The first comprehensive effort to estimate household income and its distribution in China according to standard international definitions was made for the year 1988 by an international group of economists working with members of the host institution, the Economics Institute of the Chinese Academy of Social Sciences (CASS). The sample survey designed by this team produced estimates of household income that were substantially different from those of the State Statistical Bureau (SSB), based on its annual surveys, with different implications for both the average standard of living and the degree of inequality of income distribution. The study, first reported in the pages of this journal in December 1992, found that per capita household income was both substantially higher and more unequally distributed than suggested by the SSB estimates. It also provided insights into sources of inequality in China that were unobtainable from the published official data on income and its distribution.
1. Household income means the same thing as the more commonly used term personal income. The two terms are used interchangeably in this paper.
2. Main results were first published in Khan, A. R., Griffin, K., Riskin, C. and Renwei, Zhao, “Household income and its distribution in China,” The China Quarterly, No. 132 (12, 1992), pp. 1029–1061; a longer version appears as Khan, A. R., Griffin, K., Riskin, C. and Renwei, Zhao, “Household income and its distribution in China,” Chapter 1 of Griffin, K., and Renwei, Zhao (eds.), The Distribution of Income in China, (London: Macmillan, 1993).
3. A detailed discussion of the sampling methodology for 1988 can be found in Eichen, Marc and Ming, Zhang, “The 1988 household sample survey—data description and availability,” in Griffin, and Zhao, (eds.), Distribution of Income. While selection of provinces for the 1995 survey changed as described in the text, selection of households proceeded as in 1988.
4. Our rural survey contained a question designed to elicit the household's net income as defined by the SSB. This was known because our sample was drawn from the SSB's larger sample. No similar question was included in the urban survey.
5. This somewhat incongruous combination was designed to maintain consistency with the 1988 survey. However, that survey under-counted non-wage income from enterprises. Therefore, the estimated increase from 1988 to 1995 is probably too high. This income item can be thought of as representing rural individual entrepreneurial income in 1995, since payments from the collective welfare fund comprised a very small part of it (averaging less than 2.5 yuan out of 140 yuan per capita).
6. In 1988, it was not possible to distinguish net farm income from net income of non-farm activities, since purchased inputs could not be differentiated by sector of use. Nor could they be differentiated by use between commercial and subsistence output, which led us to arbitrarily allocate them entirely to the former. Therefore, we ended up with net cash income from the sale of all farm and non-farm products, and gross value of consumption of farm products. Together these two items added up to the sum of net income from fanning and non-farm activities.
7. It may be noted that in 1988 market rent was estimated indirectly, first, by indirectly estimating the replacement value of the house and, next, estimating market rent as 8% of the replacement value. See Griffin, and Zhao, , Distribution of Income, Chapter 1. In the 1995 survey, respondents directly estimated the market rent of their abodes. This difference in estimation method should be kept in mind when assessing the differences in estimated housing subsidy between the two dates.
8. Thus, for example, our finding, reported later in this paper, that inequality increased between 1988 and 1995 does not imply that it rose steadily throughout that period. We suspect it might in fact have peaked some time before 1995.
9. We did not have a separate estimate of net income from household farming for 1988 (see note 6). We had only the sum of the gross value of self-consumption of farm products and the net value of sale of farm products. Of these two, the first alone accounted for 41% of rural income in 1988, strongly suggesting that total farm income as a proportion of total income was far above the 1995 ratio of 46%.
10. “Household member's net individual income from private, individual and/or jointly operated enterprises.”
11. “Entrepreneurial income” differs from “net income from household non-farm activities” in that the former arises from activity of individuals and the latter from household operations. While in practice the borderline may be somewhat blurred – e.g. some respondents might regard a particular activity as individual, others as a family operation – we are confident that there is no overlap in the data from the two categories.
12. Reported property income is several times smaller than that implied by data on bank savings deposits and relevant interest rates (an observation we owe to an anonymous referee). Yet our estimate appears to be consistent with that of the SSB (which lumps property income from its household survey together with transfers; the combined estimate is virtually identical to ours. Whether or not the data on bank deposits from alternative sources shows that both the SSB and we have missed a part of this income is an issue on which we would like to reserve judgement, given the ambiguity of the evidence. For instance, it is apparently not uncommon for enterprises to place their funds in individual accounts, which bear higher interest rates. Nor is it certain that reported official interest rates are applicable to all deposits, especially those held by local savings institutions.
13. This is a Paasche index with 1985 as the base. The “Rural CPI” for 1995 (220.09) is actually the value of the SSB's rural CPI for 1995 as a percentage of the same for 1988. The Paasche formula is known to understate the rate of increase in cost of living (see Allen, R. G. D., Index Number, Theory and Practice (London: Macmillan, 1975)). In this case, when neither of the two years compared is the index's base year, the use of the Paasche formula makes it impossible even to give a clear interpretation to the “index” showing the change in CPI between 1988 and 1995.
14. For instance, the costs of a kilocalorie of food energy in both urban and rural areas increased at substantially higher rates than the official CPIs. If these are combined with the lowest of other available indices of price change for non-food goods, the resulting CPIs would be much higher than the official ones.
15. However, because non-wage income from enterprises was under-counted in 1988, growth of receipts from enterprises is very probably overstated.
16. The SSB estimates are 0.6 per cent of income in 1988 and 2.4 per cent in 1995, both somewhat below ours. This is surprising, since our estimates of income, the denominator, are substantially higher than the SSB's.
17. Urban property income may be under-reported. Our estimate is 80% of the income from interest, dividends and rent as estimated by the SSB (see SSB, Statistical Yearbook of China (Beijing: Statistical Publishers, various years)). We do not know how much of this difference is due to definition and classification of components and how much to possible bias in our sub-sample of the SSB parent sample. Both our estimate and the SSB's fall far short of an alternative estimate that could be derived from data on bank deposits and interest rates. The significance of this difference is unclear for reasons given above in note 12.
18. This item was probably substantially under-reported in both years. Moreover, there were too few non-zero observations for private enterprise income in 1995 to generate a statistically significant estimate for this item alone.
19. These include cash income from other jobs, unemployment benefits, income in kind, income received for being a village cadre and other cash income not from household activities. The statement that the SSB evidently excludes these items is based upon both the lack of any explicit mention of them in SSB, Statistical Yearbook (1997), p. 313 and the close similarity between the SSB's estimate for “labourers' remuneration” and ours for regular plus non-regular wage income. We include pensions in wage-type income, whereas the SSB probably includes them in transfers. It is possible that some or all of our additional categories are indeed included in “labourers' remuneration,” in which case there is an inexplicably large gap between our respective estimates.
20. We asked for the total of all production costs, including labour costs, and subtracted them from gross revenues. Possibly, this resulted in under-enumeration of costs.
21. The SSB has two different concepts of urban income, “per capita income” and “per capita income available for living” (see SSB, Statistical Yearbook of China (Beijing: Statistical Publishers, 1996), Table 9–5). From the meagre explanation of concepts provided by the SSB, it is impossible to know what the difference between them is. In a table of comparative urban and rural per capita incomes, the SSB shows the latter of the two urban measures along with “per capita net income of rural households,” suggesting that the two are comparable (SSB, Statistical Yearbook of China (1996), Table 9–4). After a careful comparison of components, we have decided that “per capita income” is the relevant measure of income and this has been used throughout as representing the SSB estimate.
22. The saving rate for rural China was 17% (on the basis of the data in SSB, Statistical Yearbook of China (1996), p. 300) and for urban China 17.5% (on the basis of the data in SSB, Statistical Yearbook of China (1996), p. 284). Aggregate household incomes in urban and rural China are roughly equal, so that the weighted average saving rate is 17.25%. GNP estimate is also from SSB, Statistical Yearbook of China (1996).
23. The estimates for these other Asian countries are from World Bank, World Development Indicators 1997 (Washington, D.C.: The World Bank, 1997). There is, of course, the question of whether the rates elsewhere are accurately estimated.
24. This is based on the data in SSB, Statistical Yearbook of China (1996).
25. The annual growth rate in real per capita household income for China as a whole (the weighted average of rural and urban incomes) was higher than either the rural or the urban real income growth because there was a rise in the weight of the urban population – the richer of the two income groups – between 1988 and 1995.
26. This point was called to our attention by Barry Naughton.
27. The share of government revenue in GNP declined sharply between 1988 and 1995: from 15.7% to 10.9% (SSB, Statistical Yearbook of China (1997)). Thus, if the shares of both the household and government sectors of the economy have decreased, then business sector income must have risen during the period in question.
28. Thus, the Theil Index, which, unlike the Gini ratio, can be decomposed, is sensitive to the sample size and is not amenable to intuitive interpretations such as one can make of the Gini ratio. The Atkinson Index is very sensitive to the subjective value of the inequality aversion parameter which is essentially arbitrary. Estimates of none of these other indices are nearly as widely available, for purposes of comparison, as the Gini ratio. For a comparison of different measures of inequality see Sen, Amartya, On Economic Inequality (Oxford and New York: Oxford University Press, 1997).
29. See Khan, et al. , “Household income,” The China Quarterly), p. 1038, for a more detailed discussion of Gini and concentration ratios.
30. Decile shares of 1988 income and its components are shown in Khan, et al. , “Household income,” Distribution of Income.
31. Like other users of Gini ratios, we have not tried to measure their standard errors. We have adopted the convention of designating any change in Gini ratio of 10% or greater as significant, but we cannot establish statistically the significance of a 10% difference or, indeed, of a larger difference, for that matter. We believe the plausibility of individual estimates of increased inequality is enhanced by the broad range of such increases, and by the meagreness of examples of counter movements. Readers are of course free to arrive at their own judgements.
32. See Khan, et al. , “Household income,” Distribution of Income, for comparative data for other Asian countries.
33. If property income is underestimated (see note 12), then so is the Gini ratio, albeit by a small margin, because property income is such a highly disequalizing source of income.
34. This pattern, and its exacerbation by fiscal decentralization and government's declining share of GDP, is discussed in Wong, Christine, Heady, C. and West, L., Financing Local Development in the People's Republic of China (Oxford and New York: Oxford University Press, 1997). Its effect on health care is discussed in World Bank, Financing Health Care (Washington, D.C.: The World Bank, 1997).
35. Throughout this paper a change in the inequality of distribution of a component of income is measured by a change in its concentration ratio, not by a change in its Gini ratio.
36. Decile shares of urban income and its components for 1988 are shown in Khan, et al. , “Household income,” Distribution of Income.
37. See ketizu, Guowuyuan yanjiushi (Study Group of State Planning Commission Research Office), “Guanyu chengzhen jumin geren shouru chajude fenxi he jianyi” (“An analysis and proposal concerning income inequality among urban residents”), Jingji yanjiu (Economic Research), No. 8 (08 1997), p. 3.
38. Despite the apparent equality of distribution of the combined category of private and individual enterprise income, we would conjecture that, if not underestimated, this category would be disequalizing. This is because the chief source of underestimation is private enterprise income, which can be substantial and which accrues to very few individuals.
39. This paper does not take the position that ration coupons, with their allegedly adverse consequences for efficiency, should have been retained. The point is that they were not replaced by alternative measures to offset the adverse distributional consequences of their abolition.
40. See text above and note 18.
41. If property income is underestimated, as we suggested earlier it might be, then the Giniratio is also slightly underestimated, since property income is such a disequalizing source of income.
42. In 1995 rural China represented 71% of total population of China. The share of rural population in the survey was just under 62%. We therefore drew a 50% random sample of rural households in the sample and added it to the sample. This raised the share of rural population in the survey to about 71%.
43. Decile shares of income and its components for 1988 are shown in Khan, et al. , “Household income,” Distribution of Income.
44. International comparison of Gini ratios is subject to many problems. One has to be particularly careful about the variable for which it is measured. The Gini ratio of the distribution of per capita expenditure is typically lower than the Gini ratio of the distribution of per capita income. The Gini ratio of per capita income distribution in Pakistan was 0.407 in 1990/91 (Amjad, R. and Kemal, A. R., Macro Economic Policies and their Impact on Poverty Alleviation in Pakistan (Manila: ILO/South-East Asia and Pacific Multidisciplinary Advisory Team, 1996)). The Gini ratio of expenditure distribution was 0.338 for India (1992) and 0.317 for Indonesia (1993) (World Bank, World Development Indicators). It seems implausible that the Gini ratios for income distribution in India and Indonesia would be so much higher than their Gini ratios for expenditure distribution as to be equal to China's Gini ratio. For the Philippines, the Gini ratio of distribution of expenditure was 0.43 (Balisacan, Arsenio, What is the Real Story on Poverty in the Philippines? A Re-examination of Evidence and Policy, manuscript (Manila: School of Economics, University of the Philippines, 1996)).
45. See World Bank, Sharing Rising Incomes: Disparities in China (Washington D.C.: The World Bank, 1997), for the first estimate, and World Bank, World Development Indicators for the second, both presumably based on official data and definitions.
46. Their estimated Gini for personal income in 1994 was 0.434, very close to our own estimate for 1995. Their estimate for rural income was 0.411, again close to our own; for urban income it was 0.377, well above our estimate (personal communication from State Planning Commission, Department of Social Development, 11 1997).
47. Net taxes (negative net transfers), receipts from TVEs and other enterprises, property income and rental value of owned housing are other components of rural income that became more disequalizing and thus contributed to the decline in the equalizing character of rural income as a whole.
48. See Khan, et al. , “Household income,” Distribution of Income for comparative evidence.
49. The index of terms of trade (the ratio of farm and sideline product price to rural retail price of industrial products) increased by 4.5%. Gross value of agricultural production increased by 10.9%. These data are from SSB, Statistical Yearbook of China (1996).
50. Assuming migrant income should be assigned to the urban category, its neglect almost certainly raises average urban income. If one thinks it belongs in rural income, its neglect probably lowers average rural income. In either case, the measured urban-rural gap is higher than it should be. We are indebted to an anonymous referee for raising the issue of treatment of migrant income.
51. The range is the ratio of the average per capita income of the richest province to that of the poorest. The coefficient of variation is the standard deviation of the distribution of provincial average per capita incomes divided by the mean provincial per capita income. It thus compares the absolute dispersion of the provincial average incomes with their mean, which has of course risen.
52. The standard deviation of rural provincial per capita income rises by 3.5 times while the mean doubles.
* We are greatly indebted to Professor Zhao Renwei of the Economics Institute, Chinese Academy of Social Sciences, whose leadership made possible the study on which this paper is based; and to Professor Li Shi, also of the Economics Institute, whose involvement in every phase of the project contributed immeasurably to its success. We thank Barry Naughton, Scott Rozelle, Terry Sicular, James Wen, and two anonymous referees for perceptive comments on an earlier draft of this paper. Acknowledgement is due to the Asian Development Bank and The Ford Foundation for their financial support of the project. Finally, we express appreciation for the co-operation of the State Statistical Bureau, which implemented the household income survey on the results of which this report is based.
Email your librarian or administrator to recommend adding this journal to your organisation's collection.
Full text views reflects the number of PDF downloads, PDFs sent to Google Drive, Dropbox and Kindle and HTML full text views.
* Views captured on Cambridge Core between September 2016 - 23rd July 2018. This data will be updated every 24 hours.