This research was carried out with a grant from the Andrew P. Mellon Foundation. Earlier drafts have benefited from comments from Sandra Archibald, Melvin Fuss, Paul Hohenberg, Donald McCloskey, Ramon Myers, Thomas Wiens, and two anonymous referees, and from participants at seminars at the University of California, Davis, and the University of Toronto.Google Scholar

¹ Huang, Phillip C. C., Peasant Economy and Social Change in North China (Stanford, 1985).Google Scholar For India, see Chandra, Bipan, “Reinterpretation of 19th Century Indian Economic History,” The Indian Economic and Social History Review, 5 (03 1968), pp. 35–76. esp. pp. 50–51.CrossRefGoogle Scholar

² Berry, Albert and Cline, William, Agrarian Structure and Productivity in Developing Countries (Baltimore, 1979).Google Scholar

³ This review appears in Economic Development and Cultural Change, 35 (04 1987), pp. 670–82.CrossRefGoogle Scholar

⁴ It should be pointed out that the data are primarily drawn from areas that were relatively commercialized. By the 1930s this segment of the rural Chinese economy constituted between 40 and 50 percent of the rural population, or upwards of 200 million people. My analysis does not necessarily hold for the remaining half.Google Scholar

⁵ According to Buck, on small farms the percentage of net income from other-than-farm sources was approximately three times that on larger farms. These estimates appear in Buck, John, Land Utilization in China: Statistical Volume (Chicago, 1937), p. 311.Google Scholar

⁶ This article focuses primarily on local factor markets. Implicit in my analysis is a similar assumption about product markets. I examined some of the changes in Chinese product markets in “Chinese Agriculture and the International Economy, 1870s–1930s: A Reassessment,” Explorations in Economic History, 22 (05 1985), pp. 168–93.CrossRefGoogle Scholar

⁷ The two factors that contributed most to this process were the secular rise in the terms of trade between the 1890s and the late 1920s and new domestic and international market opportunities.Google Scholar

⁸ For a related “revisionist” view on the influence of commercialization on the prewar Japanese rural economy, see Smethurst, Richard, Agricultural Development and Land Disputes in Japan, 1870–1940 (Princeton, 1986).CrossRefGoogle Scholar

⁹ Data compiled by Ramon Myers for various villages in East-Central China in the 1930s show the same phenomenon. See his “The Commercialization of Agriculture in Modern China,” in Willmott, W. E., ed., Economic Organization in Chinese Society (Stanford, 1972).Google Scholar

¹⁰ More formally, I tested the relationship between farm size and land productivity by regressing gross farm output per unit of cultivated area (GFOCA) on the log of cultivated area (In CA) using the pooled data. (Lacking local price indices, the data for Wukuan have been deflated by the Nankai wholesale price index for agricultural goods for Tientsin, a major outlet for the market surplus of the region. This index appears in Nan-ching ta-hsueh ching-chi yen-chiu-so, eds.Google Scholar, Nan-k'ai chih-shu rzu-liao hui-pian, 1913–1952 [Beijing, 1958], p. 12.) The t–statistics are in parenthesis. GFOCA = 33.3_legacy1 + 1.31(In CA) (5.05) (0.63) R ² = 0.07 n = 57 The small t–value for the coefficient on In CA and the low explanatory power of the model (as measured by R²) do not support a systematic relationship between farm size and land productivity in these localities.Google Scholar

¹¹ By comparison, Huang found in his sample of 14 farms that net profits (and, therefore, net profits per unit of cultivated area) were negative for 4 of the 5 small (less than 30 mou) farms, but positive for the remaining 9 farms. He attributes this to the greater use of labor on small farms, of which more is said below, and uses this finding to support the view that smaller farms were less efficient.Google Scholar

¹² On the basis of oral testimony he obtained, Huang noted that a wage laborer worked a longer day at greater intensity than family members did on their own farms. According to Buck, on the other hand, women and children performed between 20 and 30 percent of the work on farms in North China. On some of these smaller farms the adult male hired out as a monthly or annual laborer, so remaining household members performed most of the work on the family farm.Google Scholar

¹³ Buck, John L., Chinese Farm Economy (Nanking, 1930).Google Scholar

¹⁴ Buck used relative prices to convert nongrain crops into their grain equivalents.Google Scholar

¹⁵ Even if there is not a well-developed rural labor market that offers off-farm wage opportunities, we would still expect land to be leased until differences in the marginal product of labor across farm sizes disappeared. Only in the case where neither set of markets is working well would we find small peasant farms using land more intensively. More formally, if imperfections are present in at least two of the factor markets (that is, markets for land, labor, capital, and draft animals), the factor price ratios that peasant households implicitly face will differ. Assuming profit maximization, this implies that optimal factor combinations will differ among farm households, as will output/input ratios.Google Scholar

¹⁶ See, for example, Rawski, Thomas, China's Republican Economy: An Introduction, Joint Center of Modern East Asia, University of Toronto-York, Discussion Paper No. 1 (1978);Google Scholar and Myers, Ramon, The Chinese Economy: Past and Present (Belmont, 1980).Google Scholar After arguing to the contrary in earlier work, Albert Feuerwerker noted in his recent contribution to the Cambridge History of China that few studies have been able to document the presence of monopolistic or monopsonistic elements in local markets. See his “Economic Trends, 1912–1949,” in Cambridge History of China, Republican China 1912–1949, Part I (Cambridge, 1983).Google Scholar

¹⁷ Farm households in North China were not only aware of local opportunity costs, but were equally informed of and profoundly influenced by interregional wage differences. According to Thomas Gottschang, these same households “year in and year out weighed the information they received about job possibilities and wage levels in Manchuria against local conditions, with an eye to sending off a son or a brother when the difference promised a positive return to their investment.” See Gottschang, Thomas, “Economic Change, Disasters, and Migration: The Historical Case of Manchuria,” Economic Development and Cultural Change, 35 (04 1987), pp. 461–90.CrossRefGoogle Scholar

¹⁸ Buck, Land Utilization in China, pp. 55–56.Google Scholar

¹⁹ Huang, in Peasant Economy and Social Change, arrived at the same conclusion, but did not try to relate this finding to the fact that a market for draft-animal services existed.Google Scholar

²⁰ The surveys do not actually provide data on prices received, but such an interpretation is consistent with the data on crop selection.Google Scholar

²¹ See Yotopoulus, Pan and Lau, Lawrence, “A Test for Relative Economic Efficiency: Some Further Results,” American Economic Review, 63 (03 1973), pp. 214–23, and a number of earlier papers cited therein.Google Scholar

²² Thirty mou can be used to set off smaller farms that relied primarily on family members for labor from those that depended more heavily on hired farmhands. In North China at that time an adult male could farm approximately 15 mou by himself. Assuming that the average household had 2 male equivalents, it would take 30 mou to absorb the supply of family labor. Any “average-sized” household with landholdings exceeding 30 mou would presumably have to hire in some labor, while any smaller would probably have tried to hire out. My experimentation with cutoff points on both sides of 30 mou yielded qualitatively and quantitatively similar results.Google Scholar

²³ In the estimation results reported, capital services are defined as the sum of expenditure on fertilizer, seed, and feed, plus depreciation and maintenance on fixtures and implements. The results are robust to the use of alternative measures of either capital services or the capital stock.Google Scholar

²⁴ Once family labor has been valued, the daily wage rate, W, can be found by dividing the sum of wages paid in cash and kind to hired labor and the imputed value of family labor by the total number of hired and family-member work days.Google Scholar

²⁵ On the first of these points, see fn. 12.Google Scholar

²⁶ The rationale for using 85 percent is as follows. In the cotton-growing areas of North China, women and children provided slightly less than 30 percent of self-supplied labor. Imputing each day of this labor at the commonly used rate of 70 percent of the value of its adult male counterpart, and assuming an additional 10 percent difference in the value of a labor day supplied by the family and one supplied by hired labor, I arrived at a weight in the vicinity of 0.85.Google Scholar

²⁷ To reply to an inquiry from an anonymous referee, weighting family labor on smaller farms at 70 percent would eliminate most of the differences in reported labor use with larger farms. Because I have no independent basis for identifying those households on which the contribution of women and children exceeded the average, I have decided to use only one weight for all farms.Google Scholar

²⁸ See Zellner, Arnold, “An Efficient Method for Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias,” Journal of the American Statistical Association, 57 (06 1962), pp. 348–68.CrossRefGoogle Scholar

²⁹ This increases my confidence in the tests for absolute price efficiency that are discussed below. In other similar studies the large standard errors of the coefficients on the price variables in the profit equation make it very unlikely that a null hypothesis of absolute price efficiency could ever be rejected. See, for example, Kahn, Mahmoud and Maki, Dennis, “The Effects of Farm Size on Economic Efficiency,” American Journal of Agricultural Economics, 61 (02 1979), pp. 64–69;CrossRefGoogle Scholar and Trosper, Ronald, “American Indian Relative Economic Efficiency,” American Economic Review, 68 (09 1978), pp. 503–16.Google Scholar

³⁰ Once I failed to reject a particular hypothesis, I reestimated Model I with that restriction imposed in order to increase the efficiency of estimation. These estimates are provided in columns I_B and 1_C of Table 10. Their use is explained below.Google Scholar

³¹ In an earlier version of this article, I carried out tests of absolute price efficiency for both groups of farms separately for each input and also tests (discussed below) of equal absolute and equal relative price efficiency. I would like to thank Melvin Fuss of the University of Toronto for demonstrating the need to carry out these tests jointly for the two variable inputs.Google Scholar

³² A type II error is the acceptance of the null hypothesis when the alternative is true.Google Scholar

³³ The results of these tests are in general agreement with earlier work by Myers, Ramon and Dietrich, Scott, “Resource Allocation in Traditional Agriculture: Republican China, 1937–1940,” Journal of Political Economy, 79 (07/08 1971), pp. 887–96;Google Scholar and Wiens, Thomas, “The Microeconomics of Peasant Economy: China, 1920–1940” (Ph.D. dissertation, Harvard University, 1973). Myers and Dietrich found, on the basis of the same data, that the allocation of factors of production among crops was reasonably efficient, while Wiens, using only the data for Michang village, arrived at a similar conclusion.Google Scholar

³⁴ See Yotopoulus and Lau, “A Test for Relative Economic Efficiency,” esp. pp. 217–18.Google Scholar

³⁵ Estimates of the same elasticities can be obtained directly from the production function by ordinary least squares but are generally statistically inconsistent because of the existence of simultaneous equation bias.Google Scholar

³⁶ By comparison, Kyers and Dietrich, “Resource Allocation in Traditional China,” found in the estimation of production functions for each crop with two inputs (land and labor) that in Michang and Matsun, cotton, corn, kaoliang (sorghum), wheat, and millet generally experienced increasing returns to scale, but that in Wukuan, cotton, wheat, and millet all reflect decreasing returns. Their estimation, however, made no adjustments to labor.Google Scholar

³⁷ Some support for this view appears in Dennis Chinn's work. Utilizing Buck's aggregate cross-sectional data, Chinn found, in estimating a Cobb-Douglas production function that included both conventional inputs and indices for land fragmentation and tenancy, that fragmentation was negatively related to output in four of Buck's eight regions, including the two that contain the villages I examined. See his “Land Utilization and Productivity in Prewar Chinese Agriculture: Preconditions for Collectivization,” American Journal of Agricultural Economics, 59 (08 1977), pp. 559–64.CrossRefGoogle Scholar

³⁸ Berry and Cline, Agrarian Structure, address some of the key issues. For a more recent perspective on the subject, see Alston, Lee, Datta, Samar, and Nugent, Jeffrey, “Tenancy Choice in a Competitive Framework with Transactions Costs,” Journal of Political Economy, 92 (12 1984), pp. 1121–33.CrossRefGoogle Scholar

³⁹ Trosper, Ronald, “American Indian Relative Ranching Efficiency,” American Economic Review, 68 (09 1978), pp. 503–16.Google Scholar

⁴⁰ Unlike the earlier tests for the influence of farm size on relative economic efficiency, the present ones can not test separately for the effects of tenancy on price and technical efficiency, but only on overall economic efficiency. With more observations on pure tenants the sample could be divided and tests for relative economic efficiency between tenants and owners carried out along the same lines as before.Google Scholar

⁴¹ These test results can be interpreted as tests for differences in efficiency with any influence of tenancy now controlled for.Google Scholar

⁴² Michelle McAlpin has argued that in India commercialization extended substantial benefits to some of rural society's poorest members and those least well endowed (in terms of land) by increasing the demand for labor relative to land and ultimately the share of output going to labor. See “The Effects of Markets on Rural Income Distribution in 19th Century India,” Explorations in Economic History, 12 (05 1975), pp. 289–302.CrossRefGoogle Scholar

⁴³ The surveys included Michang village of Feng-jun hsien in Hopei province for the three years, 1937, 1938, and 1939; Matsun village of Huo-lu hsien, also in Hopei, for 1939; and Wukuan village of Chang-te hsien in Honan province for 1940.Google Scholar See Minami Manshu Tetsudo Kabushiki Kaisha, Nōka keizai chōsa hōkoku, Hokushi keizai shiryoō no. 5 (Feng-jun hsien, Michang village, 1937);Google ScholarHokushi keizai shiryō no. 12 (Feng-jun hsien, Michang village, 1938);Google ScholarHokushi keizai shiryō no. 36 (Feng-jun hsien, Michang village, 1939);Google ScholarHokushi keizai shiryō no. 25 (Chang-te hsien, Wukuan village, 1940);Google ScholarHokushi keizai shiryō no. 32 (Huo-lu hsien, Matsun village, 1939).Google Scholar

⁴⁴ Full citations are offered in the article, fns. 1, 33.Google Scholar

⁴⁵ On the basis of a sample of over 500,000 households taken by the National Land Commission, the average amount of land cultivated per farm household in North China was 21.5 mou. See National Land Commission, Ch'uan-kuo t'u-ri riao-ch'a pao-kao kang-yao (Nanking, 1937), p. 24.Google Scholar