1、 Copyright UNU-WIDER 2007 *UNU-WIDER, Helsinki. This study has been prepared within the UNU-WIDER project on Inequality and Poverty in China. UNU-WIDER acknowledges with thanks the financial contributions to its research programme by the governments of Denmark (Royal Ministry of Foreign Affairs), Fi
2、nland (Ministry for Foreign Affairs), Norway (Royal Ministry of Foreign Affairs), Sweden (Swedish International Development Cooperation AgencySida) and the United Kingdom (Department for International Development). ISSN 1810-2611 ISBN 92-9190-944-0 ISBN 13 978-92-9190-944-5 Research Paper No. 2007/0
3、5 Regional Income Inequality in Rural China, 1985-2002 Trends, Causes and Policy Implications Guanghua Wan* January 2007 Abstract This paper depicts the trend of regional inequality in rural China for the period 1985-2002. The total inequality is decomposed into the so-called within- and between-com
4、ponents when China is divided into three regional belts (east, central and west). A regression-based accounting framework is then used to explore root sources of the rising inequality. Policy implications are discussed. Keywords: China, inequality, spatial decomposition, rural JEL classification: O1
5、8, P2, D63 The World Institute for Development Economics Research (WIDER) was established by the United Nations University (UNU) as its first research and training centre and started work in Helsinki, Finland in 1985. The Institute undertakes applied research and policy analysis on structural change
6、s affecting the developing and transitional economies, provides a forum for the advocacy of policies leading to robust, equitable and environmentally sustainable growth, and promotes capacity strengthening and training in the field of economic and social policy making. Work is carried out by staff r
7、esearchers and visiting scholars in Helsinki and through networks of collaborating scholars and institutions around the world. www.wider.unu.edu publicationswider.unu.edu UNU World Institute for Development Economics Research (UNU-WIDER) Katajanokanlaituri 6 B, 00160 Helsinki, Finland Typescript pre
8、pared by Lorraine Telfer-Taivainen at UNU-WIDER The views expressed in this publication are those of the author(s). Publication does not imply endorsement by the Institute or the United Nations University, nor by the programme/project sponsors, of any of the views expressed. 11 Introduction It is wi
9、dely recognized that regional inequality in China has been on the rise since economic reforms were initiated in the late 1970s (Kanbur and Zhang 2005; Wan 2005). In addition to its repercussions on social and political stability, such a rise has hampered poverty alleviation (Ravallion and Chen 2004;
10、 Zhang and Wan 2006) and is found to be detrimental to long-run economic growth (Wan et al. 2006). Many Chinese scholars also consider high inequality as a major contributor to the sluggish domestic demand in China. It is thus not surprising to witness a broad and growing interest in Chinas regional
11、 inequality. Earlier studies largely focused on the measurement of regional inequality. Subsequent efforts were devoted to breakdown total inequality into various components, either by population subgroups (Tsui 1991) or by factor components (Wan 2001). Recently, the technique of regression based de
12、composition has gained popularity (Fields and Yoo 2000; Morduch and Sicular 2002; Wan 2002) and has been applied to China (Morduch and Sicular 2002; Wan 2004; Wan and Zhou 2005). Despite a large volume of literature on regional inequality in China, few existing studies constructed time profiles of i
13、nequality among rural regions in China.1This is surprising given that a dominant proportion of Chinas population live in the countryside and, as discussed later, rural inequality is a large component of the overall regional inequality. More importantly, rural inequality is fundamentally different fr
14、om the urban counterpart in terms of causes, trends and policy implications. For example, geography is much more important in driving rural inequality than the urban inequality. When encompassing weather, infrastructure and other natural resources, the geography variable would account for a very sig
15、nificant share of total rural inequality. This is not necessarily the case for the urban sector despite the probable relevance of location as a determinant of wages. Needless to say, distribution of arable land is relevant to rural inequality but not to the urban counterpart in China. Clearly, while
16、 studies on whole China or urban China are important, there exist obvious justifications for a separate paper focusing on rural China. This paper will fill such a gap in the literature by providing a time profile of regional income inequality in rural China for the period 1985-2002. Such a time prof
17、ile was appealed for by Rozelle (1994). Another purpose of the paper is to identify the components of rural regional inequality. This is accomplished by undertaking conventional as well as regression-based inequality decompositions. These decompositions offer different insights into the determinants
18、 of the total inequality. Policy implications, also, will be explored. 1A search in Econlit using keywords China, region, rural and inequality produced 59 journal article entries and only a few of them touched on, but did not focused on, rural regional inequality. 2The paper is organized as follows.
19、 The next section describes the data and the time profile of rural regional inequality, wherein conventional and newly proposed methods will be employed to decompose total inequality into two broad components: between regional belts (i.e., east-central-west China) and within these belts. Section 3 a
20、pplies the regression-based decomposition to rural China, which helps reveal the root sources of regional inequality. Finally, we discuss policy implications in Section 4. 2 Data and preliminary analysis As a precursory note, it is useful to mention that a substantial proportion, in the order of 25
21、per cent or so, of Chinas regional inequality is attributable to the urbanrural gap.2The remaining is due to inequalities within urban and rural regions. In accounting for the total regional inequality for China as a whole, these so-called within-components are given by their respective Theil-L inde
22、x estimates, weighted by their population shares. Since the unweighted rural regional inequality is found to be larger than the urban counterpart (Wan 2005) and a dominant percentage of population lives in the rural areas, the contribution of rural regional inequality to the total must be substantia
23、l. To accomplish research objectives of this paper, most of our data are compiled from the China Rural Household Survey Yearbook (NBS various years) for the period 1985-2002. Earlier data are incomplete. Ideally, rural population should be used as our income observations are for rural residents. How
24、ever, we failed to find consistent rural population series for all regions. Instead, agricultural population statistics are used. They are expected to be highly correlated with the rural population and are available from the China Rural Statistical Yearbook (NBS various years). Excluding Hong Kong,
25、Taiwan and Macao, there are 31 regions (provinces, autonomous regions or metropolitan cities) in China. However, our sample contains data for 28 regions with Hainan merged with Guangdong, Congqing merged with Sichuan, and Xizang (Tibet) excluded. Data for Tibet are not complete. As argued by Wan (20
26、01), such exclusion is not expected to distort the analytical results. All data in value terms are deflated by regional rural consumer price indices (CPIs) as well as the regional price indices compiled by Brandt and Carsten (2004). The deflated regional income data are plotted in Figure 1 against y
27、ears; for each year the plot contains per capita real incomes for all 28 regions. The Figure shows that while real income has been increasing over time, its dispersion is also on the rise as indicated by the expanding height of the plots over time. According to Figure 1, regional incomes had been in
28、creasing over 1985-89. After a setback for most regions in 1989-90, the 2If differences in price levels and inflations between urban and rural areas were not considered, this proportion would be over-estimated as in Kanbur and Zhang (1999). See Sicular et al. (2007). 3Figure 1: Regional per capita r
29、eal income, by year increasing trend resumed. The setback is probably caused by the austerity programme initiated by the Chinese government in 1989 (Wan 2001). Reasonably assuming no changes in the composition of regions in the rich and poor groups, the poor (lower segments of the plots) consistentl
30、y experienced slower income growth than the rich (upper segments of the plots). In fact, real income declined or was stagnant before 1996 for the poor regions as a group. Even after a small rise in 1996, this groups income rose little over 1996-2002. In contrast, income increased from 1985 to 1988 f
31、or the rich regions. For this group, there was a small drop in income in 1989, but the increasing trend resumed right after and continued strongly until 2002. Judging from these observations, one may conclude that regional income inequality in rural China has been increasing in both absolute as well
32、 as relative terms, respectively indicated by the expanding height of the plots over time, and by the differing growth rates for the poor and rich regions. Examination of Figure 1 reveals that the gaps between the income groups (top, middle, and bottom segments of the plots) seem to have expanded mo
33、re than those within these groups or segments. This clustering of regional income in recent years implies some forms of polarization in China. In other words, income has been diverging more 15035055075095011501984 1986 1988 1990 1992 1994 1996 1998 2000 2002YearIncome(Yuan)4between income groups tha
34、n within income groups. Nevertheless, Figure 1 may be misleading as far as gathering inequality trend is concerned because the expansion in income dispersion had been accompanied by changes in income levels. It is known that an identical income growth for all regions can also result in increased dis
35、persions, as in Figure 1, but such growth leaves inequality unchanged as long as relative rather than absolute inequality measures are used. A formal way to analyze inequality is to construct Lorenz curves and conduct stochastic dominance analysis. For this purpose, Lorenz curves are obtained for ea
36、ch of the 18 years. Although there is a tendency for the Lorenz curve to move downwards over time, any first-degree stochastic dominance is not clearly visible when they are all displayed in one diagram. On the one hand, this may be caused by too much informationmany curves are squeezed onto one dia
37、gram. On the other, this is understandable as inequality changes are usually small from one year to the next. To reduce distractions caused by too much information, we average Lorenz curves over a 3-year interval and present these curves in Figure 2. Unfortunately, even Figure 2 does not exhibit any
38、 first-degree stochastic dominance very clearly. As a consequence, pair-wise comparisons of these curves have to be done and they indicate that nine out of the 15 pairs of the Lorenz curves cross, mostly at the top or bottom ends of the distributions. Figure 2: Lorenz curves 00.10.20.30.40.50.60.70.
39、80.910 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 11985-87 1988-901991-93 1994-961997-99 2000-025When Lorenz curves cross, they cannot be used to rank income distributions. In this case, second or higher degree of stochastic dominance can be introduced. Alternatively, summary inequality measures could be u
40、sed instead (Fields 2001). To minimize possible sensitivities to inequality measures,3we compute most relative inequality indices that are commonly in use, namely the Gini, Theil-L, Theil-T and half CV2.4Let Z denote the target variable, denote the mean of Z, j index observations (j = 1, 2, , N), th
41、e following formulae can be used: Atkinson = jNjZ/11, Theil-L = jjZLnN1, and Theil-T = jjjZLnZN 1. The Atkinson index is not considered here because it can be expressed as a monotonic transformation of Theil-L (Shorrocks and Slotjje 2002). The computed values are tabulated in Table 1 (left panel). S
42、ince CV2violates the principle of transfer, values in the last column are reported only for comparison purpose as there are many studies in China using the measure CV2. Results in Table 1 show that all measures are consistent in demonstrating a rising trend in regional inequality in rural China. In
43、particular, the inequality increased rather dramatically until 1995-96. After that, the increasing trend became moderate. This finding is consistent with Figure 1, which shows that income of poor regions experienced little growth before 1995-96 but some improvement afterwards while the rich regions
44、exhibited growth throughout the period. The slowing down in inequality increases after 1995-96 may be caused by the implementation of the grain price support policy which benefited poor regions more (Zhang 2005). Since the trend was only moderated, not really reversed, other forces must be stronger
45、than the policy change in pushing up the long-run inequality trend. Identifying these other forces is crucial for policymakers if rural regional inequality is needs to be brought down. As a by-product, we calculated inequality values using undeflated data (see the right panel in Table 1). As is expe
46、cted, not taking into consideration inflation and regional price levels leads to upward biases in inequality measurement. The biases are quite substantial. What is interesting, and perhaps surprising, is that the biases are larger in 3Different measures imply different social welfare functions and d
47、ifferent aversions to inequality (see Dagnum 1990). 4There exist alternative ways to calculate the Gini coefficient. We follow Silber (1989) by defining Gini = PQI, where P is the vector containing population shares and I is the vector containing income shares, both sorted in ascending order by the
48、per capita income variable. Q is a square matrix with 0 on the diagonal, 1 above the diagonal and -1 below the diagonal. 6the early years, a finding consistent with Brandt and Carsten (2004). Also, the biases are less severe when using Gini relative to other measures, possibly due to the differing s
49、ensitivities of these measures to different sections of the underlying Lorenz curves. Table 1: Regional inequality in rural China Deflated data Undeflated data Gini Theil-L Theil-T CV2Gini Theil-L Theil-T CV21985 0.109 0.020 0.019 0.037 0.152 0.038 0.042 0.095 1986 0.123 0.025 0.024 0.047 0.171 0.047 0.050 0.114 1987 0.129 0.027 0.026 0.052 0.180 0.052 0.056 0.127 1988 0.134 0.029 0.028 0.057 0.187 0.056 0.061 0.138 1989 0.137 0.030 0.029 0.059 0.194 0.060 0.065 0.148 1990 0.141 0.032 0.031 0.062 0.198 0.063 0.069 0.162 1991 0.142 0.032 0.032 0.065 0.
