ALTERNATWE MODELS OF REGIONAL COMPARATIVE ADVANTAGE IN THE UNITED STATES Thesis fer the Degree of Ph. D. MICHEGAN STATE UNIVERSITY THfiMAS ALBERT KLAASEN 1969 0-169 This is to certify that the thesis entitled ALTERNATIVE MODELS OF REGIONAL COMPARATIVE ADVANTAGE IN THE UNITED STATES presented by THOMAS ALBERT KLAASEN has been accepted towards fulfillment of the requirements for Ph.D. degree in Economics A ./3. Ammo (7 Major professor ‘7 Due May 27; 1969 F—k LIBRARY Michigan State University 'M"=m'w BINDING BY HDAG 8: SONS’ SS3 Fd-“fi‘ I . _ £922 A” MAR 1 42007 ABSTRACT ALTERNATIVE MODELS OF REGIONAL COMPARATIVE ADVANTAGE IN THE UNITED STATES BY Thomas Albert Klaasen The goal of the research undertaken in this disser- tation has been to test empirically the Heckscher-Ohlin and Classical trade models. The uniqueness of these tests is that United States regional data were employed rather than international data. Two sets of comparative regions were used: South-non-South and New England-non-New England. Incorporating the regional approach into the two models, they could be stated in a form leading directly to empirically testable hypotheses. The Heckscher-Ohlin model brings together a combination of relative factor endowments and relative factor intensity in production as determinants of comparative advantage. Specifically, the model predicts that a region tends to specialize in producing those goods requiring intensively the use of the relatively abundant factor of that region. Stated as an empirically testable hypothesis: industry rankings of concentration in the South will be negatively correlated with industry capital-labor ratios. Thomas Albert Klaasen For actual testing, capital-labor ratios were found by dividing year-end book value of capital assets by total employees for 71 Standard Industrial Classification three- digit industries for 1957-1958, while concentration in the South was found by dividing value added in the South for each industry by value added in the nation for each corres- ponding industry. Data were available in the Annual Survey of Manufactures, 1957, and the Census of Manufactures, 1958. Different measures of the basic variables were used in the tests. They were: gross capital, net capital, unweighted labor, and labor weighted by a wage index. The Classical model, using the labor theory of value, bases comparative advantage on relative labor productivity advantage. With the inclusion of wages, the determinant of comparative advantage becomes relative average labor cost. Both labor variables were considered in the study, the em— pirically testable hypotheses being that ratios of labor productivity in the South to that in the non-South will be positively correlated with concentration in the South; while South-non-South average labor cost ratios will be negatively correlated with concentration in the South. Average labor productivity is found by dividing value added by total employees, while average labor costs are found by dividing the average annual wage (total payroll divided by total employees) by average productivity. Thomas Albert Klaasen Two broad conclusions can be drawn from these tests. First, an already industrially developed region can be ex- pected to display patterns of specialization in those indus- tries which have a comparative advantage with respect to labor productivity as well as those industries whose pro— duction functions require relatively more of the relatively abundant factor of that region. Second, for a newly developing region, initial at- traction of industries is likely to be based directly on sources of raw materials and in the endowment of natural resources of that region. As development proceeds, however, there will be a relatively higher growth in those industries which can achieve a comparative advantage based on labor productivity, or on intensive utilization of the relatively abundant and therefore relatively cheap factor of production. ALTERNATIVE MODELS OF REGIONAL COMPARATIVE ADVANTAGE IN THE UNITED STATES BY Thomas Albert Klaasen A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Economics 1969 (O'Copyright by J" Thomas Albert Klaasen 1970 To my father, Dr. Adrian J. Klaasen, whose love of teaching inspired me to pursue graduate study ACKNOWLEDGMENTS Especial appreciation is given to Dr. John R. Moroney for his unselfishness, guidance, and understanding in directing this study. The writer would also like to acknowledge the members of his committee, Dr. Mordechai Kreinin and Dr. Anthony Koo, for their helpful suggestions; and to the Economics Department at Michigan State University for underwriting the computer time that was used. Gratitude is also expressed to the author's wife, family, and friends for their encouragement. iv TABLE OF CONTENTS ACKNOWLEDGMENTS O O O O O O O O O O O O O O O O O 0 LIST OF TABLES O O O O O O O O O O O O O O O O O O 0 LIST OF FIGURES O O O O O O O O O O O O O O O O O 0 LIST OF APPENDICES O O O O O O O O O O O O O O O O 0 Chapter I. II. III. IV. V. BIBLIOG APPENDI THEORETICAL MODELS OF TRADE AND COMPARATIVE ADVAN TAGS O O O O O O O O O O O O I O O I O 0 Introduction . . . . . . . . . . . . . . . The Classical Model. . . . . . . . . . . . The Factor Proportions Model . . . . . . . Summary and Preview of Following Chapters- TESTS OF THE HECKSCHER-OHLIN MODEL . . . . . Introduction and Review of the Literature. The Case for Regional Tests. . . . . . . Tests in the South . . . . . . Tests in New England . . . . . Conclusions. . . . . . . . . . TESTS OF THE CLASSICAL MODEL . . . . . . . . Introduction . . . . . . . . . . . . . . Previous Tests . . . . . . . . . . . . . Preliminary Tests of the Classical Model Tests of an Alternative Classical Model. Some Comparative Static Tests. . . . . . Conclusions. . . . . . . . . . . . . . . THE ROLE OF DEMAND AND NATURAL RESOURCES . . Introduction . . . . . . . . . A Test for the Role of Demand. The Role of Natural Resources. Regression Analysis. . . . . . SUMMARY AND CONCLUSIONS. . . . . . . . . . . RAPHY. O O O O O O O O O O 0 O O O O O O O 0 CBS 0 O O O O O O O O C O O O O O O O O O O O Page iv vi vii viii 03030)?“ I-’ 109 115 121 Table 10. 11. LIST OF TABLES Linear programming allocation of labor inputs between industries and regions for maximization of national value added . . . . . . . . . . . . Ranks of average labor cost ratios and percent- age changes in relative concentration in the SOUth, 1947-19580 0 o o o o o o o o o o o o o o Ranks of labor productivity ratios and percent- age changes in relative concentration in the SOUth, 1947-19580 0 o o o o o o o o o o o o o o . . . C Ranks of relative concentration ratios, 5, n labor productivity ratios, and average labor cost ratios for industries for which at least 50 percent total costs are labor costs. . . . . Rank correlation coefficients and levels of significance by industries for tests between regional concentration ranks and regional demand ranks. . . . . . . . . . . . . . . . . . Comparison of results between full sample tests and non-market-oriented tests: South . . . . . List of industries in ascending order according to the coefficient of resource dependency . . . Regression variables, equations, and results for modified Classical model: South. . . . . . Regression variables, equations, and results for modified Classical model: New England. . . Regression variables, equations, and results for modified Heckscher-Ohlin model: South. . . Regression variables, equations, and results for modified Heckscher-Ohlin model: New England 0 O O O O O O O O I O O O O O O O O O 0 vi Page 66 72 73 80 85 87 89 92 98 102 105 LIST OF FIGURES Figure Page II. 0 O O O O O O I O O O O O O O O O O O O 0 14 III. 0 O O O O O O O O O O O O O O O O O O O O 15 IV. 0 O O O O O O O I O O O O O O O O O O O O 63 vii Appendix I. II. III. IV. VI. VII. LIST OF APPENDICES Standard Industrial Classification of 71 Three-Digit Industries . . . . . . . . . List of Rankings, in Ascending Order, of Variables for each Heckscher-Ohlin Test. Rank Correlation Test Results for the HeCkSCher-Ohlin Medelo o o o o o o o o 0 List of Rankings, in Ascending Order, of Variables for each Classical Model Test. Rank Correlation Test Results for the Classical Model. . . . . . . . . . . . . Analysis of Eight SIC Three-Digit Industries Ranks of SIC Three-Digit Industries by Co- efficient of Resource Dependency . . . . viii Page 122 124 134 136 148 151 156 CHAPTER I THEORETICAL MODELS OF TRADE AND COMPARATIVE ADVANTAGE Introduction Man has traded goods and services since means of communication and transportation emerged between societies, and the consequent writings by early economists dealt with the gains and/or losses of trade and its effects on the domestic economy. Adam Smith, presenting a free trade argument in his Wealth of Nations, suggests a reason for trade which involves a comparative cost theory. Smith writes, "It is the maxim of every prudent master of a family, never to attempt to make at home what it will cost him more to make than to buy. . . . What is prudence in the conduct of every private family, can scarce be folly in that of a great kingdom."l Also primarily concerned with the gains or losses from trade,/David Ricardo set forth the first exposition of the comparative cost doctrine in his Principles of § 1Adam Smith, An Inquiry into the Nature and Causes of the Wealth of Nations, ed. Edwin Cannan (New York: Ran- dom House (The Modern Library), 1937), p. 424. Political Economy and Taxation.2 The basis of Ricardo's analysis was the labor theory of value. In his "two good, two country" example, labor input was used as the measure of absolute costs, making commodity prices proportional to labor costs. ‘Let us consider Ricardo's example. The countries involved are Portugal and England; the goods are wine and cloth. England's cost in producing one unit of wine is the labor of 120 men for one year; for one unit of cloth, it is 90 men for one year. Portugal's cost in wine production is 80 men for one year; for cloth, it is 90 men for one year.3 .Although Portugal has an absolute advantage in producing both goods, she will purchase her cloth from England in ex- change for wine. As Ricardo states, "Though she [Portugal] could make the cloth with the labour of 90 men, she would import it from a country where it required the labour of 100 men to produce it, because it would be advantageous to her rather to employ her capital in the production of wine, for which she would obtain more cloth from England, than she could produce by diverting a portion of her capital from _—A‘ w— 2David Ricardo, Principles of Political Economy and Taxation (London: J. M. Dent and Sons, Limited, 1911). 3In this particular example, Ricardo does not spec- ify any physical quantity of wine or cloth. Later, he re- fers to a "pipe” of wine and a "certain quantity" of cloth. Ibid., p. 84. the cultivation of vines to the manufacture of cloth."4 The concept in the above passage has become known as the classical theory of comparative advantage. It is based on relative labor cost differences which in turn lead .. _.'\1 to relative commodity price differences. The key is the . {wreiative~price- concept, for if all money prices in each *3; \I \" .n—u-WM‘ d” .. .—-- country, although different absolutely, differed in the (2", same proportion, no trade would occur. ’J A The Classical Model To prepare the way for empirical testing of the classical comparative cost theory, an updated restatement of the theory is desirable.5 The assumptions are: (1) perfect competition in factor and product markets, (2) no artificial barriers to trade, (3) no transportation costs, (4) perfect factor mobility within countries but complete immobility between countries, (5) linearly homogeneous pro- duction functions for all goods, and (6) production func- tions for a given commodity vary between countries. The last assumption provides the basis for comparative 41bid., p. 82. SJagdish Bhagwati, "The Pure Theory of International Trade," Economic Journal, LXXIV (March, 1964), pp. 1- 64. Richard E. Caves, Trade and Economic Structure (Cambridge: Harvard UniversityL Press, 1960)— M. O. Clement et al., Theoretical Issues in Inter- national Economics (Boston: Houghton-Mifflinl Company, 1967). cost differences as it is derived from the idea that equal combinations of the factors used in the production of a given commodity would yield different quantities of that commodity in different countries.6 Labor costs were assumed to contain all influences of an economy on the production of goods. Factors not convertible to labor costs were as- sumed to be used in constant proportions with labor in all uses.7 For simplicity, one can restate the assumption as: goods in any one country are produced with the same capital- 1abor ratio, and capital-labor ratios differ between coun- triesfi~f \”\i For trade to occur, relative prices must differ between countries. The reason for price differences is the real unit cost differences between countries. In the classical theory, these costs are expressed as labor costs per unit of output, or its reciprocal, thus making the aver- .\ age product of labor the key to cost differences./) Using an example, we can show how labor productivity determines trade specialization.8 Assume two countries, A and B, each producing two goods, x and y. Under classical 6Clement et al., p. 4. 7Caves, p. 12. 8J. L. Ford, "On the Equivalence of the Classical and the Factor-proportions Models in Explaining Interna- tional Trade Patterns," The Manchester Schogl of Economic and Social Studies, XXXV (May, 1967), p. 185. 5 conditions, 6:9 = 63 in, say country A, where K and L X Y represent the amount of capital and labor respectively re— quired to produce one unit of output of either x or y. Then, -%—-and -%—-represent the average products of the two fac- tors. Under competition, product prices equal production costs such that P = rK + wL and P = rK + wL , where x x x y y y r and w are the prices of capital and labor inputs respec- tively. In either country, say A, the money cost ratio be- r w E a tween x and y can be expressed as Kx + Lx . Because i x Kyr + Lyw = 6% , we can let ky = °LKX and Ly = obe. Dividing the Y 1 + wa cost ratio by er, we get er =‘7i7 . Because oL.= L w x «,1 + Kr) x L L :1, the cost ratio equals :53 and is completely independent X Y of the factor price ratio -¥—u It is then, a function solely of the average product of labor in the production of the two goods. If trade conditions exist, that is Px £ P y A P L L (PX , they are a result of Lx :4 Lx , which repre- y B y A y B sents unequal labor productivity ratios between countries A and B. Say cost ratios are such that @A< @B, or Lx) < Lx . The reciprocals of L and L are the L L x Y y A y B average products of labor in producing x and y. The cost 1 l L L ratios can then be written as -jfL- <: -Ix—' , which yields A I"x B L x APL APL APL APL . _l < _l , or _§ > x . This last re- APL APL APL APL X A B y A y B lation results in different relative price ratios between countries which gives a basis for trade between those coun- P P tries. Specifically, if (if?) < GE) , country A will A y B export good x to country B and import good y from B. Both countries will tend to specialize in the production of their respective export goods. We have established that the pre-trade commodity price ratio within a country is a function only of the aver- age productivity of labor in the two industries. A country will have a comparative cost advantage in manufacturing that good in whose production its labor productivity is relatively higher. This is the essence of the classical theory of com- parative advantage. The Factor-Proportions Model An alternative theory of comparative advantage and trade was provided by two Swedish economists, Eli Heckscher 9 and Bertil Ohlin. Like Ricardo, Heckscher did not undertake 9Eli Heckscher, "The Effect of Foreign Trade on the Distribution of Income," Readings in the Theory of Inter- national Trade, eds. H. 8. Ellis and L. A.‘Metzler (Homewood, his paper to explain trade flows, but rather to find the influence of foreign trade upon the prices of factors of production. At the outset it was necessary for him to es- tablish reasons for differences in comparative costs among countries.10 These reasons, in Heckscher's model, are sub- stantially different from those in the classical model. Heckscher assumes constant and immobile factor supplies within each country; that each commodity is produced accord- ing to the same linearly homogeneous production function in all countries; that the production functions differ among all commodities in the specific sense that, given the same factor price ratios, the capital-labor ratios differ between any commodities x and y; and perfect competition in factor and commodity markets. Heckscher then suggests two reasons for comparative advantage: first, factor endowments differ between countries; these differences giving rise to inter- country differences in relative costs of labor and capital; and second, given the presumed differences in factor inten- sities in the production of different goods, the money costs of production of any specific commodity differ between coun- tries. Illinois: Richard D. Irwin, Inc., 1949), pp. 272- 300. Bertil Ohlin, Inter-regional and International Trade (Cambridge: Harvard University; Press, Harvard Economic Studies, 1967). loHeckscher, p. 277. As a student of Heckscher, Ohlin expanded upon the work of his teacher with the stated purpose of constructing a theory of international trade.11 The basic framework of Ohlin's book was designed to answer the problem of how com— modity price ratios were determined and how they differ be- tween countries. Ohlin suggests four determinants of commodity price ratio differentials: consumer tastes, distribution of fac- tor ownership, supply of factors, and production functions. The last determinant can be eliminated by assuming that pro- duction functions are the same in all countries for each good. This is not to say that Ohlin ignored possible dif- ferences in production functions between countries, but rather that he relegated any differences to a subordinate role in determining patterns of commodity prices. The first two determinants can be combined under the heading of consumer demand. Interregional or inter- national differences in factor supplies are crucial deter- minants of differences in costs of production. Yet as long as the demand element remains, it could offset the factor supply influence on prices. After discussing demand, Ohlin warns, "But one must be careful to remember the qualifica- tion implicit in the possible influence of differences in ."12 demand conditions. . This effect was considered lthlin, Preface. lzIbid., p. 10. remote, however, and the demand element has been essentially dropped.l3 Thus, the essence of the Heckscher-Ohlin trade model lies in factor supply conditions. The crucial assumption is that different relative factor supplies or "endowments" exist between countries. Although there are differences of opinion as to how to measure "relative abundance" of factors, the ultimate effects on costs are the same as long as all other assumptions hold. If factors are measured in terms of physical units, the opportunity costs of producing a unit of the good that uses relatively intensively the abundant factor are lower in that country than elsewhere. If relative factor supplies are measured as factor price differences, then by definition, the relatively cheap factor is the "abundant" factor. Any good which requires the rela- tively cheaper resource more intensively in production will have relatively a lower cost of production and price. A country involved in trade will tend to export that good and specialize in its production. A better understanding of the Heckscher-Ohlin model may be gained by examination of a "two good, two factor, two country" example. Assume competition prevails in both factor and commodity markets, free trade exists between countries, and there are no transportation costs. In addi— tion, production functions are assumed linearly homogeneous l3Caves, p. 11. 10 and are the same for each good across countries, but differ between goods within each country. Factor supplies are fixed within countries and are immobile internationally. Assume country A to be relatively capital abundant and good x to be relatively capital intensive. The condi- tion for trade between countries A and B is the inequality P P of commodity price ratios, that is, -§%> £ (35%) B' This Y A Y relation can exist only when the cost ratios in the two coun- tries are unequal. Given the assumptions of the model, these ratios are a direct function of factor price ratios. Under the given factor supply conditions, capital is cheaper rela- tive to labor in country A compared to country B; that is, @A < 6:? 8' Capital intensive good x can then be pro- ~ duced at a lower unit cost in country A, and competition Px Px ensures that T < —£-,— . A The model can be analyzed further by use of the fol- lowing example. Different factor price ratios between coun- tries A and B indicate different relative factor endowments; say @A > GE; B resulting in g) A< 63) B' In Figure I, we have isoquants for goods x and y in both countries. Because of the assumption of linearly homogeneous production functions, these isoquants are rep— resentative of all isoquants for each of the two goods in both countries. In addition, goods x and y are capital and ll labor intensive respectively, irrespective of factor price ratios. The factor price ratio in country A is shown by the slope of line PSRQ (with sign changed). Under the given different factor endowments, the factor price ratio for country B has a lesser slope and is represented by lines MNU and DET. K P M s D X Y L O Q T U Figure I By finding the relative costs of producing x and y in the two countries, we know relative commodity prices. Dividing total cost by the units of output gives us average cost. Line PSRQ is the total expenditure line for factors of production, and the total cost of producing each good can be expressed in terms of either of the two factors. 12 Distance OP represents the cost, in terms of capital, of producing n_units of x or y, given factor prices as they would be if used in the proportions OS and OR. Because total cost and units of output are equal for x and y in country A, average costs are also equal. In country B, using the same cost measure with factor proportions ON and OE, OM represents the total cost of producing x while OD represents total cost of y. Because OM >> OD, the ACX:> ACy in country B. Comparing country A with country B, we find rela- AC AC tive average costs are such that -—5- <: -—5- . Because ACy A ACY B commodity prices directly reflect production costs, country A will sell good x at a relatively lower price, export it to country B, and specialize in its production. Country B will export and specialize in good y. The statement about trade flows assumes similar demand structures between A and B. To summarize, the Heckscher—Ohlin model predicts that a country tends to specialize in producing those goods re— quiring intensively the use of the relatively abundant fac- tor of that country. A question often raised concerning the Heckscher- Ohlin theory is whether a reversal of factor intensity in 14 production is possible. Factor intensity reversal would l4Clement et al. Romney Robinson, "Factor Proportions and Comparative 13 occur if the relative capital—labor intensities in the pro- duction of two goods changed as a result of a change in rela- tive factor prices. When such a reversal occurs, the goods obviously can no longer be classified categorically as either capital or labor intensive. Reversal is most likely to occur as a result of wide differentials in factor price ratios between countries, coupled with different elasticities of substitution between capital and labor in the production of x and y. Under the Heckscher-Ohlin assumptions, relative factor price ratios are reflected in commodity price ratios before trade. As free trade opened, demand would rise for the relatively abundant factor and fall for the relatively scarce factor. Thus, in the example considered above, 69 A would rise and (Eva B would fall, and the two ratios would tend to equality. Equalization would occur, however, only if, say, good x were always capital intensive in both countries regardless of any change in relative factor prices. If factor—intensity reversal occurred, it would be possible for a capital abun— dant country to have a comparative advantage in a labor in- tensive good and the Heckscher—Ohlin theory would break down as an explanation of trade. Figure II can be used to illustrate the above point. Advantage," Quarterly Journal of Economics, Part I, LXX (May, 1956), pp. 1 - 2. Caves. 14 Let x and y be two isoquants representing given output rates of goods x and y. The factor price ratio for country A is shown by line CEFG, indicating that capital is relatively cheap. Equilibrium points of optimum output are at E and F, showing that good x is relatively capital intensive. The factor price ratio for country B is shown by line MNPR where capital is relatively expensive. Equilibrium points are at N and P, and by comparing factor proportion lines 61-1?) y’ and 69 x’ good x is found to be relatively labor intensive. Hence the relative intensities of x and y are reversed be- tween countries A and B. ,t‘lx x' \ ,‘k‘ a) ) G R Figure II 15 From Figure II, we can derive Figure III. Here factor-ratio curves show changes in the capital-labor ratio for the two goods as relative factor prices change. Below factor price ratio M, good x is relatively capital intensive, while for factor price ratios above M, good y is relatively capital intensive. _L PC M - ------ L 0 K Figure III If the factor price ratios of the two countries lie on either side of M, that good which is relatively labor intensive in one country is relatively capital intensive in the other and factor reversal exists.15 In general, the possibility of factor reversal was left open due to the vagueness of Heckscher's assumption 15M. Michaely, "Factor Proportions in International Trade: current State of the Theory,“ K klos, XVII (1964), Fasc. 4, pp. 529-50. 16 of different factor intensities for different goods. Samuelson, in proving factor price equalization as a result of trade under the Heckscher-Ohlin conditions, restated the assumption as the strong factor-intensity hypothesis.16 This hypothesis simply states that goods will maintain their relative factor intensity regardless of factor price ratios. The hypothesis is derived by beginning with the two key Heckscher-Ohlin assumptions: (1) different production func- tions between goods, but always exhibiting constant returns to scale, and (2) different factor prices due to different factor endowments. It follows then that for optimum resource allocation, the two goods will have two different factor proportions in production, irrespective of relative factor prices.l7 Summary and Preview of Following Chapters The two theories of international trade under review are without doubt the two most prominent theories of trade, and therefore it is important that their empirical usefulness 16P. A. Samuelson, "International Trade and Equali- sation of Factor Prices," Economic Journal, LVIII (June, 1948), pp. 163-84. P. A. Samuelson, "International Factor Price Equali- sation Once Again," Economic Journal, LIX (June, 1949), pp. 181-97. P. A. Samuelson, "A Comment on Factor-Price Equali- sation," Review of Economic Studies, XIX, No. 2 (1951-52), pp. 121—22. 17R. W. Jones, "Factor Proportions and the Heckscher- Ohlin Theorem," Review of Economic Studies, XXIV, No. 1 (1956-57), pp. 1-10. 17 be tested. The purpose of this dissertation will be to under- take these tests. A number of empirical tests have been done, but all, with a single exception, have used international data.18 The tests to be performed in this dissertation will be based on interregional data within the United States. There are several reasons why interregional data may be more suitable for testing the theories than inter- national data. In particular, consider the two crucial as- sumptions common to both theories: (1) free trade and (2) the absence of transportation costs. Interregional trade within the United States fully satisfies the first. And the second assumption may be more nearly applicable to inter- regional than international trade. Both theories assume comparable factor quality between trade areas. Less diver- sity in cultures and technology between regions in the United States than between nations justifies the notion that capital and labor quality are more nearly uniform across regions in the United States than between countries. The assumptions concerning production conditions differ between the two theories, and evaluation of the use— fulness of interregional versus international data is dif- ficult. The Heckscher-Ohlin model is based on similarity of production functions between trading areas, but not 18Previous empirical tests are reviewed in Chapters II and III. 18 between goods, while the Classical model relies on the con- trary assumption of similarity of production functions within a region but dissimilarity across regions. There appears to be some support for preferring interregional data for the Heckscher-Ohlin test. Two studies of capital-labor substitution using international data found that different countries producing the same goods were operating on differ- ent production functions.19 Gallaway, however, rejected the hypothesis of dissimilar production functions, as an explanation for regional wage differences. A final argument for using interregional data is that the potential problem of factor intensity reversal does not seem to be present. If such reversal occurs, it becomes impossible to classify goods unequivocally as either labor or capital intensive. A recent test of the "strong factor- intensity" hypothesis that involved rank correlation tests of capital-labor ratios for two-digit Standard Industry Clas- sification (SIC) industries among the nine census regions 19K. Arrow et al., "Capital-Labor Substitution and Economic Efficiency," The Review of Economics and Statistics, XLIII (August, 1961), pp. 225-50. Victor R. Fuchs, “Capital-Labor Substitution: A Note,“ The Review of Economics and Statistics, XLV (November, 1963), pp. 436-38. 20Lowell E. Gallaway, "The North-South Wage Differ- ential," The Review of Economics and Statistics, XLV (August, 1963), p. 270. 19 of the United States did not reject the hypothesis.21 A test using international data, however, yielded somewhat inconclusive results.22 In addition, based on theoretical considerations, smaller differentials in factor-price ratios between regions give less reason to expect reversal within the United States. In testing both theories in this thesis, the loca- tion of industries will be used as an indicator of compara- tive advantage. Both theories predict that trade will lead to specialization in export products; thus, areas of concen- tration of production of a good are assumed to exist because the areas possess a comparative cost advantage in the produc- tion of that good. In fact, Heckscher explicitly states that "[in the absence of mobility] . . . the different kinds of production will be located where the necessary factors of production are present."23 The use of location rather 2J'John R. Moroney, "The Strong-Factor-Intensity Hy- pothesis: A Multisectoral Test," The Journal of Political Econom , LXXV (June, 1967), pp. 241-49. 22B. S. Minhas, "The Homohypallagic Production Func- tion, Factor-Intensity Reversals, and the Heckscher-Ohlin Theorem," The Journal of Political Economy, LXX (April, 1962), pp. 138-56. ' Wassily Leontief, "An International Comparison of Factor Cost and Factor Use," The American Economic Review, LIV, No. 4 (June, 1964), pp. 335-45. David Stafford Ball, "Factor-Intensity Reversals in International Comparison of Factor Cost and Factor Use," The Journal of Political Economy, LXXIV (February, 1966), pp. 77-80. 23Heckscher, p. 289. 20 than export and import flows also eliminates the need for assuming equivalent demand functions within each region.24 A test of the Heckscher-Ohlin hypothesis concerning regional production concentration is presented in Chapter II. The test involves finding the rank correlation between industry concentration ratios and capital-labor ratios for 71 three-digit SIC industries. Chapter III presents a test of the Classical theory of trade. The Classical model postulates that comparative cost advantages result from higher relative labor productiv- ity. The hypothesis to be tested is that relative labor productivity should be positively correlated with industry concentration ratios in each region. Chapter IV takes into consideration the role of nat- ural resources and an industry's dependency on external sources of raw materials. It is hypothesized that a high level of dependency on external sources of raw materials will influence industry concentration and may override either the factor proportions or labor productivity determinants of trade. The final chapter consists of a summary and review of the conclusions resulting from the tests. 24Jones, p. 6. CHAPTER II TESTS OF THE HECKSCHER-OHLIN MODEL Introduction and Review of the Literature Only since the early 1950's has a concentrated ef— fort been made to test empirically the Heckscher-Ohlin hy- pothesis. The earliest test was part of an extensive United States-Britain trade study by MacDougall.l His purpose was to determine whether United States exports were relatively more capital intensive than British exports. If this was the case, the United States should show a larger share of the world market, relative to the United Kingdom, in rela- 2 He found, however, tively capital intensive commodities. that Britain's largest export industries, for exports to third countries, had capital-labor ratios above the average for Britain and the United States; while United States ex- port industries, for exports to third countries, had capital- labor ratios below the average. He thus concluded that his evidence rejected the Heckscher-Ohlin hypothesis. 1G. D. A. MacDougall, "British and American Exports: A Study Suggested by the Theory of Comparative Costs," Egg- ggmic Journal (Part I: December, 1951, pp. 697-724; Part II: September, 1952, pp. 487-521). 2Clement et al., p. 99. 21 22 The most controversial test of the Heckscher-Ohlin theory was conducted by Wassily Leontief using 1947 input- output data for the United States.3 His purpose was ". . . to find out whether it is true that the United States exports commodities the domestic production of which absorbs relatively large amounts of capital and little labor and imports foreign goods and services which, if we had pro- duced them at home, would employ a great quantity of indig- enous labor but a small amount of domestic capital."4 With the available data, Leontief determined the capital and labor needed to produce a desired dollar value of some output. He then considered a one million dollar decrease in exports and competing imports, all goods being reduced in equal pro- portion. In order to replace the competing imports by do- mestic production, using resources from the reduced export good production, Leontief found that less labor, but more capital would be required than would be released from export production. In other words, United States exports were labor intensive relative to import substitutes produced 3W. W. Leontief, "Domestic Production and Foreign Trade; the American Capital Position Re-examined," Proceed- ings of the American Philosophical Society, XCVII (September, 1953), pp. 332-49. W. W. Leontief, "Factor Proportions and the Struc- ture of American Trade: Further Theoretical and Empirical Analysis," The Review of Economics‘and Statistics, XXXVIII, No. 4 (November, 1956), pp. 386-407. 4Leontief, Proceedings, p. 339. 23 in the United States. This conclusion contradicts the Heckscher-Ohlin hypothesis and is known as the Leontief scarce-factor paradox. Leontief's explanation was that because United States labor productivity exceeded that of the rest of the world by approximately 200 percent, the United States could be considered as having three times as much labor using the world productivity standard. The United States is, in this sense, a labor abundant country, and it "resorts to foreign trade to save its capital and to dispose of its relative surplus labor."5 Immediately after the Leontief paper was published, a rash of critiques emerged, generally arguing that the Heckscher-Ohlin hypothesis was not credited and that Leontief had mis-interpreted his results, selected a poor year for his tests, or erred in applying input-output anal- ysis to international trade. One of the earliest criticisms of Leontief was by P. T. Ellsworth.6 Ellsworth felt that to determine the relative factor intensity of import substitution goods, their production coefficients should be compared to those of the same goods in the foreign country, not to other goods SIbid., p. 344. 6P. T. Ellsworth, "The Structure of American Foreign Trade: A New View Examined," The Review of Economics and Statistics, XXXVI, No. 3 (August, 1954), pp. 279-85. 24 within the United States. Thus, labor intensive import sub- stitutes relative to other goods in the United States may not be contradictory to the Heckscher-Ohlin hypothesis, as these goods may be capital intensive relative to actual im- port goods from foreign countries. This observation receives some empirical support from later tests. (See pages 26-28.) He also offered an explanation of higher United States labor productivity in terms of a more abundant supply of comple- mentary factors such as entrepreneurship, natural resources, and capital. Kravis7 also showed little concern about the so-called paradox. He argued that many goods are imported because they are products of natural resources which have become relatively scarce in the United States. To assure a con- tinuing supply of these products, the United States financed the construction of facilities abroad with the result that these goods are produced under capital intensive conditions and therefore explain some of the capital intensive imports. A group of writers including Kenen, Becker, Colberg, and Swerling criticized the paradox conclusion by arguing that capital was poorly defined by Leontief because he 7Irving B. Kravis, "'Availability' and Other Influ- ences on the Commodity Composition of Trade," The Journal 2f Political Economy, LXIV, No. 2 (April, 1956), pp. 143- 55. 25 excluded human capital.8 Colberg perhaps expresses their position best: "The simplest explanation of the paradox may be that the term '1abor' has included too much, while the term 'capital' has comprehended too little of our pro- ductive resources."9 Kenen finds that the paradox does in fact disappear in the limiting case where skill differentials of labor are assumed to be due to the quantity of capital invested in man.10 A last group of writers is concerned with Leontief's contention that United States labor productivity is three times that of the world average. From data obtained by means of a questionnaire, Kreinin found Leontief's labor produc- tivity differential too high, and maintains that United States labor productivity is l%-to li-times as great as foreign labor.11 8Peter B. Kenen, "Nature, Capital, and Trade," The Journal of Political Economy, LXXIII, No. 5 (October, 1965), pp. 437- 60. Gary Becker, "Investments in Human Capital: A The- oretical Analysis," The Journal of Political Economy Supple- ment (October, 1962), pp. 9-49. Boris C. Swerling, "Capital Shortage and Labor Sur- plus in the United States?“ The Review of Economics and Statistics, XXXVI, No. 3 (August, 1954), pp. 286- 89. Marshall R. Colberg, "Human Capital as a Southern Resource," Southern Economic Journal, XXIX (January, 1963), pp. 157-66. 9Colberg, p. 158. 10Kenen, p. 457. llMordechai E. Kreinin, "Comparative Labor Effective- ness and the Leontief Scarce-Factor Paradox," The American Economic Review, LV, No. 1 (March, 1965), pp. lSI-89. 26 He concludes that such a small margin of superior productiv- ity is insufficient to make the United States a "labor abun- dant" country. Studying the same problem, Diab and Bhagwati obtained conflicting conclusions, although both estimated capital- 1abor ratios with Cobb-Douglas production functions. Diab, holding capital productivity constant over all countries, agrees with Leontief's conclusions, while Bhagwati, holding labor productivity constant, agrees with Kreinin.12 In an effort either to lend more substantive support or to reject the Leontief paradox, four input-output studies have been undertaken. The first used a "replacement” approach 13 An index similar to Leontief's, applying it to Japan. of comparative capital-labor intensities was computed, and it was found that an average one million yens worth of ex- ports embodies more capital and less labor than is required to replace, domestically, one million yens worth of compet- itive imports. 12M. A. Diab, The United States C_pital Position and the Structure of its Forei n Trade (Amsterdam: North- Holland Publishing Co., 1956). Jagdish N. Bhagwati, "Some Recent Trends in the Pure Theory in International Trade," International Trade Theory, in a Develo in World, eds. Roy Harrod and Douglas Hague (New York: St. Martin's Press, 1963). 13Masahiro Tatemoto and Shinichi Ichimura, "Factor Proportions and Foreign Trade: The Case of Japan,” The Review of Economics and Statistics, XLI, No. 4 (November, 27 If Japan is assumed to be relatively labor abundant, then those results appear at variance with the Heckscher- Ohlin model. Nonetheless, because only 25 percent of her trade is with developed nations while 75 percent is with underdeveloped nations, Japan can be considered capital abundant relative to underdeveloped countries and would 14 The therefore export relatively capital intensive goods. opposite would be true with developed nations. And, in fact, the capital-labor ratio of her exports to the United States is lower than that for all other exports. Thus, when Japan's trade is broken down with respect to her trading partners, the Heckscher-Ohlin theory is supported, and the Leontief paradox is rejected. A second study using East German data was done by Stolper and Roskamp.15 Their findings showed East German exports to be relatively capital intensive. Because East Germany is probably the most capital abundant of the East Bloc countries, with which she carries on 75 percent of her trade, this study also supports the Heckscher-Ohlin hypoth- esis. ”£141., p. 445. 15Wolfgang F. Stolper and Karl w. Roskamp, "An Input- Output Table for East Germany with Applications to Foreign Trade," Bulletin of the Oxford University Institute of Sta— tistics, XXIII, No. 4 (November, 1961), pp. 379-52. 28 A third study concerned the Canadian trade structure.16 The results were that Canada's exports were found capital intensive while imports were relatively labor intensive. This held for total exports, exports to the United Kingdom, and exports to the United States. Such results tend to re- ject the Heckscher-Ohlin hypothesis. A fourth study considered the structure of Indo- United States trade.17 The hypothesis tested was that "In- dian exports to the United States absorb in their production relatively more labor than her competitive imports from the United States which, if produced at home [in India] would 18 The findings of the require relatively more capital." study support the Leontief paradox as Indian exports were found more capital intensive than imports. Finally, two "indirect" tests of the Heckscher-Ohlin theory imply support for the hypothesis. In the first, Kravis considered wage rates in export and import industries.19 16Donald F. Wahl, "Capital and Labour Requirements for Canada's Foreign Trade," The Canadian Journal of Economic and Political Science, XXVII, No. 3 (August, 19617, pp. 349— 58. 17R. Bharadwaj, "Factor Proportions and the Struc- ture of Indo—United States Trade, " The Indian Economic Journal, X, No. 2 (October, 1962), pp. 105- 16. laIbid., p. 105. 19Irving B. Kravis, "Wages and Foreign Trade, " The Review of Economics and Statistics, XXXVIII, No. 1 (February, 1956), pp. 14- 30. 29 He found that a relatively high share of United States ex- ports are produced by high-wage industries, and a relatively high share of competing imports consist of goods produced domestically by low-wage industries. Hypothesizing that the higher wages are due to a greater supply of capital and therefore higher productivity, this would tend to support the Heckscher-Ohlin hypothesis. No data on capital per unit of output was offered, however. In the second indirect test, Tarshis analyzed rela- tive commodity prices in hopes of drawing some conclusions about trade flows.20 He found that price ratios of capital intensive goods relative to labor intensive goods were lower in the United States, while the opposite held for less cap- ital abundant countries. The implications of these relative price ratios for trade are consistent with the Heckscher-Ohlin hypothesis. Two implications of these tests exist for this dis- sertation. First, because of the inconclusiveness of the results using international data, tests using regional data may be preferred. The second implication stems from the fact that in virtually all of the studies some comments exist about the 20Lorie Tarshis, "Factor Inputs and International Price Comparisons," The Allocation of Economic Resources, ed. M. Abramovitz (Stanford, California: Stanford Univer- sity Press, 1959), pp. 236-44. 3O drawbacks and problems of the tests such as the influence of different demand conditions between countries, the pos- sibility of different production functions between countries, the differences in the quality of factors between countries, and the influence of tariffs and other trade restrictions. As indicated in Chapter I, these are essentially eliminated with the use of regional rather than international data. Only one previous study has dealt with regional data of the United States.21 Dividing the United States into South and non-South regions and using location rather than trade flows as indicators of comparative advantage, Moroney and Walker hypothesized that: "There is an inverse rank ordering between capital-labor ratios and location quotients" in the South.22 The rank correlation was positive, however, although not highly significant. This result gives some indication that the South has a comparative advantage in producing relatively capital intensive goods, a conclusion inconsistent with the Heckscher-Ohlin hypothesis. The authors then eliminated certain "natural resource" oriented indus- tries from their tests, but the results were still not con- sistent with the Heckscher-Ohlin hypothesis. 21John R. Moroney and James M. Walker, "A Regional Test of the Heckscher-Ohlin Hypothesis," The Journal of Political Economy, LXXIV (December, 1966), pp. 573-86. 221bid., p. 581. 31 The Case for Regional Tests The tests of the Heckscher-Ohlin model presented in this chapter are similar to those undertaken by Moroney and Walker. Using regional data in the United States, the country was divided into two sections called the South and the non-South. This division is based on the United States Census Bureau classification of areas. The South is composed of the South Atlantic, East South Central, and West South Central census regions. This regional division is also convenient because several studies pertaining to wage differentials have been based on the same South-non—South division. If a wage dif- ferential exists between the South and non-South, and the cost of capital differential does not offset it, there is a presumptive evidence of differential relative factor sup- plies in the two regions. The evidence of wage differentials is clear. Moroney and Walker computed an index of wage differentials and found that the average hourly wage of production workers in the South was 78 percent of the non-South average, while average annual non-production salaries were 87 percent of the non— South average.23 A second study found Southern skilled maintenance wages to be 83 percent to 94 percent of the national median 23Ibid., p. 577. 32 and unskilled plant labor wages to be 67 percent to 79 per- cent of the national median.24 A third study points out that since 1947, relative earnings in the South have remained 20 percent to 25 percent below the national average.25 Evidence of regional differences in the cost of cap— ital is scarce. However, a survey of interest rates on a geographical basis from the Federal Reserve suggests that the cost of capital in the South is at worst, equal to that in the non-South, and may even be lower.26 The evidence of relatively lower wages in the South suggests that the South is relatively labor abundant. Can one assume with confidence that these wage differentials result mainly from labor supply differences? In general, the answer is probably "yes." Fuchs and Perlman suggest the differences exist due to low-wage industry mix in the 27 South plus relatively lower earnings for similar work. Gallaway also feels that wage differentials imply lower 24Toivo P. Kanninen, "Wage Differences Among Labor Markets," Monthly Labor Review, XLIV (June, 1962), p. 616. 25Victor Fuchs and Richard Perlman, "Recent Trends in Southern Wage Differentials," The Review of Economics and Statistics, XLII (August, 19607, p. 295. 26Board of Governors of the Federal Reserve System, Federal Reserve Bulletin, XLIV (January, 1958, and April, 19585, pp. 34, 312. 27Fuchs and Perlman, p. 293. 33 capital—labor ratios in the South.28 In a critique of the Moroney and Walker study, Estle suggests that in fact the South may be relatively capital abundant.29 Estle found that some industries in 1957 have higher capital-labor ratios in the South than in the non- South, where the capital-labor ratio is measured as gross book value of capital per man year. Nonetheless, it seems that his finding is attributable mainly to the relatively more recent investment in plant and equipment in the South, rather than to a higher relative overall regional capital endowment. Therefore, Estle's study might suggest that the assumption of identical production functions between regions does not hold. There will be a further discussion of this when the results are evaluated. The absence of overall regional capital stock esti- mates requires that regional factor endowments be defined in terms of relative factor prices. Thus, if the wage rate is lower in a given region relative to another, the impli- cation is that the low wage region is relatively labor- abundant. A potential difficulty in testing the Heckscher-Ohlin hypothesis using SIC three-digit industries is that regional 28Gallaway, ”The North-South . . .," p. 270. 29Edwin F. Estle, "A More Conclusive Regional Test of the Heckscher-Ohlin Hypothesis," The Journal of Political Econom , LXXV (December, 1967), pp. 886e88. 34 capital stock estimates are not available. Hence national capital—labor ratios must be used to rank the industries according to capital-intensity of production. The strong factor—intensity hypothesis, which seems to have a solid empirical basis in the United States,30 ensures that the national ranking is preserved among regions. Thus the use of national ratios should not lead to ambiguous test results. In this thesis the Heckscher-Ohlin hypothesis is tested in two sets of tests as follows. Firstly, two re- gions, the South and New England, are each identified as being relatively labor abundant by comparison with the rest of the nation. Secondly, it is well-known that a ranking of commodities according to a region's "abundant-non-abundant" input ratios provides a corresponding ranking by order of 31 Hence the research hypothesis is comparative advantage. that there is a negative correlation between industry capital- 1abor ratios and concentration of production in each of these regions. By the nature of the data used in the subsequent tests, the capital-labor ratio is a sufficient determinant of a commodity's intensive factor. That is, each industry's measure of output is value added, and thus current factor input proportions determine the factor intensity. 3OMoroney, "The Strong—Factor Intensity. . . ." 31Jones, "Factor Proportions . . .," p. 6. 35 Tests in the South To test the above hypothesis, gross and net capital- labor ratios were computed for 71 Standard Industry Classifi— cation three-digit manufacturing industries (see Appendix I). Data were taken from the Census of Manufactures and the Ag: nual Survey of Manufacturers.32 Capital-labor ratios were computed by dividing book value of assets by employees for 1957. As mentioned earlier, national capital-labor ratios were used. These ratios are felt to be adequate for two reasons: first, the assumption of similar production functions between regions appears to be reasonable; second, under the strong factor-intensity hypothesis, rankings of capital-intensity nationally give identical regional rankings. Concentration ratios for each industry were computed by dividing value added in the South, vi, by value added in the nation, Vi. digit industries for the year 1957 are not published, so Regional value added data for SIC three- 1958 value added figures are used. This change should not have any significant influence on the results as capital- labor ratios are for the end of year 1957, and would not change to any significant degree in 1958. 32U.S. Bureau of the Census, Annual Survey_of Manu— factures,,l957 (Washington: U.S. Government Printing Office, 1959)} U.S. Bureau of the Census, Census of Manufactures, 1958 (Washington: U.S. Government Printing Office, 1961). 36 The capital-labor and concentration ratios are then ranked in ascending order (see Appendix II, Tables 1 and 2, for rankings). Kendall's 75 is used to show the degree of rank correlation. (A summary of all results of tests of the Heckscher-Ohlin model is shown in Appendix III.) Using the gross capital-labor ratios,‘27 is +.O632, not signifi- cant at the ten percent level. Using the net capital-labor ratios,'?7 is +.0664, not significant at the ten percent level. The sign of the coefficient in both cases was "wrong"; that is, the concentration ratios are somewhat higher in the South for high capital-labor industries. Clearly, the hypothesis fails to predict industry location based on rela- tive factor endowment. Several reasons for these results are possible. First, the model tested contains only two factors of pro— duction. Obviously, more factors play a role in production, and Heckscher and Ohlin both considered the range of possi— bilities. Heckscher, for example, states: "It must be stressed at this point that the term 'factor of production' does not refer simply to the broad categories of land, cap- ital, and labor, but to the different qualities of each of these."33 In addition to such differences in quality, nat- ural raw materials and climate conditions are potentially important. 33Heckscher, p. 279. 37 In order to compensate for quality differences in labor inputs, a second test was made after new capital-labor ratios were computed using labor input figures adjusted for productivity differences. Assuming competitive conditions in the labor market, wage differences will reflect produc- tivity or skill differences. By eliminating these differ- ences, one more closely approaches the condition of homogene— ous factor inputs. For each industry, an annual average wage was com- puted by dividing total annual payroll by total employees. The source of data was the Annual Survey of Manufacturers.34 Next, an index was derived by taking a ratio of each indus- try average annual wage to a national all-industry average wage. Finally, the original labor input figures for each industry were weighted by the relative wage index. Kendall's Wt'was computed as a measure of rank cor- relation. Using both gross and net capital, 1? was not sig- nificant at the ten percent level in either case, although the coefficients were somewhat higher than in the earlier tests. 1k in both cases was positive, the opposite of that hypothesized. Using gross capital, Qf‘was equal to +.166, while with net capital,‘27 was equal to +.135. (See Appen- dix II, Tables 3 and 4, for rankings.) A second possibility is that the assumption of 34U.S. Bureau of the Census, Annual Survey. . . . 38 complete factor immobility between regions may not hold. Because of the difference in "natural" factor endowments, that is, climate and natural resources, there will exist goods which the non-South will be unable to produce, but will demand. If capital is not available in the South, it may come from the non-South with the result that some South- ern industries will become capital intensive. This condi- tion is not a complete contradiction of the Heckscher—Ohlin hypothesis, however, because the source of capital was a capital abundant region. It merely indicates that one of the Heckscher-Ohlin assumptions does not hold, a possibility that seems particularly strong when interregional rather than international data are used.35 A third possible explanation is that the assumption of homogeneous factors does not hold. This is most likely to be true for labor, where differences in quality will re- sult in differences in labor productivity. An attempt was made to eliminate these differences but the disappointing results of the Heckscher-Ohlin test were not influenced to any degree. A final possible explanation for the test results is that the South is a highly atypical region, having lagged behind the non-South in industrial development. To explore this possibility, a new region, New England, was chosen for 35Kravis, "'Availability' and . . . ." 39 comparative purposes. New England's industrial structure is well established; and relative to the rest of the United States, it is a labor abundant region (see below). The greater importance of recent Southern industrial development can be seen by comparing the range and direction of percentage changes in relative concentration between 1947 and 1958 in the South and New England. For the South the range was from -21 percent to +325 percent, with nine indus- tries showing decreases and 47 showing increases. For New England, the range was from -85 percent to +238 percent, with 28 industries showing decreases and 25 showing increases. The South, then, was clearly in a developmental and growth stage; and one might not expect tests performed in a static framework to yield significant results. To see if industrial growth patterns in the South were consistent with the Heckscher-Ohlin hypothesis, percent- age changes in concentration were ranked with gross capital- labor ratios. (See Appendix II, Table 5, for ranking.) A coefficient of -.256, significant at the .005 level, was found, indicating that Southern industrial development did take place more strongly in relatively labor intensive in- dustries. This result is in agreement with that obtained by Moroney and Walker using a sample of two-digit industries. Tests in New England The basis for establishing New England as a relatively labor intensive region is the same as that used for the South, 40 that is, lower relative wage levels. Average annual wage levels for New England and non-New England for a 68 industry sample, show that the New England level is 97 percent of the non-New England level. In addition, a study by Eisen- menger found that the average hourly wage per employee man- hour in New England, 1958, is less than 100 percent of the United States average in 15 of 18 two-digit SIC industries.36 By performing the same rank correlation tests as were done for the South using static concentration for 1958, it is found that the structure of specialization in New Eng- land can be explained by the Heckscher-Ohlin model. This offers further evidence that the South is unique because of its comparatively recent industrial development. The original sample of 71 SIC three-digit industries used in the South tests is reduced to 68 since New England data were not available for Industries 206, 322, and 333. It is felt that this slight difference will not invalidate any comparison between test results of the two samples. (See Appendix II, Tables 6 and 7, for rankings.) Rank correlation tests between gross and net capital- labor ratios and concentration ratios yielded coefficients of -.228 and -.243, significant at the pg .01 and p g‘ .005 levels respectively and of the sign hypothesized. These 36Robert W. Eisenmenger, The Dynamics of Growth in New En land's Econom 1870-1964 (Middletown, Connecticut: Wesleyan University Press, 1967), p. 28. 41 results in three-digit industries are in agreement with those obtained by Estle in a sample of two-digit industries. For New England, then, because it is relatively labor abun- dant, industries with relatively low capital-labor ratios tend to be more highly concentrated there. Although New England did not go through a period of latent industrial development as did the South, changes in relative industry concentration in New England between 1947 and 1958 took place in a pattern as would be predicted by the Heckscher-Ohlin hypothesis. That is, those industries with relatively low capital-labor ratios showed a tendency toward increasing relative concentration. (See Appendix II, Table 8, for ranking.) A rank correlation test between capital—labor ratios and percentage changes in concentra- tion yielded a coefficient of -.129, not significant, but of the sign hypothesized. It might be noted that Estle ob— tained a similarly weak result in a corresponding test among two-digit industries. Conclusions One's impression of the comparative analysis between the two sets of tests is that the Heckscher-Ohlin model has greater explanatory power for an already industrially devel— oped region. New England was the earliest industrially de- veloped region in the United States. As industry expanded into other regions, industries with comparative advantages stayed and grew in New England. Those with comparative 42 disadvantages shifted into other regions where a comparative advantage existed for them. Thus, the tendency for labor intensive industries to be concentrated in and to continue to grow in New England is consistent with the factor propor- tions theory of comparative advantage. Other forces, how- ever, appear to dominate the existing pattern of industry concentration in the South, yet developmental trends appear to follow the factor proportion hypothesis. CHAPTER III TESTS OF THE CLASSICAL MODEL Introduction The Classical explanation of trade flows, as empha- sized in Chapter I, is based on comparative labor costs. Although Ricardo failed to specify a complete model, his use of the labor theory of value led to the expression of comparative costs in terms of relative labor productivity. Beginning with the contention that trade occurs because of different relative prices between countries, it can be shown that these price differences are a result of differences in relative labor productivities. Recall from Chapter I that under competition, total revenue for goods x and y were said to equal costs: er + wLx and rKy + wLy.l With the assumption of equivalent capital-labor ratios in the production of both goods within each country, the average products of capital and labor were equated between goods by the use of a scalar, at, such that Ky = aLKx and Ly = dLLx. Dividing the cost ratio, er+wa, by er, a new cost ratio expressed only in terms K r+L w Y Y 1See page 5. 43 44 of labor productivity is found. The use of the scalar per- mitted the reduction of a two-factor model to a one-factor model of the Ricardian type. Thus, an hypothesis can be derived for a two country-two commodity model which states that countries A and B export (to one another) goods x and y (PX P respectively, because 1535- B’ which is a direct result (‘ “:)> (“75-33% y Most tests of the Classical model have been based on the above productivity concept. Labor productivity is not the only factor determining labor costs, however, since the price of a unit of labor is a crucial determinant of cost. Thus, in some tests, the influence of wages has been included. This influence can be added into the model so that Lx wx x wx the condition for trade is written as L; W; A< r; w; B’ that is, one assumes that internal price ratios are propor- tional to internal wage bill ratios. Either labor costs or labor productivity can be used to form the conditions for comparative cost advantage. This is in line with the traditional Classical attachment to the. labor theory of value. Its weakness, of course, is that it ignores capital costs. Previous Tests The first test of the Classical model was made by 4S MacDougall.2 Using a two country, g_commodity model, he hypothesized that each country "will export those goods for which the ratio of its output per worker to that of the other exceeds the ratio of its money wage rate to that of the other."3 Using productivity data for the year 1937, MacDougall found the United States weekly wage to be two times that of the United Kingdom. According to the hypoth- esis, in those industries where United States productivity is more than two times United Kingdom productivity in the same industries, the United States should have a larger share of the export market to third countries. The converse is true when United States productivity is less than twice as high as in the United Kingdom. lThese relationships held in 20 out of 25 industries. Relative wages were then explicitly included as MacDougall computed relative wage costs per unit of output for each of the 25 industries. In general, these costs were found to be less in the United States in those industries where United States productivity exceeded that of the United Kingdom by more than two times. In addition, relative wage costs per unit of output were inversely related to relative 2G. D. A. MacDougall et al., "British and American Productivity, Prices and Exports: An Addendum," Oxford Economic Papers (October, 1962), pp. 297-304. 3MacDougall, "British and American . . .," p. 697. 46 export shares.4 Finally, MacDougall related price ratios to relative export shares with regressions for each year from 1913 to 1948. The results were favorable with the lowest correla- tion coefficient being -0.73. Despite comparative advantage to third countries, United States and United Kingdom exports to these third countries and to one another were not complete. That is, the Classical consequence of comparative advantage, complete specialization, was not found to exist. MacDougall attrib- uted this to relative tariff rate differences, transporta— tion costs, imperfect markets, and non-homogeneous goods.5 Another study of the Classical theory was made by Robert Stern.6 As productivity data were updated, Stern in effect expanded upon and further strengthened the MacDougall study. Stern's purpose was to find the "extent to which differences in the relative labour productivity and production costs . . . are reflected in differences in the relative export performance of the two countries."7 For the year 1950, Stern found that United States 4Ibid., p. 698. 5Ibid., p. 699. 6Robert M. Stern, "British and American Productivity and Comparative Costs in International Trade," Oxford Eco- nomic Papers, XIV (October, 1962), pp. 275-96. 7Ibid., p. 275. 47 weekly wages average 3.4 times those in the United Kingdom. A productivity difference of more than 3.4 times was required if the United States was to have the larger share of exports to third markets. Twenty out of twenty—four industries con- formed to expectations. Stern then undertook three correlation studies. First, relative productivity and relative export shares were correlated, yielding a coefficient of +.44; positive as hypothesized. Second, unit labor costs were correlated with relative export shares resulting in a coefficient of -.43; negative as hypothesized. Finally, net cost ratios were correlated with relative export shares, where it was assumed that these cost ratios were indicators of compara- tive resource productivity rather than labor productivity alone. The coefficient obtained was -0.36; negative as hy- pothesized.8 A third study of the Classical theory was done by Balassa,9 and it too followed the pattern set by MacDougall. The first part consisted of correlating labor productivity ratios with export ratios for 1951. The countries involved were again the United States and the United Kingdom. 81bid., p. 293. 9Bela Balassa, "An Empirical Demonstration of Clas- sical Comparative Cost Theory," The Review of Economics and Statistics, XLV (August, 1963), pp. 231-38. 48 Assuming a linear relationship, the £_coefficient was +.80, positive as hypothesized, while using a logarithmic relation- ship between variables yielded an £_coefficient of +.86.lo Both coefficients strongly supported the Classical hypoth- esis. Next, Balassa considered wage ratios as an additional variable in the regression equation. The £_coefficient, assuming a linear relation, was little changed from that found in his first test. The partial correlation coeffici— ent between wage ratios and export ratios was only .24, pos- itive as hypothesized, but not significant at the five per- cent level. Transformation to a logarithmic relation did not improve the results.11 Finally, Balassa correlated export ratios with net unit cost ratios finding £.coefficients of -.60 and -.71 for linear and logarithmic relations respectively.12 In general, all of the above tests yielded very good results, indicating substantial evidence in support of the Classical model. These studies are not without weaknesses. Bhagwati's critique is probably the most extensive.13 Bhagwati is primarily concerned with the tenuous loBalassa, p. 235. llIbid., p. 236. 12Ibid., p. 237. 13Bhagwati, "The Pure Theory. . . ." 49 relationship between the hypotheses tested by MacDougall, Stern, and Balassa, and what he feels are the original "Ricardian" hypotheses which reflect differences in relative productivities or relative unit labor costs between countries. The breakdown occurs, according to Bhagwati, because "the assumption that the relative prices of exported goods will be lower than those of imported goods is now replaced by the postulation of some relationship between (United States- United Kingdom) price ratios of third-market exports and (United States-United Kingdom) shares in third markets."l4 Specifically, he questions the use of cross-section investi- gation to analyze the relation between third market export ratios between the United States and the United Kingdom, and their price ratios for any one industry. Bhagwati also considered a problem common to any test of the Classical model. The derivations of the two "Ricardian“ hypotheses suggested at the beginning of the chapter rely on the assumption that prices are closely re- lated to labor productivities and/or unit labor costs. This assumption was also implicit in the tests of MacDougall, Stern, and Balassa. Bhagwati tested this prOposition using data from the three previously mentioned studies and found that these data do not support the required assumption. He concluded that "a fullblooded test of these [Ricardian] l41bid., p. 11. SO hypotheses, directly examining the ranking of [bilateral] exports and imports by comparative labour productivities and/or unit wage-cost ratios, is impossible to carry out with this information. . . ."15 Preliminary Tests of the Classical Model The tests of the Classical hypotheses to be pre- sented in this chapter involve United States regional data rather than international data. Relative production con- centration is used because relative export flows between regions are not known. The regional divisions of the United States and the production concentration ratios are the same as those used in the previous chapter to test the Heckscher- Ohlin hypothesis. Data for the year 1958 are taken from the U. S. Census of Manufactures.16 The first hypothesis tested is that those industries with higher labor productivity in the South relative to the non-South will be more concentrated in the South relative to the nation as a whole. Specifically, the research hy- pothesis is that productivity ratios will be positively correlated with production concentration ratios. This hy- pothesis arises from the expression of the Classical model where the initial condition for trade to take place between, l51bid., p. 14. l6U.S. Bureau of the Census, Census of Manufac- tures. . . . 51 P P say, countries A and B, that is, Pl) A<> APf—’ B' Rela— Y tive labor productivities are assumed to be representative of relative costs and therefore of relative commodity prices. Productivity, measured in each industry by dividing regional value added by regional employment, is expressed as dollar value of output per man year. To test the hypothesis, Kendall's 15 was computed to measure rank correlation between production concentration ratios and productivity ratios for 71 three-digit SIC indus- tries, where productivity ratios are the ratio of productivity in the South to productivity in the non-South. Rankings of concentration ratios and each industry's corresponding pro- ductivity ratio for all Classical tests are shown in Appen- dix IV and all test results are listed in Appendix V. '2: was found to be -.202, significant at the five percent level but of the "wrong" sign. In seeking an explanation for these perverse re— sults, it seems reasonable to first consider the conditions specified for the hypothesis tested. Of primary interest is the relationship between productivity and costs. Is productivity a legitimate proxy of costs and therefore of prices? The answer depends a great deal on the role of wages in costs. If interindustry wage differences exist 52 and wages reflect skill differences as they would under com- petitive conditions and moderate labor mobility, high wages are paid to highly skilled, and therefore productive workers. Relatively high productivity in isolation does not insure relatively low unit costs, however. Thus, the strict Ricardian model expressed in terms of productivity may be insufficient to explain trade flows or industry location concentration. Opinions differ as to the relative role of wages and productivity. Forcheimer feels that wages may play a significant role in the structure of comparative costs.17 In consideration of the important determinants of compara- tive advantage, he suggests wages, productivity, and the ratio of average total costs to average wages. If relative wage differences are to play a leading role, the other two items must have minor effects or offset one another. Under certain conditions this will occur. When manufacturing in- dustries are considered, productivity differences due to "natural" conditions may be minor, allowing wage differences to exert the primary net effect on total costs. In addition, industries whose transportation costs are low relative to total costs and whose purchase of raw materials can be made at world prices are likely candidates for wages to dominate l7Karl Forcheimer, "The Role of Relative Wage Dif— ferences in International Trade," The Quarterly Journal of Economics, LXII (November, 1947), pp. 1-30. 53 cost determination. Specifically, Forcheimer feels that light manufacturing industries seem to fit these conditions. Kravis, on the other hand, feels that wage differ- ences are not likely to alter the productivity determinants of comparative advantage, and in fact shows that export in— dustries in the United States pay relatively high wages.19 In addition, by comparing hourly earnings of different in- dustries between the United States and Japan, he finds evi- dence that wage structures of noncompeting groups are sim- ilar in different countries, and therefore wage differences between industries have little effect on comparative advan- tage between countries.20 Kravis also finds that the average wage level in a country is representative of average productivity in that country, and therefore differences in industry costs between countries are more apt to be a function of productivity dif- ferences between industries.21 Because of the possibility that wage differentials may have influenced the "pure" productivity tests, a second test of the Classical model is undertaken, in which concen- tration ratios are ranked with average labor cost ratios 18Ibid., p. 24. 19Kravis, "Wages and Foreign. . . ." 20Kravis, "'Availability' . . .," p. 146. 21Ibid. 18 54 for each industry. Average labor costs are computed by dividing the average annual wage per man by productivity, that is, by value added per man year. Costs can then be expressed by stating that each $1.00 of value added per man year costs SX.in wages. Specifically, the hypothesis tested is that average labor cost ratios will be negatively correlated with con- centration ratios. Using Kendall's rank correlation test, 15 is +.116, not significant, but of the "wrong" sign. Both models considered thus far fail to explain relative production location in the South and non-South. One possible explanation is that average labor cost is an insufficient cost concept to be a price proxy. Capital and raw material costs certainly are a part of the average total cost or the marginal cost of producing any good. Thus, it is possible that neither labor costs nor labor productivity by themselves are sufficient to indicate comparative advan- tage in the production of any one good between regions. In order to reduce the influence of other factors of production and to more closely approximate the condition implied in the Classical labor theory of value, additional tests were made which included only labor intensive indus- tries; that is, those industries in which labor costs account for 60 percent or more of total costs. Sixteen industries are tested for rank correlation between concentration ratios and average labor cost ratios and labor productivity ratios. 55 Using cost ratios, Kendall's 7& is -.183, negative as hypoth- esized but not significant. Using productivity ratios, 1b is +.033, positive as hypothesized but also not significant. Although these results are not statistically significant, the fact that the signs were reversed in both tests gives some indication that in non-labor intensive industries other variables override labor cost differences. A weakness in all of the preceding tests is that they were performed in a framework that is rigorously sug- gested by the Ricardian "two country, two commodity" model. Hence the model is not strictly appropriate for multi-com- modity tests. Tests of an Alternative Classical Model Frank Graham's effort to expand trade theory to a multi—country, multi-commodity setting while still basing comparative advantage on labor costs yields several ideas for a more comprehensive testing of the Classical model.22 In his article “The Theory of International Values Re-exam- ined," Graham states that, "It is to the assumptions of trade between two countries only and in but two commodities that attention will here be drawn in an endeavor to show 22Frank D. Graham, The Theory_of International Values (Princeton: Princeton University Press, 1948)? Frank D. Graham, “The Theory of International Values Re-examined," Quarterly Journal of Economics, XXVIII (Novem- ber, 1923), pp. 54-86. 56 that to construct a theory of international values in this piecemeal way is a method so faulty as to have issued in wholly unwarranted inferences."23 With that statement, Graham launched into a series of numerical examples indicating gains from trade, the role of demand, and the basis behind relative ranking of more than two commodities according to comparative advantage. These were, in effect, general equilibrium models whose 24 In solutions were points of competitive equilibrium. these models, Graham assumed labor to be the sole source of productive power and that all goods were produced at constant labor costs.25 In Graham's more complete model, a country or re- gion, rather than specializing in only the one good in which it had a comparative advantage under the two commodity case, is now faced with a problem of optimal allocation of its labor among several uses. This problem is analogous to that of a firm choosing the optimal product mix in order to max- imize profits, subject to the constraint of resource limita- tion, and where each product requires factors in different 23Graham, "The Theory of . . ." (1923), p. 55. 24Lionel W. McKenzie, "Specialization and Effici- ency in World Production," Review of Economic Studies, XXI, NO. 1 (1954), pp. 165-800 25Lloyd Metzler, "Graham's Theory of International galues," The American Economic Review, XL (June, 1950), pp. 01-220 S7 proportions. The problem is simply a case in which two or more activities are competing for limited resources. If it can be assumed that all relationships are linear, then the op- timal solution can be found by solving the problem as a linear program. This is, in effect, a trial and error ap- proach. First, some initial obvious and feasible output combination is stated. For example, all resources may be allocated to the production of the good whose profit per unit is the highest. It is likely, however, that not all resources will be fully used, and an additional product will be included and a new combination of outputs considered. This search process continues until an optimal solution is reached, that is, one that maximizes profits. The important fact is that Graham had this general approach in mind. Thus, in a multi-commodity example, goods with the highest comparative advantage are more intensively produced and traded first, while those with lower compara- tive advantage are added in and are profitable only after the demand for the initial goods has been sufficiently sat- isfied to lower their gains from trade.26 Following this approach, the relevant concentration index for rank correlation tests should show the share of each industry in its region's output relative to that industry's 26Graham, "The Theory of . . ." (1923), p. 64. 58 share for the rest of the nation. Thus, a testable hypoth- esis will read: The South has a comparatively larger share of its own regional value added in those industries in which the South has the largest labor productivity advantage. For the two region study of this paper, the alter- native concentration measure would be a ratio of relative output concentration between the South and non-South. Thus, if the South has a labor productivity advantage over the non-South in a given industry, a higher percentage of re- sources in the South should be allocated to this industry than in the non-South, resulting in value added being rela- tively higher in the South. This concentration can be com- puted as the percent of value added in the South by an in- dustry, 2;, divided by the percent of value added in the v5 i i non—South by that industry, Egg, Letting :§_equal Cs and vi vns Vs vii-equal Cn’ the concentration measure is gs, For testing n purposes, one can hypothesize a positive rank correlation between relative industry concentration and labor productiv- ity ratios, and a negative rank correlation when labor costs are used. Rank correlation tests of the Classical model were made using the original sample of 71 industries and the con- centration concept suggested above. The results show that ’b’is of the "wrong" sign for both labor variables used and is significantly different from zero at the five percent 59 level for labor productivity (-.l99) and the ten percent level for labor costs (+.l3l). These results are clearly contrary to those predicted by the hypotheses. The tests show that there is a statis- tically significant indication that industries having rela- tively low labor costs in the South are more heavily concen- trated in the non-South, and those industries having rela— tively high labor productivity in the South are more heavily concentrated in the non-South. It becomes obvious then, that factors other than labor costs and productivity play a dominant role in deter— mining relative industry concentration between the two re- gions. Other possible factors are: the combination of de— mand and high transportation costs for the output of these industries, the dependency on raw materials from external sources and the location of these raw materials, and differ- ential rates of industrial development between the two re— gions. Because of the difficulty in finding any strong re- lationship between labor costs or labor productivity and some measure of production concentration, it seemed useful to attempt to determine in which industries and to what ex— tent comparative advantage should exist in the South under the Classical conditions. With this in mind, consideration was given to several articles which, based on Graham's works, are concerned with expanding the Classical model beyond the 6O two-country, two-commodity stage and with putting it in a form more conducive to empirical analysis.27 A solution of production specialization can be put in geometrical terms by use of a world production transfor- mation curve and the world price ratio line. Once the prob- lem goes beyond the three commodity stage, however, diagrams become impossible. With £_commodities, the production trans— formation curve becomes an g.dimensional polyhedron and the optimal solution is a point of tangency with the price hyper- plane. The intriguing thing about this model is that the optimal solution can be obtained by the application of lin- ear programming. Whitin, giving credit to Graham as his source of inspiration, suggests an objective function of maximizing the value of world trade where labor is the sole 28 source of factor inputs. McKenzie demonstrates the appli- cation of "activity analysis," the goal of which is ”the selection of productive processes which can be used to pro- . . 2 vide a max1mum output from given resources." 9 Jones' 27McKenzie. T. M. Whitin, “Classical Theory, Graham's Theory, and Linear Programming in International Trade," Quarterly Journal of Economigg, LXVII (November, 1953), pp. 520-44. Ronald W. Jones, "Comparative Advantage and the Theory of Tariffs: A Multi-Country, Multi-Commodity Model," ngiew of Economic Studies, XXVIII (June, 1961), pp. 161- 175. 28whitin, p. 542. 29McKenzie, p. 165. 61 approach is similar to that of McKenzie as he suggests solv— ing for the pattern of complete specialization lying on the world efficiency locus, although he considers minimizing labor inputs as well as maximizing output as a goal.30 Following the suggestion by the above writers that the Graham-Classical model can be solved through linear pro- gramming, an attempt was made to compare the optimal output predicted under the strict labor productivity theorem with the actual value added data for 71 SIC three-digit indus- tries for 1958. The problem, then, becomes one of finding the optimal allocation of labor between industries and re- gions so as to maximize total value added. It was assumed that constant costs prevail, and that labor is the sole in- put factor. The objective function to be maximized is v=gbijLij’ where V is value added for the nation; bij is labor produc- tivity, value added per man year, for all i industries and j regions, and where L is the number of man years allocated in each industry in each region. The function is subject to two sets of constraints. First, in each region, the sum of labor used in the indus- tries where output activity occurs cannot exceed the total labor supply available for that region, that is, g: Lij‘S-Lj° 3 Second, a minimum value added must be produced in 3OJones, p; 164. 62 each industry in the nation in order that the demand for the output from each industry be satisfied. This can be written as: vi 2;vi, where vi is the actual value added in the nation for the ith industry in 1958 and is used as the demand indicator for each industry's output. If this constraint were not imposed, all labor would go to the one most efficient industry in each region and only one "prod- uct" would be produced. The revised simplex method of solving for the ob- 31 The method required that the jective function was used. problem be put in matrix form, where each row represented either the objective function equation or a constraint equa- tion. The first row gave the objective function and was therefore named MAX. The second and third rows contained the labor supply constraints for the South (STH) and the non-South (NSH) respectively. Since this constraint is expressed as a "less than or equal to" condition, a posi— tive slack variable was inserted in these two rows to per- mit them to be treated as equalities during the solution process. The remaining rows were the output constraints (Cg). Expressed as "greater than or equal to" conditions, they required the insertion of a negative slack variable. The equations, in the order they appeared in the matrix are found in Figure IV. 31931cu1ationfigf_£inear Programming Problems on the AESLP, AESLPED, and EDITLP Routines, Michigan State Univer- sity Agricultural Experiment Station, 1968. 63 >H magmas mmsuumspsa as Hap mom m m >N. o + schmn + o +......... o+ ma ms + o ..H '00. O+O+CHACHQ+ 0000000000 0+0...qu HQ W30." U qwcsfibq oooo +HHNQH+HHH1H+00000000000 +0170 Boummz Ma m. l m2. ......+ mmq+ A soumam AVOOOOOO +O+O+ ‘Hooo CHhAch Cm CN CH CH man was mm mm ma ma n+..+ q h+ q n+ a n+..+ a n+ q n sou x4: 64 In the MAX row, the b coefficients were labor productivity data as used in earlier tests in the paper. The L coeffi- cients were the unknowns, that is, the labor allocation for which the problem was being solved. In the labor constraint rows, the L coefficient is again the unknown, and the only entry was a coefficient of 1. For the demand constraint rows, labor productivity was again as b, while L is still unknown. There were 143 columns in the fully written matrix. One hundred forty—two columns represented all combinations of the 71 industries and the two regions, while the 143rd was the "right hand side" column containing the values of the constraint equations. The solution yielded the number of man years of labor which should be allocated to the vari— ous industries in the two regions so as to maximize value added for the nation, while at the same time operating within the constraints specified. To anticipate the results, two factors were noted. First, the suggestion by Bhagwati, that in a Ricardian model expanded beyond two commodities, “there will be a chain in which all commodities are ranked in terms of their compara- tive factor-productivity ratios such that it will always be true that each of a country's exports will have a higher factor-productivity ratio than each of its imports."32 Second, a comparison of labor productivities between 32Bhagwati, "The Pure Theory. . .," p. 5. 65 the two regions shows that the South has an absolute advan- tage in only 12 of the 71 industries, that is, where the ratio of productivity in the South to that in the non-South is greater than one. For the remaining 59 industries, the productivity ratios ranged from +.993 to +.617. The produc- tivity ratios and the optimum allocation of labor are shown in Table 1. As might be expected, the South's labor was first allocated to those industries in which an absolute advantage existed. Thus, the South was shown to specialize in the production of those goods and supply the entire amount de- manded by the nation. The remaining labor in the South was then allocated according to the ranking of productivity ra- tios. First, enough labor was given to SIC industry 366, whose productivity ratio was .993, to satisfy total national demand. The next allocation went to industry 356 with a ratio of .988. This continued until the labor supply in the South was exhausted. The result was that two industries in the South were allocated labor based on "pure“ compara- tive advantage, that is, with no absolute advantage already existing. Of these, industry 366 output will be entirely produced in the South, while industry 356 output will be divided between the South and non-South. The actual output data for the year 1958 are quite different from the results of the linear programming solu- tion. Both regions produce in all 71 industries. And as 66 Table 1. Linear programming allocation of labor inputs between industries and regions for maximization of national value added Labor Produc- Units of Labor Industry tivity Ratio Allocated Region 201 .750 291693.9 N 202 .904 284318.1 N 203 .652 202966.2 N 204 .716 110130.1 N 205 .935 296689.9 N 206 .871 27332.0 N 207 .688 76670.0 N 208 .803 194090.4 N 209 .740 124168.6 N 225 .644 169781.9 N 228 .814 92404.1 N 229 .786 64510.7 N 231 .929 122522.9 N 232 .730 232260.9 N 233 .670 345419.9 N 234 1.020 109705.2 S 236 .737 76178.4 N 238 .751 57409.0 N 239 .850 124758.4 N 243 .634 12l601.6 N 244 .787 35056.9 N 249 .774 53501.7 N 251 .788 228898.1 N 252 .801 22855.6 N 253 .640 13694.2 N 265 .938 179170.6 N 273 .617 65682.4 N 278 .962 39883.1 N 279 .906 41604.2 N 283 .940 95494.2 N 284 .893 613551.4 N 285 .967 58487.6 N 287 .865 35837.3 N 295 .854 22453.3 N 299 .793 9505.4 N 314 1.004 226108.l S 317 .949 35551.8 N 322 .971 91390.2 N 325 .790 65726.1 N 326 .785 42419.2 N 327 .830 141151.? N 329 .984 91099.5 N 332 .927 l79844.9 N Table 1 (continued) 67 Labor Produc- Units of Labor Industry tivity Ratio Allocated Region 333 1.513 37308.6 3 335 1.058 152662.4 S 339 1.136 46118.3 5 342 .902 l34884.7 N 343 .771 68843.6 N 344 .831 327279.4 N 346 .874 124315.1 N 348 .977 55338.4 N 349 1.069 129617.9 S 351 .643 95525.1 N 352 .787 105995.9 N 353 .974 198765.8 N 354 1.030 226757.3 S 355 .909 160135.1 N 356 .988 l90149.5 N 21343.7 5 357 1.504 82129.9 S 362 1.295 123066.4 S 366 .993 216348.8 S 369 1.180 67286.3 S 371 1.147 509292.7 S 372 .908 754943.1 N 373 .887 137219.2 N 384 .871 41362.2 N 391 .679 41557.3 N 394 .926 97949.4 N 395 .903 28737.4 N 396 1.090 51841.5 S 399 .926 325353.4 N 68 seen in earlier tests, the relative intensity of production between industries and regions is not correlated with pro- ductivity ratios as the program results indicate they should be. Of the 14 industries that should produce solely in the South according to Classical optimization, only two have very high rankings in the concentration ratios computed. A list of ranks for the relative regional concentration concept, SE: is shown below: C n Concentration Industry, Ranking (out of 71) Labor Productivity Ratios 234 57 1.020 314 31 1.004 333 69 1.513 335 35 1.058 339 4 1.136 349 36 1.069 354 5 1.030 356 8 .988 357 6 1.504 362 17 1.295 366 27 .993 369 14 1.180 371 16 1.147 396 7 1.090 The poor predictive content of the Classical model indicates that other factors play a dominant role in indus- trial location. This conclusion seems to be highly plaus- ible in industries where total labor costs, as measured by total wage bill, make up only a small portion of the total costs of value added. To deal with this circumstance, reconsideration was given to the Graham-modified Classical model for industries 69 in which at least 50 percent of total costs could be attrib- uted to labor. Remaining was a sample of industries where non-labor costs such as capital and raw material costs would play a subordinate role. From the original sample of 71 industries, 38 met this condition and were subjected to the same rank correlation tests as performed earlier. Using both labor cost and productivity ratios, both correlation coefficients were of the "wrong" sign (.013 and -.180 re- spectively). The level of significance fell in both tests, however, thus indicating some improvement over the full- sample tests. One explanation of the apparent randomness of asso- ciation between labor cost or productivity ratios and rela- tive concentration can be based on the concept of differen- tial rates of industrial development between the South and non-South. In the earlier linear programming solution, only 12 industries had an absolute advantage in the South based on labor productivity. (See page 68.) Of these, only two actually show high concentration in the South. Of the re- maining ten, seven could be classified as being involved in heavy industrial output. They are: 335, nonferrous metal rolling and drawing; 339, primary metal industries, n.e.c.; 349, fabricated metal products, n.e.c.; 354, metal- working machinery; 362, electric industrial apparatus; 369, electrical products; 371, motor vehicles and equipment. Because of the more recent industrial development 70 in the South, these industries have more modern capital equipment, making their labor inputs more productive. These same industries in the non-South, however, must allow for depreciation of older equipment before replacing it with newer machinery or even a new scale of operations. In ad- dition, because the non-South did develop earlier, the sources of demand for these products are still mostly lo- cated in the non-South; and thus, despite the labor produc- tivity disadvantage, most of the nation's output in those industries is still produced in the non-South. Some Comparative Static Tests In an effort to investigate further the idea of dif- ferent regional development rates, percentage changes in the relative concentration ratios were computed for the period between 1947 and 1958. If during this period, in- dustry in the South had been developing at a faster rate than in the non-South in those industries in which the South had a comparative labor cost advantage, the above explana- tion of the earlier test results could have some validity. To test this, rank correlation tests were performed between the rankings of percentage change in E§_and both average n labor cost ratios and labor productivity ratios. The hypoth- esis was that industries with relative cost or productivity advantages would be positively correlated with percentage changes in relative concentration in the South. In the case 71 of average labor costs, the rank correlation coefficient should be negative, while with labor productivity, it should be positive. Due to changes in SIC classifications between 1947 and 1958, 21 of the industries had to be eliminated as data were not available or not comparable between the two years, leaving a sample of 50 SIC three-digit industries. The re— spective coefficients of the two tests were -.231, signif- icant at the 6 percent level, and .062, not significant. Both were of the hypothesized sign. Ranks are shown in Tables 2 and 3. Analysis of the rankings spotlights several inter- esting points. First, Industry 273, book printing and pub- lishing, performs very poorly, regardless of whether labor cost ratios or labor productivity ratios are compared with concentration changes. That is, the data show a large per- centage increase in relative concentration in the South for Industry 273, despite a high labor cost ratio and low pro- ductivity ratio for the South. However, when Industry 273 is considered in the context of static concentration for 1958 alone, it performs very well. In a sample of 71 indus- tries, it ranks 70th in labor cost ratios and only 13th in concentration, a definite negative relationship as hypoth- esized. Thus, despite a substantial percentage shift to the South over the period covered, the industry remained primarily 72 Table 2. Ranks of average labor cost ratios and percentage changes in relative concentration in the South, 1947-1958 - Percentage Percentage Changes in Changes in Average Relative Average Relative SIC Labor Cost Concentration SIC Labor Cost Concentration Code Ratio Rank Ratio Ranks Code Ratio Ranks Ratio Ranks 279 1 18 267 26 ' 27 354 2 44 251 27 24 317 3 50 205 28 15 314 4 30 228 29 2 384 5 42 231 30 28 348 6 35 249 31 20 234 7 48 295 32 29 229 8 40 225 33 25 342 9 45 394 34 41 332 10 31 344 35 14 349 ll 46 204 36 6 395 12 38 209 37 11 327 13 19 284 38 9 208 14 7 238 39 33 355 15 36 346 40 12 244 16 32 373 41 8 399 17 34 243 42 5 201 18 22 326 43 4 287 19 l 203 44 21 202 20 26 343 45 17 285 21 13 236 46 49 322 22 10 233 47 43 278 23 23 232 48 37 325 24 16 352 49 39 239 25 3 273 50 47 73 Table 3. Ranks of labor productivity ratios and percentage changes in relative concentration in the South, 1947-1958 Percentage Percentage Changes in Changes in Labor Relative Labor Relative SIC Productivity Concentration SIC Productivity Concentration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 273 l 47 287 26 l 243 2 6 384 27 42 225 3 25 346 28 12 203 4 21 373 29 9 233 5 43 284 30 5 204 6 7 342 31 45 232 7 37 395 32 38 236 8 49 202 33 26 209 9 11 279 34 18 201 10 22 355 35 36 238 11 33 394 36 41 343 12 17 399 37 34 249 13 20 332 38 31 326 14 4 231 39 28 229 15 40 205 40 15 352 16 39 267 41 27 244 17 32 317 42 50 251 18 24 278 43 23 325 19 16 285 44 13 208 20 8 ~ 322 45 10 228 21 2 348 46 35 327 22 19 314 47 30 344 23 14 234 48 48 239 24 3 354 49 44 295 25 29 349 50 46 74 located in the non-South as predicted by the static hypoth- eses. The reason for the shift can be attributed to changes in income and population, movement to an area of absolutely lower labor costs, and a small shift in paper industries to the Soutn.33 With Industry 273 not included in the tests, the coefficients were -.245, significant at the five percent level, and +.146, significant at the 15 percent level. A second point of interest is the performance of the apparel industry group, made up of Industries 231, 232, 233, 234, 236, 238, and 239. Four of these industries, 232, 233, 236, and 238, perform very poorly in the tests. All four have relatively high labor cost ratios and low labor productivity ratios, yet show a relatively high movement to the South, a condition contrary to the hypothesis. To explain this, several characteristics of the ap- parel industry must be noted. First, these four are quite labor intensive industries; that is, at least 60 percent of their total cost is attributed to labor. Second, although the productivity ratios are low for these industries, the absolute level of productivity is also low for these industries in the non-South relative to all other industries in the non-South. Thus, although the non-South may have an absolute advantage over the South 33Victor Fuchs, Changes in the Location of Manufac- turin in the United States Since 1929 (New Haven: Yale University Press, 1962), p. 254. 75 in these industries, they are characterized by very low pro- ductivity throughout the non-South. Third, all apparel in— dustries are generally regarded as requiring unskilled labor.34 The relatively large shift to the South, then, can be attributed to several interacting conditions. The apparel industries require large amounts of unskilled and therefore low productivity labor. Because of the relative shortage of this type of labor in the non-South, wages are higher. At the same time, improved technology in agriculture has freed much unskilled labor in the South. Thus, the attrac- tion of a substantial supply of unskilled labor has been a major cause of the movement to the South.35 An additional factor is that the source of raw ma- terials, Industry 22, textile mill products, is concentrated in the South and has shown signs of further movement to the South.36 The other three apparel industries, 231, 234, and 239, show mixed results in the rank tests. Industry 234 definitely supports the hypotheses, 239 probably does, while 231 is difficult to judge. The reason these three vary from 34Ibid., p. 172. 35Ibid., pp. 24, 25, 172. 36For sub-industries 225, 228, and 229, the percent— ages of output produced in the South are 46, 68,and 25. Industries 225 and 228 rank 35th and 38th (out of 50) in movement to the South. 76 the others is explained by their difference in the intensity of labor required. Industry 234 has the lowest labor require- ment of all apparel industries, 48 percent, and therefore was not under as much pressure to seek new sources of un- skilled labor. Industries 239 and 231 have labor require- ments of 56 and 62 percent of value added. After omitting the seven apparel industries, the new rank correlation coefficients were —.349, significant at the .0005 percent level, and +.181, significant at the ten percent level. These results constitute very strong evidence that although the composition of the industrial structure in the South in 1958 did not conform to that which would be expected under the Classical hypothesis, it was due in part to the differences in the vintage of capital employed between the South and the non-South, and not because the Classical model in general has no predictive power. In fact, changes in the industrial structure of the South did take place in ac- cordance with expectations derived from the Classical model. Because of the latent industrial development of the South, a comparative set of tests was performed between New England and non-New England. The same variables are used and rankings are shown in Appendix IV. For a sample of 68 industries, the rank correlation test between labor productivity ratios and relative concen— tration ratios yielded a coefficient of +.221, significant 77 at the one percent level and positive as hypothesized. This result shows strong evidence that those industries with rela- tively high labor productivity in New England tend to be relatively highly concentrated there. This is, of course, contrary to the relation between those two variables in the South, providing further evidence that because the South did not have a fully developed industrial structure, its industry concentration could not be explained by either the Heckscher-Ohlin or the Classical model. When average labor cost ratios are substituted for labor productivity ratios, the coefficient is -.076, not significant, but of the hy- pothesized sign. Although the industrial structure of New England is regarded as being relatively well established, changes during the 1947-1958 period took place in a manner expected under the Classical hypothesis. That is, those industries with relatively high labor productivity in New England gen— erally experienced increases in relative concentration in New England. A coefficient of +.249 is obtained when labor productivity ratios are ranked with percentage changes in relative concentration, and a coefficient of —.l38 is found for average labor costs. The former is significant at the one percent level, while the latter is not significant. Conclusions From these comparative tests, several conclusions can be drawn. In a rapidly developing region, the Classical 78 model does not predict accurately the relative industry con- centration at any one point in time. It can, however, pre- dict which industries have shown and will continue to show relatively higher growth rates as indicated by increases in their relative concentration in that region. On the other hand, for a region with an historically established indus- trial structure, the Classical model predicts with tolerable accuracy the relative concentration at any point in time as well as changes in relative concentration over time. CHAPTER IV THE ROLE OF DEMAND AND NATURAL RESOURCES Introduction To this point, neither the Heckscher-Ohlin nor the Classical hypothesis has very well explained the static levels of industry concentration in the South. That the South had not yet reached industrial maturity seemed to offer a partial explanation; however, it is felt that other variables might play a significant role. In an effort to isolate these other variables, re- consideration was given to the rank correlation test for the 38 labor intensive industries discussed in the previous chapter. Rankings of relative concentration ratios and labor productivity ratios are shown in Table 4. A sample of industries selected in such a way as to give the Clas— sical model every chance of indicating some predictability of relative industry concentration in the South is shown. The test result indicated, however, that the rankings were distributed in a random manner, and therefore the model had no explanatory power. Eight industries which clearly vio- lated the hypothesized relation are subjected to a more detailed analysis. The industries are: 243, 253, 225, 232, 278, 354, 356, and 339. 79 80 Table 4. Ascending ranks of relative concentration ratios, C Eé" labor productivity ratios, and average labor n cost ratios for industries in which at least 50 percent of total costs are labor costs L Relative Labor Average SIC Industry Concentration Productivity Labor Classification Ranks Ratios Ranks Cost Ranks 351 l 4 36 391 2 7 34 339 3 38 2 354 4 35 l 356 5 32 4.5 273 6 l 37 233 7 6 32 352 8 15.5 35 278 9 30 11 394 10 25.5 22.5 399 11 25.5 9 366 12 33 20 314 13 34 3 355 14 24 7 238 15 11 25 372 16 23 27 335 17 36 16 349 18 37 6 343 19 12 30 332 20 27 4.5 326 21 14 29 265 22 29 14 236 23 9 31 353 24 31 21 231 25 28 17.5 239 26 21 12.5 243 27 2 28 344 28 20 24 201 29 10 10 253 30 3 38 249 31 13 19 325 32 18 12.5 251 33 17 15 373 34 22 26 225 35 5 22.5 232 36 8 33 244 37 15.5 8 228 38 19 17.5 81 These industries fall into two distinct groups. The first four have relatively low labor productivity ratios, yet show relatively high concentration in the South. Some other influence appears to be offsetting the low labor pro- ductivity such that it is profitable for these industries to produce quite intensively in the South. The second group has the opposite relation; that is, relatively high labor productivity ratios, yet relatively low concentration in the South. Two variables felt most likely to influence concen- tration are the location of demand for the output of an in- dustry and the location of an industry's sources of raw ma- terials. In some cases, sources of demand or raw materials were specifically spelled out in the Census of Manufactures.1 More often, however, a look at four or five-digit sub-indus- tries gave a clue as to potential sources of raw materials and to other industries which use the output of the indus- try under consideration as an input and thus create a demand for it. The detailed analysis for the eight industries ap- pears in Appendix IV, and only the conclusion drawn for each industry will be presented here. An examination of the in- dustries seemed to indicate that an important role is played 1U.S. Bureau of the Census, Census of Manufactures, 1958, Vol. II, Parts 1 and 2 (Washington: U.S. Government Printing Office, 1961). 82 by the location of demand and of raw material sources. For Industry 243, millwork and related products, both sources of demand and raw materials are reasons for locating in the South. For Industry 253, public building furniture, a lack of concrete relationships existed, and no explanation is offered for its relative rankings. For Industry 225, knitting mills, and Industry 232, men's and boys' furnishings, the sources-of raw materials are the major determinants of regional concentration. For Industry 278, bookbinding and related work, un— certainty about the sources of both demand and raw materials makes any judgment difficult. A lack of strong demand in the South could be important. For the remaining three in- dustries, 356, general industrial machinery; 3S4, metal- working machinery; and 339, primary metal industries, n.e.c.; high demand and raw material concentration in the non-South explains the high production concentration in the non-South. Because the more detailed examination of the eight industries seemed to indicate that an important role is played by the location of demand and of raw material sources, further investigation was undertaken. If demand location does influence the relative concentration of some industries, those industries' elimination from the sample could cause the hypothesized relationships between variables within the two models to be more closely approached. 83 A Test for the Role of Demand In an effort to classify which of the 71 three-digit industries are "market-oriented," a characteristic of these industries as described by Victor Fuchs was used as a start- ing point. According to Fuchs, "their [market-oriented in— dustries] distribution throughout the country tends to con- form to the distribution of income and population."2 This statement implicitly assumes that one demand structure exists across the United States, and therefore within each of the nine census regions. Thus, demand for the output of each of the 71 industries exists in all nine regions, where the level of demand is a function of per cap— ita income. A demand-oriented industry can then be defined as an industry that is located in all nine regions and whose relative level of output in each region is the same as each region's relative level of demand. To find demand-oriented industries, each region is ranked according to per capita income, weighted by a population index. Then, for each in- dustry, each region is ranked according to its percentage of national value added for that industry. Industries whose regional output ranks closely approximate (Kendall's ’b'of at least +.666) the regional demand indicator ranks are con- sidered to be demand-oriented industries. Of the 71 2Fuchs, Changes in the Location . . ., p. 152. 84 industries, 30 meet these requirements. Industries, ‘& coefficients, and level of significance are listed in Table 5. These industries are then eliminated from the sam- ple, and new tests are performed. For the reduced sample, rank correlation tests between measures of concentration and labor productivity ratios, average labor cost ratios, and gross capital-labor ratios, yielded no significant im- provements over the same tests when the full sample of 71 industries was included. The comparative results are shown in Table 6, while the relative rankings for the limited sample tests are listed in Appendix III and IV. The Role of Natural Resources The location of sources of raw materials remains a potentially important explanatory variable. A measure of the relationship between value added and cost of mate- rials would appear to give some indication as to how sensi- tively industries depend upon raw materials from sources outside the plant. Industries that depend heavily on raw material inputs are apt to have their concentration more strongly related to the concentration of their sources of these materials than industries that are not so "raw mate- rials-oriented." The determination of the sensitivity of an indus— try's dependence on raw materials is formulated from three accounts kept by the United States Bureau of the Census and 85 Table 5. Rank correlation coefficients and levels of sig- nificance by industry for tests between regional concentration ranks and regional demand ranks SIC Industry Kendall's Level of Classification 7? Significance 201 .611 .025 202 .833 .001 203 .889 .001 204 .555 .025 205 .778 .005 206 .555 n.s 207 .722 .005 208 .778 .005 209 .778 .005 225 .444 .100 228 .000 n.s 229 .333 .200 231 .555 .025 232 .389 .100 233 .722 .005 234 .389 .100 236 .389 .100 238 .555 .025 239 .722 .005 243 .722 .005 244 .278 .200 249 .611 .025 251 .500 .050 252 .722 .005 253 .722 .005 265 .833 .001 273 .722 .005 278 .778 .005 279 .833 .001 283 .611 .025 284 .778 .005 285 .833 .001 287 .333 .200 295 .833 .001 299 .666 .010 314 .278 .200 317 .278 .200 322 .555 .025 325 .500 .050 326 .611 .025 327 .778 .005 329 .666 .010 332 .611 .025 86 Table 5 (continued) SIC Industry Kendall's Level of (Slassification 2: Significance 333 -.166 n.s. 335 .333 .200 339 .389 .100 -342 .555 .025 343 .611 .025 344 .889 .001 346 .666 .010 348 .722 .005 349 .500 .050 351 .167 n.s. 352 .500 .050 353 .555 .025 354 .555 .025 355 .666 .010 356 .611 .025 357 .389 .100 362 .666 .010 366 .555 .025 369 .611 .025 371 .666 .010 372 .500 .050 373 .222 n.s. 384 .389 .100 391 .444 .100 394 .889 .001 395 .722 .005 396 .500 .050 399 .722 .005 87 Table 6. Comparison of results between full sample tests and non-market oriented tests: South Explanatory Concentra- Sample Sign Level of Variable tion Ratios Size Hypothesized ‘17 Significance Labor CS Productivity C—' 71 + -.l99 5% n Labor CS Productivity E- 41 + -.l93 10% n 1&verage CS .Isabor Cost E—- 71 - +.136 10% n .}\verage Cs :Igabor Cost E—- 41 - +.144 n.s. n 1( vi —-G —5 71 - 078 I; i +. n.s. vN i is v5 41 037 L T "' +0 nos. VN 88 published in the Census of Manufactures.3 One measure of the value of output for a final consumption good is the "value of shipments." This concept is defined as "received or receivable net selling values, f.o.b. plant, after dis- counts and allowances, and excluding freight charges and excise taxes."4 To find the net contribution of any one industry, that is, its value added, an account of that industry's cost of materials is required. Included in the cost of materials account are total delivered costs of all raw materials, semi- finished goods, parts, components, scrap, containers, sup- plies, electrical energy, fuel, and contract work.5 Sub- tracting the cost of materials from the value of shipments yields value added. Using data from the Eggsus of Manufactures, 1958, ratios of value added to value of shipments are computed. These ratios, hereafter called coefficients of resource de- pendency, indicate the percentage of value of shipments by an industry attributable to value added. Hence, the lower the ratio, the more dependent the industry is on external resources a In Table 7, industries for which data were available 3U.S. Bureau of the Census, Census of Manufactures, Vol. I, 1958. 41bid., p. 11. 5Ibid. 89 Table 7. List of industries in ascending order according to the coefficient of resource dependency Coefficient Coefficient SIC Industry of Resource SIC Industry of Resource Classification Dependency Classification Dependency 201 .157 353 .505 206 S .230 317 N .507 209 S .257 391 N .510 204 .269 369 - .517 202 .284 205 .518 287 S .301 372 .523 203 .342 349 .525 228 S .349 394 .525 335 .360 314 .530 229 S .362 396 N .538 299 .364 253 .541 295 .369 395 .545 239 .380 329 .547 207 .394 373 S .547 243 .410 355 .551 265 .418 284 .563 232 S .421 356 N .564 244 S .427 357 N .575 285 .429 342 N .585 344 .448 384 N .585 233 .451 332 .588 236 .452 362 .593 234 .465 252 .598 238 .465 273 .606 327 S .467 322 .646 249 .468 354 N .648 339 N .476 278 .663 251 S .483 326 .677 343 .487 325 .680 231 .496 283 N .703 279 .834 S denotes one of the ten industries with highest concen- tration in the South. N denotes one of the ten industries with highest concen— tration in the non-South. 90 are listed in ascending order according to the coefficient of resource dependency. If the earlier considered relation between dependency on raw materials and concentration holds, we should hypothesize that industries highly concentrated in either region would tend to cluster at the end of the industry rankings where cost of materials were the highest percentage of value of shipments, that is, at low values of the coefficient of resource dependency. This does not occur, and in fact the distribution is quite evenly spread. An interesting pattern developed, however. Those industries with relatively high concentration in the South had a tendency to cluster at the low end of the ranking, while the reverse is true for those industries with rela- tively high concentration in the non-South. To further an- alyze this pattern, a rank correlation test between relative . O C O 0 concentration ratios, 5, and coeffiCients of resource de- C n pendency was made. This test seems warranted for the following reason. Some output is produced in all industries in the South de- spite some comparative disadvantages in terms of labor pro- ductivity and labor costs. A high dependency on raw mate- rials purchased from external sources might be the overrid- ing influence. Thus, an industry at a disadvantage in terms of labor productivity in the South may depend enough on ma- terials available in the South that the industry may be highly concentrated in the South. If this is true, a negative 91 relation would exist between relative concentration ratios and coefficients of resource dependency. And in fact, the rank correlation coefficient was -.365, significant at P13 .0001, indicating that those industries more dependent on raw materials tend to be concentrated in the South. Regression Analysis To this point, several empirical variables have been suggested as explanatory factors in the determination of relative industry concentration between the South and the non-South. In order to see the interaction of these vari- ables, and to compare further the South with New England, multiple regression analysis was undertaken. The variables included were: relative concentration ratios for the South and New England, percentage changes in concentration ratios between 1947 and 1958 for the South and New England, aver- age labor cost ratios, labor productivity ratios, and co- efficients of resource dependency. The static and dynamic concentration variables are, of course, the dependent vari- ables. Regression equations and results for the South are listed in Table 8. The test of significance for the re- gression coefficient is a test of the null hypothesis that bi=0; that is, that the independent variable, Xi, does not account for any variation in Y, the dependent variable. A criterion of significance of P;g .10 is used to reject the null hypothesis. 92 Table 8. Regression variables, equations, and results for Modified Classical Model: South Variables: Y C _E. C n Relative concentration ratios, Percentage change in relative concentration, 1947-1958 Average labor cost ratios Labor productivity ratios Coefficient of resource dependency Matrix of Simple Correlations: Yl 1.00000 Y -0033037 1000000 2 X 0.10703 0.00580 1.00000 1 X2 -O.17319 0.11913 -0.60510 1.00000 X3 —0.35518 0.15228 -0.30866 0.36897 1.00000 Y Y 1 Equation I: Y1 2 r .1262 X1 X2 X3 bo + b X +b X +e l 1 3 3 1 Standard error of estimate = 1.4465 Regression b1 53 t test for b1 b3 coefficients (standard errors) -0.0310 (1.6585) -4.1149 (1.7738) regression coefficients : .985 level of significance .025 level of significance Table 8 (continued) ¥ Partial correlation coefficients x1 x3 Equation II: Y1 ao+a2X2+a3X3+e2 2 r .1282 -0.00292 -0.34064 Standard error of estimate = 1.44484 Regression coefficients (standard errers) ' a2 93 -0.67112 (2.1583) -3.89663 (1.8131) t test for regression coefficients a22.757 level of significance a3:.038 level of significance Partial correlation coefficients x2 X3 Equation III: Y2 = do+lel+d3X3+e3 2 r .0263 Standard error of estimate = 109 0892 Regression coefficients (standard errors) d 1 8 3 45.38514 (125.9972) 141.63216 (134.7502) 94 Table 8 (continued) t test for regression coefficients dl!.721 level of significance d33.299 level of significance Partial correlation coefficients x1 X2 .05617 .16198 Equation IV: Y2 = g6*92X2*93X3*e4 2 r .0278 Standard error of estimate = 109.807 Regression coefficients (standard errors) 72.14239 (164.02995) A 92 104.28977 (137.79403) 6:. t test for regression coefficients 923.662 level of significance 933.453 level of significance Partial correlation coefficients X 2 .06853 X .11738 3 95 As in the rank correlation tests, neither average labor cost ratios nor labor productivity ratios can be said to be important factors in determining relative industry location, and the null hypothesis is not rejected in either case. The coefficient of resource dependency, however, is a significant variable and the null hypothesis can be re- jected with a high degree of significance. In seeking an explanation for the change in relative concentration, none of the variables included make a signif- icant contribution. This is interesting from the standpoint that results of earlier rank correlation tests indicated that average labor costs had a significant rank correlation with concentration changes. Such contradictory results also appear in regression analysis for New England using the Clas- sical variables (see Table 9). The two types of tests do not, of course, have to yield the same results. Maurice Kendall, the pioneer in rank correlation methods, states that "by a replacement [of variates] with ranks we effectively standardize the scale of the variate and fix the mean, a procedure which might in some instances lead us astray."6 In this instance, the difference in assumptions for the two tests leads to differ- ent results. The rank correlation test is non-parametric, that 6Maurice G. Kendall, Rank Correlation Methods (New York: Hafner Publishing, 1955), p. 125. 96 is, it makes no assumptions about the distribution of the sample variables. Consequently, the range and distribution of the variables is unimportant. In regression analysis, however, the distribution of the dependent variable is im— portant. For example, let the independent variable change in small proportions and in an even way. If the dependent variable for the same observation changes in a very volatile manner so that the deviations from the mean will be much larger, a regression test will suggest that the independent variable has little explanatory power. Indeed, the standard error of the estimated regression coefficient is ‘g§:_ 8x12 where 32 is the variance of residuals and Ex: is variation of the regressor. If the latter is relatively small, SE is quite large and the regression coefficient will not be significant. In a rank correlation test, however, this would make no difference. In the regression variables used, the percentage changes in relative industry concentration are often very large. But the industry variation in labor productivity or labor cost ratios is comparatively small. In view of these considerations, the rank correlation test may be the more appropriate method of determining the role of relative labor productivity or labor cost. In summary, the substan- tive empirical contribution of the regression analysis is confirmation of the importance of resource dependency. 97 The relation between the coefficient of resource dependency and concentration suggests that the dominant basis for initial location of an industry may be the source of raw materials. Once the nation's location pattern is established according to these "natural” conditions, how- ever, relative regional growth rates of industries depend upon relative average labor costs between regions. As in previous chapters, a comparison between the South and New England yields some interesting results. Be- cause New England concentration could be explained by rela- tive labor productivity ratios and by the combination of factor abundance and intensity, it might be expected that the coefficient of resource dependency would have little influence on relative concentration in New England. This is confirmed by evidence that there exists an inverse rela- tion between an industry's dependency on raw materials and its relative concentration in New England. A rank correla- tion test between the two variables shows a positive coef- ficient, +.187, significant at the 5 percent level. Regression analysis for New England using Classical variables fails to indicate any explanatory variable for static concentration. For changes in concentration, how- ever, average labor cost ratios are shown to be weakly sig— nificant, while labor productivity ratios are highly signif— icant. These results are shown in Table 9. Regression analysis was also undertaken using the 98 Unable 9. Regression variables, equations, and results for Modified Classical Model: New England L; 'Variables: Y Relative concentration ratios, Cne l c nne Y Percentage change in relative concentration, 1947-1958 X Average labor cost ratios X Labor productivity ratios X3 Coefficient of resource dependency Matrix of Simple Correlations: Yl 1.00000 Y2 -0.00896 1.00000 Xl 0.14956 -0.23674 1.00000 X2 0.09112 0.43615 -0.74894 1.00000 X3 0.16600 0.13822 -0.02961 0.04423 1.00000 Y1 Y2 X1 X2 X3 Equation 1: Yl=bo+blxl+b3x3+el r2 = 0.0514 Standard error of estimate = 1.9202 Regression coefficients (standard errors) 0') ll 2.7461 (2.6703) 2.6441 (2.3305) 00 II t test for regression coefficients bl : .310 level of significance b : .263 level of significance U) 99 Table 9 (continued) g Partial correlation coefficients X1 = 0.15672 X3 = 0.17244 Equation II: Yl=ao+a2X2+a3X3+e2 r2 = 0.0346 Standard error of estimate = 1.9372 Regression coefficients (standard errors) a2 = 1.4223 (2.5715) a3 = 2.5156 (2.3524) t test for regression coefficients a2 : .583 level of significance a3 : .291 level of significance Partial correlation coefficients X 2 X3 0.08503 0.16281 Equation III: Y2=do+lel+d3X3+e3 r2 = 0.0733 Standard error of estimate = 71.9467 Regression coefficients (standard errors) 1 -156.7754 (100.0539) Q) 0» 77.1648 (87.3213) 3 100 Table 9 (continued) t test for regression coefficients d1 : .125 level of significance d3 : .382 level of significance Partial correlation coefficients X1 = -0.23501 X3 = 0.13511 1‘3 qu ation IV: Y2 = go+gZX2+g3X3+e4 r2 = 0.2044 Standard error of estimate = 66.6628 Regression coefficients (standard errors) 276.7611 (88.4902) A 92 63 70.0165 (80.9520) t test for regression coefficients 92 : .003 level of significance g3 : .392 level of significance Partial correlation coefficients X 2 x3 0.43463 0.13229 101 variables from the earlier Heckscher-Ohlin model tests for both the South and New England. The variables included are: i i concentration ratios, :5 and vne, percentage changes in these V1 V1 N N ratios for the period 1947-1958, gross capital-labor ratios, net capital-labor ratios, and coefficients of resource de- Pendency. The results are shown in Tables 10 and 11. The regression results lend strong support to the conclusions reached by the application of rank correlation tests to the Heckscher-Ohlin model. For the South, the co- Efficient of resource dependency plays a significant role (P g .025) in determining relative industrial concentration, while factor proportions had no significant influence. These two variables reversed roles, however, when changes in con- centration for the South were considered. That is, the co- efficient of resource dependency had no apparent effect on the determination of industrial growth in the South between 1947 and 1958, while factor proportions were the primary influence. For New England, the existing industrial structure Of 1958 had been established according to the factor pro- Portion hypothesis, and the coefficient of resource depend- ency had no significant effect. Neither variable, however, offered any significant explanation for changes in concen- tration for New England prior to 1958. This is as one would exPect due to the essentially equilibrium status of the New England industrial structure for the period of 1947 to 1958. 102 Table 10. Regression variables, equations, and results for Modified Heckscher-Ohlin Model: South a P' Variables: Yl Concentration ratios, v 2146 Y2 Percentage change in concentration, 1947-1958 X1 National gross capital-labor ratios X2 National net capital-labor ratios X3 Coefficient of resource dependency Matrix of Simple Correlations: Y]_ 1.00000 Y2 -0.34411 1.00000 X1. 0.17356 -0.38169 1.00000 )(2 0.18353 -0.37543 0.97003 1.00000 X3’ —0.37292 0.19507 —0.28996 —0.24714 1.00000 Y1 Y2 X1 x2 x3 Equation I : Yl=bo+le1+b3X3+el r2 = 0.1437 Standard error of estimate = 0.10889 Regression coefficients (standard errors) 31 = 0.00266 (0.00562) b3 = —0.32211 (0.13809) t test for regression coefficients b1:.639 level of significance b3:.025 level of significance Tabl 103 Table 10 (continued) Partial correlation coefficients X1 = 0.07368 x3 = -0.34227 8 qu ation II: Y1=ao+a2X2+a3X3+e2 r2 = 0.1480 Standard error of estimate = 0.10862 Regression coefficients (standard errors) 0.00661 (0.01011) -0.31906 (0.13606) DJ D) II t test for regression coefficients a2:.517 level of significance a3:.024 level of significance Partial correlation coefficients X 2 x3 0.10162 Equation III : Y2=do+dlxl+d3x3+e3 r2 = 0.1535 Standard error of estimate = 77.18056 Regression coefficients (standard error) 8, = -9.41801 (3.98350) 8 = 60.07074 (97.88109) 3 104 Table 10 (continued) t test for regression coefficients dl:.023 level of significance d31.543 level of significance Partial correlation coefficients X 1 x3 -O.34638 0.09541 3 Chi ation IV: Y2=go+gZX2+g3X3+e4 r2 = 0.1521 Standard error of estimate = 77.24309 Regression coefficients (standard errors) -16.88451 (7.19023) 32 63 = 71.02057 (96.75331) t test for regression coefficients 92:.024 level of significance 932.467 level of significance Partial correlation coefficients X 2 x3 -0.34431 0.11389 105 ‘TEiIJle ll. Regression variables, equations, and results for Modified Heckscher-Ohlin Model: New England Variables: Y Concentration ratios, v1 1 ne 1 vN Y2 Percentage change in concentration, 1947-1958 X1 National gross capital-labor ratios X2 National net capital-labor ratios X3 Coefficients of resource dependency Matrix of Simple Correlations: Y1 1.00000 Y 2 0 .08311 1 .00000 ><1_ -0.34518 -0.18050 1.00000 x2 -0.31800 -0.16341 0.97003 1.00000 X3 0.16128 0.21073 —0.28996 -0.24714 1.00000 Y1 Y2 x1 x2 x3 Equation I : Yl=bo+blxl+b3x3+el r2 = 0.1232 Standard error of estimate = 0.081832 Regression coefficients (standard errors) 31 = -0.00900 (0.00422) 33 = 0.04538 (0.10378) t test for regression coefficients b1:.039 level of significance b3:.664 level of significance 106 Table 11 (continued) Partial correlation coefficients x1 X3 0.06813 Equation II: Yl=ao+a2X2+a3X3+e2 r2 = 0.1084 Standard error of estimate = 0.08252 Regression coefficients (standard errors) 32 = -0.01495 (0.00768) 33 = 0.05982 (0.10336) t test for regression coefficients a2:.058 level of significance a3:.566 level of significance Partial correlation coefficients X = -0.29085 2 X3 = 0.09001 Equation III: Y2=do +le1+d3X3+e 3 r2 = 0.0600 Standard error of estimate = 57.89128 Regression coefficients (standard errors) 31 = -2.46176 (2.98793) A d3 = 80.24545 (73.41825) 107 Table 11 (continued) t test for regression coefficients dl:.415 level of significance d :.281 level of significance 3 Partial correlation coefficients 1 X3 0.16826 Equation IV: Y2=go+gZX2+gBX3+e4 r2 = 0.0576 Standard error of estimate = 57.96398 Regression coefficients (standard errors) -4.089085 (5.39562) A 92 63 84.18623 (72.60463) t test for regression coefficients 92:.453 level of significance 93:.253 level of significance Partial correlation coefficients X2 x3 0.17819 108 The combinations of all tests indicate that rela- tively highly concentrated industries in New England, unlike those in the South, do not rely on external sources of raw materials and therefore their location is based on labor productivity advantages and a combination of relative fac- tor abundance with relative factor intensity. The conclusion from the work undertaken in this chapter is that the two models of comparative advantage are insufficient to predict regional industrial location patterns in all cases. There seem to be two dominant rea— sons. First, the assumption of costless trade, including zero transportation costs, does not hold between regions. Second, it has been confirmed empirically that additional factors besides labor costs, labor productivity, or factor proportions, exert a significant influence. In particular, it has been shown that the coefficient of resource depend- ency is a crucial determinant of industrial location in the South. CHAPTER V SUMMARY AND CONCLUSIONS The goal of the research undertaken in this disser- tation has been to test empirically the Heckscher—Ohlin and Classical trade models. The uniqueness of these tests is that United States regional data were employed rather than international data. There were several reasons for under- taking regional tests. First, with the exception of one case, all previous empirical tests of the Heckscher-Ohlin model have used international data. In addition, there has not been intensive regional testing of the Classical model. Second, certain assumptions of the two models are more closely approximated when regional data are used. For example, free trade between trade areas is realized and the condition of zero transportation costs is more closely ap- proached. In addition, the possibility of factor intensity reversal, a potentially serious hazard to testers of the Heckscher-Ohlin model, seems to be eliminated using United States interregional data. One potential problem encountered in using regional data is that trade flows between regions are not available. Relative industry concentration is therefore used as a proxy. Industry concentration should be a good indicator of which 109 110 industries possess comparative advantages since both models suggest a trend toward specialization within each region for such industries. Two sets of comparative regions were used: South-non-South and New England-non-New England. Incorporating the regional approach into the two models, they could be stated in a form leading directly to empirically testable hypotheses. The Heckscher-Ohlin model brings together a combination of relative factor endowments and relative factor intensity in production as determinants of comparative advantage. Specifically, the model predicts that a region tends to specialize in producing those goods requiring intensively the use of the relatively abundant factor of that region. Studies showing that relative wages are lower in the South constitutes presumptive evidence that the South is relatively labor abundant. It should therefore possess a comparative advantage in the production of labor intensive goods; that is, goods whose production requires a relatively low capital-labor ratio. Stated as an empirically testable hypothesis: industry rankings of concentration in the South will be negatively correlated with industry capital-labor ratios. The Classical model, resting on the labor theory of value, bases comparative advantage on relative labor productivity advantage. With the inclusion of wages, the determinant of comparative advantage becomes relative 111 average labor cost. Both labor variables were considered in the study, the empirically testable hypotheses being that ratios of labor productivity in the South to that in the non-South will be positively correlated with concentration in the South, while South-non-South average labor cost ra- tios will be negatively correlated with concentration in the South. I Two measures of production concentration were used. The first is the same as was used in testing the Heckscher- Ohlin hypothesis. The second is a ratio of the percent of total value added in the South contributed by each industry, to the percent of total value added in the non-South contrib- uted by each corresponding industry. Data for all variables were obtained from the Census of Manufactures, 1958. Tests of the above hypotheses permit the following tentative conclusions. Both models failed to predict indus- try concentration in the South. In fact, rank correlation coefficients were of a sign opposite of that hypothesized. These coefficients were not significant for the Heckscher- Ohlin hypotheses, but significant for the Classical tests. These results held in the full sample of 71 industries. For New England, using a sample of 68 industries, both models predicted with tolerable precision the relative industry concentration. Tests of the Heckscher-Ohlin hy- pothesis, using both gross and net capital, yielded coeffi- cients of the hypothesized sign and significant at P‘s .01. 112 For the Classical hypotheses, the rank correlation coeffi- cient is also significant at P £..01 when labor productivity is used. In tests using average labor costs, the coeffici- ent was not significant, but was of the hypothesized sign. The difference in test results between the two re- gions was attributed essentially to one basic difference in the characteristics of the two regions. The South has been experiencing during the past three decades very rapid industrial development relative to the rest of the nation. It therefore does not have a sufficiently well established equilibrium industrial structure within which the two models can be properly tested. New England, on the other hand, is much more nearly in an equilibrium state for the postu- lated tests. * The unique characteristic of the South does not mean that the tests and subsequent analysis of that region is irrelevant. To the contrary, it has prompted the search for other variables which might determine comparative advan- tage and has led to some important conclusions concerning the applicability of the two models and concerning the na- ture of regional industrial growth. The basic industrial structure of the South appears to be a function of the location of sources of raw materials. Thus, industries with a relatively high degree of dependence on external sources of raw materials are more highly concen- trated in the South. The rank correlation coefficient 113 betnneen the coefficient of resource dependency and relative (Knicentration in the South is significant at P S_.0001. It is important to note, however, that recent (1947- 1958) changes in relative concentration in the South have ‘baken place in a pattern so as to suggest the eventual es- tablishment of an industrial structure as predicted under the two models. For example, the rank correlation coeffi- cient between gross capital-labor ratios and the percentage changes in relative concentration is of the hypothesized sign and significant at P S .005. The evidence is strong that the industries experiencing relatively higher growth rates in the labor abundant South are those with relatively low capital-labor ratios. The same relatively higher growth rates also hold for industries with labor cost advantages, although the level of significance is somewhat lower. The results of the rank correlation tests are, for the most part, strongly substantiated by multiple regression analysis. This is particularly true when the Heckscher- Ohlin model variables were considered for both the South and New England. The variables which were significant in the rank correlation tests have regression coefficients also significant at the P‘S.°10 level. Two broad conclusions can be drawn from these tests. First, an already industrially developed region can be ex- pected to display patterns of specialization in those indus- tries which have a comparative advantage with respect to 114 latmor productivity as well as those industries whose produc- tican functions require more of the relatively abundant fac- tor of that region. Second, for a newly developing region, initial at- 1ucaction of industries is likely to be based directly on sources of raw materials and on the endowment of natural .resources of that region. As development proceeds, however, there will be a relatively higher growth in those industries which can achieve a comparative advantage based on labor productivity or on intensive utilization of the relatively abundant and therefore relatively cheap factor of production. Further work along these lines should be interesting. At present, the models of regional comparative advantage are not complete. But the empirical results in this thesis suggest some avenues for increased theoretical sophistica- tion. In addition, some of the empirical procedures devel- oped in this thesis might be applied with some success to international statistics. The principal finding of this study seems to be a confirmation that the basic trade models do have relevance to regional location analysis. An additional variable, the coefficient of resource dependency, was incorporated into the models and found to be a relevant factor. Although this is a step forward in the inclusion of the influence of nat— ural resources into trade models, an explicit and rigorously formulated theoretical model is still lacking. This area would seem to be a fruitful one for future research. BIBLIOGRAPHY 116 Arrow, K., et al. ”Capital-Labor Substitution and Economic Effi ciency," The Review of Economics and Statistics, XLIII (August, 19617, pp. 225- 50. Balwaissa, Bela. "An Empirical Demonstration of Classical Comparative Cost Theory," The Review of Economics and Statistics, XLV (August, 1963), pp. 2§_?§8. Ba]_21.,.David Stafford. "Factor-Intensity Reversals in Inter- national Comparison of Factor Cost and Factor Use," The Journal of Political Economy, LXXIV (February, 1966), pp. 77460. Beczlcer, Gary. "Investments in Human Capital: A Theoretical Analysis," The Journal of Political Economy Supple— ment (October, 1962), pp. 9-49. Bhagwati, Jagdish N. "Some Recent Trends in the Pure Theory in International Trade." In Roy Harrod and Douglas Hague (eds.), International Trade Theory in a Develop— ing World. New York: St. Martin' 5 Press, 196 Briamgwati, Jagdish. "The Pure Theory of International Trade," Economic Journal, LXXIV (March, 1964), pp. 1-64. Bilanradwaj, R. "Factor Proportions and the Structure of Indo- United States Trade, " The IndianL Economic Journal, X, No. 2 (October, 19627, pp. 105- 16. Board of Governors of the Federal Reserve System. Federal Reserve Bulletin, XLIV (January, 1958, and April, 1958), pp. 34,312. Calculation of Linear Programming Problems on the AESLP, AESLPED, and EDITLP Rout nes, Michigan State Uni- versity Agricultural Experiment Station, 1968. Caves, Richard E. Trade and Economic Structure. Cambridge: Harvard University Press, 1960. Clement, M. 0., Pfister, Richard L., and Rothwell, Kenneth J. Theoretical Issues in International Economics. Bos- ton: Houghton—Mifflin Company, 1967. CollDerg, Marshall R. "Human Capital as a Southern Resource," Southern Economic Journal, XXIX (January, 1963), pp. 157-66. 0131:), M. A. The United States Capital Position and the Structure of its Forei n Trade. Amsterdam: North- Holland PuEIIsfiing Co., I956. 117 Eisenmenger, Robert W. The Dynamics of Growth in rNew Eng: land's Economy, 1870-1964. Middletown, Connecticut: Wesleyan University Press, 1967. Ellsworth, P. T. "The Structure of American Foreign Trade: A New View Examined,” The Review of Economics and Statistics, XXXVI, No. 3 (August, 1954), pp. 279- 85. Estle, Edwin F. "A More Conclusive Regional Test of the Heckscher-Ohlin Hypothesis, " The Journalf of Polit- ical Econom , LXXV, No. 6 (December, 1967), pp. 886-58. Forcheimer, Karl. "The Role of Relative Wage Differences in International Trade," The Quarterlnyournal of Economics, LXII (November, 1947), pp. 1- 30. Ford, J. L. "On the Equivalence of the Classical and the Factor-proportions Models in Explaining Interna- tional Trade Patterns, " The Manchester School of Economic and Social Studies, XXV, No. 2 (May,l 1967) pp. l§5-98. Fuchs, Victor R. ”Capital-Labor Substitution: A Note," The Review of Economics and Statistics, XLV (Novem- ber, 1963), pp. 436-38. Fuchs, Victor. Changes in the Location of Manufacturing in the United States Since 1929. New Haven: Yale University Press, 1962. Fuchs, Victor, and Perlman, Richard. "Recent Trends in Southern Wage Differentials," The Review of Eco- nomics and Statistics, XLII (August, 1960?, pp. 292-355. Gallaway, Lowell E. "The North-South Wage Differential, " The Review of Economics and Statistics, XLV (August, W1 3 , pp. 2644-2‘72. Graham, Frank D. The Theory of International Values. Princeton: Princeton UniversityT Press, 1948. Graham, Frank D. ”The Theory of International Values Re- examined," QuarterlyAJgurnal of Economics, XXVIII (November, 1 23), pp. 54-86. Heckscher, Eli. "The Effect of Foreign Trade on the Distri- bution of Income." In H. 3. Ellis and L. A. Metzler (eds.), Readings in the Theoryfand International Trade. Homewood, Illinois: Richard D. Irwin, Inc., 1949. 118 Jones, Ronald W. “Comparative Advantage and the Theory of Tariffs: A Multi-Country, Multi-Commodity Model," Review9£_Economic Studies, XXVIII (June, 1961), pp. 161-75. Jones, Ronald W. "Factor Proportions and the Heckscher- Ohlin Theorem," Review of Economic Studies, XXIV, Kanninen, T. P. "Wage Differences Among Labor Markets," Monthly Labor Review, XLIV (June, 1962), pp. 614-20. Kendall, Maurice G. Rank Correlation Methods. New York: Hafner Publishing, 1933. Kenen, Peter B. "Nature, Capital, and Trade," The Journal of Political Economy, LXXIII, No. 5 (October, 1965), pp. 437-600 Kravis, Irving B. "'Availability' and Other Influences on the Commodity Composition of Trade," The Journal of Political Economy, LXIV, No. 2 (April, 1956), pp. 143-55. Kravis, Irving B. "Wages and Foreign Trade," The Review of Economics and Statistics, XXXVIII, No. l (Febru— ary, 1956), pp. 14-30. Kreinin, Mordechai E. "Comparative Labor Effectiveness and the Leontief Scarce-Factor Paradox," The American Economic Review, LV, No. 1 (March, 1965), pp. 131-39. Leontief, Wassily. "An International Comparison of Factor Cost and Factor Use," The American Economic Review, LIV, No. 4 (June, 1964), pp. 335-45. Leontief, Wassily. "Domestic Production and Foreign Trade; the American Capital Position Re-examined,” Egg: ceedings of the American Philosophical Society, XCVII September, 1953), pp. 332-49. Leontief, Wassily. "Factor Proportions and the Structure of American Trade: Further Theoretical and Empir- ical Analysis," The Review of Economics and Statis- tics, XXXVIII, No. 4 (November, 19567, pp. 386-407. MacDougall, G. D. A. "British and American Exports: A Study Suggested by the Theory of Comparative Costs," Economic Journal (Part I: December, 1951, pp. 697- 724; Part II: September, 1952, pp. 487—521). 119 MacDougall, G. D. A., et a1. "British and American Produc- tivity, Prices and Exports: An Addendum," Oxford Economic Papers (October, 1962), pp. 297—304. McKenzie, Lionel W. "Specialization and Efficiency in World Production," Review of Economic Studies, XXIII, No. 1 (1954-1955), pp. 165-800 Metzler, Lloyd. "Graham' 5 Theory of International Values,” The American Economic Review, XL (June, 1950), pp. 301- 22. Michaely, M. "Factor Proportions in International Trade: Current State of the Theory," K klos, XVII (1964), pp. 529-50. Minhas, B. S. "The Homohypallagic Production Function, Factor-Intensity Reversals, and the Heckscher-Ohlin Theorem," The Journal of Political Economy, LXX (April, 1962), pp. 138-56. Moroney, John R. "The Strong-Factor-Intensity Hypothesis: A Multisectoral Test," The Journal of Political Econom , LXXV (June, 1967), pp. 241-49. Moroney, John R., and Walker, James M. "A Regional Test of the Heckscher-Ohlin Hypothesis," The Journal of Political Economy, LXXIV (December, 1966), pp. 573- 86. Ohlin, Bertil. Inter-regional and International Trade. Cambridge: Harvard University Press (Harvard Eco- nomic Studies), 1967. Ricardo, David. Principles of Political Economy and Taxa- tion. London: J. M. Dent and Sons, Limited, 1911. Robinson, Romney. "Factor Proportions and Comparative Ad- vantage," Qua arterly Journal of Economics, Part I, LXX (May, 1956), pp. 169- 92. Samuelson, P. A. ”International Trade and Equalisation of Factor Prices, " Economic Journal, LVIII (June, 1948), pp. 163- 84. Samuelson, P. A. "International Factor Price Equalisation Once Again," Economic Journal, LIX (June, 1949), pp. 181-97. Samuelson, P. A. "A Comment on Factor-Price Equalisation," Review of Economic Studies, XIX, No. 2 (1951- 52), pp. 121-220 120 Smith, Adam. An Inguiry into the Nature and Causes of the Wealth of Nations. In Edwin Cannan, ed. The Modern Library. New York: Random House, 1937. Stern, Robert M. "British and American Productivity and Comparative Costs in International Trade," Oxford Economic Papers, XIV (October, 1962), pp. 275-96. Stolper, Wolfgang F., and Roskamp, Karl W. "An Input-Output Table for East Germany with Applications to Foreign Trade, " Bulletin of the Oxford University Institute of Statistics, XXIII, No. 4 (November, 1961), pp. 379- 92. Swerling, Boris C. "Capital Shortage and Labor Surplus in the United States?“ The Review of Economics and Sta- tistics, XXXVI, No. 3 (August, 1954): pp. 286-89. Tarshis, Lorie. "Factor Inputs and International Price Com- parisons." In M. Abramovitz et al. (eds.), The Allo- cation of Economic Resources. Stanford, California: Stanford University Press, 1959, pp. 236-44. Tatemoto, Masahiro, and Ichimura, Shinichi. "Factor Propor- tions and Foreign Trade: The Case of Japan," The Review of Economics and Statistics, XLI, No. 4 (November, 1959), pp. 442-46. U. 5. Bureau of the Census. Annual Survey of Manufactures, 1957. Washington: U. S. Government Printing Office, 1959. U. S. Bureau of the Census. U. S. Census of Manufactures, 1958. Washington: U. 5. Government Printing Office, 1961. Valavanis-Vail, Stefan. "Leontief's Scarce Factor Paradox," The Journal of Political Economy, LXII, No. 6 (De- cember, 1954), pp. 523- 28. Wahl, Donald F. "Capital and Labour Requirements for Canada's Foreign Trade,“ The Canadian Journal of Economic and Political Science, XXVII, No. 3 (August, 1961), pp. 349-58. Whitin, T. M. "Classical Theory, Graham's Theory, and Lin- ear Programing in International Trade," Quarterly Journal of Economics, LXVII (November, 1953 , pp. APPENDICES APPENDIX I STANDARD INDUSTRIAL CLASSIFICATION OF 71 THREE-DIGIT INDUSTRIES { , H 1, 1 o, o .1. ._ .1 " L Ohlin Tests C‘las ical Model Tests Static Static Static South- °’ Change Non-market Ne ew England % Chan e in South— % Change Non—market 60% Labor 50% Labor New England- % Chan e in SIC non-South in South Oriented non—New England New England non—South in South Oriented Intensive Intensive non— —New England New England SIC Code Industry Titlea Tests Tests in South sts Tests Bests sts ‘ sts Test Cod 201 Heat products ' ‘ ‘ ‘ ' ‘ ' 0 ' 201 262 Dairies O O O O t t O O C O 202 203 Canned and frozen foods ‘ ‘ ' ' ' ' ’ ° ’ ' 203 204 Grain mills ‘ ‘ ' ‘ ' ' ‘ ‘ 204 205 Bakery products ’ ‘ ‘ ‘ ‘ ‘ ' ’ ‘ ’ 205 206 Sugar ' ' ' ' 206 207 Candy and related products ‘ ‘ ‘ ‘ ‘ ‘ ' ‘ 207 208 Bevera aeg o o o o o o o o o u 208 209 Miscellaneous foods and kindred products ’ ‘ ‘ ’ ‘ ‘ ‘ ‘ ‘ ' 209 225 Knitting mills ‘ ° ‘ ' ° ‘ ' ' ° 225 228 Yarn and thread mills ‘ ‘ ‘ ‘ ‘ ' ' 228 229 Miscellaneous textile goods ‘ ' ‘ ‘ ‘ ‘ ‘ ‘ 229 231 Men's and boys' suits and coats ‘ ' ‘ ° ' ' ' ' ' 0 231 232 Men's and boys' furnishings ' ' ' ' ‘ ‘ ' ' t 232 233 Women's and misses' outerwear ' ‘ ’ ’ ' ' ‘ ‘ ' ° ‘ 0 233 234 Women's and children's underwear ' ' ' ' ' ‘ ' t 234 236 Children's outerwearl ‘ ' ‘ ‘ ' ‘ ' ' ' 0 236 238 Miscellaneous appare ' ‘ ‘ ‘ ‘ ‘ ' ' ' ‘ 238 239 Fabricated textiles,ln nc.e. . ‘ ' ‘ ' ' ‘ ‘ ‘ ' t t 239 243 Millwork and related products ‘ ‘ ' ‘ ‘ ‘ ‘ ‘ ' ° ‘ ' 243 244 Wooden container ' ' ‘ ‘ ' ' ' ' ' ' 244 249 Miscellaneous wood products ‘ ' ‘ ’ ‘ ' ' ' ' 249 251 Household furniture ’ ‘ ’ ' ‘ ‘ ' ' ' 251 252 Office furniture ' ‘ ' ‘ ‘ ‘ ' ‘ ‘ ‘ 252 253 Public building furniture‘ ' ‘ ‘ ' ‘ ‘ ‘ 253 265 Paperboard containers and boxes ' -‘ ' ‘ ‘ ‘ ‘ ‘ ' ‘ ' 265 273 Books a o o o o o o o o o o 273 278 Bookbinding and related work ‘ ' ' ' ‘ ‘ ' ‘ ‘ ‘ ‘ ' 278 279 Printing trades services ’ ‘ ‘ ‘ ’ ‘ ‘ ‘ ‘ ‘ 279 283 Drugs and medicines ‘ ‘ ‘ ‘ 283 284 Cleaning and toilet goods ‘ ' ’ ‘ ‘ ‘ ' ‘ ‘ ' 284 285 Paint and allied products ‘ ' ' ‘ ’ ‘ ‘ ‘ ‘ ‘ 285 287 Agricultural chemica s ‘ ' ‘ ‘ ' ' ‘ ' 287 295 Paving and roofing materials ' ‘ ’ ' ‘ ‘ ‘ ‘ 295 299 Petroleum and coal products, n.e.c. ' ‘ ‘ ‘ ‘ ‘ ‘ ‘ ‘ ° 299 314 Footwear, except rubber ‘ ‘ ‘ ‘ ‘ ' ‘ ‘ ‘ 314 317 Purses and small leather goods ‘ ‘ ‘ ' ‘ ‘ ‘ ‘ 317 322 Pressed and blown glassware ‘ ' ‘ ' 322 325 Structural clay products ’ ‘ ’ ‘ ' ‘ ' ‘v ' 323 326 Pottery and related products ‘ ' ‘ ' ‘ ‘ ‘ ‘ ' ’ 326 329 Nonmetallic mineral products ' ‘ ‘ ' ‘ ‘ ‘ ' 329 327 Concrete and plaster products ‘ ‘ ‘ ' ’ ‘ ‘ ‘ ‘ ' 327 332 Iron and steel foundries ' a t O . o . o u . 332 333 Primary nonferrous metals ' ‘ - o 333 335 Nonferrous rolling and drawing ‘ ' o o o 335 339 Primary metal industries, n. e. c. ' 0 . o o 339 342 Cutlery, hand tools, hardwar ‘ ‘ ' t . . . o 342 343 Plumbing and heating, except eelectric ' ° ‘ 0 - . . o o 343 344 Structural metal produ usct ‘ ‘ ' ' t 0 - . . . . o 344 346 Metal stampings ' 0 0 t . . o o . . 346 348 Fabricated wire products, n.e.c. ‘ ‘ ' ' 0 t o o o o 348 349 Fabricated metal products, n.e.c. ' t a o a o . o o 349 351 Engines and turb nes ' t ' s o o o . 351 35 Farm machinery and equipment ' ' ' 0 . c o o . 352 353 Construction and like equipment ' ‘ t t - - . - o 353 354 Metalworking machinery ‘ ' ’ o o 354 355 Special industry mac chin ' ' ' ' o s . t - . o 355 356 General industrial machinery ‘ ' . o o 356 357 Office machines, n.e. c. ‘ ’ ' ' ‘ ‘ 357 362 Electrical industrial apparatus ‘ ‘ ‘ ‘ ' ‘ 362 366 Communication equipmen ‘ ‘ ‘ ‘ ‘ 366 369 Electrical products, n.e.c. ' ' ‘ ‘ 369 371 Motor vehicles and equipment ‘ ' ‘ ‘ ‘ ' 371 372 Aircraft and par ’ ' t o . . 372 373 Ship and boat building ‘ ' ' t t o . . o . 373 384 Medical instruments and supplies ’ ' ‘ ‘ ' ' . . 384 391 Jewelry and silverwares ‘ ‘ 0 t c . . a . . 391 394 Toys and sporting good ' ‘ ‘ ‘ ’ ‘ t . a o a 394 395 Pens, penci s, and office supplies ' ‘ ‘ ‘ ' t t a - t 395 396 Costume jewelry and notions ‘ ’ ‘ ‘ ‘ ' 396 399 I1 ‘rpI'lanpnus ‘ ‘ ' ‘ O ' 0 v c n o 399 aSIC code and industry title from Census of Manufacturesl 1958 ".3 Department of Commerce, Bureau of Census. APPENDIX II LIST OF RANKINGS, IN ASCENDING ORDER, OF VARIABLES FOR EACH HECKSCHER-OHLIN TEST 125 Table 1. Ranks of gross capital-labor ratios and concentra- i tion ratios,I:§, for the South VN Gross Concen- Gross Concen- SIC Capital-Labor tration SIC Capital-Labor tration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 233 l 18 356 36 8 317 2 3 207 37 24 236 3 42 353 38 43 232 4 67 342 39 10 231 5 44 369 4O 19 314 6 31 355 41 32 238 7 34 203 42 46 234 8 57 332 43 39 239 9 45 229 44 58.5 372 10 33 343 45 38 394 ll 21 265 46 41 278 12 22 354 47 4 391 13 2 228 48 71 396 14 7 351 49 l 251 15 64 322 50 55 244 16 68 357 51 6 399 17 25 352 52 20 249 18 56 325 S3 61 253 19 49.5 349 54 36 366 20 27 202 SS 52 225 21 66 339 56 5 279 22 23 299 57 30 373 23 65 371 58 15 243 24 47 329 59 28 395 25 29 285 60 37 384 26 9 204 61 49.5 252 27 14 283 62 12 344 28 48 327 63 62 273 29 13 209 64 58.5 362 30 16 335 65 35 205 31 53 284 66 17 348 32 26 208 67 60 326 33 40 295 68 S4 346 34 11 287 69 70 201 35 51 206 70 63 333 71 69 126 Table 2. Ranks of net capital-labor ratios and concentra- 1 tion ratios, :3, for the South vi N Net Concen- Net Concen- SIC Capital-Labor tration SIC Capital-Labor tration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 233 1 18 369 36 19 236 2 42 346 37 11 317 3 3 356 38 8 231 4 44 353 39 43 232 5 67 355 40 32 238 6 34 201 41 51 314 7 31 351 42 1 234 ,8 57 354 43 4 239 9 45 343 44 38 394 10 21 332 45 39 372 ll 33 203 46 46 391 12 2 229 47 58.5 244 13 68 228 48 71 278 14 22 352 49 20 396 15 7 322 50 55 249 16 56 265 51 41 279 17 23 299 . 52 30 251 18 64 325 53 61 225 19 66 202 54 52 399 20 25 349 55 36 373 21 65 357 S6 6 253 22 49.5 339 57 5 366 23 27 329 58 28 243 24 47 285 59 37 395 25 29 371 60 15 273 26 13 209 61 58.5 384 27 9 204 62 49.5 205 28 53 335 63 35 344 29 48 327 64 62 362 30 16 283 65 12 326 31 40 208 66 60 348 32 26 295 67 ' S4 252 33 14 284 68 17 207 34 24 287 69 70 342 35 19 206 70 63 333 71 69 127 Table 3. Ranks of gross capital-weighted labor ratios and i concentration ratios, XE; for the South i VN Gross Capital Concen— Gross Capital Concen- SIC Weighted-Labor tration SIC Weighted-Labor tration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 233 l 18 355 36 32 231 2 44 201 37 51 317 3 3 342 38 10 236 4 42 351 39 1 314 5 31 225 40 66 232 6 67 369 41 19 238 7 34 326 42 40 372 8 33 332 43 39 234 9 S7 343 44 38 279 10 23 357 45 6 239 ll 45 265 46 41 399 12 25 352 47 20 278 13 22 207 48 24 391 14 2 349 49 36 394 15 21 322 50 55 373 16 65 229 51 58.5 366 17 27 339 52 5 253 18 49.5 299 53 30 251 19 64 371 54 15 396 20 7 202 55 52 344 21 48 325 56 61 361 22 16 329 S7 28 252 23 14 285 58 37 395 24 29 203 59 46 384 25 9 283 6O 12 273 26 13 284 61 17 243 27 47 222 62 71 356 28 8 335 63 35 244 29 68 204 64 49.5 348 30 26 295 65 54 353 31 43 208 66 60 249 32 56 327 67 62 354 33 4 209 68 58.5 205 34 53 287 69 70 346 35 11 206 70 63 333 71 69 128 Table 4. Ranks of net capital-weighted labor ratios and i concentration ratios, v5, for the South VN Net Capital Concen- Net Capital Concen- SIC Weighted-Labor tration SIC Weighted-Labor tration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 233 1 18 355 36 32 231 2 44 342 37 10 317 3 3 369 38 19 236 4 42 225 39 66 232 5 67 346 40 11 314 6 31 326 41 40 238 7 34 201 42 51 234 8 57 332 43 39 372 9 33 343 44 38 279 10 23 299 45 30 239 11 45 352 46 20 391 12 . 2 207 47 24 278 13 22 229 48 58.5 394 14 21 322 49 55 373 15 65 267 50 41 399 16 25 357 51 6 396 17 7 349 52 36 251 18 64 339 S3 5 244 19 68 202 54 52 273 20 13 325 55 61 249 21 S6 371 56 15 366 22 27 285 57 37 344 23 48 329 58 28 253 24 49.5 203 S9 46 361 25 16 335 60 35 384 26 9 222 61 71 395 27 29 283 62 12 243 28 47 204 63 49.5 354 29 4 209 64 58.5 205 30 53 295 65 S4 351 31 l 284 66 17 348 32 26 208 67 60 356 33 8 327 68 62 353 34 43 287 69 70 252 35 14 206 70 63 333 71 69 129 Table 5. Ranks of gross capital—labor ratios and percentage i changes in concentration ratios, XE, in the South, 1947-1958 vi N Percentage Percentage Change in Change in Gross Concen- Gross Concen- SIC Capital-Labor tration SIC Capital-Labor tration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 287 55 1 299 46 29 239 9 2 201 30 30 249 16 3 203 35 31 353 32 4 295 54 32 204 49 5.5 278 11 33 243 20 5.5 202 45 34 228 40 7 267 39 35 208 53 8 231 5 36 373 19 9 332 36 37 284 52 10 314 6 38 322 41 11 232 4 39 209 51 12 238 7 40 329 47 13 399 15 41 206 56 14 395 21 42 346 29 15 355 34 43 285 48 16 229 37 44 205 26 17 333 57 45 326 28 18 352 42 46 325 43 19.5 394 10 47 344 24 19.5 384 22 48 327 50 21 252 23 49 391 12 22 348 27 50 343 38 23 233 1 51 225 17 24 342 33 52 207 31 25 349 44 53 251 13 26 317 2 54 279 18 27 234 8 55 244 14 28 273 25 56 236 3 57 130 Ranks of gross capital—labor ratios and concentra- for non-market oriented industries SIC Capital-Labor Gross Ratio Ranks Concen- tration Ratio Ranks Table 6. tion ratios, XE: i vN — q Gross Concen- SIC Capital-Labor tration Code Ratio Ranks Ratio Ranks Code 317 1 3 342 236 2 23 369 232 3 37 332 231 4 25 229 314 5 15 343 238 6 17 354 234 7 30 228 372 8 16 351 391 9 2 322 396 10 7 357 251 11 34 352 244 12 38 325 249 13 29 349 366 14 14 339 225 15 36 204 373 16 35 283 384 17 9 335 326 18 22 287 201 19 27 206 356 20 8 333 353 21 24 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 10 12 21 31 20 4 41 1 28 6 13 32 19 5 26 11 18 4O 33 39 131 Table 7. Ranks of gross capital-labor ratios and concentra- i tion ratios, vne, in New England vi N Gross Concen- Gross Concen- SIC Capital—Labor tration SIC Capital-Labor tration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks .233 1 35 201 35 15 317 2 61 356 36 57 236 3 32 207 37 42 232 4 20 353 38 6 231 5 30 342 39 64 314 6 66 369 40 24 236 7 53 355 41 62 234 8 44 203 42 19 239 9 37 332 43 14 372 10 38 229 44 65 394 11 54 343 45 11 278 12 45 267 46 40 391 13 68 354 47 56 396 14 67 228 46 59 251 15 27 351 49 52 244 16 26 357 so 55 399 17 47 352 51 1 249 16 56 325 52 ’ 4 253 19 46 _349 53 43 366 20 so 202 54 29 225 21 31 339 55 51 279 22 34 299 56 6 373 23 63 371 57 2 243 24 13 329 56 49 39s 25 39 285 59 16 364 26 41 204 60 7 252 27 10 263 61 9 344 26 21 327 62 22 273 29 46 209 63 12 362 30 16 335 64 60 205 31 25 264 65 33 348 32 36 208 66 17 326 33 5 295 67 26 346 34 23 287 68 3 132 Table 8. Ranks of net capital-labor ratios and concentration 1 ratios, vne, in New England vi N Net Concen- Net Concen- SIC Capital-Labor tration SIC Capital-Labor tration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 233 1 35 342 35 64 236 2 32 369 36 24 317 3 61 346 37 23 231 4 30 356 38 57 232 5 20 353 39 6 238 6 53 355 40 62 314 7 66 201 41 15 234 8 44 351 42 52 239 9 37 354 43 58 394 10 54 343 44 11 372 11 38 332 45 14 391 12 68 203 46 19 244 13 26 229 47 65 278 14 45 228 48 59 396 15 67 352 49 l 249 16 56 267 50 40 279 17 34 299 51 8 251 18 27 325 52 4 225 19 31 202 53 29 399 20 47 349 54 43 373 21 63 357 55 55 253 22 48 339 56 51 366 23 50 329 57 49 243 24 13 285 58 18 395 25 39 371 59 2 273 26 46 209 60 12 384 27 41 204 61 7 205 28 25 335 62 60 344 29 21 327 63 22 362 30 16 283 64 9 326 31 5 208 65 17 348 32 36 295 66 28 252 33 10 284 67 33 207 34 42 287 68 3 133 Table 9. Ranks of gross capital-labor ratios and percentage changes in concentration ratios in New England, 1947-1958 Percentage Percentage Change in Change in Concen- Gross Concen- Gross SIC tration Capital-Labor SIC tration Capital-Labor Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 299 1 46 384 27 23 287 2 53 351 28 40 343 3 38 249 29 17 244 4 15 284 30 51 352 5 42 279 31 19 399 6 16 314 32 6 232 7 4 344 33 25 394 8 10 285 34 47 357 9 41 251 35 14 346 10 30 396 36 13 348 ll 28 204 37 48 349 12 44 231 38 5 332 13 36 201 39 31 205 14 27 209 40 50 342 15 33 208 41 52 203 16 35 327 42 49 234 17 8 225 43 18 243 18 21 273 44 26 278 19 11 202 45 45 355 20 34 395 46 22 267 21 39 233 47 1 353 22 32 236 48 3 391 23 12 238 49 7 239 24 9 317 50 2 229 25 37 326 51 29 325 26 43 252 52 24 373 53 20 APPENDIX III RANK CORRELATION TEST RESULTS FOR THE HECKSCHER-OHLIN MODEL 135 Explanatory Concentra- Sample Sign Level of Variable tion Ratios Size Hypothesized ‘2: Significance K i G Vs L_' -e- 71 - +.O78 n.s. v1 N KN v: _ — 71 "' +0091 nos. L Vi N KG V: I7' -e- 71 — +.l74 10% w v:L N KN v1 i.— -—?— 71 — +.168 10% w v1 N K Vi G s i— _i 41 "' +0037 nos. VN non-market oriented % change in KG v1 E— -—~?" 51 - -.256 .576 V1 N KG vie V N i K V L—“— 4.19- 68 - -.243 .594 v1 N % change in i K v E2. ne 53 - -.129 n.s. APPENDIX IV LIST OF RANKINGS, IN ASCENDING ORDER, OF VARIABLES FOR EACH CLASSICAL MODEL TEST 137 Table 1. Ranks of average labor cost ratios and concentration i ratios, :5, for the South Vi N Average Average SIC Labor Cost Concentration SIC Labor Cost Concentration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 279 l 25 325 36.5 61 357 2 6 239 36.5 45 362 3 20 265 38 41 333 4 69 299 39 30 371 5 15 251 40 64 396 6 7 205 41 53 354 7 4 335 42 35 317 8 3 228 43.5 71 339 9 5 231 43.5 44 369 10 19 249 45 56 314 11 31 295 46 54 384 12 9 366 47 27 348 13 26 353 48 43 329 14 28 394 49.5 21 283 15 12 225 49.5 66 234 16 57 344 51 48 229 17 58.5 204 52 49.5 342 18 10 209 53.5 58.5 332 19.5 39 284 53.5 17 356 19.5 8 238 55 34 349 21 36 207 ' 56 24 395 22 29 346 57 11 327 23 62 373 58 65 208 24 60 372 59 33 355 25 32 243 60 47 244 26 68 326 61 40 399 27 25 203 62 46 201 28 51 343 63 38 287 29 70 236 64 42 202 30.5 52 233 65 18 285 30.5 37 232 66 67 322 32 55 391 67 2 206 33 63 352 68 20 252 34.5 14 351 69 1 278 34.5 22 273 7O 13 253 71 49.5 138 Table 2. Ranks of labor productivity ratios and concentration i ratios, :3, for the South 1 VN SIC Productivity Concentration SIC Productivity Concentration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 273 243 253 351 225 203 233 391 207 204 232 236 209 201 238 343 249 326 229 244 352 251 325 299 252 208 228 327 344 239 295 287 206 384 346 \qummbwml‘ 13 47 49.5 1 66 46 18 2 24 49.5 67 42 58.5 51 34 38 56 40 58.5 68 20 64 61 30 14 6O 71 62 373 284 342 395 202 279 372 355 394 399 332 231 205 265 283 317 278 285 322 353 348 329 356 366 314 234 354 335 349 396 339 371 369 362 357 333 36 37 38 39 4O 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 65 17 10 29 52 23 33 32 21 25 39 44 53 41 12 3 22 37 55 43 26 28 8 27 31 57 4 35 36 7 5 15 19 16 6 69 139 Table 3. Ranks of average labor cost ratios and relative concentration ratios,IE§, for the South C n Average Relative Average Relative SIC Labor Cost Concentration SIC Labor Cost Concentration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 279 1 22 325 36 61 357 2 6 239 37 45 362 3 17 267 38 41 333 4 69 299 39 29 371 5 16 251 40 64 396 6 7 205 41 S3 354 7 5 335 42 35 317 8 3 228 43.5 71 339 9 4 231 43.5 44 369 10 14 249 45 56 314 11 31 295 46 54 384 12 9 366 47 27 348 13 26 353 48 43 329 14 28 394 49 23 283 15 12 225 50 66 234 16 57 344 51 48 229 17 59 204 52 49 342 18 10 209 53 58 332 19 39 284 54 18 -356 20 8 238 55 33 349 21 36 207 56 24 395 22 30 346 57 11 327 23 62 373 58 65 208 24 60 372 59 34 355 25 32 243 60 47 244 26 68 326 61 40 399 27 25 203 62 46 201 28 50 343 63 38 287 29 70 236 64 42 202 30 52 233 65 19 285 31 37 232 66 67 322 32 55 391 67 2 206 33 63 352 68 20 252 34 15 351 69 1 278 35 21 273 70 13 253 71 51 140 Table 4. Ranks of labor productivity ratios and relative concentration ratios, 5;, in the South C n Labor Relative Labor Relative SIC Productivity Concentration SIC Productivity Concentration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 273 l 13 373 36 65 243 2 47 284 37 18 253 3 51 342 38 10 351 4 1 395 39 30 225 5 66 202 40 52 203 6 46 279 41 22 233 7 19 372 42 34 391 8 2 355 43 32 207 9 24 394 44 23 204 10 49 399 45 25 232 11 67 332 46 39 236 12 42 231 47 44 209 13 58 205 48 53 201 14 50 267 49 41 238 15 33 283 50 12 343 16 38 317 51 3 249 17 56 278 52 21 326 18 40 285 53 37 229 19 59 322 54 55 244 20 68 353 55 43 352 21 20 348 56 26 251 22 64 329 S7 28 325 23 61 356 58 8 299 24 29 366 59 27 252 25 15 314 60 31 208 26 60 234 61 57 228 27 71 354 62 5 327 28 62 335 63 35 344 29 48 349 64 36 239 30 45 396 65 7 295 31 54 339 66 4 287 32 70 371 67 16 206 33 63 369 68 14 384 34 9 362 69 17 346 35 11 357 70 6 333 71 69 141 Table 5. Ranks of average labor cost ratios and percentage changes in relative concentration in the South, 1947-1958 Percentage Percentage Changes in Changes in Average Relative Average Relative SIC Labor Cost Concentration SIC Labor Cost Concentration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 279 1 18 267 26 27 354 2 44 251 27 24 317 3 50 205 28 15 314 4 30 228 29 2 384 5 42 231 30 28 348 6 35 249 31 20 234 7 48 295 32 29 229 8 40 225 33 25 342 9 45 394 34 41 332 10 31 344 35 14 349 11 46 204 36 6 395 12 38 209 37 11 327 l3 19 284 38 9 208 14 7 238 39 33 355 15 36 346 40 12 244 16 32 373 41 8 399 17 34 243 42 5 201 18 22 326 43 4 287 19 1 203 44 21 202 20 26 343 45 17 285 21 13 236 46 49 322 22 10 233 47 43 278 23 23 232 48 37 325 24 16 352 49 39 239 25 3 273 50 47 142 Table 6. Ranks of labor productivity ratios and percentage changes in relative concentration in the South, 1947-1958 Percentage Percentage Changes in Changes in Labor Relative Labor Relative SIC Productivity Concentration SIC Productivity Concentration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 273 1 47 287 26 1 243 2 6 384 27 42 225 3 25 346 28 12 203 4 21 373 29 9 233 5 43 284 30 5 204 6 7 342 31 45 232 7 37 395 32 38 236 8 49 202 33 26 209 9 11 279 34 18 201 10 22 355 35 36 238 11 33 394 36 41 343 12 17 399 37 34 249 13 20 332 38 31 326 14 4 231 39 28 229 15 40 205 40 15 352 16 39 267 41 27 244 17 32 317 42 50 251 18 24 278 43 23 325 19 16 285 44 13 208 20 8 322 45 10 228 21 2 348 46 35 327 22 19 314 47 30 344 23 14 234 48 48 239 24 3 354 49 44 295 25 29 349 50 46 143 Table 7. Ranking of relative concentration in the South, SE, n labor productivity ratios, and average labor cost ratios for non-market oriented industries Relative Labor Concentration Productivity Average Labor SIC Code Ratio Ranks Ratio Ranks Cost Ratio Ranks 351 1 l 41 391 2 3 39 317 3 27 5 339 4 38 6 354 5 34 4 357 6 4O 1 396 7 37 3 356 8 3O 15 384 9 20 9 342 10 22 13 283 11 26 10 369 12 39 7 352 13 14 40 366 14 31 28 314 15 32 8 238 16 8 32 372 17 23 34 335 18 35 24 349 19 36 16 343 20 9 36 332 21 24 14 326 22 11 35 236 23 6 37 353 24 29 29 231 25 25 26 204 26 4 31 201 27 7 18 322 28 28 20 249 29 10 27 234 30 33 11 229 31 12 12 325 32 16 22 206 33 19 21 251 34 15 23 373 35 21 33 225 36 2 30 232 37 5 38 244 38 13 17 333 39 41 2 287 40 18 19 228 41 17 25 144 Table 8. Ranks of average labor cost ratios and relative concentration ratios, Cne , in New England nne Average Relative Average Relative SIC Labor Cost Concentration SIC Labor Cost Concentration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 253 1 50 332 35 14 357 2 55 369 36 24 399 3 47 362 37 16 366 4 49 295 38 27 352 5 1 317 39 61 201 6 15 284 40 32 373 7 63 229 41 65 204 8 8 353 42 6 239 9 37 355 43 62 225 10 31 349 44 43 278 11 46 238 45 53 326 12 5 395 46 39.5 342 13 64 391 47 68 351 14 52 243 48 12 273 15 45 394 49 54 232 16 20 314 50 66 234 17 44 372 51 38 236 18 33 231 52 30 356 19 57 203 S3 19 343 20 11 205 54 25 344 21 21 209 55 13 265 22 39.5 371 56 2 279 23 34 249 57 56 339 24 51 207 58 41 285 25 18 348 59 36 327 26 22 299 60 7 354 27 58 325 61 4 228 28 59 287 62 3 251 29 28 244 63 26 335 30 60 384 64 42 208 31 17 396 65 67 202 32 29 283 66 10 233 33 35 329 67 48 252 34 9 346 68 23 _l‘lull ‘lll‘llllu‘l' ‘I 3.1111111 .l 145 Table 9. Ranks of labor productivity ratios and relative C concentration ratios, Cne , in New England nne Labor Relative Labor Relative SIC Productivity Concentration SIC Productivity Concentration Code Ratio Ranks Ratio Ranks Code Ratio Ranks Ratio Ranks 346 1 384 2 283 3 287 4 252 5 371 6 209 7 203 8 329 9 231 10 243 11 349 12 362 13 205 14 244 15.5 394 15.5 249 17 325 18 353 19 208 21 332 21 355 21 369 23 207 24 354 25 299 26 335 27 233 28 391 29 285 30 279 31.5 348 31.5 202 33 395 34 23 42 10 39.5 265 343 238 352 356 396 372 314 295 229 284 339 342 344 239 234 273 251 357 236 278 399 317 366 201 204 327 351 326 232 228 225 373 253 35 36 37 39.5 11 53 1 57 67 38 66 27 65 32 51 64 21 37 44 45 28 55 33 46 47 61 49 15 8 22 52 5 20 59 31 63 50 146 Table 10. Ranks of average labor cost ratios and percentage changes in relative concentration in New England, 1947-1958 Average Percentage Average Percentage SIC Labor Cost Change in SIC Labor Cost Change in Code Ratio Rank Concentration Code Ratio Rank Concentration 357 1 6 332 28 10 399 2 38 317 29 50 352 3 4 284 30 26 201 4 36 229 31 21 373 5 53 353 32 20 204 6 31 355 33 14 239 7 22 349 34 8 225 8 42 238 35 47 278 9 16 395 36 45 326 10 51 391 37 17 342 11 9 243 38 15 351 12 25 394 39 5 273 13 43 314 40 28 232 14 34 231 41 33 234 15 12 203 42 13 236 16 48 205 43 11 343 17 2 209 44 39 344 18 37 249 45 27 265 19 18 348 46 19 279 20 29 299 47 49 285 21 30 325 48 23 327 22 41 287 49 1 251 23 32 244 50 3 208 24 40 384 51 24 202 25 44 396 52 35 233 26 46 346 53 7 252 27 52 147 Table 11. Ranks of labor productivity ratios and percentage changes in relative concentration in New England, 1947-1958 Percentage Percentage Labor Change in Labor Change in SIC Productivity Concentration SIC Productivity Concentration Code Rank Rank Code Rank Rank 346 1 7 343 28 2 384 2 24 238 29 47 287 3 l 352 30 .4 252 4 52 396 31 35 209 5 39 314 32 28 203 6 13 284 33 26 231 7 33 229 34 21 243 8 15 342 35 9 349 9 8 344 36 37 205 10 11 239 37 22 394 11 5 234 38 12 244 12 3 273 39 43 249 13 27 251 40 32 325 14 23 357 41 6 353 15 20 236 42 48 208 16 40 278 43 16 332 17 10 399 44 38 355 l8 14 317 45 50 299 19 49 201 46 36 233 20 46 204 47 31 391 21 17 327 48 41 285 22 30 351 49 25 279 23 29 326 50 51 348 24 19 232 51 34 202 25 44 225 52 42 395 26 45 373 53 53 265 27 18 APPENDIX V RANK CORRELATIONS TEST RESULTS FOR THE CLASSICAL MODEL Explanatory Concentra- 149 Sample Level of Variable tion Ratios Size Hypothesized 2: Significance vi Labor 5 Productivity vi 71 -°202 5% N Average i Labor vS 71 +.116 n.s. Cost -f V N i Labor vs 16 Productivity -3- 60% Labor +.O33 n.s. VN intensive 1 Average VS 16 Labor -v- 60% Labor -.l83 n.s. 1 . . Cost VN 1ntens1ve Labor Cs Productivity CE. 71 -'199 5% Average CS Labor E—- 71 +.136 10% Cost n C Labor _g, 38 Productivity Cn 50% Labor -.183 10% intensive Average E§_ 38 Labor Cn 50% Labor +.013 n.s. Cost intensive Labor SE. 41 Productivity Cn Non-market -.193 10% oriented Average SE, 41 Labor Cn Non-market +.144 n.s. Cost oriented 150 Explanatory Concentra- Sample Sign Level of Variable tion Ratios Size Hypothesized g: Significance % change in Labor Cs Productivity 53' 50 * *0062 n.s. % change in Average Cs Labor E—- 50 - -.231 6% Cost n Labor Cne . 68 + +.221 1% Productiv1ty Cnne Average Cne Cost nne % change in Labor ne ' '—__' 53 + +.249 1% Productivity Cnne % change in Average ne Labor C 53 - -.138 n.s. Cost nne APPENDIX VI ANALYSIS OF EIGHT SIC THREE-DIGIT INDUSTRIES 152 ANALYSIS OF EIGHT INDUSTRIES TO DETERMINE SOURCES or DEMAND AND RAW MATERIALSa 1. Industry 243-—Mi11work and related products Sub-industries: millwork plants, veneer and plywood plants, prefabricated wooden build- ings and structural members Raw Material Sources: Industry 2421: sawmills and planing mills; 23 per- cent of total output in the South; mills produce rough lumber, dressed lumber, and softwood cut stock Industry 3553: woodworking machinery; 20 percent of total output in the South Demand for Industry4243 Products: Industry 244: wooden containers; 47.5 percent of total output in the South Conclusion: Both sources of demand and raw materials are reasons for locating in the South. 2. Industry 253--Public building furniture Raw Material Sources: Difficult to discern; possibly Industry 243, just analyzed, which has a high South concen- tration Demand for Industry 253 Products: No industrial demand as these goods are sold for direct use, not as raw materials. Conclusion: No explanation for its ranking. 3U. S. Bureau of the Census, Census of Manufactures, 1958, Vol. II, Parts 1 and 2 (Washington: U. S. Government Printing Office, 1961). 153 3. Industry 225--Knitting mills Sub-industries: hosiery, knit outerwear and underwear, knit fabrics Raw Material Sources: Knit products from yarns Industry 228: yarn and thread mills; 68 percent of total output in the South Demand for Industry 225 Products: Non-industrial de- mand Conclusion: Sources of raw materials are the major factor in the regional concentration of this industry. 4. Industry 232-~Men's and boys' furnishings Sub-industries: dress shirts, underwear, neckwear, trousers, and work clothing Raw Material Sources: Goods manufactured from pur- chased woven or knit fabric Industry 2256: knit fabric mills; 27 percent of total output in the South Industry 2211: weaving mills, cotton; 91 percent of total output in the South Industry 2221: weaving mills, synthetics; 70 per- cent of total output in the South Demand for Industry 232 Products: Non-industrial de- mand Conclusion: Sources of raw materials are the major factor in the regional concentration of this industry. 5. Industry 278--Bookbinding and related work Sub-industries: blankbooks and looseleaf binders; bronzing, gilding, and edging; map and sample mounting Raw Material Sources: Difficult to condense Industry 262: paper mills; 29 percent of total output in the South 154 Industry 264: paper and paperbound products Demand for Industry7278 Products: Industry 273: publishing and printing of books; 7.4 percent of total output in the South Industry 2761: manifold business forms; 16 percent of total output in the South Industry 2771: greeting cards; 1.2 percent of total output in the South Industry 275: general commercial printing; 13 per cent of total output in the South Conclusion: Uncertainty about the sources of both demand and raw materials makes any judg- ment difficult. A lack of strong demand in the South could be important. 6. Industry 356--General industrial machinery Sub-industries: pumps and compressors, ball and roller bearings, blowers and fans, power transmission equipment, and industrial ovens and furnaces Raw Material Sources: Industry 34: fabricated metal products; 13 percent of total output in the South Demand for Industry 356 Products: The general level of industrial activity is prob- ably the best indicator due to the diversity of products in Industry 356. Thus, most demand is in the non-South. Conclusion: Relative concentration of both demand and raw material sources in the non- South explains output concentration in the non-South. 155 7. Industry 354--Meta1working machinery Sub-industries: metal cutting and forming machine tools; special dies and tools; ma- chine tool accessories Raw Material Sources: Industry 331: steel rolling and finishing; 17 per- cent of total output in the South Industry 335: nonferrous rolling and drawing; 14 percent of total output in the South Demand for Industry 354 Products: Industry 34: fabricated metal products; 13 percent of total output in the South Conclusion: High demand and raw material concentra- tion in the non-South explains the high production concentration in the non-South. 8. Industry 339--Primary metal industries, n.e.c. Sub-industries: iron and steel forgings and non- ferrous forgings Raw Material Sources: Industry 331 and 335 as analyzed above Demand for Industry 339 Products: Industry 34 as analyzed above Conclusion: High demand and raw material concentra- tion in the non-South explains the high production concentration in the non-South. APPENDIX VII RANKS OF SIC THREE-DIGIT INDUSTRIES BY COEFFICIENT OF RESOURCE DEPENDENCY 157 Table l. Ranks of coefficients of resource dependency and relative concentration ratios in the South Coefficient Relative Coefficient Relative SIC of Resource Concentration SIC of Resource Concentration Code Dependency Ratios Code Dependency Ratios 201 1 43 353 31 36 206 2 55 317 32 2 209 3 51 391 33 l 204 4 42 369 34 12 202 5 45 205 35 46 287 6 60 372 36 27 203 7 39 349 37 29 228 8 61 394 38 19 335 9 28 314 39 24 229 10 52 396 40 6 299 ll 22 253 41 44 295 12 47 395 42 23 239 13 38 329 43 21 207 14 20 373 44 57 243 15 40 355 45 25 265 16 34 284 46 15 232 17 58 356 47 7 244 18 59 357 48 5 285 19 30 342 49 9 344 20 41 384 50 8 233 21 16 332 51 32 236 22 35 362 52 14 238 23 26 252 53 13 234 24 50 273 54 11 327 25 54 322 55 48 249 26 49 354 56 4 339 27 3 278 57 17 251 28 S6 326 58 33 343 29 31 325 59 53 231 30 37 283 60 10 279 61 18 158 Table 2. Ranks of coefficients of resource dependency and relative concentration ratios for New England Coefficient Relative Coefficient Relative of Resource Concentration of Resource Concentration SIC Dependency Ratio SIC Dependency Ratio Code Ranks Ranks Code Ranks Ranks 201 1 11 231 29 24 209 2 10 353 30 4 204 3 6 317 31 50 202 4 23 391 32 57 287 5 1 369 33 18 203 6 14 205 34 19 228 7 48 372 35 30 335 8 49 349 36 35 229 9 54 394 37 43 299 10 5 314 38 55 295 11 21 396 39 56 239 12 29 253 40 40 207 13 33 395 41 32 243 14 9 329 42 39 265 15 31 373 43 52 232 16 15 355 44 51 244 17 20 284 45 25 285 18 13 356 46 46 344 19 16 357 47 44 233 20 28 342 48 53 236 21 26 384 49 34 238 22 42 362 50 12 234 23 36 252 51 7 327 24 17 273 52 37 249 25 45 354 53 47 339 26 41 278 54 38 251 27 22 326 55 3 343 28 8 325 56 2 279 57 27 1’ I II «131111311311 11113111171