¹ DeCanio, Stephen, “Cotton ‘Overproduction’ in Late Nineteenth-Century Southem Agriculture,” Journal of Economic History XXIII (September 1973) 608–633CrossRefGoogle Scholar; DeCanio, S., Agriculture in the Postbellum South (Cambridge, 1974), pp. 240–261Google Scholar.

² The price of cotton divided by the price of com averaged 153 in New Orleans during 1849–1860, and 148 in Charleston. For 1876–1880, the on-farm relative price was 145 in Louisiana, 144 in South Carolina; computed from Cole, A. H., Wholesale Commodity Prices in the United States 1700–1861, Statistical Supplement (Cambridge, 1938)CrossRefGoogle Scholar; and USDA Statistical Bulletin No. 16, June 1927. To a surprising extent economists have been willing to assert that “falling transportation costs” must explain increasing regional specialization, without showing that the required effect local prices actually occurred. Most recently, see Brown, William W. and Reynolds, Morgan, “Debt Peonage Re-examined,” Journal of Economic History XXXIII (December 1973), p. 871Google Scholar. The appeal to “general market behavior” leaves a lot to the imagination. Similarly, the end of the Civil War cannot be invoked to explain an increase in specialization relative to the pre-war period.

³ See particularly the “letters from prominent cotton-growers” in U.S. Senate Report 986, Report of the Committee on Agriculture and Forestry on Conditions of Cotton Growers in the United States (1895). Hereinafter referred to as “Senate Report 986.”

⁴ The aggregate swing to cotton cannot be explained by geographic shifts. Of the leading 386 counties in the Cotton South—excluding Texas and Arkansas—fully 85.9 percent show a rise in the cotton/com ratio, while 80.9 percent show a rise in the cotton/hogs ratio.

⁵ Ransom, Roger and Sutch, Richard, “Debt Peonage in the Cotton South after the Civil War,” Journal of Economic History XXXII (September 1972), p. 665Google Scholar; Brown and Reynolds, “Debt Peonage Re-examined,” pp. 867–868; Joseph D. Reid, Jr., “Sharecropping as an Understandable Market Response—The Postbellum South” and Higgs, Robert, “Race, Tenure, and Resource Allocation in Southern Agriculture, 1910,” both in Journal of Economic History XXXIII (March 1973) 106–130, 149–169.Google Scholar

⁶ Brown and Reynolds are quite mistaken in their inference that “before the war, as well as after, small farms were less self-sufficient than large farms,” and they are still more mistaken in contending that “about 35 percent of the total decline (in corn output per capita) can be ‘accounted for’ by the decline in farm sizes” (“Debt Peonage Re-examined,” p. 869). Small farms had relatively low outputs per capita primarily because they were small (I) and not because they were “less self-sufficient.” The figures on hogs per capita show a clear negative correlation with farm size. Hence, using 1860 parameters, the decline in farm size cannot explain any of the decline in hogs per capita, and Table 2 makes it clear that it cannot explain any of the shift of resources into cotton.

⁷ See primarily Gallman, Robert, “Self-Sufficiency in the Cotton Economy of the Antebellum South,” Agricultural History XLIV (January 1970)Google Scholar, and Hutchinson, W. K. and Williams, S. H., “The Self-Sufficiency of the Antebellum South,” Journal of Economic History XXXI (September 1971)Google Scholar.

⁸ The argument that small fanners attempt to avoid market-place risks by devoting a large share of their acreage to food crops appears in Behrman, J. R., Supply Response in Underdeveloped Agriculture (Amsterdam, 1968)Google Scholar; Boussard, J. M. and Petit, M., “Representation of Farmers' Behavior with a Focus-Loss Constraint,” Journal of Farm Economics XLIX (November 1967)Google Scholar; Lipton, M., “The Theory of the Optimizing Peasant,” Journal of Development Studies IV (April 1968)Google Scholar. “Safetyfirst” models are categorized and discussed in detail in J. Roumasset, Risk and the Efficiency of Peasant Agriculture (North-Holland, forthcoming).

⁹ Explicit statements are provided below.

¹⁰ The theoretical aspects of this model are explored in H. Kunreuther and G. Wright, “Safety-First, Gambling, and the Subsistence Farmer” (unpublished).

¹¹ This statement is based on analysis of the Parker-Gallman 1860 sample. For farms of a given size, a higher cotton-com ratio is positively correlated with higher t tal output value per worker or per acre in every part of the Cotton South.

¹² See, in addition to Gallman, Moore, John H., Agriculture in Antebellum Mississippi (New York, 1958), p. 58Google Scholar; Phillips, U. B., American Negro Slavery (New York, 1918), pp. 207–208Google Scholar; Davis, Charles S., The Cotton Kingdom in Alabama (Montgomery: Alabama State Department of Archives and History, 1939), p. 66Google Scholar; Parker, W. N., “A Note on Regional Culture in the Corn Harvest,” Agricultural History XLVI (January 1972), pp. 181–189Google Scholar. For the postbellum period, see statements in J. R. Dodge's survey in “Report of the Statistician,” United States Agriculture Department Annual Report (1876). For example, “they use the same labor [on food crops] which is employed on cotton, but at seasons which do not interfere” (p. 149).

¹³ For example, Texas Agricultural Experiment Station, “Cost of Cotton Production and Profit per Acre,” Bulletin No. 26 (Bryan, Texas: March 1893); USDA, “The Cost of Cotton Production,” Division of Statistics, Miscellaneous Series, Bulletin No. 16 (1899).

¹⁴ Masson, R. T., “The Creation of Risk Aversion by Imperfect Capital Markets,” American Economic Review LXII (March 1972), pp. 77–86Google Scholar. See also Masson, , “Utility Functions with Jump Discontinuities: Some evidence from Peasant Agriculture,” Economic Inquiry XII (December 1974)Google Scholar.

¹⁵ This effect would help to explain the fact that com output per capita is higher on large farms, even though m is clearly lower. For twentieth-century evidence that the “elasticity of the poverty line” with respect to income is about 0.6, see Kilpatrick, R., “The Income Elasticity of the Poverty Line,” Review of Economics and Statistics LV (August 1973), 327–332CrossRefGoogle Scholar.

¹⁶ See Wagner, H. M., Principles of Operations Research (Englewood Cliffs, New Jersey, 1969), pp. 792–795Google Scholar, or Hillier, F. and Lieberman, , Introduction to Operations Research (San Francisco, 1967), pp. 370–377Google Scholar.

¹⁷ The conversion of output data into acreage shares does raise a technical difficulty. The assumption of normally distributed error is clearly inappropriate where the dependent variable is a bounded fraction and many observations occur at one endpoint (zero cotton). Since the zero-cotton farms are important to our argument, we nave chosen to include them anyway, but the qualitative conclusions of Tables 3 and 4 are not changed when these farms are excluded, nor when (in addition) the dependent variable is converted into the logistic form m′ = log (m/l-m). For the postbellum regressions using county data, this problem is ignored because there are few observations near the endpoints.

¹⁸ Fogel, R. and Engerman, S., Time on the Cross (Boston, 1974), esp. Vol. I, 199–209Google Scholar, and Vol. II, 138–149.

¹⁹ “Debt Peonage,” 658–664.

²⁰ The best statement of the Ransom-Sutch position on the crop mix is “The ‘Lock-in’ Mechanism and Overproduction of Cotton in the Post-Bellum South,” Agricultural History XLIX (April 1975)Google Scholar. By “overproduction” of cotton Ransom and Sutch mean that cotton cultivation was pushed beyond the point of equality between the marginal rate of transformation and the price ratio.

²¹ USDA, Agricultural Marketing Service, Cotton and Cotton Seed (Statistical Bulletin No. 164, June 1955); “Revised Estimates of Corn Acreage, Yield and Production, 1866–1929” (mimeo, 1934); Prices of Farm Products Received by Producers (Statistical Bulletin No. 16, June 1927).

²² “The Cost of Cotton Production,” Division of Statistics, Miscellaneous Series, Bulletin No. 16 (1899).

²³ Input costs are adjusted for each year by the Warren-Pearson price index, to achieve comparability with the agricultural prices for earlier years. For the years in which on-farm cotton prices are not available (1866–1875, 1877, 1881), we used the New York price, reduced by an “adjustment factor” equal to the average New York-on-farm price differential during the first ten years of observations.

²⁴ Ransom and Sutch (“The ‘Lock-in’ Mechanism,” Table 1) present figures for retail cash prices (as well as credit prices) for Georgia, which average 17 percent higher than the farmgate prices for the years 1878–1890. An adjustment of this size would leave the conclusion unaffected for all states but Alabama.

²⁵ Ransom, and Sutch, , “The Ex-Slave in the Post-Bellum South,” Journal of Economic History XXXIII (March 1973)Google Scholar, Tables 2 and 3. An inverse correlation between yields and farm size was also reported by J. R. Dodge in USDA Annual Report for 1876, pp. 130, 139–147, and is indeed virtually a universal cross-section finding. See Bachman, K. L. and Christensen, R. P., “The Economics of Farm Size,” in Southworth, H. M. and Johnston, B. F., editors, Agricultural Development and Economic Growth (Ithaca, 1967)Google Scholar.

²⁶ The calculation of Ransom and Sutch (“The ‘Lock-in’ Mechanism”) that corn was “more profitable” than cotton rests in major part on their assertion that the release of one cotton acre allowed the cultivation of two acres of corn. Regardless of the labor requirements of the two crops, we view this assumption as inappropriate because it ignores the land constraint, which we have put at the center of our argument.

²⁷ “Productivity and Income Distribution,” p. 437; Agriculture in the Post-bellum South, pp. 176–180.

²⁸ “The ‘Lock-in’ Mechanism,” Tables 1 and 2. The variance in annual state average yields may not, of course, accurately reflect the yield variance at the farm level. This consideration applies to both cotton and corn yields, and should not greatly affect the comparison between the variances of x and y. Experimentation with the time-series figures suggests that no reasonable alteration of the yield variances could affect the conclusions.

²⁹ This is the objective function proposed by Roy in his classic statement of the safety-first principle: “Safety-First ana the Holding of Assets,” Econometrica XX (July 1952), 431–449Google Scholar.

³⁰ Kunreuther and Wright, “Safety-First, Gambling, and the Subsistence Farmer.”

³¹ For general discussions of tenancy as a temporary status, held by those trying to accumulate experience and financial resources, see Bogue, Allan G., From Prairie to Corn Belt (Chicago, 1968), pp. 53, 65–66Google Scholar; Brooks, R. P., “The Agrarian Revolution in Georgia, 1865–1912,” Bulletin of the University of Wisconsin, History Series, Volume III (Madison, 1914), 59Google Scholar; Winters, Donald L., “Tenant Fanning in Iowa, 1860–1900,” Agricultural History XLVIII (January 1974)Google Scholar. For numerous additional references, and a critical view of the reality of this “agricultural ladder” (as opposed to the motivation), see Cox, LaWanda, “Tenancy in the United States, 1865–1900,” Agricultural History, XVIII (July 1944)Google Scholar.

³² Comments to this effect may be found in the 1895 Senate Report No. 986, 323 and elsewhere, and in Hammond, M. B., The Cotton Industry (New York: American Economic Association, 1897), pp. 121–122, 136Google Scholar. See also the long quotation from W. J. Northen (1889) in DeCanio, Agriculture in the Postbellum South pp. 108–109. DeCanio also quotes (pp. 103–104) from DeBow's Review of the late 1860's, showing the apprehension with which observers viewed the concentration on cotton. For graphic complaints about the effects of the rapid price fall, see Cox, LaWanda and Cox, John H., editors, Reconstruction, the Negro and the New South (New York, 1973), pp. 331–333Google Scholar.

³³ In this interpretation, we attribute the significant positive coefficient for this variable in regression (14) to the strong correlation between tenancy at this size class and at other size classes in a county.

³⁴ “Debt Peonage,” p. 665.

³⁵ We are indebted to Professor Sutch for this result, though he is not to be held responsible for the inference which we draw.

³⁶ DeCanio, “Cotton ‘Overproduction’,” p. 612; Agriculture in the Postbellum South, p. 96.

³⁷ Senate Report 986, p. 25.

³⁸ USDA, “The Cost of Cotton Production,” p. 62.

³⁹ Hammond, The Cotton Industry, p. 173.

⁴⁰ For demand curve estimates, see Wright, G., “Cotton Competition and the Post-Bellum Recovery of the U.S. South,” Journal of Economic History XXXIV (September 1974)Google Scholar. This study finds some increase in the demand elasticity after the Civil War, but argues that such a result is in part an artifact. For estimates of very low elasticities for later years, see Schultz, Henry, The Theory and Measurement of Demand (Chicago, 1938), chapter viiiGoogle Scholar.

⁴¹ DeCanio, Agriculture in the PostbeUum South, p. 104.

⁴² Texas Agricultural Experiment Station, Bulletin No. 26, p. 310.

⁴³ Senate Report 986, p. 286. Emphasis added.

⁴⁴ Ibid., XLIII. The next sentence reads: “Such diversification will also tend to increase the price by diminishing the supply.”

⁴⁵ USDA Annual Report for 1876, pp. 148–151.

⁴⁶ Quoted in DeCanio, Agriculture in the Postbellum South, p. 99.

⁴⁷ Senate Report 986, p. 302. Hammond stated plainly that the “unprofitability” of cotton dated only from the period since 1890, which is to say the early 1890's (The Cotton Industry, p. 167).

⁴⁸ Hammond, The Cotton Industry, p. 149.

⁴⁹ USDA Annual Report for 1876, p. 149.

⁵⁰ Senate Report 986, p. 330. Similar statements appear on pp. 144, 292, 410.

⁵¹ USDA Annual Report for 1876, p. 151.

⁵² Senate Report 986, p. 317.

⁵³ Ibid., p. 393.

⁵⁴ Hammond, The Cotton Industry, p. 190. Emphasis added. The attack by Charles H. Otken on the “credit monopoly” is similarly ambiguous at many points on how this monopoly causes “overproduction” of cotton. See The Ills of the South (New York, 1894), esp. pp. 56–57, 61. In his policy prescriptions, it is clear that what Otken is calling for is “two years of self-denial and economy” (p. 30).

⁵⁵ See Bull, Jacqueline P., “The General Merchant in the Economic History of the New South,” Journal of Southern History XVIII (February 1952), pp. 41–42Google Scholar; Clark, Thomas D., “The Furnishing and Supply System in Southern Agriculture since 1865,” Journal of Southern History XII (February 1946)Google Scholar.

⁵⁸ DeCanio, “Cotton ‘Overproduction’,” pp. 624–631; Agriculture in the Postbellum South, pp. 241–261. Because the “speed or adjustment” is defined in terms of a percentage of the long-run shift, DeCanio's estimates in fact imply that the cotton supply response was below that for wheat in every year after the first. DeCanio acknowledges the possible inappropriateness of a symmetrical supply elasticity (Agriculture in the Postbellum South, pp. 309–311).

⁵⁷ 6Johnson, D. Gale, “Resource Allocation Under Share Contracts,” Journal of Political Economy LVIII (April 1950)Google Scholar.

⁵⁸ Senate Report 986, p. 298.

⁵⁹ ibid., p. 289.

⁶⁰ Lurie, Jonathan, “Speculation, Risk and Profits: The Ambivalent Agrarian in the Late Nineteenth Century,” Agricultural History XLVI (April 1972), 269–278Google Scholar. The “Uncle Remus” quotation may also be found in this article.

⁶¹ Wright, “Cotton Competition,” pp. 629–635.

⁶² Hammond, p. 122.