^1. Of the 30 administrative divisions of China only Tibet and Xinjiang (together accounting for 1.5% of China's population) have been excluded from the rural sample. Needless to say, Taiwan has also been excluded. The administrative divisions included in the urban sample are Beijing, Shanxi, Liaoning, Jiangsu, Anhui, Henan, Guangdong, Yunnan, Gansu and Hubei. (Beijing, Shanghai and Tianjin are centrally-administered municipalities with provincial status; we use “province” to include these municipalities.)

^2. This is the proportion of the non-agricultural population implicit in the SSB's figures for per capita consumption of the national, agricultural and non-agricultural populations in 1988. See SSB China Statistical Yearbook, 1989 (Beijing, 1990), p. 619.Google Scholar

^3. The SSB urban panel under-represents towns, which are generally poorer than cities. Our urban sample, which is drawn from the SSB panel, inherits and somewhat amplifies this bias. In particular, towns were entirely eliminated from the urban sample in Liaoning province. As a result, measured urban inequality will tend to be biased downward, a factor that should be kept in mind when considering the discussion below of the level of urban income, the overall urban distribution, the rural-urban gap, and the urban inter-regional distribution.

^4. Production is estimated by adding the value of sales and self consumption of output produced. This means that stocks are assumed to remain unchanged.

^5. Of the 10,258 households in the rural sample, 86 have negative disposable income. Negative income is invariably due to a large negative income from production activities and/or a large tax payment. Many of those with negative incomes have the characteristics of well-to-do households. Barring the falsification of accounts, the explanation of their negative income is a large addition to stocks or a temporary loss sustained by their family enterprises. In this article income and its components refer to the averages for all households, including those with negative incomes. To exclude the latter, even though there is evidence that they are mostly well-to-do, would be arbitrary. The same factors that rendered the incomes of these households temporarily negative must also have affected the ranking of the remaining households. As long as one is limited to observed income over a given time period, rather than measures of “normal” or “permanent” income, such anomalies are inevitable. If the households with negative income are excluded, per capita rural income would be 1.3% higher than that shown in Table 1.

^6. This is confirmed by the fact that for the households in our sample we were able to obtain the estimate of household income according to the SSB definition. Since our sample was drawn from the SSB panel, it was possible for the enumerators to have access to the SSB questionnaire for the same year and to record household income according to the SSB definition. The average for our sample using the SSB definition of income is 548 yuan as compared to the average of 545 yuan for all the households in the SSB survey.

^7. The stock of productive capital owned by Chinese peasants is small, less than 20% of income on average. Its annual depreciation, as a proportion of gross value of output is 1.6% of income, assuming a ten-year life for machines and equipment and a 20-year life for buildings.

^8. 40% of rural households are exclusively farming households with no non-farm production activity. Their average input coefficient is higher than the average input coefficient for the remaining 60% with diversified production activities. The very low input coefficient that is implied for non-farm activities of diversified households suggests that many of these activities consist of services or wage-type contract income rather than manufacturing output.

^9. Strictly comparable data for other countries are hard to find. Rural surveys for Bangladesh show that the contribution of household farming activities to income is well below half. The proportion is believed to be lower for most other Asian countries.

^10. In using the contribution of family farm income to total household income as an index of diversification of China's rural economy over time one must of course allow for the change in agriculture's terms of trade. Thus any comparison with the period before 1979 must make allowance for the sharp improvement in agriculture's terms of trade in 1979 and later years.

^11. See Riskin, C., China's Political Economy (Oxford: Oxford University Press, 1987)Google Scholar, ch. 9, for a discussion of these policies.

^12. Of the 9,009 households in the urban sample four have negative or zero income. In this article income and its components refer to the averages for all households, including those with zero or negative income. The exclusion of the households with non-positive income would raise per capita urban income by an insignificant 0.03%.

^13. Of the urban households in the sample, 84.5% live in publicly-provided housing, 13.7% live in their own houses and 1.7% rent private housing.

^14. For example, in Indonesia the ratio was 1.66 in 1987. In Bangladesh the highest observed ratio during the 1980s was 1.85 whereas the typical ratio was closer to 1.5. Admittedly the household surveys on which these estimates are based make a less than full accounting of subsidies and incomes in kind. But these components in Indonesia and Bangladesh are tiny compared to China. For Indonesia the estimate is from World Bank, Indonesia, Poverty Assessment and Strategy Report (Washington, D.C.: The World Bank, 05 1990), p. 8.Google Scholar For the Bangladesh estimates see Khan, A. R. and Hossain, M., The Strategy of Development in Bangladesh (London: Macmillan, 1989), p. 150.CrossRefGoogle Scholar

^15. Redistributive intermediation by the government is not limited to taxes, subsidies and payments in kind. It also works through government intervention in determining agricultural purchase prices and the prices of the products of public sector enterprises. What the effect of the elimination of all such intermediation on urban-rural income disparity would be is hard to predict.

^16. According to official estimates non-agricultural private enterprise employed 23 million workers in China in 1988. Presumably this figure excludes employment in individual enterprise, which is defined as those privately-owned enterprises which employ no more than four workers including the owner (Beijing Review, No. 4, 1990).Google Scholar It is not known how many of those employed in private enterprise are located in the rural areas.

^17. When we say that the ith source of income has an equalizing effect on income distributions if C_i<G, we simply mean that a rise in the share of i in total income would reduce the Gini ratio. Whether raising the share of i in total income is a feasible or desirable income redistributive policy is of course a more complex question. Thus consider two components of income, urban subsidies and rural self-consumption (production for subsistence). The former has a disequalizing effect and many policymakers would consider a reduction in its share a prime instrument of income redistributive policy. The latter has an equalizing effect, but few policy-makers would consider a greater degree of subsistence orientation a desirable income redistributive policy.

^18. Although 45 women were said to be retired, only four reported receiving pensions, and these averaged 60.75 yuan, or 73% of that of the male retirees with pensions.

^19. 93% of our urban sample households report having one or more household members with regular wages. However, the development of a “floating population” of itinerant workers, officially estimated to average from 60 to 80 million, and the concomitant growth of an “informal sector” in Chinese cities and towns, is a relatively recent phenomenon that is undoubtedly under-represented in our sample. We can only speculate at this stage about how the inclusion of this population might affect the egalitarianism of the urban distribution.

^20. It has been noted that the urban sample includes 38% of all persons in the rural and urban samples together, whereas, in reality, the urban area from which the urban sample was drawn represents only about 21% of the population of China. To get round the problem of over-representation of urban areas, we have counted each rural household twice in the aggregate sample for all China. This makes the urban share of the all-China sample 24%, much closer to the actual urban share of population. Doing this leaves the Gini ratio and the concentration ratios of rural income unchanged.

^21. The other three provinces for which our ranking differs from that of the SSB by more than six (out of a total of 28 provinces) are Inner Mongolia (ours being lower), Jilin (lower) and Guangxi (higher). In the case of Inner Mongolia the difference is from a low share of rental value of housing (5% as compared to 10% for China). For Jilin the reasons are the same as for Heilongjiang, though to a lesser extent: the shares of self-consumption and net subsidies are lower. For Guangxi the difference is largely explained by a high share of net subsidies (—0.8%).

^22. It might appear anomalous that housing subsidies should contribute a lower proportion of income in rich Beijing (24%) than in poor Gansu. In fact, urban residences in Gansu have 20% more living area and 41% more total area than those in Beijing. Beijing's smaller area per capita is partly offset by higher value per square metre.

^23. However, it is almost certainly under-reported in free-wheeling Guangdong, perhaps to a greater extent than elsewhere.

^24. SSB, China Statistical Yearbook, 1990 (Beijing, 1991), p. 36.Google Scholar

^25. See above for a discussion of the somewhat anomalous ranking of Heilongjiang.

^26. According to SSB figures, the provincial range for per capita consumption in 1988Google Scholar was twice as great for the agricultural population (4.1) as for the non-agricultural (2.0). See Zhongguo tongji nianjian, 1990, p. 292.Google Scholar

^27. Ningxia is the only province in which net taxes are significantly progressive.

^28. But see footnote 3.

^29. Note that per capita urban income in Gansu, a very poor province, is very high. This is because state sector wages and pensions are determined by a national scale (with premiums for hardship or remote locations) and by public subsidies. Clearly local resource endowments play an insignificant role in the determination of urban income.

^30. The World Bank, China: Socialist Economic Development (Washington, D.C.: The World Bank, 1983), Vol. I, pp. 83–95.Google Scholar

^31. The World Bank, China: Long-Term Issues and Options (Baltimore: Johns Hopkins University Press, 1985), pp. 29–30.Google Scholar

^32. These estimates, made by a World Bank Working Group on Poverty in Developing Countries, are quoted in Ahmad, Ehtisham and Wang, Yan, Inequality and Poverty in China: Institutional Change and Public Policy 1978–1988 (London: The Development Economics Research Programme, London School of Economics, CP No. 14), p. 46.Google Scholar

^33. The Gini ratio for rural Indonesia for 1987 refers to that for expenditure, not income, which would be higher. It is however, believed that inequality in rural Indonesia declined between 1976 and 1987. For 1987 the Gini ratio for rural income distribution would be higher than 0.28 but lower than the 0.4 which it was in 1976.