Six
Land Concentration and Income Distribution in Republican China

Loren Brandt and Barbara Sands

There is reason to think that in the (last) twenty years . . . occupying ownership has lost ground. . . . The fact of concentration can hardly be questioned. . . . Over a large part of China, tenancy of one kind or another undoubtedly predominates and it appears everywhere to be increasing .
—RICHARD TAWNEY, AGRARIAN CHINA (1938)

These remarks by the British historian Richard Tawney typify much of the thinking both in China and in the West about the growing concentration of landholdings in late nineteenth- and early twentieth-century rural China. Perceptios of widening disparities in land ownership in turn underlie a nearly endless list of critical assessments of the changing distribution of incomes and economic welfare during the same period. Carl Riskin, for example, argues: "Chinese rural society was hardly egalitarian with respect to the distribution of income and wealth: and although characterized by substantial social mobility, the predominant direction for the latter was downward, at least in relative terms."^[1] Harry Harding similarly notes: "Though still prone to drought, flood, and plague, China's rural economy could, in most years, also produce the surplus to support one of the world's most advanced urban civilizations and most highly bureaucratic states. But the distribution of that output was increasingly unequal."^[2]

Quite often these concerns of widening inequality are accompanied by observations of increasing population pressure and the lack of technological change; indeed, the received economic story of late nineteenth- and early twentieth-century agrarian China is one of increasing immiseration for the populace in general and a relative worsening for the rural poor. Political events in the first half of the twentieth century involving two major regime

[1] Carl Riskin, China's Political Economy (New York, 1986), p. 32.

[2] See Harry Harding, China's Second Revolution: Reform after Mao (Washington, D.C., 1987), p. 13.

changes have only strengthened these views of China's recent troubled past, if not fired them directly.^[3]

Empirical support for this history is, however, quite limited. This is particularly true for questions concerning the distribution of land ownership and economic welfare. Despite the keen interest in the development literature concerning questions of distribution and their obvious historical relevancy to the case of China, they have been largely neglected.^[4] Moreover, China's distributive record prior to 1949 has rarely been put in any kind of comparative perspective.

The purpose of our paper is to reexamine a number of issues relating to the distribution of land ownership and economic welfare in the pre-1949 economy. Because the historical data on distribution are spotty at best and China itself is so diverse, this paper is largely exploratory, and some of its findings preliminary in nature. Despite these limitations, based on data presently at our disposal, we make the following points.

First, we have failed to find convincing evidence that land ownership became more concentrated during the late nineteenth and early twentieth centuries. Second, even if concentration increased, it cannot be inferred that the increase necessarily heralded increasing immiseration or welfare inequality. A number of alternative processes can underlie increasing concentration in land ownership with opposing interpretations for the behavior of economic welfare. Third, drawing on survey materials for the 1930s, some of them covering all of China and others relating only to a select number of North China villages, we have calculated measures of income distribution that suggest income inequality was much lower in rural early twentieth-century China than has been previously inferred on the basis of data on land distribution alone; in fact, when compared with other low-income countries, China actually appears to be on the moderate side. Moreover, the estimates of income inequality that we have obtained fall within the limits some Chinese officials have recently recommended for their rural sector.

These observations suggest that the current view of China's modern economic history is uncertain at best and yet to be firmly established by contem-

[3] Philip C. C. Huang, in fact, argues that these imbalances underlay the social and political upheavals of the nineteenth and twentieth centuries. See his Peasant Economy and Social Change in North China (Stanford, 1985), p. 293. Exceptions to this conventional wisdom include two books: Thomas G. Rawski, Economic Growth in Prewar China (Berkeley and Los Angeles, 1989), and Loren Brandt, Commercialization and Agricultural Development: Central and Eastern China, 1870–1937 (Cambridge, 1989).

[4] C. Robert Roll's unpublished Ph.D. dissertation (Harvard University, 1974) represents the only empirical investigation of income distribution in pre-1949 rural China that we are aware of. Ramon H. Myers, in The Chinese Peasant Economy: Agricultural Development in Hopei and Shantung, 1890–1949 (Cambridge, Mass., 1970), on the other hand, examines changes in land distribution in North China from the 1890s to the 1940s.

porary quantitative evidence. In the sections below, we take each of the above points in turn and discuss our findings.

Increasing Concentration of Landholdings?

Concerns over a rise in the concentration of landholdings and growing tenancy were voiced in China prior to the early twentieth century. During the late Ming and Qing, for example, government officials regularly expressed alarm over what appeared to be an increasing accumulation of land by the wealthy and a decline in the holdings of smaller farm households. These developments are frequently linked to the commercialization and market development of the period.^[5] Are data consistent with views of a secular rise in land concentration?

Reasonably good data on land ownership exist for the early twentieth century. On the basis of data compiled by the National Land Commission in 1934 on almost 1.75 million rural households residing in 16 provinces, Table 6.1 presents an estimate of the size distribution of landholdings for rural China.^[6] These aggregate data are complemented by many village-level surveys undertaken during the 1920s and 1930s.^[7] Both kinds of data confirm what others have long argued—the distribution of land ownership in rural China was highly unequal—and on the surface are consistent with the view that distribution may have been worsening. For the distribution in Table 6.1, the Gini coefficient, a frequently used measure of inequality, is 0.72.^[8] The top 1 percent of rural households owned approximately 18 percent of the land;

[5] See, for example, Yang Yi, "Qingchao qianqi de tudi zhidu," Shehui yuekan 7 (1958): 21–26, and Fu Yiling, "Guanyu Mingmo Qingchu Zhongguo nongcun shehui guanxi de xin guji," Xiamen daxue xuebao 6 (1959): 57–70.

[6] These data suffer from a number of omissions that have been overlooked. These problems are discussed in more detail in Brandt, Commercialization and Agricultural Development , chap. 6.

[7] The results of 47 surveys carried out between 1931 and 1941 are reported in Chao Kang [Zhao Gang] and Chen Chung-yi [Chen Zhongyi], Zhongguo tudi zhidu shi (Taibei, 1982), pp. 234–38. A majority of these are the product of the research efforts of the South Manchurian Railway Company and can be found in Minami Manshu Tetsudo Kabushiki Kaisha, Mantetsu chosa geppo .

[8] The Gini coefficient is equal to 1 minus 2 times the area under the Lorenz curve. The Lorenz curve, based on an arrangement of households in ascending order of income, relates the cumulative proportion of households to the cumulative proportion of income received. The Gini coefficient can assume a value between 0 and 1 with a measure of 0 implying a uniform distribution of incomes (or assets) across the population and the measure 1 implying extreme inequality. It can also be interpreted as the expected gain—measured as a percentage of the mean for the population—of a lottery in which each person is allowed to draw an income at random from the population and compare it with his or her own. Thus, when incomes are uniformly distributed, the expected gain is 0, as would be the Gini coefficient. The more concentrated incomes are, the larger the expected gain will be. See Graham Pyatt, "On the Interpretation and Disaggregation of Gini Coefficients," Economic Journal 86 (June 1976): 243–55.

the top 5 percent owned nearly 39 percent; and the top 10 percent owned around 53 percent. By contrast, slightly more than a quarter of all rural households were landless, and an additional quarter owned less than 5 mu . Similar percentages are suggested by the village-level data.

The problem is that comparable information on the previous distribution of landholdings, be it 25 years earlier or 200, presently does not exist. For the Qing we are limited to a handful of observations at the county and village level that simply are too few to generalize about. Table 6.2, nonetheless, documents one such case: the distribution of landholdings in the early eighteenth century for several subdivisions in Huailu County, Hebei, and village-level data in the same county 200 years later.^[9] At least for this North China village the distributions are very similar, though, not unexpectedly, average farm size in 1939 was slightly lower than in either 1706 or 1736.

While data on landholdings are also seriously deficient for the late nineteenth century, additional information on such things as the percentage of land that was rented out and the hiring of long-term labor, either or both of which typically accompany an increasing concentration of landholdings,

[9] Additional pre-twentieth-century data can be found in Kang Chao, Man and Land in Chinese History (Stanford, 1986), chap. 6.

suggests no long-term rise in the half-century between the 1880s and 1930s. Estimates of land rental, for example, reveal that in both periods approximately a third of cultivated acreage was rented, two thirds of which was owned by absentee landlords.^[10] We have not uncovered estimates of the number of long-term agricultural laborers for the 1880s—though qualitative evidence suggests their numbers were substantial—but of the 1.75 million rural households surveyed by the National Land Commission in the 1930s, only 1.57 percent were classified as landless, laboring households, that is, households earning a living as long-term contract agricultural laborers.^[11] Finally, over the 25-year period between 1912 and 1937, the percentage of households classified as owner-operators failed to decrease, as we might expect had tenancy been on the rise.^[12]

On the basis of these very limited data we do not want to push any view very far, especially one suggesting that the degree of concentration of landholdings remained constant for several hundred years. As we will explain

[10] Chung-li Chang has estimated that about a third of all land was rented in the 1880s, or roughly the same percentage as in the 1930s. See Chung-li Chang, The Income of the Chinese Gentry (Seattle, 1962), p. 145.

[11] Tudi Weiyuanhui, Quanguo tudi diaocha baogao gangyao (Nanjing, 1937), p. 34.

[12] Nongqing baogao 5, no. 12 (Dec. 15, 1937): 330.

below, over the long run there were a number of alternative processes simultaneously influencing land concentration; and indeed, it would be odd if these processes offset one another exactly. Nonetheless, the point needs to be made that data presently at out disposal simply cannot support the popularly held view that the high degree of concentration of landholdings observed in the 1930s was the product of several decades (or centuries) of increasing inequality.

Finally, although in the 1930s 10 percent of the population owned over half the farmland in China, can this be considered extreme compared to landholding patterns in other countries? Interestingly, data compiled by Peter H. Lindert suggest that it was not extreme. Concentration in China is markedly less than in Mexico in the 1920s, where 10 percent of the households owned an estimated 64–99 percent of the land and 79 percent were landless, or in Victorian England (excluding London), where 85 percent of all households owned no land and 10 percent owned 82 percent of the land; and it is actually slightly better than that estimated separately for farming households and the entire rural population (whites only) in the United States—certainly not noted for its extreme concentration—in both 1798 and 1860.^[13] In many respects, land distribution in China in the 1930s is very similar to that estimated for post-independence India.^[14]

Historical Processes

From the perspective of many observers, land sales by indebted smallholders were the major underlying cause for the increasing concentration of landholdings. Farm households borrowed for a variety of reasons: to obtain working capital, to meet consumption needs in times of emergency, and in some cases to help cover major ceremonial expenses, such as weddings and funerals. To secure these loans, land was frequently pledged as collateral. In still other cases, the rights of land use were ceded to the lender, with the income generated from the land serving as the interest. Once the loan was repaid, rights of land use reverted back to the original owner. This land was commonly referred to as diandi .

In many cases, borrowers were unable to service these loans. Han-seng Chen's description is typical: "When a poor peasant in China mortgages his bit of land, he has practically no hope of ever getting it back. Everything conspires against him in his frantic effort to meet the interest charges, and eventually he loses not only the land but also this additional fruit of his

[13] Peter Lindert, "Who Owned Victorian England? The Debate over Landed Wealth and Inequality," Agricultural History 61, no. 4 (1987): 25–51.

[14] I. Z. Bhatty, "Inequality and Poverty in Rural India," in Poverty and Income Distribution in India , ed. T. N. Srinivasan and P. K. Bardhan (Calcutta, 1974).

labor."^[15] These difficulties were invariably attributed to declining economic fortunes caused by high taxes and land rents, local disaster, depressed agricultural conditions, and so forth. If land was used as collateral, it might simply be forfeited; if an asset other than land secured the loan, land still might have to be sold to make repayment. If land-use rights were ceded to the lender, the possibility of redeeming the land in the near future became even more remote.^[16] With the loss of their land, small cultivators were invariably forced either to rent back the same land from the new owners or possibly to hire out as wage laborers. This explains the link implicit in the opening quote from Tawney between rising tenancy (or perhaps wage labor) and a growing concentration of landholdings.

Although some small farm households were unquestionably forced to relinquish ownership of their land during times of economic duress, this analysis suffers from several weaknesses; moreover, there are a number of equally compelling explanations for increasing concentration of landholdings and a rise in tenancy with alternative implications for the secular behavior of incomes.^[17]

For example, in asserting that the loss of land through debt-sales gives rise to an increasing concentration of land ownership, there is an implicit assumption of who the borrowers and lenders were, the lenders typically being larger landowners/landlords, frequently absentee, the borrowers, on the other hand, being primarily poorer, small farm households that had earlier exhausted all personal financial resources. Although larger landowners were usually heavily involved in the credit market, a detailed examination of land and credit arrangements in three diverse villages in North China reveals that lending activity was more diversified among households than previously believed; in other words, lending activity was not limited to a class of big landowners. In fact, in a majority of those cases involving diandi , farm households with less than 30 mu were the creditors. The borrowers were frequently larger landowners themselves in need of capital.^[18] Data compiled by John Lossing Buck on lending activity in 151 localities in 16 provinces are very

[15] Han-seng Chen, Landlord and Peasant in China (New York, 1936), p. 95.

[16] In some cases, the rights of land use were actually rented back to the original owner. According to Madeleine Zelin, failure to fulfill the rental agreement could mean the total alienation of the original owner from the land. See Zelin, "The Rights of Tenants in Mid-Qing Sichuan," Journal of Asian Studies 45, no. 3 (May 1986): 499–526.

[17] Please recall, however, our earlier remarks regarding the uncertainty over the inter-temporal behavior of land distribution.

[18] Based on an examination of contracts involving diandi in the surveys cited in n. 33. In the villages we examined, the amount of land that villagers had obtained the farming rights to through diandi exceeded the amount of land that they had ceded the rights to . The likely explanation for this is that some of the former was owned by absentee landlords or perhaps village members who had migrated yet wanted to retain some claim to village resources should they decide to return.

consistent with the village-level data for North China: for only 8 percent of all loans were landlords and other wealthy households cited as the source of credit. Much more important were relatives and neighbors.^[19]

Observations such as these make generalizations about the relationship between forced land sales and increasing concentration difficult without much more careful examination of credit and land markets in rural China. Moreover, even if it is true that land ownership was increasingly concentrated, this does not necessarily mean that income became more concentrated as well. John R. Shepherd has recently pointed out that there are alternative historical processes capable of simultaneously generating an increasing concentration of landholdings and a rise in tenancy which do not necessarily imply increasing immiseration of the rural poor.^[20] These include an expansion in cultivated area through investment by wealthy households in land reclamation, migration of landless households into such areas because of better economic opportunities, and shifts in forms of landlord farm management. These processes could operate during periods of economic growth and expansion just as well as in times of economic decline. More generally, a rise in real wages and an increase in labor's share of national income could offset the influence on income distribution of any increased concentration of landholdings.

Each process can be adduced for many parts of China. Between the seventeenth and nineteenth centuries, for example, the increase in the demand for rice in the Lower Yangzi drew the capital of wealthy landowners into land reclamation projects in the Dongting Lake area of Hunan. Immigrant households supplied much of the labor for reclaiming and later for tenanting this land.^[21] This process contributed to a relatively high degree of land concentration and tenancy in the area, which persisted into the 1930s.^[22] Similar kinds of investments occurred in the Jiangnan area and in some of the more commercialized areas of the North China Plain. During the last half of the Qing, cultivated area increased by roughly half, much of it through investments such as these.^[23]

[19] John L. Buck, Land Utilization in China: Statistical Volume (Chicago, 1937), p. 404. If the loans extended by landlords and other wealthy households were larger than average, they may have been more important than their percentage of the total number of loans otherwise suggests.

[20] John R. Shepherd, "Rethinking Tenancy: Spatial and Temporal Variation in Land Ownership Concentration in Late Imperial and Republican China," Comparative Studies in Society and History 30.3 (July 1988): 403–31.

[21] See Peter C. Perdue, Exhausting the Earth (Cambridge, Mass., 1987).

[22] According to the National Land Commission survey, 47.8 percent of cultivated area in Hunan was rented. The high degree of land rental and estimates of the concentration of cultivated holdings provided in Table 6.3 suggest a highly concentrated land ownership.

[23] By the early twentieth century, if not before, much of the potential for expanding cultivated area in China proper (excluding Manchuria) had been exhausted. In eastern Jiangsu, however, an estimated 5 million to 6 million mu of land in abandoned salt farms (or a 25 percent increase over existing cultivated area in eastern Jiangsu) was reclaimed for agricultural use. In these counties we typically find a relatively high degree of tenancy. Information on tenancy and the landholdings of land reclamation companies in the area can be found in the Ministry of Industry survey of Jiangsu for 1933. See Zhongguo Shiyebu, Guoji Maoyiju, Zhongguo shiye zhi: Jiangsu sheng (Shanghai, 1933), pp. 231–32.

There was also migration, most notably in the eighteenth century into Sichuan, in the mid-to-late nineteenth century into highly fertile areas of the Lower Yangzi depopulated by the Taiping Rebellion, and in the twentieth century into Manchuria.^[24] In each case, individuals or entire households moved into these areas to take advantage of better economic opportunities. Even without land exchanging hands, migration under these conditions can lead to an increasing concentration of landholdings (because of an increase in the number of landless households) and a rise in tenancy in the area experiencing the in-migration.^[25] In fact, one observes a higher degree of concentration of land ownership in Manchuria for the early twentieth century than in other parts of China.^[26]

The historical role of these alternative processes throughout much of China points to the shortcomings in many earlier interpretations of land concentration and tenancy in the early twentieth century. The concentration of landholdings cannot simply be viewed as the cumulative product of a single process. Rather, the degree of concentration observed in a locality at any given time needs to be seen as the legacy of several, possibly interrelated, economic/historical processes including migration, reclamation, and commercialization as well as forced land sales that may or may not have been weakened over time by other forces operating on the rural economy.^[27]

[24] On the migration to Manchuria, see Thomas R. Gottschang, "Economic Change, Disasters, and Migration: The Historical Case of Manchuria," Economic Development and Cultural Change 35.3 (Apr. 1987): 461–90.

[25] A possible reduction in concentration in the area experiencing the out-migration may have kept inequality, in the aggregate, from worsening.

[26] For a sample of villages settled after the turn of the century, Myers has found an average Gini coefficient for landholdings equal to 0.86. This is 15–20 percent higher than the average for the 47 villages cited in n. 7. See Ramon H. Myers, "Socioeconomic Change in Villages of Manchuria during the Qing and Republican Periods: Some Preliminary Findings," Modern Asian Studies 10, no. 4 (1976): 591–620. Interestingly, the Gini coefficient for villages settled before the turn of the century is not markedly less than for those settled afterward, thus suggesting the legacy of the initial pattern of settlement.

[27] One of these other forces we have in mind here is family division. Since Adam Smith, there has been a presumption that over the course of several generations family division helps to reduce the inequality of landholdings. Lavely and Wong argue that this need not be the case and demonstrate how the impact of partible succession on land concentration is influenced by both inter- and intra-landclass differences in reproductivity. See William Lavely and R. Bin Wong, "Family Division, Reproductivity, and Landholding in North China," Research Report no. 84–65, Population Studies Center, University of Michigan, Nov. 1984.

Land and Income Distribution in Rural China

Data on land distribution have long served as the key indicator of the degree of economic welfare inequality in the rural sector. Yet how accurate are landholdings as such a measure? There exist data sources from the 1930s that allow estimates of the degree of dispersion of both land and income in the rural sector. These include the survey of the National Land Commission used earlier that provides provincial-level data, and village-level surveys carried out by the South Manchurian Railway Company in North China. Although these data sources differ in a number of key respects and the latter are confined to North China, together they provide new insights into the relationship between the degree of concentration of incomes and land ownership in China's rural sector.

In estimating the degree of inequality in an economy, several important methodological questions immediately arise. These are related to the selection of the measure of economic well-being and the unit of observation (individual, family, income earner, etc.).^[28] We have selected income as our measure of well-being, although other indicators of welfare, such as consumption or expenditures, could also be used and might actually be preferable. All the income data are provided at the household level, but rather than rank households by total household income as has frequently been done, we have elected to rank them by per capita household income. In this regard, we are following Simon Kuznets, who concludes in a comparison of the two methods:

It makes little sense to talk about inequality in the distribution of income among families or households by income per family or household when the underlying units differ so much. . . . Before any analysis can be undertaken, size distributions of families or households by income per family must be converted to distributions of persons (or consumer equivalents) by size of family or household income per person (or per consumer).^[29]

Kuznets's remarks would seem to have a particular relevancy for rural China, where the range of household sizes was enormous. With many households consisting of one or two people and others including up to 20, the potential for differences in measures of inequality based on household rather than individual income or wealth is obviously large.

Before reporting provincial-level estimates of the degree of income in-

[28] These issues are discussed in R. Albert Berry, "Evidence on Relationships among Alternative Measures of Concentration: A Tool for Analysis of LDC Inequality," Review of Income and Wealth 33.4 (Dec. 1987): 417–30.

[29] Simon Kuznets, "Demographic Aspects of the Size Distribution of Income: An Exploratory Essay," Economic Development and Cultural Change 25, no. 1 (Oct. 1976): 87. See also Gautam Datta and Jacob Meerman, "Household Income or Household Income per Capita in Welfare Comparisons," Review of Income and Wealth 26, no. 4 (Dec. 1980): 401–18.

equality, we first look at the concentration of landholdings reported in the National Land Commission survey. Unfortunately, the summary report of the National Land Commission only provides a distribution of landholdings for the country as a whole (see Table 6.1). Provincial-level information, however, is provided on the distribution of operational holdings, the sum of owned and rented land that a household cultivates. The survey reports the number of households in each of 13 size categories. We have modified these data slightly to include households that did not farm, that is, had no operational holdings, and then calculated Gini coefficients for the new distributions.^[30] These estimates are provided in Table 6.3.

It is widely agreed that land rental increased access to land; thus, the distribution of operational holdings should be more equal than the distribu-

[30] Most computational formulas for the Gini coefficient based on grouped data such as these underestimate the degree of inequality because they ignore intragroup inequality. We have followed Kakwani in correcting for this, using the midpoint of each interval as an estimate of its mean. See Nanak Kakwani, "On the Estimation of Income Inequality from Grouped Observations," Review of Economic Studies 43, no. 3 (Oct. 1976): 483–92.

tion of land ownership. This is borne out by the data: the Gini coefficient for landholdings is 0.72, while that for cultivated holdings is 0.62. For a majority of provinces the Gini coefficient for operational holdings is between 0.53 and 0.60.

In general, we expect the difference between the Gini coefficient for owned and cultivated holdings to be positively correlated with the percentage of land rented; in other words, the more land that is rented out to tenants, the lower the concentration of operational holdings associated with any given level of inequality of land ownership. Since the Gini coefficient for operational holdings for the provinces of North China and the Yangzi Valley are almost the same, the greater prevalence of tenancy in the Yangzi Valley than in North China would suggest that land ownership was more widely distributed in North China than in the Yangzi Valley.^[31] The Gini coefficient for landholdings in the Yangzi was probably in excess of 0.75, while that for North China was nearer to 0.65.

But what about incomes? Were they just as concentrated as land ownership? Again drawing on data compiled by the National Land Commission, we have calculated Gini coefficients for household incomes in each of the 16 provinces. These estimates also appear in Table 6.3. Without exception, they reveal a degree of income inequality much lower than that suggested by the land data alone. For a majority of the provinces the Gini coefficient for household income was in the vicinity of 0.40–0.45, while the Gini coefficient for the entire sample is 0.46.^[32] These estimates are 35 percent to 40 percent lower than what the provincial Gini coefficients for landholdings probably are. Geographically, the degree of income dispersion appears to be slightly below the national average in the more densely populated areas of the Yangzi and South China and modestly higher in the North and Northwest.

Data from North China

In a number of respects these estimates are not entirely satisfactory. On the one hand, it is not exactly clear from the summary report which measure of income was used: net or gross, cash or total (which includes cash income plus income in kind). C. Robert Roll has argued that households probably re-

[31] Approximately 45 percent of land was rented out in the Yangzi region, as compared to 15–20 percent in North China. These estimates are based on Tudi Weiyuanhui, Quanguo tudi diaocha , pp. 26–27.

[32] In calculating the Gini coefficient for the entire sample, we have pooled the data by taking a weighted average of the percentage of households in each province in each size category, using as weights the percentage of the total rural population in each province. Rural population estimates are based on provincial population estimates contained in Dwight H. Perkins, Agricultural Development in China, 1368–1968 (Chicago, 1969), p. 212, and estimates of the percentage of population classified as rural are provided in Wu Baosan, Zhongguo guomin suode, 1933 nian (Shanghai, 1947), 1:151.

ported their total net income, yet the difficulty of accurately estimating all in-kind income and production expenditure cannot be underestimated. On the other hand, if household incomes and family size are systematically related, the distribution of household income may provide a misleading indicator of income distribution on a per capita basis. From a welfare perspective, it is the latter distribution with which we are concerned. Again, we need to know how the distributions for household income and household income on a per capital basis are related to each other.

Three village surveys undertaken by the South Manchurian Railway Company in 1936 in Hebei Province provide important information on each of these questions.^[33] These surveys are unique because they enumerate every household in the village and give information on both in-kind and cash incomes and expenditures, thus allowing comprehensive estimates of household incomes.^[34] The village data allow us to compare the degree of inequality in the distribution of assets and income by households and in per capita terms. The difference between these two measures of village inequality can be combined with the information in Table 6.3 to make a tentative estimate of the degree of income inequality in rural China at the household level on a per capita basis, which can then be compared with estimates for other low-income countries. These surveys also provide important clues as to why the concentration of incomes observed in Table 6.3 was so much lower than that suggested by the data on land ownership alone.

The three villages selected by the South Manchurian Railway Company for detailed examination include Michang Village in Fengrun County, Dabeiguan in Pinggu County, and Qianlianggezhuang (Lianggezhuang for short) in Changli County. All three were located in northeastern Hebei within 100 kilometers of each other and were less than several hundred kilometers from Tianjin. Despite some obvious similarities, there were marked differences between the villages, some of which are captured in Table 6.4.

Dabeiguan was a relatively poor, only moderately commercialized farming village that was first settled in the early Ming dynasty (1368–1644). At the time of the survey it was made up of 98 households and had a total population of 601. Much of its acreage was devoted to food crop production

[33] See Minami Manshu Tetsudo Kabushiki Kaisha, Kito Noson Jittai Chosahan, Dainiji kito noson jittai chosa hokokuso: tokeihen ; Dai ichiban: Heikoku ken ; Dai sanban: Hojun ; Dai yonban: Shorei ken (Dairen, 1937). These surveys are part of an enormous collection of materials on social and economic life in China compiled by the South Manchurian Railway Company during the late 1930s and early 1940s. For an introduction to these materials, see Philip C. C. Huang, Peasant Economy .

[34] An earlier estimate of the degree of income inequality in a North China village made by Marc Blecher suffers in both regards because it is based on a stratified sample in which larger farms are overrepresented and because household income is measured simply by the gross agricultural output of each household. See Marc Blecher, "Income Distribution in Small Rural Chinese Communities," China Quarterly 68 (Dec. 1976): 795–816.

for village consumption, although approximately 15 percent of the total cultivated area was used to grow cotton and sesame primarily for market sale. Slightly less than 15 percent of agricultural output was marketed. Animal husbandry represented an important, but perhaps declining, sideline for some households.^[35]

Every household in the village was engaged in agriculture, and all but five actually owned some land. Farms averaged almost 25 mu , or 1.67 hectares (15 mu equals 1 hectare). Only five percent of the land in the village, however, exchanged hands through the rental market. In contrast, the labor market

[35] According to the preface to the survey, population pressure had reportedly begun to crowd out animal husbandry. Some households (the poorer ones) were forced to abandon animal husbandry altogether, whereas others switched to rearing smaller animals.

was very active. Villagers employed substantial amounts of labor from inside and outside the village on an annual and seasonal basis. Villagers also worked outside the village primarily on a daily basis. In this regard, the labor market appears to have been the primary mechanism through which households offset imbalances in resource endowments. On the whole, Dabeiguan was a small net exporter of labor. In all likelihood because of more lucrative opportunities outside the village, Daibeguan was, at the same time, a net lender of capital.

Michang, in contrast, typifies the richer, more highly commercialized villages of North China. It was made up of 114 households and had a total population of 753. Because of its rail links to the region's major urban centers, agriculture had become highly specialized in the course of the early twentieth century. In the process, farm households replaced the cultivation of basic foodgrains, like sorghum and millet, with cash crops, such as cotton, and purchased large amounts of grain from outside sources. By the mid-1930s more than a third of the cultivated area was in cotton, and almost half of agricultural output was marketed. Commercial fertilizers also began to be widely used and represented a significant expenditure item for almost all farming households.

Although agriculture was the primary source of income in Michang, 15 households were engaged solely in nonagricultural activities. An additional 22 households were engaged in nonfarm sidelines. These activities included peddling, carpentry, food processing, and local government service. Three households reported no income and were classified by surveyors as "beggar households." A total of 19 village members had also migrated (primarily to Tangshan and Tianjin, and mostly as shopkeepers) and, through remittances, represented a key source of income for some households.

Cash-cropping and the commercialization of the local economy substantially increased productivity and the demand for labor.^[36] Only a portion of this increase, however, was accommodated by an increase in labor supply from within the village, and Michang looked increasingly outside its boundaries for labor. In the 1930s, Michang was a net importer of labor. This was complemented by a very active land rental market, in which almost a third of all land exchanged hands. Yet most of this exchange was due to villagers cultivating land owned by village outsiders (absentee landlords). Within Michang, households cultivated most of the land they owned. Unlike Debeiguan, Michang was a net borrower of capital.

Situated on the Bei-Ning Railroad, Lianggezhuang was also a highly commercialized village, although more closely tied to Manchuria than to North

[36] The annual labor requirement for cotton, for example, was 12 days per mu , or double the requirement for grain crops. The demand for marketing and auxiliary services also increased labor demand.

China. With 101 families and a total population of 574, it was slightly smaller than either Michang or Dabeiguan. Through the first three decades of the twentieth century, Manchuria had provided a rapidly expanding market for farm products and employment for some village members. Lianggezhuang's chief agricultural export was pears. Much of the profit from these exports was reportedly reinvested in land, and by the 1930s almost one third of total acreage was in pear orchards.

With the establishment of the puppet state of Manchoukuo in 1932, however, exports and seasonal migration were sharply curtailed. These difficulties were further compounded by a series of poor harvests. In 1936 pear output contributed only one-seventh to the gross value of agricultural output despite representing almost a third of cultivated area. This had profound ramifications for the labor market, for this one crop had formerly absorbed over one half of agricultural household labor and had drawn in much labor from outside the village.^[37] As these opportunities dried up, many villagers now sought short-term work in the nearby prefectural capital, and Lianggezhuang actually became a net exporter of labor. Reflecting these difficulties, in 1936 agriculture contributed only 58 percent of total gross village income.

What can we say about the distribution of land and income in these three villages, and how does it compare to the aggregate estimates for Hebei? Table 6.5 reports Gini coefficients for asset holdings (land, draft animals, and labor power) and incomes, both on a household and household per capita basis for each of the three villages. Mean values of each variable are reported in parentheses.^[38] In Michang and Lianggezhuang average landholdings were smaller, and land ownership much more concentrated, than in Dabeiguan. The same is also true of household ownership of draft animals. In the two more highly commercialized villages the degree of concentration of landholdings was 20–25 percent higher than in Dabeiguan. The average Gini coefficient for the three villages was 0.70, only slightly higher than the average (0.65) for 32 villages also surveyed in Hebei in 1936.^[39] In all three villages, however, land rental helped to equalize access to land among households; this effect is most pronounced in Michang and Lianggezhuang, where a third and a half of the land were leased to tenants, respectively. In Michang, for example, the Gini coefficient for landholdings was 0.75, while that for operational holdings was 0.55; in Lianggezhuang the coefficients were 0.75 and 0.53, respectively. The mean operational holding in Michang was also much larger than the mean landholding because of the substantial

[37] This was the opinion of the surveyors as expressed in the preface to the survey.

[38] Here, and throughout this paper, we apply the term "per capita" to data on landholdings, incomes and other variables that we obtain by dividing household figures for these variables by the number of persons per household.

[39] See Mantetsu chosa geppo 18, no. 1 (1939): 39–73 and 18, no. 4 (1939): 21–31.

holdings of absentee landlords. In Lianggezhuang the situation was the exact opposite, because a quarter of the land was rented out to nonvillagers, who were not included in the survey.

Table 6.5 also shows Gini coefficients for household income and household income on a per capita basis.^[40] The Gini coefficients for household income are very consistent with the estimate suggested by the National Land Commission data for Hebei. As reported in Table 6.3, the Gini coefficient for the distribution of rural incomes in Hebei was 0.46. We also observe that for two of the three villages net household incomes were much more evenly distributed than cash incomes.

Yet for reasons spelled out above, it is the distribution of household income on a per capita basis that concerns us. And here we observe in every case a much lower degree of inequality on a per capita basis than on a household basis: Michang, 0.488 versus 0.391; Lianggezhuang, 0.409 versus 0.346;

[40] The details of the income calculations are included in a lengthy appendix available from either author on request.

and Dabeiguan, 0.436 versus 0.349. The average difference between the Gini coefficients is 0.08. If this relationship held throughout the province, it would suggest a Gini coefficient for per capita rural household incomes in Hebei of 0.38, that is, 0.46 – 0.08. We also note that when measured on a per capita basis, the differences in the degree of income inequality between the villages are very small. Moreover, although overall inequality was slightly higher in Michang, average per capita income was also more than a fifth higher than in either of the other villages. With differences in consumption expenditure per capita even smaller, much of the higher income in Michang obviously took the form of higher savings.

Several questions immediately arise: (1) why is the inequality of incomes on a per capita basis much lower than the inequality at the household level? (2) why are household incomes distributed so much more equitably than landholdings? and (3) what diluted the influence of the much higher inequality in landholdings in Michang and Lianggezhuang and produced a dispersion in incomes similar to that observed for Dabeiguan?

Until the recent work of Kuznets, the influence of the size distribution of households on income disparities was neglected. Most estimates of income inequality were calculated using income per household. Since households differ in size, this can provide a very misleading indicator of the degree of dispersion of incomes on a per capita basis. In general, family size and family incomes are highly positively correlated. In the three villages we examined, the households with the highest incomes are usually those with ten or more members. The most obvious reason for this is that larger families typically have more members of working age who can add to the household's income. In China these households were typically extended, multigenerational households that had not recently divided.

If incomes increase more than proportionally with household size, the dispersion of incomes on a per capita basis will actually exceed their dispersion on a per household basis. But this is typically not the case; rather, incomes increase less than proportionally, and family size and per capita incomes are negatively associated. Indeed, some of the poorest households on a per capita basis in these villages are among those identified as having the highest total household incomes. Although it need not always be the case, under such conditions the dispersion of household incomes on a per capita basis can be less than that for total household incomes. This is in fact what we observe for all three villages.^[41]

[41] The tendency for the dispersion of household incomes on a per capita basis to be lower than their dispersion on a household basis appears to be much more common for developing than for developed countries. For a sample of six developing countries (Colombia, India, Malaysia, Nepal, Sri Lanka, and Taiwan), the average difference between the Gini coefficient for household incomes or expenditures and the same measure on a per capita basis was 0.05. The largest difference was observed for Malaysia, where the Gini coefficient for households ranked by incomes was 0.52, whereas that for households ranked on a per capita basis was 0.43. The data also suggest a slightly larger difference between the two measures in rural than in urban areas. These estimates are taken from Pravin Visaria, "Demographic Factors and the Distribution of Income: Some Issues," in International Union for the Scientific Study of Population, Economic and Demographic Change: Issue for the 1980's (Helsinki, 1979), 1:289–320; and Berry, "Evidence on Relationships." On the relationship more generally between household size and income distribution, see Simon Kuznets, "Size of Households and Income Disparities," Research in Population Economics 3 (1981): 1–40.

The second and third questions concerning the link between land ownership and income earnings benefit from a brief examination of what we mean by income. Conceptually, income can be thought of as a stream of earnings generated by all factors of production owned by an individual or household (current technology defining what is and is not a factor of production). In a primarily agrarian economy, the key factors of production are land, labor, and capital (the last including such things as farm implements and draft animals). Letting H, L , and K represent these three factors, and r, w , and t their returns, for household i , income, Y , is figured by the equation

where Hⁱ , Kⁱ , and Lⁱ denote household i 's ownership of land, labor, and capital. A household's income, therefore, is positively related to how much of each factor it owns and to the return to each factor.

Yet where do r, w , and t come from? The returns to the various factors of production depend on the nature of the rural economy—most importantly, existing production technology, factor proportions, and the workings of markets or exchange. If all households had access to the same production technology, but each household simply used the resources that it possessed, rⁱ , wⁱ , and tⁱ would differ among households, with the return to each factor negatively correlated with its relative availability within the household. Exchange alters all this. Allowing households to buy and sell the services of the factors of production frees them from their particular endowments and allows them access to all factors of production within the village. Exchange effectively coordinates all households' demands for and supplies of all factors of production, so that the return to each factor is determined by its relative availability within the village rather than within the household.^[42] That is, r, w , and t would be the same across households.

Looking at incomes from this perspective, it becomes clear why many observers thought income distribution within China was highly unequal. For a majority of rural households agriculture was the primary source of income, only modestly supplemented by income from subsidiary activities, such as

[42] If factors are mobile, their returns will depend on relative availability within an even wider marketing area, not just availability within the village.

handicrafts. Thus, household incomes in the rural sector were systematically tied to income from agriculture and land ownership. At the same time, farming in China was highly labor-intensive; the land-labor ratio, in fact, was one of the lowest in the world. Modern inputs into agriculture had yet to be introduced in any quantity. As a result, the marginal productivity of labor, the level of rural wages, and the percentage of output captured by labor tended to be low relative to the return to land. A direct result of the limited off-farm opportunities, low relative returns to labor, and high concentration of land ownership was an unequal distribution of rural incomes.^[43]

Yet perceptions were that income inequality was becoming more severe. Why? On the one hand, both Chinese and Western observers believed that land ownership had become much more concentrated; on the other hand, they observed a falling land-labor ratio and associated with that a marked decline in wages relative to land rents.^[44] This fall was only exacerbated by the imperfectly working factor markets, which prevented households from rationally redistributing factors among themselves.^[45] A concurrent decline in off-farm employment opportunities, which small farm households disproportionately relied on, was also perceived to have occurred.^[46] What all this seemed to mean was that landowners, who were becoming an even smaller minority, appeared to be capturing an increasing percentage of the net product of the rural sector.

[44] Estimates by Dwight H. Perkins suggest that the land-labor ratio in China proper declined by roughly 15 percent between the 1890s and the 1930s. See Agricultural Development in China , pp. 212 and 236.

[45] See Philip C. C. Huang, Peasant Economy , esp. chaps. 4–12.

[46] See, for example, Fan Baichuan, "Zhongguo shougongye zai waiguo zibenzhuyi qinruhou de zaoyu he mingyun," Lishi yanjiu 3 (1962): 88–115.

The evidence on land and income distribution presented earlier belies much of this story. To see why, recall that income distribution depends not only on the distribution of land ownership but also on the distribution of the other factors of production in the rural economy, most importantly, labor power and, to a lesser extent, draft animals and other forms of capital. It also depends on the returns to these factors and on the correlation of factor ownership across households. A high correlation of landholdings with ownership of other factors of production would only reinforce the influence of a high concentration of landholdings on incomes. The distributive impact of a high degree of concentration in land ownership could, however, be partly or entirely offset by the effects of more even distribution of ownership for other factors of production, by a rise in the share of total income earned by other factors, or by a weak or inverse correlation between ownership of different productive factors by different households.

In Table 6.5 we observe that labor power (and therefore income from labor) was indeed much more evenly distributed among households than was land. The Gini coefficients for labor power per household were 0.33, 0.32, and 0.24, for Michang, Lianggezhuang, and Dabeiguan, respectively. Draft animal ownership, on the other hand, was about as unequally distributed as landholdings in the three villages.

Yet the strength of this influence clearly depends on the portion of output that labor captures. If labor power is evenly distributed among households, but the portion of output going to labor is relatively small, the equalizing effect would itself be modest. In rural North China, land captured about half of agricultural output, labor a third, and the remainder represented the return to draft animals and other forms of farm capital.^[47] These estimates are misleading, however, because agriculture was not the only source of income. In all three villages there were a variety of sidelines that offered subsidiary sources of incomes that primarily represented returns to labor. Our calculations, based on the three village surveys, show that land rents (including rents attributed to owner-operated farm land as well as cash or in-kind payments by tenants to landlords) amounted to about a third of total village income, while labor incomes accounted for more than half of all income.

Intervillage differences along these same lines helped to diminish the influence of the much higher concentration of landholdings on incomes in Michang and Lianggezhuang. In Michang, land rents averaged between 4 and 5 yuan per mu , while a long-term laborer could expect to earn 45–50 yuan in cash, plus an in-kind component of perhaps equal value. In Dabeiguan, by contrast, land rents were modestly less, at 3–3.50 yuan per mu , but wages paid long-term agricultural laborers were only about half the rate in Michang. In

[47] These estimates are taken from Loren Brandt, "Farm Household Behavior, Factor Markets, and the Distributive Consequences of Commercialization in Early Twentieth-Century China," Journal of Economic History 47.3 (September 1987): 731.

Lianggezhuang, land rents averaged even less, between 2 and 2.50 yuan per mu , while wages (cash component only) paid long-term agricultural laborers were 30–35 yuan . The same relationship is reflected in output shares; the share of agricultural output going to land in Michang and Lianggezhuang was roughly 40 percent, while in Dabeiguan it exceeded 50 percent.^[48] The significantly higher return to labor relative to land in both Michang and Lianggezhuang helped to reduce the influence of a more concentrated land ownership in these villages and produce a degree of income inequality not much different from that in Dabeiguan.

Nonagricultural activities and remittances played a similar role. In both Michang and Lianggezhuang, 40 households were engaged in nonagricultural activities either as a sideline or on a full-time basis.^[49] From both villages there was also substantial out-migration—37 household members from Michang, and 19 from Lianggezhuang had migrated—and the out-migrants in turn typically remitted a healthy portion of their earnings back to the villages. The higher the percentage of income generated from nonagricultural sources and received in the form of remittances, the weaker the influence of land concentration on the overall dispersion of incomes. In Lianggezhuang, nonagricultural activities and remittances produced 42 percent of gross household income, while in Michang they amounted to 17.4 percent; in Dabeiguan they totaled only 6.1 percent.

Finally, Table 6.6 reveals that the correlation between land ownership and the other factors of production was also much weaker in Michang and Lianggezhuang than in Dabeiguan. The same is also true for the relationship between labor power and draft animal ownership. While factor ownership was positively correlated, the probability that a household with substantial landholdings simultaneously had more than average labor power and draft animals was significantly lower in either Michang or Lianggezhuang than in Dabeiguan. This partially reflects the higher degree of specialization in the local economy of the first two. As a result, even though land was distributed more unequally in Michang and Lianggezhuang than in Dabeiguan, the

[48] Average land rents were computed from detailed land contract information provided by the survey. Land's share was then calculated by dividing average land rent by average output per unit of land.

[49] The percentage of households in Michang and Lianggezhuang engaged in nonagricultural activity on a full-time basis (3.5 percent and 7.4 percent, respectively) was low compared to that in other North China villages. The average for 30 villages examined by the South Manchurian Railway Company was 13.7 percent. In a number of localities, 20 percent or more of households was classified solely as nonagricultural. These include Macun in Huailu County, 21.8 percent; Zhongliangshan in Changli County, 26.8 percent; Huzhang in Ninghe County, 31.2 percent; and Dongjiao, located just outside Shijiazhuang, 46.8 percent (see Huang, Peasant Economy , pp. 314–20). In a sample of 22 villages in southern Manchuria, 16 percent of all labor (not households) was classified as nonagricultural; see Kokumuin Jitsugyobu Rinji Sangyo Chosakyoku, Kotoku gannendo noson jittai chosa , 4 vols. (Changchun, 1936).

superior market access and non-farm employment opportunities available to residents of Michang and Lianggezhuang produced distributive outcomes that displayed no more inequality than we observe in Dabeiguan.

These calculations show, then, that the distribution of landholdings can be a very misleading indicator of the degree of income inequality in the rural economy. There is no direct line of causality running from higher concentration of landholdings to a higher concentration of income. What is true for several villages at a single time is probably true for a single village over a long period of time. Unfortunately, we do not have earlier data on distribution for these villages, and so the origins of the higher concentration of landholdings in Michang and Lianggezhuang remain a bit of a mystery. Nevertheless, even if landownership in the latter two villages became more concentrated, no widening of income differentials can be inferred.

Income Distribution in China in a Comparative Perspective

If household-level data for our small sample of North China villages are an accurate indication of the difference between the Gini coefficient for household incomes and household incomes on a per capita basis in other parts of

the country—the Gini coefficient for the rural sector as a whole was probably in the vicinity of 0.38.^[50] What tentative conclusions can we draw about the degree of inequality in the rural sector and China more generally on the basis of these data? In other words, how does the distribution of incomes in China compare with that in other low-income countries? Are there reasons to believe that incomes in rural China were exceptionally unequal? Given recent suggestions in the People's Republic of China that the degree of rural inequality, as measured by the Gini coefficient, be controlled within 0.30 to 0.40, perhaps not.^[51]

Although much more work is required before these questions can be answered exhaustively, some comments are in order. Ideally, we would like to compare the distribution of income observed for China with that estimated for other low-income countries in the 1930s. Short of this, we would like to compare our estimate with contemporary estimates for low-income countries similar in key respects to the China of the 1930s. India is an obvious candidate. Comparisons such as these are hampered by data limitations, however.

On the one hand, estimates of the degree of dispersion of rural incomes are almost always for households rather than households on a per capita basis. As we argued earlier, the latter is the preferred basis for measuring the distribution of economic welfare. Unless one is willing to assume that the influence of the size distribution of households on income distribution is constant over time and space, comparisons on a household basis can be misleading.

With this caveat in mind, in Table 6.7 we report Gini coefficients for rural household incomes in several Asian countries. And here we observe a degree of dispersion very similar to that estimated for China. Only in the case of Taiwan, which after World War II benefited from an income-leveling land reform, is it consistently markedly lower. Moreover, for nine of the 16 provinces that we provide estimates for (see Table 6.3), the degree of concentration compares even more favorably. In light of this finding, it seems highly unlikely that inequality was rising over time; if it was, extrapolation would imply a degree of concentration of incomes before the turn of the century that would be considered exceptionally low by international standards.

[50] This estimate was obtained by subtracting the average difference between the Gini coefficient for household incomes and household incomes on a per capita basis for the three Hebei villages from the Gini coefficient for the pooled sample in Table 6.3. One indication that in the aggregate household incomes on a per capita basis were more evenly distributed than household incomes is that cultivated area per household on a per capita basis was more evenly distributed than cultivated area per household. This is reflected in data provided by Buck on the size distribution of farms. The fact that Buck oversampled larger farms is not a point here. While the largest farms were almost 20 times larger than the smallest, on a per capita basis the difference was only one third of this. The same is true for data on land ownership. These relationships also hold separately for each of the eight regions Buck surveyed. See Buck, Land Utilization in China: Statistical Volume , pp. 289, 291, 300.

[51] "Nongcun you meiyou liangji fenbu?" Banyuetan 22 (1987): 20–21.

On the other hand, the few estimates that we have for the dispersion of household incomes on a per capita basis are for the rural and urban sectors combined, while our estimate for China is for the rural sector alone. This is not as great a problem as at first it might seem. In the 1930s, the rural sector—the focus of the National Land Commission survey—constituted between 85 percent and 90 percent of the entire population. The urban sector, although growing at a rate of 2–3 percent per annum, was still very small. The critical question is, how does the addition of data for household incomes in the expanding urban sector influence our overall assessment of income distribution in the Chinese economy? Is there any reason to believe that the distribution for the entire economy differs markedly from that for the rural sector alone?

It was once widely believed that income distribution tends to worsen initially with the growth of the modern/urban sector. Kuznets cites two primary reasons for this belief: (1) a tendency to pay substantially higher wages in the modern/urban sector than in the rural sector and (2) an increasing concentration of capital in the modern/urban sector.^[52] More recent work on a host of countries casts some doubt on this relationship. Nevertheless, for China Kuzent's hypothesis would seem to imply that incomes were more concentrated nationally than they were in the rural sector.

Data on the distribution of incomes in the urban sector presently do not exist. Nonetheless, it seems to us that any upward revision to the estimates obtained separately for the rural sector would actually be relatively modest. Despite the absolute doubling of China's urban population in the early twentieth century, it still amounted to no more than 10–15 percent of the total. As for incomes, it is important to remember here the small size of China's modern sector, almost all of which was located in the cities. Estimates by Ta-chung Liu and Kung-chia Yeh suggest that even as late as the mid-1930s the modern sector constituted no more than 6–8 percent of the gross domestic product.^[53] Because agriculture's contribution to total GDP was two thirds and rural households earned an additional 20–25 percent of their income from nonfarm sources, the percentage of national income going to the urban population could not have been much more than 20 percent and was probably slightly less.^[54] Moreover, an examination of budgetary surveys for

[52] Simon Kuznets, "Economic Growth and Income Inequality," American Economic Review 45 (Mar. 1955): 1–28.

[53] Ta-chung Liu and Kung-chia Yeh, The Economy of the Chinese Mainland (Princeton, 1965), p. 66.

[54] Assuming that 10 percent of income from agriculture went to absentee landlords residing in urban areas (based on the assumptions that 35 percent of land was rented, 60 percent by absentee landlords, and that rental payments constituted 45 percent of output on rented land), then 60 percent of total gross domestic product would have been going to rural households in the form of income from farming. If we add to this the income earned by farming households from nonagricultural activities—estimated by Buck to have been approximately 15 percent of total household income—plus the income earned by the 10–15 percent of the rural population classified as nonagricultural (under the assumption that their average incomes equaled that of the farming population), the percentage of total income going to the urban sector would have been roughly 1 - {((0.90*0.67) / 0.85) / 0.125} = 0.165, or 16.5 percent. With slightly in excess of 10 percent of the population classified as urban, the ratio of average urban-to-rural incomes would have been approximately 1.5 to 1.

working-class households in such cities as Shanghai, Wuxi, and Tianjin reveals a level of per capita income very similar to that observed for the rural sector.^[55] This is consistent with empirical work of Thomas B. Wiens, who found that "despite considerable regional variation in the monetary wage levels of farm laborers and industrial workers in various industries . . . nationally . . . farm wages appear to be slightly below the level of wages of the least skilled urban occupations, as might be expected."^[56]

Although examples can be provided of accumulation of great wealth by some Chinese families, the above considerations lead us to believe that the Gini coefficient for household incomes on a per capita basis for all of China was probably not much in excess of 0.40. How does this compare to that for other countries? In Table 6.7, we report estimates of the Gini coefficients for the distribution of household incomes on a per capita basis for several countries. The sample is small, but the degree of concentration of incomes in China does not appear to be high by comparison.

Finally, what can we say about the concentration of incomes and landholdings in the aggregate? Excluding the holdings of absentee landlords, the Gini coefficient for rural landholdings was 0.72. Making allowances for these landholdings by assuming an average holding of 100 mu suggests a Gini coefficient for rural landholdings in the vicinity of 0.80. This, of course, excludes urban landholdings, which, by all indications, were even more concentrated, implying a still higher degree of concentration. Compared to the Gini coefficient for incomes, it becomes obvious that even in a primarily agrarian economy like China's, the link between land concentration and income distribution is tenuous at best.

Summary

In this paper we have reexamined the links between land concentration and income distribution in rural China. From our perspective, much of the conventional wisdom regarding distribution in late nineteenth- and early

[55] See Bureau of Social Affairs, City Government of Shanghai, Shanghaishi gongren shenghuo chengdu (Shanghai, 1934); Bureau of Statistics, Ministry of Industry, Wuxi gongren shenghuofei ji qi zhishu (Nanjing, 1935); H. N. Feng, "An Enquiry into the Family Budget of Handicraftsmen in Tientsin, 1927–28," Quarterly Journal of Economics and Statistics 1, no. 3 (Sept. 1932).

[56] Thomas B. Wiens, "The Microeconomics of Peasant Economy: China, 1920–40" (Ph.D. diss., Harvard University, 1973), p. 169.

twentieth-century China is badly in need of reexamination. Certainly on the basis of the exploratory examination we have carried out, the views of Tawney and many other observers now seem very difficult to sustain. The many questions that do remain can only be answered through more detailed empirical work of the kind we have attempted here.

Income distribution in rural China has always been a highly charged issue. Perhaps inevitably, views on the subject are difficult to divorce from individual perceptions of the events leading up to and including the 1949 Communist takeover. Yet to say that incomes were no more concentrated—perhaps even less so—than they were in other low-income countries is not, of course, to deny that there was extreme poverty in China or that the number of people living on the margin of subsistence might have possibly increased during the early twentieth century. Given that between the 1890s and 1930s the Chinese population expanded by almost 125 million, an increase in the absolute number of poverty-stricken people can be reconciled almost as easily with the view that incomes became less concentrated as it can be with the view that income distribution worsened. Nonetheless, once we free ourselves from the established paradigm of a secular worsening of inequality and begin to entertain notions of more complicated behavior for land and income distribution, it will be easier to analyze objectively the long-term evolution of the Chinese economy, including the post-1949 changes in the rural sector.