روند توزیع درآمد در برزیل و چین: ارزیابی رشد اقتصادی مطلق و نسبی
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|11199||2008||11 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : The Quarterly Review of Economics and Finance, Volume 48, Issue 2, May 2008, Pages 359–369
Over the past two decades real per capita income has increased significantly in Brazil and spectacularly in China. Relative inequality in the distribution of income, however, has remained high in Brazil and has worsened in China. This paper uses a “stochastic dominance” approach to evaluate the welfare effects of a combination of rising mean per capita income in the context of worsening relative inequality. It concludes that by this criterion economic welfare in Brazil increased slightly in the 1981–2002 period. In China, the rapid increase in mean per capita was more than sufficient to overcome significantly increased relative inequality. Between 1985 and 2001 economic welfare thus increased substantially. The overall increase in welfare in both countries, however, is more complex when analyzed by consideration of specific time periods or by rural–urban decomposition.
The recent economic experience of a number of major developing economies has raised the concern that the price of high per capita income growth may be an accompanying worsening inequality in the relative distribution of income. The controversy about both the positive and the negative features of the economic growth process may have been deepened by the fact that each feature is often discussed independently. This is in part due to the fact that aggregate statistics used to characterize each trend, such as per capita income or consumption as a measure of overall growth, and the Gini coefficient or the Lorenz curve as indicators of relative income inequality, may indeed be calculated and examined in isolation from each other. The difficulty with considering each trend in isolation is that it easily leads to arbitrary value judgments about the success or failure of the development process. A more satisfactory approach must consider per capita income growth and relative inequality in the distribution of income simultaneously. Such an evaluation, however, requires an explicit consideration of the preferences that underlie judgments that one kind of economic growth outcome is preferred to another. Economists have traditionally summarized these kinds of preferences using utility functions, while seeking to make the resulting judgments as general as possible. This is the approach of this paper, which uses a utility-based or “stochastic dominance” approach to examine the effects of economic growth in recent decades in Brazil and in China. The use of stochastic dominance rules to compare income distributions is not new.1 This paper extends the approach, however, by using stronger and computationally simpler criteria to make such comparisons. Section 2 is summarizes the intuition, assumptions, and methodology of the stochastic dominance comparison. Section 3 applies the method to data from Brazil's Pesquisa Nacional por Amostra de Domicílios (PNAD), or National Household Sample Survey, from 1981 to 2002, noting some of the sample-specific and country-specific features of these data. Section 4 is a similar application to China's State Statistical Bureau Annual Household Survey, using data from 1985 and 2001. The paper concludes in Section 5 with a discussion of some of the empirical and theoretical limitations faced by any utility-based evaluation of income trends.
نتیجه گیری انگلیسی
Both Brazil and China provide interesting examples of economies in which increasing per capita income is accompanied by rising relative inequality. Judgments made about either trend in isolation from consideration of the other, however, are more likely to fuel ideological debates than to inform good policy. It is clear from examination of both countries’ experience in recent decades that growth has raised the welfare of the poor as well as the rich—even if those already better off have benefited much more. When countries have stumbled, as Brazil did between 1981 and 1992, it is the poor who have suffered most. The stochastic dominance approach outlined in Section 2 suggests that for theoretically plausible utilities—especially that class of utilities which may be used to derive SD rules up to any degree—increases in the likelihood of extremely low or unfavorable outcomes weigh much more heavily and adversely on welfare than would negative shifts in the income distribution of equal magnitude occurring at higher initial levels of income or wealth. If this is so, an income fall of equal absolute magnitude is a much more serious event for a low income individual than for a high income one. The reverse side of this coin is that even modest increases in the incomes of the very poor may do much more for aggregate welfare than would similar absolute income increases for those already better off. The policy implication for both Brazil and China is that an emphasis on lower income groups in the provision of public goods is likely to have larger welfare effects than would equal expenditures that go largely to higher income groups. Expenditure on primary school and basic health services, for example, would therefore have larger impacts on welfare than would expenditures of equal size on universities or higher-end medical services. Another implication of the stochastic dominance approach, at least for third and higher degree rules, is a ceteris paribus preference for greater positive skewness. Alternatively, this result may be interpreted as a strong argument for avoiding negative skewness, which is likely when the probabilities of very low outcomes increase. This may occur when the safety net provided by both family networks and government policy is compromised. This point appears particularly relevant to contemporary China. The household survey data suggest that all income groups have benefited from the high rates of growth of the past several decades. But in the same period both demographic and policy changes have made a number of people more vulnerable to a possible slowing of China's growth than might have been the case in the past. The fall in family size—a trend that was well underway even before the widespread implementation of the “one child” policy for the Han Chinese—means that many elderly people have no family member to assist in their support. This is a sharp departure from many centuries of Chinese cultural and social tradition. In addition, the relaxation of residence permits and registration requirements have permitted many younger and working-age Chinese to migrate from the rural areas and the jobs to which they once would have been tied, often leaving the elderly behind. Simultaneous changes in government policy have also increased the potential vulnerability of the poor, especially in rural areas and in provinces away from the high growth coastal areas. Virtually guaranteed jobs in the pre-Deng Xiaoping era provided a minimum level of subsistence—popularly known as the “iron rice bowl”. The dismantling and termination of many state subsidized or run services and the privatization of a number of large state enterprises have taken away the iron rice bowl. Personal disasters, such as accidents or medical emergencies can be devastating for the poor. In contrast to arbitrary definitions of a “poverty line”, the stochastic dominance approach used here avoids any sharp discontinuity between the “poor” and the “non-poor”. This is important for policy purposes, since improvements in the income of the very poor may not move them above the poverty line, yet would have a greater impact on welfare than would similar increases for slightly less poor. There are some important limitations to the stochastic dominance way of making income distribution comparisons. First, it provides only a partial ordering over a set of distributions, since small chances of extremely low outcomes in an otherwise desirable income distribution will make it less preferred by those who are extremely risk averse. This may be viewed as an advantage, however, if we recognize that in such a case the impossibility of ordering is telling us that the kinds of orderings produced by comparisons of means, or of means and variances, or of shares of “poor” and “non-poor” may in fact not be legitimate for some reasonable utilities. Secondly at an operational level, full implementation of the stochastic dominance rules requires the estimation of the moments of the underlying distributions. This appears easy when dealing with empirical distributions, for which the moments are defined. But if the underlying moments of the theoretical distributions are not defined, as is the case for some distributions, then the apparent empirical moments may be unstable. Although this is probably not the case when dealing with income distribution data, which may be well described by a number of theoretical distributions for which the moments exist, it may limit the applicability of SD rules (or any moment-based method) in some cases.13 A third difficulty with the stochastic dominance approach, as is the case with almost any utility-based method of choice, is that we have assumed that utility is a function of a single argument, usually income or consumption. This ignores a number of possible complications. It is possible that the income or consumption level of others—for example family members outside the household—is a secondary argument in the utility function. Alternatively, expected future income or consumption might be an argument. These potential complications may be addressed in a stochastic dominance framework, but at the cost of added complexity. The possibility that the income or consumption of others may be an argument in one's own utility raises an interesting question in a Brazilian, or especially, in a Chinese context. An increase in inequality due to a rise in the probability of extremely unfavorable outcomes, or a “downward transfer” in the terminology of Section 2, is clearly what motivates the stochastic dominance approach, especially for higher degree rules. But in China in the past decade the increase in the dispersion or spread of the income distribution has been an “upward transfer”, in which the number of households enjoying incomes at unprecedented levels has increased markedly. Since the higher level SD rules give positive weight to increases in the upper tail they are more informative than would be measures like variance that penalize any increase in dispersion of the distribution about the mean. But even the SD rules may underestimate the impact of an “income breakout” like that of contemporary China on those who are not yet in the upper tail. Like people who are excited by knowing that someone won big in the lottery, the fact that others have had rapid income increases may be viewed favorably even by those who do not share in the increase. This may be the case if they interpret the improvement in others’ incomes as a signal that they too may share in this trend in the future.