-x a; sac copumu match Axw\xV meowmmmcmma o : mxooFocsumh exeocoom upmum mtmvcwg cowmmmswmm Pon" ppwxm cmsnz a u mums» macauommzcmz mnmp ~P~_xm ease: "mupammm amok mo apnoh acmesamii._.o~ u4mppapoe yo mmucmgmmcwn .upamms mmgm>cmq a mmpouwucp a .xgomgu mg» saw: ucmEmoemm movaUWucw < .mpnou as» we now: «cu pa coed: mcowmmmgmmg on» :_ mmpnnwtm> ovum; Loam_\pmuwamu Ago—V pcmuwmwcmwm on» we mcmwm on» men mmapm> umpamp ushm .AH mmmFu Ppwxmv mucoaxm cw umwuonem ucmucou muw>tmm Lonmpn_w~_pmcowugoaoga may .xe» ou mcwucouum mmwcpcaou mo mmcwxcoam .xcomgu one saw: mucmucoammgsou pumwgoa mmWPaew ogm~ we oucmcmwmwv < .mmcmxcme utoa5m\ugoaxm cw vwvvonEm mqum »o ovum; use acmxcmg mNTm ucmszoFQEm mcwgauumeacoe chopum: cmmzpma mucmgmmw_o¢ .aLomgu xouocuco saw: ucmpmwmcou mew muyammt we» mmuouWucw o mgmucwg cup: ucmumwmcou «so mupamoc ummu msu moumu_ucw 4 m Anm=_mmn xpcm—ws_m mv mmwgucsou cwguo mo mgopum> utoaxm mg“ cu Louum> ugoasw m.xuuc:ou ago we comwgmaeou a now one mupzmog umpnmu one away mmumu_ucw em .mupammg cormmwcmmg "Aum_-.50nmcwx wouwcz uqmuxm ~m>mp Rog wumewxogaamv ucauwwwcmwm mew: new xuomcu mcp cur: cameos sums: mmocu use mammoFu ppwxm ompnmu on» .mgmgonog Egmmiimn~> .memsonmo Egmwcoz umfipwxmcaii>fi .mm>wumgmao-i> .cosmtom new cmEmu¥mguuu>H .memmcwmcm use mumwucmwumiim note mammmFu F—_xmu .mu_:mmg uaqpaoiuaacw macomgu mzu Low pgoaaam mcocum was» .mmcwxcmc ozu on» cmmzumn mucmvcoammcgou pummcma m mmuau_ucw ogm~ do mapm> < .Axmrv mmCTxcmL xmucw Ppwxm ugoaxm use mmcwxcme unmezoucm Ppwxm pccowumc asp cmozpmn mwucmtmwwwu as» one mm3~e> cuppa» ochp .uwoe as» e_ amaspoew Hoe - x "mmeoz .~.m can m.m .m.n .p.m .m.m .n.m mmpamh Eogm umuqmu< "mumzom x NF pt 0 o > o cmumvxoa x a, a- o o > _+ aweem x A ea A :3 <- a_ m- a o > >_ .H F+ aoeox x mp o > >H .H a- aeog ommm arm< » m m e .F_xme= __exm » mmocmcw : ”emanate“ oswuw_am Az-xv xa a we - we“ Az-xV xN -ucom sueeaau Louumu outch Axw\xv ecowmmmcmmm oi: mxmofiocgumh exsocouu «Foam emucwg cowmmmemmx pct” ppwxm cogs: m ~__F¥m ease: umacwucoutn.—.o~ m4mgmm can cosmcou mm_mm .owemmmcmz memm:_mcm mo quh ...._ a 39300 E.,—m... use mLmLoaog mm>mumgmao use 9.8 ucm 392.com» ucm cm N ”..0 omumm {.0.—Emu Egon—:02 :mEmuwwLu 32:55 chowmmmwoti meuo mumwucowum mpmmzm> HH> H> > >2 HHH He H . ucmucmnwa .apeoz one new: ovate motaeoaeseaz mea_ mucmwuwwwmou .m .2 Bose umczmcmz mucmewcwzomm Louomd mpmwumEEH cam _muop mcwm: ”mcemuuma mung» .mmweuczou cmmumcvz mo mwmxpmc< co_mmmcmmm m_awu_:zii .Fim u4mcmm can cosmsom me_mm powgmmmcmz mgmmcwmcu mo maze ..— a :0:th E9; .35 mgmgonmg mm>zwgmao new Us... tam 30.5508. 2.5 cm N mo ovumm memecmu Egowcoz :mEmuwmgu Fau_gwpu pmco_mmmwocm meuo mumwucwwum o_mmpcm> 22> ~> > >_ -~ an H eeoaeoaoa uwzcwucouii.—im m4mwumgmqo new was ten qu_c;umh uco cm N mo 05mm 205me Egmacoz cmsmummuu 30758 "meowmmwwogm meuo mummucwwum QBWCM> vmacmu:00ii.pim m4m

com new cosmeom mopom pomewmoooz mtmmcwmcm we won» “— m ~92th Ea... new mLmLonmq mm>3mgono Una tea tea pmuwccumh can an N we 0.53. 2220... Egmmcoz :wEmuwmsu poutm—u pncommmmwokm Lmnuo mumwucmwum 0 v _oa_ea> coacmucouoi.—im u4m

o~mo :> Louuou Eo Loco; cu mtmconoo wu_>tmm too coawgou mmpom puwgmmocm: mgomcwmcu o mo» m m pou_oou so»; new mgmgoamo mm>wuogmoo new con new —ou_:cumh woo m cm h w mo o_uom mcoEgou Egowcoz cosmumoeu pouwgmpu pocowmmmmoco meuo mumwucmwum mpawpco> oo> o> > >o ooo oo o oeooeoooo "Iii. JUNE“ "1.1. a. Jul“ I oooemoeao--._-o moocmm new :mEmgou mopom pmwgmmocmz mcmmcwmcu no ooze o a poumoou seem too mcmgonmo mm>_uagooo com too can qupcguop new on m we owuom meoeLou Eeowcoz cosmuuogu pouwgmpu pocowmmmwogo guano mumwucowum mpnmpgo> woo oo o oo ooo oo o oeooeoaoo I‘ll umacwu=00ii.pim m4me_ meceeomceu no— as» an accommocmoo «a upe>ep mesmemweeu am as» ue woeeowocmwoc .oueeeegoooog mueoeeeso ecu see» eweooeme wee mm—neogo> Homeomoeeow ecu moueeoxe anacoee w_oseo we ecezo o.>suooeu one no e_eooee> Homeeeoee as» uezu oeueuoeco Hiaxu\xv .mucmeecmooeg peueu as» seem emgooees wee omFDewco> acmeceoeecp on» mouseoxm xgucoou oposoo we osezo o.»gu::ee ecu ow epeeoee> Homecmome one was» omuouweco hiaxuxxv .oucmEoL_:oeg _euop on» see» emeooees wee oe_oewge> ucoecmomeoo ecu "Aogeppee we meempposv wagons? mocos oueeoxo op opeeoge> pomeceome ezu peg» omueeweco hiaz-xv .mpnewge> acmecmoee on» we come as» Low eouooneo mo mope> Na eeogeoee we» uegu oeueumucv Ema "who: Aoeoeh eouoooozv oemuo_uoum moose moweoseeu .z .2 “Nuaaom Aoo.-o -- Aom.-o .A_o.mo aafieo._-o Ao~.o “mo.o Aom._-o oo.m «No. Npo.- -- opo.- Npo. N_o.- Koo. ooo. ooo.- o-Axo\xo ANo.o afieo.oo .Aofi.m-o .Ame.~o Aeo.-o Amo.-o .Noo.o Aoo.-o oN.__ NNo. om". mom. oop.- opo. ~_o.- _oo.- .oo. “mo.- . F-Axo\xo ANo.-o .Aoo.No .ANo.o-o .ANo._o Aoo.po A_o.-o Aoo.o Aoo.-o oN.__ moo. ~o.- mo.” oo.- oo. __. op.- Pm. om.- F-Az-xo eaoa_eaa . Leueeo so Loeeo op oeeceoeo euw>eem use :mseeoo om_em —e_eooe=ez mommcomcu we womb m m powwoeu seen one oeoeoeeo mw>oueemqo use use nee peu_:neoh use :e N we oruem mLmEeeo etemcez :msoowegu Peewswpu Fecowmmmmoxo emcee mumoucmwum upewpge> 11“]. it umocpucouui.pim m4m