I am grateful to Christopher Clague, David Felix, Ricardo Halperin, Bruce Herrick, Charles P. Kindleberger, Robert Levine, Daniel Schydlowsky, Stanley Stein, Jonathan Levin, Carlos Diaz-Alejandro, and especially to Maurice Wilkinson for helpful comments on earlier drafts of this article. They bear no responsibility, however, for any errors. Michael Sterling and Bruce Phillips provided highly competent research assistance. Support from a grant of the faculty research program of the Graduate Business School of Columbia University is also gratefully acknowledged.

¹ The source referred to is Octávio Dias Carneiro's “Past Trends in the Economic Evolution of Brazil, 1920–1965,” Harvard University Center for International Affairs, 1966Google Scholar (mimeographed). Dr. Carneiro's time-series use data from the census years as benchmarks and the intermediate years are estimated with economic indicators such as electricity consumption, agricultural output, and construction figures. Copies of the Cameiro paper are available in the Columbia and Harvard University libraries. Data sources for the other principal time-series used in this article (exports, imports, the balance of trade, the exchange rate, population, currency supply, price deflator, industrial output and the federal budget deficit) are presented in the Appendix.

² See, e.g., Kafka, Alexandre, “The Theoretical Interpretation of Latin American Development,” in Ellis, H. S., ed., Economic Development for Latin America (New York: St. Martins Press, 1963), pp. 1–4, 8–11. Similar views are to be found in most other writings on Brazilian development. It should be noted that Kafka at least stated that these ideas had only slender evidence in their support, and considered them only as tentative suggestions (p. 2).Google Scholar

³ Stein, Stanley J., The Brazilian Cotton Manufacture: Textile Enterprise in an Underdeveloped Area, 1850–1950 (Cambridge: Harvard University Press, 1957), p. 127.CrossRefGoogle Scholar

⁴ Leff, Nathaniel H., The Brazilian Capital Goods Industry, 1929–1964 (Cambridge: Harvard University Press, 1968), pp. 8–15.CrossRefGoogle Scholar

⁵ This figure is from Table II of Huddle's, Don L. Discussion Paper No. 30, “Notes on the Brazilian Industrialization: Sources of Growth and Structural Change” (Yale University Economic Growth Center, 1967).Google Scholar His data, in turn, are from United Nations, The Growth of World Industry, 1938–1961 (New York, 1963).Google Scholar The figure does not appear unduly high if we bear in mind that imports constituted approximately

⁶ percent of GNP in 1938 (Carneiro, “Past Trends,” Table XVA); and as noted in Table 1 of this article, raw materials and food took approximately 58 percent of total imports. This would leave imports amounting to 3.2 percent of GNP for supply of manufactured goods from abroad.

⁷ Gouveia de Bulhões, Octávio, “Agriculture and Economic Development,” in Rostow, W. W., ed., The Economics of Take-Off into Sustained Growth (New York: St. Martin's Press, 1963), p. 230.Google Scholar Concerning the dating of the coffee planters’ political eclipse in Sao Paulo politics, historians have suggested 1919 or the 1920's as a likely turning point. See Luz, Nicia Vilela, A Luta Vela Industrializacao do Brasil (São Paulo: Difusão Européia do Livro, 1961), p. 149–50Google Scholar; and Stein, Brazilian Cotton Manufacture, pp. 125–26, 135.

⁸ These figures are from a detailed analysis of the demographic census, presented in Borges, T. Pompeu Accioly and Loeb, Gustaaf, “Desenvolvimento Econ6mico e Distribuição da População Ativa,” in Contribuições à Analise do Desenvolvimento Econômteo (Rio: Agir, 1957), p. 41.Google Scholar The 1919 figure is, moreover, an underestimate, for the 1920 census did not include in “industry’ the workers in bakeries and sugar mills. See Baer, Wemer, Industrialization and Economic Development in Brazil (Homewood, 111.: Irwin, 1965), p. 251.Google Scholar If, as a rough check on these figures for the industrial labor force, we compare the extent of urbanization in 1919 and 1939, a very similar picture emerges. As early as 1919, 12.8 percent of the total population lived in Brazil's major cities, the state capitals. In 1939 the percentage was 13.8. These figures were computed from data in Baer, Industrialization, pp. 242 and 246.

⁹ This is suggested by international and historical data in Clark's, ColinThe Conditions of Economic Progress (3d ed.; New York: Macmillan, 1957), pp. 526–31.Google Scholar See also Ojala, E. M., Agriculture and Economic Progress (Fairlawn, N.J.: Oxford University Press, 1952), Table LI.Google Scholar

¹⁰ Baer, Industrialization, p. 71.

¹¹ Dean, Warren K., “São Paulo's Industrial Elite, 1890–1960” (unpublished Ph.D. dissertation, University of Florida, 1964), especially pp. 43–74.Google Scholar The Brazilian experience during World War II was similar—only the industries with elastic supply of inputs from domestic sources were able to expand rapidly under the “stimulus” of import limitations. See Leff, Nathaniel H., “Import Constraints and Development: Causes of the Recent Decline of Brazilian Economic Growth,” Review of Economics and Statistics (Nov. 1967).Google Scholar Concerning the pre-World War I origins of Brazilian manufacturing, we should also note that all giant “economic groups” of Brazilian industry—Matarrazzo, Votorontim, Jafet, Klabin, and Lundgren—began their activities well before 1914.

¹² This is computed from data in Martin, Jean-Marie, Processus d'Industrialisation et Développement Energétique du Brésil (Paris: Institut des Hautes Etudes de l'Amerique Latine, 1966), p. 346.Google Scholar

¹³ Computed from data in Stein, Brazilian Cotton Manufacture, p. 192.

¹⁴ Ibid., p. 191.

¹⁵ This is computed from data in Stein, Brazilian Cotton Manufacture, p. 191, and figures on the 1907 Census in George Wythe, “Brazil: Trends in Industrial Development,” p. 70, in Kuznets, Simon et al., eds., Economic Growth: Brazil, India, Japan (Durham: Duke University Press, 1955).Google Scholar

¹⁶ Baer, Industrialization, p. 17.

¹⁷ Ibid., p. 13.

¹⁸ In addition to tariff protection relative prices also moved to favor local producers because of the rise in the real domestic cost of imports. This occurred as the external devaluation of the milreis proceeded more rapidly than the internal inflation. Between 1912 (when a domestic price index for use as a deflator becomes available) and 1939 the milreis/AoMax exchange rate, expressed in constant milreis, appreciated at an annual trend rate of 1.6 percent. Although there were some fluctuations (the f12 of the trend equation was 0.55), the regression coefficient on the trend term was significant at the 0.001 level. Over 27 years, 1.6 percent cumulates to 54 percent.

¹⁹ On Brazilian tariff policy during this period, see Luz, A Luta Vela Industrializacao, especially pp. 18, 36, 49,118, 129; and Stein, Brazilian Cotton Manufacture, pp. 80–85. Effective rates of protection were probably lower than the nominal tariffs, since duties were also levied on many primary-product raw materials. In addition, for some important products there may also nave been “water” in the tariff. The size of the internal market was large enough to permit entry by a large number of producers who could therefore not set their prices on the basis of the world market price plus the tariff. On these points, see my Economic Policy-Making and Development in Brazil, 1947–1964 (New York: Wiley, 1968), pp. 10–11.Google Scholar

²⁰ Luz, A Luta Pela Industrialização, pp. 101–07.

²¹ For monetary and fiscal; policy before 1910, see Calógeras, J. Pandiá, A Politico Monetáia do Brasil ]trans, by Neto, Thomaz Newlands from the. 1910 edition of La Politique Monetaire du Bréail[ (Sao Paulo: Compannia Editôra National, 1960), pp. 103–114. Data on die federal government deficits and on price inflation are presented in -Baer, Industrialization, pp. 287, 289–90, 300–1.Google Scholar

²² For a detailed ‘analysis of the valorization programs, see ntônio Delfim Netto, O Problema do Café no Brasil (São Paulo, 1959).Google Scholar

²³ For a discussion of such factor-market imperfections, which were probably even more severe 80 years ago, see my “Export Stagnation and Autarkic Development in Brazil, 1947–1962,” Quarterly Journal of Economics, XLIX (May 1967), pp. 298–99.Google Scholar

²⁴ That is, consider an economy in which income grows as a function of export growth, and foreign supply of manufactured products is constrained by the rate of export growth. Let Dt denote the demand for manufactured goods; S _t (=M _t), the supply from importation; X _t(-z=M _t), annual export receipts; Y(, national income; and r, the rate of growth of exports, income, and import supply: (1) Y _t = Y _oe _rt (2) D_t = D_oe ^λrt, (3) X_t = X _oe ^rt; and (4) S _t = M _oe ^rt. We assume λ > 1, reflecting an income-elasticity of demand for manufactured products greater than unity. D ( o ), the size of the market for manufactured products, is equal in approximate terms to aggregate demand minus the demand for food and services. At low per-capita income levels, this may be some 25 percent of GNP. (See Kuznets, Simon, Modern Economic Growth ]New Haven: Yale University Press, 1966[, pp. 234–41 and 265–69).Google Scholar If this is greater than the share of imports in GNP, a market for domestically manufactured products exists initially and will grow more than proportionately with the increase in income resulting from expanding exports. As noted below, this seems to have been the case in Brazil, where imports constituted a small proportion in national product. In other cases of a relatively large foreign-trade sector, imports may initially exceed the share of manufactured products in final demand. However, as income grows because of expanding exports, with an income elasticity of demand for manufactured products greater than unity, D(t) will eventually exceed M(t). Income growth based on expanding exports will create a greater market for domestic industrial producers, of course, with a foreign trade multiplier greater than unity ( r ^vr ^∞). Similarly, if the rate of growth of imports is less than the rate of growth of exports (r ^∞r _m) because of growing foreign-capital payments. Growth of industrial production for the internal market will, of course, create a source of income growth and demand which does not generate its own import supply, and should therefore become a continuing process as long as necessary input conditions are satisfied.

²⁵ Albert Hirschman has also suggested, in somewhat different terms, the complementarity on the demand side between exports and domestic industrial expansion. See his “The Political Economy of Import-Substituting Industrialization in Latin America,” Quarterly Journal of Economics, LXXXII (Feb. 1968), p. 4.Google Scholar See also Munoz, Oscar E., “An Essay on the Process of Industrialization in Chile Since 1914,” Yale Economic Essays VIII (Fall 1968), p. 47.Google Scholar For a general theoretical statement, see Caves, Richard E., “Vent-for-Surplus Models of Trade and Growth” in Trade, Growth, and The Balance of Payments (Chicago: Rand & McNally, 1965).Google Scholar

²⁶ In the Brazilian case there is also indirect evidence of a high domestic impact of foreign trade. In the first decade of the century, the export producers applied pressure on the government to intervene in the foreign-exchange market and lower the dollar/mtZrets exchange rate. (On this undervaluation, see my Economic Policy-Making and Development in Brazil, pp. 11, 23, 129.) This experience suggests that despite all the talk about their high propensity to import, the coffee planters were in fact interested primarily in milreis earnings and local expenditure. For a similar case, in Argentina, of exchange-rate undervaluation under pressure of the exporters, see Williams, John H., Argentine International Trade Under Inconvertible Paper Money: 1880–1900 (Cambridge: Harvard University Press, 1920), p. 160.Google Scholar

²⁷ During this period the stock of currency functioned as “high-powered” money, whose variations had disproportionate effects on the private sector's liquidity. A regression of DOMf against CSt showed an elasticity of 1.98 for the period 1924–1939.

²⁸ Concerning the influence of this doctrine on Brazilian economic policy-making and its pernicious effects on the country's development in the post-World War II period, see my Economic Policy-Making and Development in Brazil, pp. 19, 86–87, 139–143. Brazil is, of course, not the only less-developed country to have followed this approach. See, for example, Myint, Hla, “The Inward and Outward Looking Countries of Southeast Asia,” Malayan Economic Review XII (Apr. 1967).Google Scholar

²⁹ Computation of the ratio of Brazilian exports to GNP raises the question of the appropriate exchange rate for conversion purposes. Using the observed exchange rate gives an average export ratio of 15.5 percent for 1920–21. There is evidence, however (Carneiro, “Past Trends,” Table XXVI), that in terms of purchasing power parity, the milreis was grossly undervalued during this period. If we use the PPP rate to obtain a valuation of Brazilian GNP at world market prices, the average export ratio for 1920–21 is 8.8 percent. The figures relate to the share of exports in the monetized sector of the economy alone. Hence the share of exports in the economy as a whole was even lower. The economy's import ratio was usually lower than the export ratio before 1939. As noted below, Brazil maintained a balance-of-trade surplus, reflecting foreign capital payments and immigrants’ remittances.

³⁰ In its relatively low export coefficient Brazil seems to resemble other countries of large population and area—e.g., the United States, Russia, China, and India. See Kindleberger, Charles P., Foreign Trade and The National Economy (New Haven: Yale University Press, 1962), pp. 32, ff.Google Scholar

³¹ For the share of exports in Brazilian GNP to have been larger in earlier years than the ratio observed in 1920–21 would require that per capita income had been growing at an annual rate larger than the 1.3 percent observed for per capita exports. This is not likely. Per capita income in 1920–21 has been estimated at $163 (1950 prices). If per capita income had been growing at even the same rate as exports during the preceding 100 years, this would imply that per capita income in 1821 had been $44. In fact, estimates which are discussed in my 1969 paper “Economic Development in Nineteenth-Century Brazil” (mimeographed) suggest 65, 45, and 79 dollars, respectively, as likely, lower-bound and upper-bound figures for Brazilian per capita income in 1821. These computations have used the PPP exchange rate for obtaining a dollar figure for Brazilian income in 1920–21. Using the observed exchange rate gives a figure of $91 for those years, which would make it even less likely that per capita income had been growing at a rate higher than exports during the preceding century.

³² Brazil's dependence on foreign trade as a source of final demand was also lessened because from 1889, when a republic was established in Brazil, to 1939, the economy was usually experiencing inflation. Hence a model of deficient aggregate demand is not relevant. The Federal government usually ran a deficit, with expenditures exceeding revenues by more than 10 percent in 33 of these 50 years. The costof- living index, available from 1912 to 1939, shows prices rising by 5 percent or more in 16 of these 28 years. As noted below through all of this period the exchange rate depreciated almost constantly.

³³ For a study of such monetary effects on a dependent economy, see Wallich, Henry C., Monetary Problems of an Export Economy: The Cuban Experience, 1914–1947 (Cambridge: Harvard University Press, 1950).CrossRefGoogle Scholar

³⁴ For a similar situation in Argentina, 1909–1934, see Halperin, Ricardo, “The Behavior of the Argentine Monetary Sector: An Econometric Study” (unpublished Ph.D. dissertation, Columbia University, 1968), pp. 240–52.Google Scholar

³⁵ For a similar situation in Argentina during the period 1909–1934, where the Banco de la Nacidn's policy was to “increase its rediscounts to commercial banks in periods of gold outflows,” see Halperin, Argentine Monetary Sector, pp. 248 ff.

³⁶ Between 1920 and 1924 the Federal deficit average 2.9 percent of GNP, and from 1930 to 1939, 2.9 percent. These figures were computed from Carneiro's data on GNP and data in Onody, Oliver, A Infiagao Brasileira, 1820–1958 (Rio, 1960), p. 199, on the Federal deficit.Google Scholar

³⁷ Such an antonomous monetary policy was nothing new for Brazil. In 1857, world demand for Brazil's exports fell sharply, the country's export receipts declined 40 percent, and virtually all specie was shipped out of t i e country. The Banco do Brasil lowered its discount rate to protect the level of internal activity. See J. Pandiá Calógeras, A Politico Monetdria do Brasil, pp. 103–14.

³⁸ For example, between 1900 and 1905 the government embarked on a stabilization program, balancing the Federal budget and reducing the currency supply in order to deal with internal inflation. This occurred despite a favorable foreign trade situation. As noted below, a stabilization program was implemented in similar circumstances in 1925–1929.

³⁹ The onset of the 1929 Depression did hit Brazil hard, as GNP fell 22 percent between 1929 and 1932. The effects of the world crisis were aggravated in this case, however, because of special internal conditions. Brazil's president from 1925 to 1930, Washington Luís, was a man of principle, and he happened to believe in monetary and fiscal orthodoxy. In 1925 the government launched a stabilization program, reducing the currency supply 14 percent over 2 years; and in 1927, balancing the Federal budget for the first time in two decades. In 1928 and 1929 the cost-of-Hving index actually fell, for the first time since 1918. The world crisis struck Brazil in the midst of this internally motivated deflation. In contrast with the experience in the 1921 world recession, when Brazilian prices had continued to rise, excess demand had been squeezed out of the economy and internal prices now fell even more sharply. Quick policy adjustment was also impossible because of domestic political conditions. The President, whose term of office was to continue through 1930, stubbornly refused to change his monetary policy. In addition, he so mismanaged coffee export policy that world coffee prices fell 45 percent between September 1929 and December 1930, a period in which the U.S. wholesale price index fell only 8 percent (See Antônio Delfim Netto, O Problema do Café no Brasil (São Paulo: Faculdade de Ciêicias Econômicas e Administrativas, Universidade de São Paulo, 1959), pp. 129–136.) Effective policy response was also inhibited because the country was in the midst of the deep internal political crisis that culminated in the Vargas revolution of October 1930. Vargas’ political control, in turn, was too tenuous for purposeful action in economic policy until after his repression of the 1932 São Paulo rebellion. Beginning thereafter, the government initiated large expenditure programs, and Brazilian GNP grew at a cumulative rate of 3.6 percent through the 1930 s despite conditions in the rest of the world economy; though, as noted earlier, industrial output grew at a rate lower than in the 1920's because of the less income-elastic supply of imports.Google Scholar

⁴⁰ In their analysis of the sources of economic growth in Chile, Arnold C. Harberger and Marcello Selowsky introduced human capital inputs and found that they accounted for a large percentage of the increase in aggregate output during the period 1940–1962. See their paper, “Key Factors in the Economic Development of Chile,” Cornell Conference on the Next Decade of Latin American Economic Development, 1966, especially pp. 24–40. I was unable to obtain extensive data on factor shares and on the marginal productivity of educational capital in Brazil that would be necessary to follow their methodology.

⁴¹ See my Economic Policy-Making and Development in Brazil, p. 90, fnn. 4 and 5.

⁴² Computed from data on the value of imports in constant dollars, available in Carneiro, “Past Trends,” Table XVB. Joseph Grunwald has also stressed the rapid rise in imports from World War II to 1960, and the fact that imports grew more rapidly than export receipts. See his “Invisible Hands in Inflation and Growth,” in Baer, Werner and Kerstenetzky, Isaac, eds., Inflation and Growth in Latin America (Homewood, 111.: Irwin, 1964), pp. 297–98.Google Scholar

⁴³ The foreign-owned railways and shipping companies were taken over (with compensation) by the Brazilian government for straightforward nationalist reasons. In retrospect, however, nationalization of these assets and their claims on the country's import capacity for dividend remissions proved to have a sound economic rationale, for the real domestic cost of foreign exchange rose through the postwar period. Strict capital controls on foreign dividend remissions might have been another approach to the problem. But as Judith Tendler points out in her study of the major remaining foreign-owned utility (Electric Power in Brazil (Cambridge: Harvard University Press, 1968), the government felt obliged to make various accommodations with the company (for example, on the rate at which it acquired foreign exchange for profit remissions), which maintained their dollar profits.Google Scholar

⁴⁴ It is hard to relate the higher growth rate of the postwar period to any inherentacceleration of the “growth process.” The increase in agricultural production and in human capital formation may have been irreversible trends, but the relatively favorable import supply conditions were not. By the early 1960's, the reflow of the capital imported in the earlier period had begun, so that coming together with poor export performance, Brazil encountered import constraints and a slackening of growth.

⁴⁵ Data in Martin, Industrialisation et De”veloppement, p. 347, indicate that between 1901 and 1910 total energy consumption (including coal, petroleum products, firewood, and electricity) increased at an annual cumulative rate of 2.8 percent, and between 1910 and 1920, at 2.4 percent. To gain an idea of the income-elasticity of total energy consumption for a country at low income levels, a double-log equation was estimated with observations for 32 countries having a per-capita income below 300 dollars. Data for this regression were from Ginsburg, Norton S., Atlas of Economic Development (Chicago: University of Chicago Press, 1961), pp. 18, 80. The results, significant at the 1 percent level, indicated an income elasticity of 0.32. Applying this coefficient to the Brazilian data would imply that Brazilian national income grew at annual rates of 8.4 percent in 1901–1910, and 7.2 percent in 1910–1920. These figures may be on the high side, but even if we reduce the annual growth rates in this period to 5 percent, the long-term growth is impressive. I am grateful to Laura Randall for bringing the Ginsburg source to my attention.Google Scholar

⁴⁶ The long-term rise in per capita income that remains after taking population growth into account cannot, of course, be regarded as identical with a similar increase in utility as perceived subjectively. The difficulties of comparing intertemporal utility are well known, especially over long periods in which new consumer goods were being introduced and new information and expectations concerning living standards diffused.

⁴⁷ Robock, Stefan H., Brazil's Developing Northeast (2d ed.; Washington: The Brookings Institution, 1965), pp. 34–36.Google Scholar

⁴⁸ By contrast, in the United States, between 1820–22 and 1855–59, exports grew at an annual cumulative rate of 4.8 percent. This was computed from data in North, Douglass C., The Economic Growth of the United States, 1790 to 1860 (Englewood Cliffs: Prentice-Hall, 1966), p. 284.Google Scholar In Argentina rapid growth led by exports occurred in the period between 1870–74 and 1895–99. The value of export receipts grew at an annual cumulative rate of 4.9 percent. This was computed from data in Scobie, James, Revolution on the Pampas (Austin: University of Texas Press, 1964), p. 174.Google Scholar Similarly, the value of wool exports from Australia grew at an annual cumulative rate of 5.2 percent between 1861 and 1890. This was computed from data in the Cambridge Economic History of Europe, Vol. VI The industrial Revolutions and After (Cambridge; 1965, p. 155.)Google Scholar

⁴⁹ The causes of Brazil's poor export experience in sugar and cotton during the nineteenth century are considered in my study, Economic Development in Nineteenth Century Brazil.

⁵⁰ Brazil may not be the only less-developed country to which this applies. For an example of revisionist economic history in Argentina, see Diaz-Alejandro, Carlos F., “The Argentine Tariff, 1906–1940,” Oxford Economic Papers, XIX (Mar. 1967),Google Scholar and Petrecolla, Albert O., “Prices, Import Substitution and Investment in the Argentine Textile Industry, 1920–1939,” (Ph.D. dissertation, Columbia University, 1968).Google Scholar The industrial sector in Argentina emerged before the Depression or World War I, and by 1900–1904, accounted for 14 percent of Argentine GDP. Aggregate industrial output grew at the highest rate in the century during the decade before World War I and slackened during the war. And there, too, industrial development proceeded together with favorable foreign trade conditions and with strong government support in the form of tariff protection. Similarly, on industrialization in Chile, see Mufioz, Oscar E., “An Essay on the Process of Industrialization in Chile Since 1914,” Yale Economic Essays, VIII (Fall 1968).Google Scholar These instances relate to revision of “nationalist” views on economic history. For an example of recent research correcting some “colonialist” economic historiography, see Green, R. H. and Hymer, Steven, “Cocoa in The Gold Coast,” THE JOURNAL OF ECONOMIC HISTORY, XXVI (Sept. 1966), 299–306.CrossRefGoogle Scholar

⁵¹ For example, Simonsen's, Roberto C.História EconÔmica do Brasil (1500–1820) (4th ed.; Sao Paulo: Compannia EditÔra Nacional, 1962), which has provided much of the empirical material for subsequent interpretation of Brazilian economic history, stopped at 1820.Google Scholar

⁵² Cf. a somewhat extreme comment by Daniel C6sio Villegas in a paper entitled “History and the Social Sciences in Latin America.” “The fact which I find most striking is that the Latin American by far prefers inventing history to studying it …” He goes on to cite with approval another writer's statement that “In Latin America, in addition to its ‘scholarly’ function, history serves a great variety of purposes, all of which, or almost all, are respectable: patriotic purposes, partisan purposes, literary purposes.” See 131 ff. in Manuel Diégues, Jr., and Bryce Wood, eds., Social Science in Latin America (New York: Columbia University Press, 1967).

⁵³ Albert Hirschman has referred to a related phenomenon, “amnesia” within the policy-making process concerning earlier efforts to deal with chronic policy problems and the consequent inability to benefit from past experience. See his Journeys Toward Progress (New York: Twentieth Century Fund, 1963), p. 320.Google Scholar