Business Enterprise and the Great Depression in Brazil: A Study of Profits and Losses in Textile Manufacturing

Transcription

1 Business Enterprise and the Great Depression in Brazil: A Study of Profits and Losses in Textile Manufacturing Stephen H. Haber The Business History Review, Vol. 66, No. 2. (Summer, 1992), pp Stable URL: The Business History Review is currently published by The President and Fellows of Harvard College. Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org. Tue Nov 6 16:38:

2 Stephen H. Haber Business Enterprise and the Great Depression in Brazil: A Study of Profits and Losses in Textile Manufacturing This article employs previously unused accounting data and manuscript censuses to determine the impact of the Great Depression on Brazil's most important cotton textile manufacturers. It argues that the Great Depression, when viewed at the level of the individual business enterprise, had far more serious consequences than the previous literature, which relied on aggregate statistical data, suggests. The analysis presented here leads to the conclusion that Brazil's major cotton firms were in serious trouble prior to the 1929 Crash and that they took longer to recover than most other studies of Brazilian industrialization have indicated. During the 1980s, economic historians of Latin America have turned their attention increasingly to the impact of the Great Depression. The numerous studies of the larger economies of the region indicate that the course of the Great Depression in Latin America was significantly different from that in the United States and Western Europe. In this view, the Depression began quite a bit earlier in Latin America than in the advanced industrial econ- STEPHEN H. HABER is associate professor of history at Stanford University. Research for this article was carried out with the support of grants kom the Latin American and Caribbean Program of the Social Science Research Council and the Stanford University Institute for International Studies and Center for Latin American Studies. Earlier versions were presented at the Conference of the All-University of California Group in Economic History and at the Stanford Workshop in Social Science History. I would like to thank the participants in those seminars for their helpful suggestions. I owe special acknowledgment to Herbert S. Klein, Kenneth L. Sokoloff, Steven Topik, and two anonymous referees of the Business History Reoiew, who made detailed and extensive comments on an earlier draft. Research assistance was provided by Catherine Barrera and Vera Guilhon Costa. Business History Reoiew 66 (Summer 1992): Q 1992 by The President and Fellows of Harvard College.

3 Stephen H.Haber / 336 omies, but it was a good deal less severe. Moreover, the recovery from the Depression was rapid: the larger economies of the region were on the rebound by 1933 and then grew at impressive rates throughout the rest of the decade. Finally, the literature suggests that during the period of recovery the engine of growth shifted from primary products to manufacturing: import-substituting industry outgrew the traditional export sector.' With few exceptions, these earlier studies have been based almost entirely on analyses of government revenues, foreign trade data, and aggregate indexes of manufacturing production derived from industrial censuses or published taxation records. We therefore know a good deal about the movement of aggregate economic indicators. Economies are not made up of statistical indexes, however, but rather of workers earning wages and capitalists earning profits. About the actual state of business enterprise during the Great Depression we know very little. The use of aggregate data has three significant limitations. First, the economic statistics published by Latin American governments during the first part of the twentieth century are not particularly reliable or consistent in their coverage. The data on manufacturing are especially suspect, because the information was originally gathered for taxation purposes, which would not have encouraged bias-free responses from factory owners. In fact, in Brazil the first industrial census (1907) was carried out by the national manufacturers' association, because it was believed that industrialists would not respond to a survey conducted by the federal government. Second, there is not always a direct correlation between the level of output and the financial performance of an industry. These variables can, and often do, move in opposite directions. Third, sole reliance on aggregate indicators may mask substantial variations among different types of industries and firms. The purpose of this article, therefore, is to shed some light on the effects of the Depression on manufacturing enterprise in Latin America by going beyond the aggregate statistics to look at the experience of individual enterprises. It concentrates on Brazil, which, in the period after the First World War, had the largest For an excellent summary of the literature, see Latin America in the 1930s: The Role of the Periphery in World Crisis, ed. Rosemary Thorp (New York, 1984).

4 Business Enterprise and the Great Depression in Brazil 1337 and most industrialized economy in the regi~n.~ Within the manufacturing sector, this study focuses on the most significant and best-developed industry, cotton textiles. No other industry approached it in size, regional diversity, or backward linkages. At the time of Brazil's 1920 industrial census, cotton textiles accounted for 24.4 percent of manufacturing value added, second only to the food-processing industries, which accounted for 32.9 per~ent.~most food processing, however, consisted of preparing sugar and coffee for export and thus did not represent the same type of import-substituting manufacturing that textiles did. If these export-related processing industries are factored out, textiles accounted for just over one-third of Brazilian manufacturing. The basic argument advanced here is that the Depression hit Brazilian manufacturing a good deal harder than the aggregate data indicate. Indeed, all three limitations of aggregate data operate in the Brazilian case to give a misleading picture of the Depression's effects on business enterprise. First, the coverage of the aggregate indicators is very irregular. The partial industrial census of 1907 was followed by a more complete census in 1920, but no other sector-wide manufacturing census was carried out until 1940, making a study of the Depression based on census data highly problematic. Researchers have relied, therefore, almost entirely on a production index derived from excise tax records.4 This series (as I shall discuss later) underestimates the impact of the Depression on the industry and overestimates the extent of the recovery. Second, firm profitability did not always move in the same direction as output. Third, there was a substantial variation among firms that is masked by the aggregate series. Indeed, some firms were pushed to the wall and underwent financial restructuring during the Depression, whereas others were turning phenomenal profits. These differences are largely attributable to the different entrepreneurial strategies followed by Brazil's mill owners before the Depression. For a discussion of the Mexican case during the Great Depression, see Stephen Haber, Industry and Underdevelopment: The Industrialization of Mexico, (Stanford, Calif., 1989), chaps. 9 and 10. Albert Fishlow, "Origins and Consequences of Import Substitution in Brazil," in International Economics and Development: Essays in Honor of Raul Prebisch, ed. Luis Eugenio Di Marco (New York, 1972), 323. The excise tax on industrial production dates back to the early 1890s. This tax was paid by the producers, who were then obligated to affix stamps to their products demonstrating that the correct tax had been paid. The federal government began to publish the receipts from the excise tax, disaggregated by product, in 1912.

5 Stephen H. Haber / 338 This article also argues that the Great Depression had important long-run consequences for the Brazilian textile industry. The analysis of the behavior of individual firms lends support to the argument made by Stanley Stein and Albert Fishlow that the Depression discouraged new investment and innovation in Brazil's consumer goods industries and that it ultimately worked to slow their rate of growth well after the Depression had ended.5 The article is organized into three sections. The first discusses the nature of the sample on which the analysis presented here is based and explains the methods used in adjusting and interpreting the data. The second examines how the sample firms were affected by the Great Depression, focusing on three key variables: profits, investment, and investor confidence. The third section discusses the broader implications of this analysis for the history of Brazilian industrial development and suggests how the results presented might cause us to reexamine recent research on the early industrial experiences of other Latin American countries. Three varieties of evidence are employed. The first consists of the balance sheets, profit-and-loss statements, and associated data contained in the annual reports of eight of the country's largest textile ~ompanies.~ These data cover the period 1925 to 1937 and were reported on a semestral basis-january to June and July to December. The second consists of data gathered from the Rio de Janeiro Stock Exchange on dividend payments and share prices for these eight firms, which were published annually in the Retrospecto Commercial do Jornal do Commercio. The third comprises data gathered from the 1927 and 1934 censuses conducted by the Cotton Textile Manufacturers Association of Rio de Janeiro Stanley Stein, The Brazilian Cotton Manufacturers: Textile Enterprise in an Underdeveloped Area, 185&1950 (Cambridge, Mass., 1957), 187; Fishlow, "Origins and Consequences," 339. Some of these annual reports were located in the Biblioteca Naciond in Rio de Janeiro, filed under the periodicals section. These sets of reports, however, were incomplete and provided extremely irregular and inconsistent coverage. Fortunately, Brazilian law required that publicly held companies publish their financial statements in public documents, so I was able to locate many of the missing reports in the Sociedades An6nimas section of the Diario OfJicial. The disadvantage of this approach was that the Diariu OfJicial did not publish a particular company's reports on the same date (or even in the same month) every year, and the Diarios are not indexed. It was therefore necessary to review them page by page for the months of January, February, March, April, July, and August (reports were almost never reprinted in other months) for each year. Researchers wanting to repeat this operation for other companies should be aware that this is not a costless operation, as it involves covering roughly 12,000 pages of Diarios for each year examined.

6 Business Enterprise and the Great Depression in Brazil (Centro Industrial de Fia~iio e Tecelagem de Alg~diio-CIlTA).~ These censuses, which have been largely ignored in other research on the Brazilian textile industry, provide detailed information on the state of the industry enumerated at the firm level. Given the nature of the evidence and the audience for which it was compiled, it can tell us relatively little about the effects of the Depression on wages and employment within firms. On the entrepreneurial side, however, it can tell us a great deal about corporate profitability, the earnings of investors, and the behavior of financial markets. The Data The firms under study are the Companhia Fiasiio e Tecidos Cometa, the Companhia Fabrica de Tecidos Siio Pedro de Alcantara, the Companhia Petropolitana, the Companhia Progreso Industrial do Brasil, the Companhia de Fiaqiio e Tecidos Allianp, the Companhia de Fia~iio e Tecidos Confianqa Industrial, the Companhia America Fabril, and the Companhia de Fia~iio e Tecelagem Industrial Mineira. The first three of these firms were located in the state of Rio de Janeiro, the next four in the Distrito Federal, and the last in the state of Minas Gerai~.~ These eight firms were among the largest cotton textile enterprises in the country. At the time of the 1927 census of the cotton textile industry, when there were some 306 cotton manufacturing companies in operation running 354 mills, the eight firms examined here accounted for 21.7 percent of the machinery in service (measured in active spindles), 12.9 percent of the work force, and 11.4 percent of national production. They ranged in size from the single-factory Companhia de FiaqBo e Tecelagem Industrial Mineira (with a paid-in capital of 1.2 million milreis, a work force of 920, and output of 4.5 million meters of cloth-a 0.7 percent 'See Centro Industrial de Fiagio e Tecelagem de Algodio, Estati'sticas da Industria, Comercio e Laooura do Algodcio Relatioos ao Anno de 1927 (Rio de Janeiro, 1928) [hereafter cited as CIFTA, Also see Centro Industrial de FiagBo e Tecelagem de AlgodZo do Rio de Janeiro, Fiaqrio e Tecelagern: Censo Organimdo pelo Centro Industrial de Fiaqrio e Tecelagem de Algodcio do Rio delaneiro e Reoisto pela Seccao de Collecta do Departamento de Estatistica e Publicidade (Rio de Janeiro, 1935) [hereafter cited as CIFTA, The Distrito Federal is Brazil's Federal District, equivalent to the District of Columbia in the United States. During the period under study, it comprised the city of Rio de Janeiro and its immediate environs.

7 Stephen H. Haber / 340 market share) to the giant, six-factory Companhia America Fabril (with a paid-in capital of 32 million milreis, a work force of 7,000, and output of nearly 28.1 million meters of cloth-a 4.0 percent market share).g The average size of the sample firms was 70,587 spindles, 1,589 looms, 2,075 workers, and annual production of 9.9 million meters of cloth. These figures were four to eight times larger (depending on the unit of measure) than the average for the industry as a whole, which was 8,444 spindles, 240 looms, 420 workers, and annual production of 2.3 million meters of cloth.10 None of the firms, however, was large enough to have raised prices over their competitive level. These firms were also among Brazil's oldest. All eight were already major enterprises at the time of the 1907 industrial census, accounting at that time for 23 percent of national textile production and 19 percent of the work force." Of the four largest firms in 1907, two-the Companhia Progreso Industrial do Brasil and the Companhia Confianqa Industrial-are included in the sample. Two of the other firms in the sample-the Companhia Petropolitana and the Companhia Allianqa-were among the oldest textile producers in the country, both dating from the early 1880s, when the nation had forty-six producers of cotton cloth, of which only sixteen appear to have been modern factories. l2 The eight firms under study here were also significantly more capital-intensive than the average Brazilian textile producer. Staffing levels indicate higher capital-intensity: the ratio of spindles to workers was 33:l in the sample mills, 20:l for the industry as a whole. The ratio of looms per worker tells much the same story:.77:1 for the sample firms,.57:1 for the industry as a whole.13 Interestingly, the data do not indicate higher productivity levels in the sample firms. Measures of physical productivity all give lower rates of output per machine and per worker for the sample firms than the average for the industry (6,256 milreis per worker and 185 milreis per spindle in the sample firms, versus 7,577 milreis During the period under study, there were approximately six milreis to the U.S. dollar. lo Calculated from data in CIFTA, " Calculated from Centro Industrial do Brasil, 0 Brazil: Suas riquezas naturaes, suas industrias, vol. 3: Industria de transportes, industria fabril (Rio de Janeiro, 1909), section on textiles. l2 Biblioteca de Associa@o Industrial, Archioo da expos&& da industria nacional de 1881 (Rio de Janeiro, 1882); Stein, Brazilian Cotton Manufacture, 191. l3 Calculated from CIFTA, 1928.

8 Business Enterprise and the Great Depression in Brazil / 341 per worker and 269 milreis per spindle for the modal firm). This suggests that the capital goods of these eight firms were probably more antiquated (explaining the lower productivity levels) than was the norm in the Brazilian textile industry and that the sample firms compensated in part for their older plant and equipment by greater than average work intensity, as indicated by the lower staffing levels per machine in the sample companies. Because these mills were among the country's oldest, they probably utilized capital goods of an older vintage than was generally the case for Brazilian textile manufacturing. In short, the firms under study here do not constitute a random sample of Brazilian cotton goods manufacturers; they were chosen because it was possible to retrieve relatively complete sets of their financial statements. Rather, they should be viewed as a random sample of the older, larger, publicly held, more capitalintensive textile firms. Such samples are not without their biases, but they also have some merits. Indeed, much of what we know about the nineteenth-century U.S. cotton textile industry is based on the study of small samples of large, Massachusetts-type mills remarkably similar to the firms under study here. l4 These types of samples allow researchers to analyze a large percentage of the activity in an industry by looking at a relatively small number of firms. How the bias in the sample affects the rate-of-return estimates developed in this article is difficult to determine. The productivity disadvantage of the sample firms as a group would tend to bias the rate-of-return results downward. On the other hand, the fact that these firms were operating with older, less costly capital goods would tend to push their rates of return upward, as the denominator in the rate-of-return equations would be smaller than if the firms had recently purchased plant and machinery. I developed estimates of earnings, net profits, capital stock book value, capital stock market value, and dividend payments for each of the firms under study. From these it was possible to cal- l4 See, for example, Paul F. McGouldrick, New England Textiles in the Nineteenth Century: Profits and Investment (Cambridge, Mass., 1968); Robert G. Layer, Earnings of Cotton Mill Operatives, (Cambridge, Mass., 1955). Also see the following articles by Lance E. Davis: "Sources of Industrial Finance: The American Textile Industry, A Case Study," Explorations in Entrepreneurial History 9 (April 1957): ; "Stock Ownership in the Early New England Textile Industry," Business History Review 32 (Summer 1958): ; "The New England Textile Mills and the Capital Markets: A Study of Industrial Borrowing, ,"]ournal of Economic History 20 (March 1960): 130.

9 Stephen H.Haber culate accounting rates of return on capital stock (net profits divided by capital stock), market rates of return (dividend payments divided by the market value of capital stock), and the ratio of market to book values of capital stock. The purpose of this study is not to calculate the exact rate of profit of any individual firm in any given year. For both theoretical and empirical reasons, such a task would, in fact, be impossible; profits are not things, but abstract concepts whose value may be calculated in any number of ways. To this conceptual issue must be added the practical problems of working with annual report data. The job of the accountants who prepare corporate financial statements is to hide losses from the stockholders during hard times and profits from the tax collectors during financially healthier periods. Changes in the method of valuation of inventories, the decision to write down a bad debt, or an acceleration in the depreciation of the company's physical plant can produce apparent losses in good years and apparent profits in bad years. To these issues must be added changes in how firms have dealt with accounting principles historically. Definitions of profits and valuation methods for physical assets have evolved over time, responding for the most part to changes in the tax code. To compensate for these conceptual and practical problems, I have developed three sets of rate-of-return estimates based on three different conceptions of what profits are and who receives them. The first is a set of estimates of accounting rates of return. This concept of profitability values capital stock (the denominator in the rate-of-return equation) at acquisition cost. It is backwardlooking, comparing current profits against the purchase price of capital, regardless of when the capital was purchased. I have adjusted these accounting rate-of-return estimates to factor out ancillary investments and the grossest of firms' accounting manipulations. This is not as daunting a task as might be assumed. Although firms can manipulate their accounts in any given year to conceal their true financial state, over time these manipulations must balance out. Firms that are consistently losing, or earning, money cannot conceal the fact over the medium term; the losses, or the profits, must eventually show up. Moreover, when a company's accounts are examined over time, the most significant accounting manipulations become apparent, making it possible to rework the accounts to correct for them. This set of estimates is presented in Table 1.

10 Business Enterprise and the Great Depression in Brazil Table 1 Estimated Rate of Return on Capital Stock, Eight Sample Companies, (percent) Semester/ Conf. S. P. Prog. Ind. Amer. Year Indl. Allian~a Alc. Petro Indl. Min. Fab. Cometa Median Average I' J J J / U / ' I' ' J Source: Relatorio da Diretoria, , for the following firms: Companhia de Fiaqio e Tecidos Confianqa Industrial; Companhia de Fiaqio e Tecidos Alliansa; Companhia Fabrica de Tecidos Sso Pedro de Alcantara; Companhia Petr~~olitana; Companhia Progreso Industrial do Brasil; Companhia de Fiaqgo e Tecelagem Industria Mineira; Companhia America Fabril; Companhia Fiaqio e Tecidos Cometa. These estimates do not, however, deal with an important peculiarity of Brazilian accounting conventions: firms did not depreciate their physical plants, except for animals and vehicles. I therefore developed a second set of book value rate-of-return estimates, which employs an accelerated depreciation schedule to adjust for this circumstance. Conceptually, this set of estimates is also backward-looking, but it differs from the first estimates in that

11 Stephen H.Haber / 344 Table 2 Estimated Rate of Return on Depreciated Capital Stock, Eight Sample Companies, " (percent) Semesterl Conf S. P. Prog. Ind. Amer. Year Indl. Allian~a Alc. Petro Indl. Min. Fab. Cometa Median Average I' I' J J J J J ' U a a J J 'Capital stock depreciated at 5.0 percent per semester for machinery and other equipment, 2.5 percent per semester for buildings and other improvements. Source: Relatorio da Diretoria, ; see Table 1 for full names of companies. it values capital stock at acquisition cost minus depreciation and profits as revenues minus expenditures minus depreciation (see Table 2). Both of these rate-of-return estimates view profits as something received by firms and compare current revenues against prior expenditures. From the point of view of shareholders, however, the value of a firm's assets is not the cost of acquiring them (the value displayed on the company's books), but their value on

12 Business Enterprise and the Great Depression in Brazil the market (the market value of their shares multiplied by the number of shares). Similarly, they conceive of profits not as the firm's income minus expenditures, but as the stream of dividends paid to stockholders by the firm. I therefore developed a set of estimates of market rates of return (dividends divided by the market value of capital), which are presented in Table 3. This way of measuring profitability is forward-looking, because a firm's market value is a statement about its ability to produce income for a defined period (the life of the capital) into the future. l5 Market rates of return provide information not only about the stream of dividends earned by investors, but also about the rate of return expected by the financial community. A company's market rate of return reflects not only the profitability of that particular company, but also the profit opportunities available in other firms or in other sectors of the economy. The stock of firms whose earnings per share tend to be below average are generally traded below their par values, and the stock of firms that pay higher than average dividends per share are usually traded at prices above their par values. Market rates of return from company to company therefore tend to equilibrate over time, as arbitrage drives down the market price of low-dividend stocks and drives up the market price of high-dividend stocks. Naturally, there will still be some differences in actual returns to compensate for differences in risk and for information asymmetries. The overall result, however, is that market rates of return are less volatile than accounting rates of return. Declines in stock prices (the denominator in the equation) tend to compensate for declines in dividend earnings (the numerator in the equation), thereby producing fairly constant l5 There is a third way to conceive of profits: the rate of return on capital's replacement cost. This concept of capital focuses on the present. It conceives of capital stock as the value of the inputs necessary to reproduce it at current factor prices. Lacking detailed data about the exact composition and vintage of the machinery and buildings for the firms in question, and lacking data on the value of used machinery, I was unable to estimate the value of the capital stock of the firms under study here at replacement cost. What effect valuing capital this way would have on the rate-of-return figures is hard to know, though it is likely that this method of valuing capital would have produced higher capital stock figures and therefore lower rates of return. Since this article advances an argument for a stronger effect of the Great Depression than is generally acknowledged, this method of capital valuation would most probably work in favor of the interpretation advanced here. For a detailed discussion of the different methods of valuing capital stock, see Robert E. Gallman, "The United States Capital Stock in the Nineteenth Century," in Long-Term Factors in American Economic Growth, ed. Stanley L. Engerman and Robert E. Gallman (Chicago, Ill., 1986),

13 Stephen H. Haber / 346 Table 3 Estimated Market Rates of Return, Seven Sample Companies, " (percent) Conf Prog. Amer. Year Indl. Petro Indl. Fab. cometa Median Average a Companhia Industrial Mineira is not included because there were not enough observations of its stock price to make a meaningful estimate of its capital stock market value. Note, however, that from 1930 to 1937 its rate of return would have been zero regardless of its market value because there were no dividends paid (see Table 4). Source: Dividend data from Relatorio da Diretoria, ; market share price data from 0 Retrospect0 Commercial do Jornal do Commercio, See Table 1for full names of companies. results. It is in this sense that market rates of return are often described as being "sticky." There are two ways to separate out the effects of movements in share prices from changes in the dividend stream in determining a firm's market rate of return. The first is to report dividend payments as a percentage of a stock's par value, which does not change over time. This series is presented in Table 4. The second method is to compare capital stock market values to book values. This ratio measures the degree to which the investment community under- or overvalues the capital stock of a company; it is an indicator of the confidence that investors have in the value of their assets. Theoretically, when the ratio is higher than one, firms will issue new shares and make new expenditures on capital stock, since they will receive greater income from the issue

14 Business Enterprise and the Great Depression in Brazil / 347 Table 4 Dividends Paid as a Percentage of Paid-In Capital, Eight Sample Companies, Semester/ Conf. S. P. Prog. Ind. Amer. Year Indl. Allian~a Alc. Petro Indl. Min. Fab. Cometa Average / / / / / J / J J W J Source: Relatorio da Diretoria, ; see Table 1for full names of companies of new shares than the actual price of the additions to capital stock. When the ratio is less than one, firms will not invest, because the income from new stock issues would be less than the cost of additions to capital stock. There are obvious problems with measuring book values and market values of capital stock, and the capital market in Brazil was thin and imperfect. My concern, therefore, is not with the ratio in any particular year, but with changes in this ratio over time. In Table 5, I have reduced the ratios to an index

15 Stephen H. Haber / 348 Table 5 Index of Market to Book Values, Seven Sample Companies, (1925 = 100) Conf. S. P. Prog. Amer. Year Indl. Alc. Petro Indl. Fab. Cometa Average "The 1931 book value was used because no data were available. The 1936 book value was used to calculate the ratio, because no 1937 data were available. Source: Derived from sources cited in Tables 1 and 3. with 1925 equal to 100 in order to chart how investors changed their perceptions of the value of their assets from 1925 to It is particularly striking that all five series move together. Though there is some stickiness in the market rates of return, all the rate-of-return and dividend series move along similar trajectories: healthy returns during the mid-1920s, falling off after 1927 and hitting bottom in the early 1930s, followed by a recovery after The market rate-of-return estimates are "sticky" because the market value to book value ratio also moved on the same trajectory as the accounting rate-of-return data. The results are therefore robust and encourage confidence. l7 l6 This ratio is based on and is similar to Tobin's q. It digers from Tobin's q in that it measures book value at acquisition cost, not at replacement value, which for historical data is not generally possible, since firms almost never (if ever) valued their assets at replacement cost in their financial statements. Indeed, most measures of Tobin's q by historians have measured book value in the way that I have here. l7 Regressions of the undepreciated series of the rate of return on capital stock against the depreciated series produced a correlation coefficient of,627. A regression of the undepreciated capital stock series against the market rate-of-return series pro-

16 Business Enterprise and the Great Depression in Brazil Adjustments to the Data To develop the first set of accounting rate-of-return estimates, I reworked the reported net profits data as follows. First, one-time write-downs of bad debt were spread out over six semesters, instead of one, on the assumption that such debts were accrued over a period of years and that their writing down should reflect a span of time. Second, various payments made to senior managers and members of the board of directors treated by the firms as costs were added back to earnings on the assumption that these charges were not in fact costs of production but transfer payments out of profits to the firm's largest stockholders. Because coverage was incomplete and because during particularly rough years firms tended not to publish profit-and-loss statements, I had to estimate the net profits of some of the firms in the sample from data contained in their balance sheets. After costs such as taxes, operating expenses, and interest payments are deducted from earnings, the remaining profit must be either distributed in the form of dividends to shareholders or allocated to any number of reserve funds as retained earnings. Both dividend payments and changes in the size of the various reserve funds can be tracked in the balance sheets. In order to estimate profits for these firms, then, additions to (or subtractions from, in the case of unprofitable years) to the reserve funds were added to dividend payments, after the payments were inflated by an additional 10 percent to account for directors' fees not enumerated in the balance sheets. l8 The adjusted profit data were then divided by the value of capital stock in order to calculate the rate of return. Capital stock (the value of land, equipment, buildings, inventories, and cash assets) was used instead of total assets as the denominator in the rate-of-return calculations to avoid the problems associated with the valuation of many types of assets held by firms. lg This concep- duced a slightly higher coefficient of,636. Neither would be strong enough to use in fitting missing observations in the undepreciated series, but they are strong enough to produce confidence in the statement that the three series follow roughly similar trajectories. ls Directors' fees of 10 percent of dividend earnings were customary among Brazilian companies during the 1920s and 1930s. The firms for which we have profit and loss statements distributed profits to their directors at this rate. l9 Capital stock dieers Gom total assets in three respects. First, debts owed to a company are not included. Carried forward from year to year in the calculation of

17 Stephen H.Haber / 350 tualization of firms' assets reduces the value of the denominator in the rate-of-return calculations and biases the results upward. Since my purpose is to make a case for the strong impact of the Depression on the textile industry, this upward bias strengthens the interpretation presented here. The rate-of-return estimates are further biased upward by the inclusion of rents paid by workers for company-owned housing in the earnings estimates (the value of the houses was also included in capital stock). Housing rents were included on the assumption that workers' housing was not a true subsidiary investment, but rather an integral part of carrying the firm's primary business. If rents were not included in earnings and the value of the houses were not included in capital stock, the rate-of-return results would have been far more volatile than those reported here. In bad years, rental receipts for some firms equaled a third of total earnings. The second set of rate-of-return estimates utilizes the profit and capital stock estimates already developed, but accounts for the physical plant depreciation omitted by Brazilian manufacturing companies. They normally valued machinery and improvements (buildings, waterworks, and roadways, for example) at acquisition cost and wrote down those investments only when they were retired from service. This accounting practice created two problems. First, improvements were never retired from service and were therefore never depreciated. Second, this method of writing down the value of machinery produced extremely low depreciation rates (on the order of less than one-half of one percent a year), because Brazilian textile manufacturers ran their machinery until it was completely unsewiceable. In 1947, for example, well over half of the machinery employed in Brazil's cotton mills had been installed prior to I therefore calculated accelerated depreciation schedules for each of the firms under study. Machinery was assets, a sizable part of the debt may not in fact be collectable. This may produce highly inflated asset figures. Second, the value of securities or bonds held in other companies is not included in capital stock. If subsidiary investments are included in capital stock and dividend earnings are included in profits, serious distortions in the rate of return on capital stock figures could result. Suppose, for example, that a firm lost money on its own operations but made money on its subsidiary investments. Estimates of the rate of return on capital stock that fail to make the appropriate adjustments would not reflect the firm's actual profitability. Third, other types of assets, the values of which are extremely difficult to determine, such as insurance policies, are not included in capital stock.

18 Business Enterprise and the Great Depression in Brazil / 351 depreciated at 5 percent per semester, improvements at 2.5 percent per semester.20 These depreciation estimates were deducted from both net profits and capital stock. New additions to physical plant, calculated from the assets data reported on the balance sheets, were not depreciated until the semester after they were placed into service. No significant adjustments were necessary in the calculation of the market rate-of-return estimates. Dividend payments declared for a particular year (regardless of when they were actually paid) were simply divided by the average market price of the company's common stock. Brazilian Textiles in the Great Depression: A Re-evaluation All three sets of estimates of the rate of return on capital indicate that the Brazilian cotton textile industry was in serious trouble before Median rates of return on capital stock fell steadily from 7 percent in the first semester of 1925 to 2 percent in the second semester of 1927 (see Table 1).The depreciated rate of return on capital stock series displays a similar trend, with a 20 These depreciation schedules were chosen for the following reasons. First, when Brazilian firms depreciated the value of animals and vehicles, they did so at 10 percent per semester. This would have been an excessive rate to depreciate machinery, so I depreciated machinery at half of this rate, and buildings at 25 percent of it. Second, I consulted a CPA from a Big Eight accounting firm who verified that depreciation schedules of 5 and 2.5 percent are consistent with depreciation schedules employed in the contemporary United States for large-scale manufacturers. Third, the depreciation rates employed here are roughly consistent with those used by Robert Gallman in computing the capital stock of the United States during the nineteenth century. See Robert Gallman, "Investment Flows and Capital Stocks: U.S. Experience in the Nineteenth Century," in Quantity and Quiddity: Essays in U.S. Economic History, ed. Peter Kilby (Middletown, Conn., 1987), Gallman used a straight-line depreciation of thirteen and seventeen years for producer durables and forty and fifty years for improvements. Over a seventeen-year period, the schedule employed here would have reduced the value of machinery to 17 percent of its acquisition cost, effectively depreciating it almost completely. Over a forty-year period, the schedule employed here to depreciate buildings and other improvements would have reduced the value of these items to 14 percent of their acquisition costs. Since the schedules employed here are flat rates, they decelerate over time (5 percent in the first semester, 4.75 percent in the second, 4.5 in the third, etc.). Although a depreciation schedule such as this mathematically cannot depreciate all the way to zero, it has several advantages. First, it means that it is not necessary to calculate separate depreciation schedules for equipment placed into service in different years. Second, the deceleration in the rate of capital consumption allows us to account for the fact that equipment tends to lose most of its value in the first few years after being placed in service.

19 Stephen H. Haber / 352 decline from 3 percent to zero over the same period (see Table 2). Market rates of return were somewhat stickier, displaying almost no significant movement from 1925 to 1927 (see Table 3). Firms depleted reserve funds in order to continue paying out dividends even when current profits did not support them. Thus, dividend payments as a percentage of par values did not decline as dramatically as did book value rates of return (see Table 4). Moreover, the market adjusted to the slightly lower dividend stream in 1926 and 1927 by bidding down the value of low-dividend company stock, thereby keeping the market rate of return constant. The result, as Table 5 shows, was a significant deterioration of the ratio of market to book values; the average of the index declined from 100 in 1925 to 62 in By 1929, with reserve funds exhausted and revenues severely depressed, most companies stopped paying dividends. The market rate of return now moved more in parallel with book value rates of return. The median book value rate of return on capital stock dropped to zero in the first semester of 1929 and stayed there through 1933, with the trough of -2 percent per semester reached in the second semester of 1931 (see Table 1). The depreciated book value rate-of-return estimates tell the same story, though more dramatically. Profits fell from zero in the second semester of 1927 to -5 percent in the first semester of 1931, hovering in the negative range through From 1928 to 1933, the median overall losses would have been 33 percent (see Table 2). As profits disappeared and firms were forced to the wall, there was nothing, except bad news, to distribute to the investors. Median dividend payments as a percentage of paid-in capital had run between 3 and 8 percent per semester from 1925 to 1928; now they stood at zero (see Table 4).21In 1929 only two firms managed to pay dividends, and by 1930 only one firm, SBo Pedro de Alcantara, continued to do so. That company alone managed to pay consistent dividends throughout the Depression; all the others ceased distributing profits to shareholders until the second semester of 1934 or paid dividends irregularly. Median market rates of return were therefore zero in four of the five years from 1929 to 1933 (see Table 3). As in the early years of the Depression, the market compensated for lower dividend payments by bidding down stock prices. Thus, the index of market to book values, which had stood Paid-in capital is the value of all shares outstanding valued at par-that the book value of stockholders' claims against the firm. is, it is

20 Business Enterprise and the Great Depression in Brazil at 62 in 1927, fell steadily to 23 in 1930 and recovered only slightly (to 34) in 1933 (see Table 5). For three of the eight firms in the sample, the Depression was a disaster. The Companhia Industrial Mineira was hit hardest. From 1930 to 1934, its combined losses totaled 33 percent of the value of its capital stock (79 percent if depreciation is taken into account). In 1934, its reserves nearly exhausted and its debts mounting, the firm was forced into financial reorganization. The Companhia Confiansa Industrial underwent financial restructuring as well. The firm went into the Depression already losing money (see Tables 1 and 2) and was therefore unable to weather the crisis. From 1931 to 1934 the mill was completely shut down, producing no cloth (or revenues) at all. With no new revenues coming in and with its cash reserves already depleted, the firm had no choice but to stop paying on its long-term bond debt. By 1934 the company was so far in arrears that it had to mortgage its physical assets in order to repurchase its bonds, make back payments on the interest due, and obtain the working capital required to restart production. A third firm, the Companhia de Fiaqfio e Tecelagem Cometa, also suffered staggering losses, though it was not forced into financial reorganization. The firm's combined losses from 1929 to 1934 totaled 30 percent (67 percent including depreciation). This is a far more dramatic story about the impact of the Depression than is presented in most other studies, which portray a slight contraction in the industry in and then a recovery beginning in The basis of this literature is the aggregate output series compiled by the Instituto Brasileiro de Geografia e Estatistica (IBGE) from consumption tax records running from 1911 to According to this series, output dropped only in 22 This series was originally published in Anucirio Estatistico do Brasil-Ano V, , It was republished in FundacBo Instituto Brasileiro de Geogrda e Estatistica-IBGE, Skries Estatisticas Retrospectivas, vol. 1: Repertbrio Estatistico do Brasil, Quadros Retrospectivos (Separata do Anucirio Estatistico do Brasil-Ano V ) (Rio de Janeiro, 1986), 39, and in IBGE, Estatisticas Histbricas do Brasil: Series Economicas, Demogrcijicas, e Sociais, 1550 a 1988 (Rio de Janeiro, 1990), With few exceptions, virtually all of the scholars who have worked on early Brazilian industrialization have used this series. For examples, see Wilson Suzigan, Industria Brasileira: Origem e Desenooloimento (SBo Paulo, 1986); Flhvio Rabelo Versiani and Maria Teresa R. 0. Versiani, "A Industria Brasileira antes de 1930: Uma Contribui@o," in Formqtio Economics do Brasil: A Experiencia da Industrializaqiio, ed. Flivio Versiani and Jose Roberto Mendon~a (So Paulo, 1977); Annibal Villanova Villela and Wilson Suzigan, Politica do Gooerno e Crescimento de Economia Brasileira, (RIOde Janeiro, 1975); Fishlow, "Origins and Consequences"; and Stein, Brazilian Cotton Manufacture.

21 Stephen H. Haber / 354 Table 6 Output of Cotton Cloth, IBGE and CIFTA Series, Output Index (thousands of meters) Percent (1929 = 100) Year lbge Series ClFTA Series Variance lbge Series ClFTA Series "The CIFTA series does not continue after 1932 Sources: IBGE series calculated from consumption tax data and published in Fundasso Instituto Brasileiro de Geografica e Estatistica (IBGE), Sdries Estatisticas Retrospectivas, vol. 1: Repertbrio Estatisitico do Brasil, Quadros Retrospectivos (Separata do Anuario Estatistico do Brasil-Ano V ) (Rio de Janeiro, 1986), 39. CIFTA series calculated from output data provided by mill owners to the Cotton Manufacturers Association (Centro Industrial de Fiago e Tecelagem de AlgodgdIFTA) and published in Fia~do e Tecelagem: Censo Organizado pel0 Centro Industrial de Fiqdo e Tecelagem de AlgodAo do Rw de Janeiro e Revisto pela Secco de Collecta do Departamento de Estatistica e Publicidade (Rio de Janeiro, 1935), ; production then stabilized in 1930 and actually increased by 33 percent in By 1934, cotton textile production was a full 50 percent higher than it had been in 1929 (see Table 6). One might be tempted to explain the difference between the firm-level data developed here and the standard IBGE output series as a function of sample bias. This interpretation, however, runs into a serious problem: the IBGE output series is at variance with the statistics compiled by the textile manufacturers themselves through their association, the Centro Industrial Fiaciio e Tecelagem de Algodiio. The CIFTA output series, not previously utilized in the literature, does not confirm the relative stability and growth of output found in the IBGE series, but instead corroborates the kind of dramatic decline revealed by the firm-level financial statistics: production declined by 9.4 percent in 1928, 12.1 percent in 1929, and 23.8 percent in In 1931 and 1932 the

22 Business Enterprise and the Great Depression in Brazil industry slowly began to stabilize, though its level of output was significantly below that from 1925 to 1927 (see Table 6). In short, this series indicates higher output for the years prior to 1930 and lower output for the years after 1930 than does the IBGE series. The more volatile behavior of the textile industry indicated by the firm-level financial data and the manufacturers' association output series is supported by contemporary commentaries on the textile industry and by statistical indicators gathered by the government of the state of Sio Paulo. Wilson Suzigan, for example, in looking at the reports of American and British consular agents for the period, came to the conclusion that "the greater part of the cotton textile factories reduced their operations to three or four days per week [in , and many closed down entirely."23 He goes on to cite an observer in SBo Paulo who declared in 1931 that "the majority of the factories were closed and those that were not were operating with dficulty."24 Data gathered by SBo Paulo's taxation authorities, analyzed by Nelson Nozoe, indicate a similar impact: the number of establishments engaged in producing cotton cloth declined from forty-seven to forty from 1929 to Scholars have therefore been aware that state-level statistical data and the available qualitative evidence point to the Depression's having had a significant effect on the textile industry. The problem has always been to account for the IBGE output series. Choosing to put more weight on the IBGE's aggregate statistics, scholars could either construct convoluted scenarios explaining how an industry that was in dire straits could be rapidly expanding production, or argue that by 1931 the crisis was largely over for the textile industry and explain the accounts of contemporaries and state-level data as caused by regional variations in the impact of the Depression. Stanley Stein, for example, followed the first course in his classic work on the Brazilian cotton goods industry. Faced with conflicting qualitative and quantitative evidence, and influenced by the manufacturers' arguments that the crisis was the product of "overproduction," he argued that firms responded to the situation by lowering their prices and increasing production in order to 23 Suzigan, Industria Brasileira, Ibid. Nelson Hideiki Nozoe, Sdo Paulo: Economia Cafeeira e Urbanim~do (SHo Paulo, 1984),

23 Stephen H. Haber / 356 cover at least part of their fixed costs and not risk losing their hard-to-replace skilled work forces. The result, according to Stein, was "an industry presumed to be in a state of overproduction while many of its factories worked beyond the normal workday."26 Wilson Suzigan and Annibal Villela, in their analysis of Siio Paulo manufacturing during the Depression, follow the second course. The local data that they present on Siio Paulo indicate that the textile industry was hit harder than the national-level aggregate statistics suggest.z7 Ultimately, however, they do not question the aggregate statistics, but instead assume that Siio Paulo's experience was different from that of the rest of Brazi1.28 The firm-level and CIFTA data presented here indicate, however, that the IBGE production series is seriously flawed: it is not, in fact, an index of output; it is really an index of sales volume, which declined far less than did the value of sales or the volume of output.29 Firms apparently engaged in heavy discounting (the average price of domestically produced cloth declined by 25 percent between 1925 and 1927) in order to keep inventories from piling up any higher than they already had.30 This strategy led to declining profit margins and pushed down corporate rates of return. Firms reacted to the situation by curtailing output in an effort to restore equilibrium to the market. Two factors apparently played a role in this contraction of the industry. The first, which operated predominantly during the later 1920s, was a shift in relative prices favoring imports. Between 1923 and 1926, the milreis appreciated because of a substantial increase in coffee exports and tight monetary policies, which drove down the effective rate of protection and favored imported goods over those domestically produced. Nondurable consumer goods imports grew by 15 percent between 1921 and Cotton textile imports grew even more, because the shift in relative prices was more pronounced in textiles than it was for imported goods as a whole. An index of imported goods prices constructed by FlAvio 26 Stein, Brazilian Cotton Manufacture, Villela and Suzigan, Politica do Gouerno, * Ibid., IBGE, Estatisticas Histdricas do Brasil, For detailed discussions of problems with the consumption tax series, see FlAvio Rabelo Versiani, "Indices de produ~io industrial para a dkcada de 1920: um reexame," Estudos Economicos 14 (1984): 4355; Flivio Rabelo Versiani, A &ca& de 20 na industrializa~cio brasileira (Rio de Janeiro, 1987), Fishlow, "Origins and Consequences," 327. Ibid.

24 Business Enterprise and the Great Depression in Brazil Versiani indicates that between 1923 and 1926 cotton textile import prices dropped by 38 percent, whereas the price for all imports decreased by 26 percent.32 Imported cotton goods then became more competitive with nationally produced goods, increasing their market share from 5.6 percent in 1921 to 17.0 percent in 1926, driving down the prices that Brazilian producers were able to charge. The average price of domestically produced cotton textiles fell by about 25 percent from 1925 to Beginning in 1927, however, relative prices began to shift in favor of domestically produced goods, causing a 17 percent increase in the cost of imported goods by With this shift and a new tar8 established in 1929, imported goods were able to take only 4 percent of the market in At this point, however, the second factor that drove down the profitability of the industry took precedence. Commodities prices, which had been declining throughout the latter part of the 1920s, now went on a downhill roller-coaster ride. This had obvious effects on the export industries (such as coffee and sugar), as well as on ancillary enterprises like the railroads, ports, and banks. As also happened in Mexico during this period, the decline in demand driven by falling commodities prices undermined the profitability of domes tic manufacturing. 34 Beneath the generalized decline in the rate of profit of the textile industry was a good deal of variation among firms. Most notable here was the Siio Pedro de Alcantara mill, which not only produced phenomenal profits from 1925 to 1927 (averaging 43.5 percent per semester, both depreciated and undepreciated, until the second semester of 1927-see Tables 1 and 2), but which managed to remain profitable even during the worst years of the Depression. In fact, during the second semester of 1930, when the other firms in the sample were either just barely breaking even or were posting losses as high as 6 percent, the Siio Pedro de Alcantara mill managed to earn a 7 percent profit. The results are even more impressive if we look at the depreciated rate-of-return data. Siio Pedro de Alcantara earned 6 percent in the second semester of 1930, while the other firms were losing from 1 to 10 percent 32 Flivio Versiani, "Before the Depression: Brazilian Industry in the 1920s," in Thorp, ed., Latin America in the 1930s, Ibid., 17S For a discussion of the effects of the Depression on Mexican cotton goods producers, see Haber, Industry and Underdevelopment, chap. 9.

25 Stephen H.Haber / 358 (see Table 2). Thus, throughout the depths of the Depression, Siio Pedro de Alcantara paid a steady dividend stream of 10 percent of par value per semester (see Table 4). The result was an extremely high market rate of return averaging 10 percent annually from 1929 to 1933 (see Table 3). Only one other firm, the Companhia America Fabril, was even moderately profitable during this period. It posted sizable losses in 1929 and 1930, but then produced profits of from 1 to 5 percent per semester during 1931, 1932, and 1933 (see Tables 1and 2). This wide variation between the Alcantara mill and its competitors is primarily explained by Alcantara's significantly higher level of physical productivity. The available data from the 1927 census of the cotton textile industry indicate that Alcantara's outputs per worker and per machine were higher than those for the other sample firms and for the industry as a whole. The value of output per worker-year in the Alcantara mill was 66 percent higher than the average for the eight sample firms and 37 percent higher than the industry average (10,400 milreis versus 6,256 milreis and 7,577 milreis, respectively).35 Data on the value of output per machine tell much the same story: 636 milreis per spindle per year for Alcantara versus 185 for the sample companies and 269 for the industry as a whole. Interestingly, this productivity advantage was evidently not the result of greater work intensity or of specialization in high-value goods. In fact, the data indicate quite the opposite. The ratio of spindles to workers at Alcantara (16:l) was half that of the other sample firms (33:l) and 20 percent lower than the industry average (20:l). In addition, calculations of the average price of goods produced and the average amounts of raw cotton consumed per unit of output do not indicate that Alcantara's advantage lay in specializing in the production of fine-weave goods. Alcantara's higher productivity (and therefore higher profits) appear to have had as their source the company's possession of newer, more efficient machinery than most of its competitors. Beginning in 1916, the firm began to invest in new machinery and other equipment, a capital modernization program that picked up speed through the 1920s: between 1925 and 1927 alone, the firm doubled the value of its physical plant. The few firms, like Alcantara, that had reinvested their earnings from the early part of the 35 Calculated from CIFTA, 1928.

26 Business Enterprise and the Great Depression in Brazil / s into new plant and equipment had a significant advantage over their competitors, the great majority of whom were running machinery of pre-1914 vintage. Beginning in 1934, the combined effects of a revision of the tariff schedule and a general recovery of the economy (driven by Keynesian fiscal policies, monetary expansion, and an increase in world coffee prices) produced an upturn in the textile industry. Median accounting rates of return on capital stock steadily climbed from 1 percent in the first semester of 1934 to 3 percent in the first semester of 1936 and to 4 percent during the first semester of The depreciated accounting rate-of-return estimates indicate a somewhat slower recovery, with the rate of return moving from zero per semester in 1934 to 3 percent per semester in 1937, but the general direction of profits was much the same as in the undepreciated series-that is, rates of return moved from negative during to moderately positive during With the revival of corporate profits came a resurgence in dividend payments to shareholders; the average dividend as a percentage of par value increased from just under 2 percent per semester during the trough of the Depression to well over 3 percent per semester from 1934 through 1937 (see Table 4). Market rates of return therefore recovered as well: the median rate of return increased from zero during to 5 percent in 1934, to 6 percent in 1935, to 4 percent in 1936, and to 7 percent in 1937 (see Table 3). Of the seven firms for which we have post-1934 data, only one, the Companhia ConfianCa Industrial, produced no dividends through These market rates of return were below their pre-depression levels but were still quite strong, especially considering that in 1933 investors began to revalue their assets upward, thereby increasing the size of the denominator in the market rate-of-return calculations. The index of market to book value ratios increased from 25 in 1931 to 34 in 1933, to 39 in 1935, and to 45 in 1937 (see Table 5). Investors, in short, were now somewhat more confident about the ability of their assets to produce a positive stream of earnings over the long run. Significantly, however, even in 1937 when the worst of the Depression was over, the ratio of market to book values still had not regained even half of its 1925 leve1.36 The 36 Since the natural tendency in an inflationary economy such as Brazil's is for market values to rise with increases in the general price level, the market value to book value ratio should be gradually biased upward over time. The estimates presented

27 Stephen H. Haber / 360 Depression did, therefore, have a significant effect on investor confidence over the medium term and served to depress new spending on plant and equipment. Several factors combined to prevent a sustained round of new investment in the industry during this period. First, the industry had been hit hard by the crisis of With their reserve funds depleted from seven years of low or nonexistent profits, Brazil's textile enterprises would have had to take on additional debt or issue new stock in order to modernize. The market would not have been receptive to such action. Share prices for textile company stocks continued to be depressed, as the market to book value data in Table 5 indicates, making the issuing of new shares a highly unlikely proposition. For the same reason, firms were also unlikely to issue long-term debt. Bond issues in the 1930s typically carried high interest rates (8-9 percent) in order to coax capital out of a very skittish investment community. Judging from the eight companies studied here, the only firms to issue new debt were those that had no choice but to do so or go out of business altogether. Second, it was not possible to import textile machinery into Brazil between 1931 and Brazil's textile industrialists had always relied on their ties to government during periods of crisis. Indeed, the first and last resort in times of trouble was state intervention. The entire industrialization project from the 1890s onward had, in fact, been intimately linked with various kinds of government support.37 Thus, the Depression produced demands on the government by the textile manufacturers to enact legislation ~rohibiting the importation of new capital goods, because the industrialists claimed that the crisis was caused by "overproduction." They also tried, unsuccessfully, to regulate the market further by limiting the number of hours that mills could be operated.38 The results of this legislation were twofold. The first was the here, therefore, overestimate the recovery of investor confidence. Since this article argues that the Depression had a significant negative effect on investor confidence, this bias in the data strengthens the interpretation advanced here. 37 For a discussion of state intervention in early Brazilian industrialization, see Steven Topik, The Political Economy of the Brazilian State, (Austin, Texas, 1987), chap. 5. For a detailed discussion of this legislation and the debates surrounding it, see Stein, Brazilian Cotton Manufacture,

28 Business Enterprise and the Great Depression in Brazil expansion of the nation's small machine shops into the production of textile machinery. The textile producers responded in typical fashion to this unexpected development: they tried to have enacted legislation prohibiting the domestic production of machinery. In this rather perverse attempt at market control, they failed.39 The argument that the development of a national machine-making industry was inimical to the country's industrial growth was seen, even by the government of Getiilio Vargas, as a transparent ploy designed only to protect the special interests of the textile industrialists. The machine shops that did move into the production of looms were fairly small, however, and could not have satisfied national demand at this juncture. The largest of them was capable of producing only 130 looms per month. The second effect of this legislation was that it discouraged technological and managerial innovation. Indeed, the textile industrialists' response to the Depression, much like their response to increased foreign competition after the Second World War, was to seek state protection as a substitute for the modernization and streamlining of production and distribution. It is striking how little money the profitable firms among the eight sample companies reinvested in new plant and equipment once the industry began to rebound. Instead, they invested large sums in lowrisk government bonds. Fully 20 percent of the assets of the Companhia America Fabril, for example, were invested in longterm government debt at the end of the Depression. The long-run impact of the lack of technological and managerial innovation was a dramatic decline in labor productivity. Between 1925 and 1944, output per worker dropped from 6,038 meters to 4,905 meters, a compound rate of decrease of better than one percent a year. This sustained decline in labor productivity was not driven by a fall in total output. The opening of markets provided by the Second World War produced boom conditions for the Brazilian textile industry, with output almost doubling between 1925 and 1944 (from 671 million meters of cloth to 1,152 million meters). The lack of new investment in the years prior to the war meant, however, that the increase in output had to be produced by hordes of workers operating an antiquated industrial plant around the clock. Indeed, in 1939 only 4,160 of Brazil's 74,!246 looms were of the modem automatic variety; over half of 39 Ibid.,

29 Stephen H. Haber / 362 the installed equipment predated the First World War. Brazil was therefore able to increase total output of cotton goods by 2.7 percent a year only at the cost of an increase in the size of the work force of 3.8 percent a year.40 Implications The firm-level analysis presented here has a number of implications for the study of Latin American industrial development. First, the data developed lend support to the argument made by Stanley Stein that the recovery from the Depression in Brazil was not accompanied by a concomitant resurgence in new investment. Instead, as Stein has pointed out, Brazil's textile industrialists turned to the state to limit new competition by freezing machinery imports, making a program of new investment difficult if not impossible.41 Why Brazil's textile industrialists chose to follow this strategy is made clear by the market to book ratios developed in Table 5: even as late as 1937, the Brazilian investment community still valued the assets of the textile industry at less than half their 1925 values. Second, the analysis underlines the important role played by state intervention in Brazil's industrial development. The modern Brazilian textile industry owed much of its early success to protectionist measures designed to insulate it from foreign competition, as well as to government loans designed to prevent many of the major firms from going bankrupt in the credit crunch of the mid- 1890~.~~ In the 1920s, Brazil's textile industrialists chose once again in a time of trouble to turn to the state to structure the market, rather than to pursue a process of reinvestment and modernization in order to become internationally competitive. As Albert Fishlow has pointed out, this experience of the Depression encouraged a model of growth in the consumer goods industries based on the exploitation of vast amounts of inexpensive labor and encouraged a mindset of short-run planning and technological and managerial obsolescence. Over the long run, this strategy of Calculated from data in Guilherme da Silveira Filho, Memoria sobre a situa~cio da industria textil brasileira (Rio de Janeiro, 1947), 23; Ministerio do Trabalho, Industria, e Comercio, ComissPo Executiva Textil, Industria textil algodeira (Rio de Janeiro, 1946), Stein, Brazilian Cotton Manufacture, Topik, Political Economy of the Brazilian State, chap. 5.

30 Business Enterprise and the Great Depression in Brazil / 363 capital-scarce industrialization perpetuated the use of antiquated technology and low-wage, labor-intensive methods of production. The strategy ultimately led "not to successful modernization, but to slow rates of growth and actual reduction of the initial labor force" in later decades once pressure for higher wages made the continuation of the labor-intensive strategy infeasible.43 Third, judging by the Brazilian case, it appears that the Great Depression had a more profound impact on Latin American manufacturing than previously thought. Certainly, the analysis of the eight cotton textile firms studied here should not be regarded as conclusive evidence that the aggregate data for all product lines in all countries are flawed. But it does suggest that scholars should be wary of relying solely on published aggregate statistics in their analyses of Latin America's early industrial experience. Indeed, a similar firm-level analysis of the Mexican case indicates that its experience was not unlike Brazil's in terms of the timing, strength, and duration of the Depression.44 The discrepancy between the published statistical series and the experience of individual firms suggests that scholars need to shift the unit of analysis from aggregate economic sectors to the individual business enterprise. The value of the firm perspective is further underlined when one considers the substantial variation in profitability among the firms measured here. Judging from this sample, those enterprises that invested in new plant and equipment during the early 1920s weathered the storm of the Depression much better than those that did not invest. Whether this was the case in all product lines or was peculiar to textiles bears examination. Why some firms followed investment and modernization strategies different from the norm is a question that deserves attention as well. Without studies that speclfy the individual business enterprise as the unit of analysis, however, these questions will remain unanswered. "Fishlow, "Origins and Consequences," Haber, Industry and Underdevelopment, chaps. 9 and 10