ELSEVIER
Journal of Development Economics
Vol. 50 (1996) 353-368
JOURNAL OF
Development
ECONOMICS
Economic reform and interprovincial
inequalities in China
Kai-yuen Tsui *
Department of Economics, Chinese University of Hong Kong, Shatin NT, Hong Kong
Received 6 July 1993; final version received 20 March 1995
Abstract
This note is an attempt to explore two issues: (1) the trend of interprovincial inequality
in the post-1978 reform era in China; and (2) the factors behind the dynamics of
interprovincial inequality. Using recently released provincial gross domestic product (GDP)
data, we have shown that interprovincial inequality decreased in the fn'st half of the 1980s,
but started to increase in the second half of the 1980s. To understand the impact of sectoral
reforms on interprovincial inequality, the overall inequality in provincial per capita GDP is
decomposed into the contributions by the primary, secondary and tertiary sectors. It seems
that the trend of interprovincial inequality has been largely induced by the reforms of the
industrial sector which make up a large share of the secondary sector. In addition, national
income data from the socialist national accounting system are used to identify the impact of
interprovincial resource flows on interprovincial inequalities. The findings seem to suggest
a relative decline in the role of redistributive budgetary transfers.
JEL classification: 05; P2; P52; R12
Keywords: China; Interprovincial inequality; Inequality decomposition
* E-mail: b040785@mailserv.cuhk.hk.
0304-3878/96/$15.00 Copyright © 1996 Elsevier Science B.V. All rights reserved.
PII S0304-3878(96)00406-3
354 K.-y. Tsui / Journal of Development Economics 50 (1996) 353-368
1. Introduction
There is a widespread fear in China that market-oriented economic reforms,
though improving economic efficiency, may increase interprovincial inequalities. 1
The spectacular economic growth of such coastal provinces as Guangdong has all
but reinforced the above impression. Some believe that China may be inexorably
entering the initial phase of WiUiamson's inverted-U process of regional development
(e.g., Zhang, 1989). In this connection, this paper is an attempt to answer the
following two questions:
1. Have interprovincial income disparities been increasing in the post-1978 reform
era?
2. What are the factors behind the dynamics of interprovincial inequalities?
With regard to the first question, there is now a small but expanding literature
which explores quantitatively the trends of interprovincial inequalities (Lyons,
1991; Tsui, 1991; Denny, 1991; Tang, 1991; Yang, 1992a,b; Wei, 1992; Dong,
1992; Jiang and Liu, 1992; Liu et al., 1993). The recent release of such provincial
income statistics as provincial gross domestic product (GDP) furnishes us with an
opportunity to take a new look at the first question. 2 While one should bear in
mind that China's GDP statistics are far from perfect, it is nevertheless a serious
attempt on the side of the Chinese to switch from the socialist national accounting
system to the United Nation's SNA system. GDP, which includes the tertiary
sector, is a more comprehensive index of the overall economic activities in a
region than such output figures as provincial net material product in the socialist
national accounting system. 3
With regard to the second question, reform measures and policy shifts may
exercise significant impact on interprovincial inequalities. It is often contended
that fiscal decentralization and the reorientation of China's regional development
1 The basic units of analysis are the provinces, centrally administered municipalities and autonomous
regions. The three centrally administered municipalities are Beijing, Tianjin and Shanghai. In the
following analysis, they are incorporated into neighbouring provinces. Guangdong does not include
Hainan. Owing to data limitation, Xizang (Tibet) and Hainan are excluded. This paper does not look
into disparities among the subprovincial units (e.g., counties).
2 The provincial GDP data used in this paper are from State Statistical Bureau (1990).
3 For more detailed discussion of the problems with respect to GDP data of China, see Guo and Han
(1991) and Tsui (1993). One referee questions the accuracy of the data for the tertiary sector since the
Chinese only started collecting data on the tertiary sector for the purpose of compiling GDP data and
the historical data were reconstructed in the 1980s. On numerous occasions, I raised this question with
officials working in the state and provincial statistical bureaux. While the Chinese GDP data are far
from perfect, it seems to me that they are not derived based on crude back-of-the-envelope calculation.
Before any better data are released, the tertiary GDP data are probably the best available. Since the
share of provincial tertiary GDP varies across provinces, it makes a difference whether the tertiary
sector is included, The referee further suggests that perhaps it is better to use net material product
(NMP), which has a longer time series. The NMP data are, however, not necessarily better for the
simple reason that the historical NMP data were only reconstructed in the 1980s.
K.-y. Tsui / Journal of Development Economics 50 (1996) 353-368 355
policy in favour of the richer coastal provinces have widened the gap between the
rich and poor provinces (see, for example, Guo and Wang, 1988; Wang and Bai,
1992; Yang, 1990). Furthermore, sectoral reform policies such as the liberalization
of the agricultural sector, the promotion of rural industries and the rapid development
of the service sector may also induce significant reinforcing as well as
offsetting effects on regional disparities. To disentangle the various forces behind
the dynamics of regional inequalities, it is useful to decompose the overall level of
inequality into contributions by the factors mentioned above.
The organization of this paper is as follows: section 2 reports the empirical
findings with respect to overall interprovincial inequality in the reform era. Section
3 attempts to explore the forces behind the changes in overall regional inequality
by decomposing the overall inequality. The relationship between reforms and the
change in overall interprovincial inequality is discussed. Finally, the prospect with
respect to interprovincial disparities for the 1990s is explored.
2. Overall interprovincial inequality
To answer question (1) above, a number of inequality indices are employed to
measure the degree of interprovincial inequality of real per capita GDP for the
period 1978-1989. 4 Presenting the results for different indices ensures that the
trend of interprovincial inequality is robust. 5 As we shall point out below, the
indices have different degrees of sensitivity with respect to transfers at different
parts of the income distribution.
Of the indices included, the Gini coefficient and the coefficient of variation
(CV) are well known. Less popular is the class of Generalized Entropy (GE)
measures which has many desirable properties (see, for example, Shorrocks,
1984). The formulae of the GE class of indices may be found in Appendix A. The
class of GE measures depends on a parameter c. In this paper, the values - 1, 0
and 1 are consecutively assigned to c; in Table 1, they are denoted by GE(- 1),
GE(0) and GE(1) respectively. When c is less than 2, the index is more
4 All the provincial data of GDP from 1978 and 1989 are from State Statistical Bureau (1990) to
ensure comparability. Data after 1989 may be collected from provincial statistical yearbooks. However,
for many provinces, the data available are those of gross national product (GNP) and not GDP. To
ensure comparability, our analysis only includes GDP data up to 1989.
5 The indices included satisfy the Pigou-Dalton transfer principle, that is, a mean- and rank-preserving
redistribution of income from the rich to the poor reduces inequality. It is well known that their
rankings of two income distributions always coincide with that of the Lorenz dominance criterion if
and only if the corresponding Lorenz curves do not intersect (see, for example, Sen, 1973). However,
when the Lorenz curves intersect, rankings by the inequality indices may be different from each other.
356 K.-y. Tsui / Journal of Development Economics 50 (1996) 353-368
Table 1
Trends of interprovincial inequality based on real per capita GDP
Gini CV GE(-1) GE(0) GE(1)
1978 0.226 0.445 0.08 0.081 0.087
1979 0,222 0.436 0.077 0.078 0.084
1980 0.222 0.434 0.077 0.078 0.083
1981 0.22 0.433 0.075 0.077 0.083
1982 0.215 0.422 0.071 0.073 0.078
1983 0.215 0.422 0.071 0.073 0.078
1984 0.216 0.416 0.072 0.073 0.077
1985 0.216 0.422 0.073 0.074 0.079
1986 0.219 0.426 0.075 0.076 0.08
1987 0.221 0.428 0.077 0.077 0.082
1988 0.221 0.429 0.078 0.078 0.082
1989 0.218 0.422 0.077 0.076 0.08
Source: author's calculation based on SSB (1990).
Notes: Gini, Gini coefficient; CV, coefficient of variation; GE( - 1), generalized entropy measure when
c = -1; GE(0), generalized entropy measure when c = 0; GE(1), generalized entropy measure when
c=l.
See Appendix A for the formulae of the inequality indices.
transfer-sensitive in the sense that the index is more sensitive to income transfers
at the bottom end of the income distribution (Shorrocks and Foster, 1987). The GE
index is equivalent to the CV when c = 2. Thus, CV is not transfer-sensitive. It
can also be shown that Gini coefficient is not transfer sensitive (see, for example,
Sen, 1973).
Table 1 and Fig. 1 report the trends of interprovincial inequality with respect to
different inequality indices (see Appendix A for the formulae of the indices). To
facilitate comparison, the values of the indices in Fig. 1 are normalized so that the
values of the indices for 1978 are all equal to 1. Without exception, the trends
1
0.98
0.96
~ 0.94
i 0.92
0.9
-~ 0.88
0,86
0.84
0.82
1978
I I I I I I I I I I I
1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989
Year
ELSEVIERสมุดรายวันของปี 50 (1996) 353-368สมุดรายวันพัฒนาเศรษฐศาสตร์การปฏิรูปเศรษฐกิจและ interprovincialความเหลื่อมล้ำทางในจีนจุ่ยไก่ยืน *ภาควิชาเศรษฐศาสตร์ มหาวิทยาลัยจีนแห่งฮ่องกง Hong Hong Kong, Shatin NTรับ 6 1993 กรกฎาคม รุ่นสุดท้ายที่ได้รับ 20 1995 มีนาคมบทคัดย่อบันทึกนี้เป็นความพยายามที่จะสำรวจประเด็นที่สอง: (1) แนวโน้มของความไม่เท่าเทียมกัน interprovincialในยุคปฏิรูป 1978 ลงในจีน และ (2) ปัจจัยอยู่เบื้องหลังของอสมการ interprovincial ใช้ล่าสุดออกผลิตภัณฑ์ในประเทศจังหวัดมวลรวม (GDP)ข้อมูล เราได้แสดงให้เห็นว่า อสมการ interprovincial ลดลงใน fn'st ในครึ่งทศวรรษ 1980แต่เริ่มต้นเพิ่มขึ้นในครึ่งหลังของทศวรรษ 1980 เข้าใจผลกระทบของรายสาขามีความไม่เท่าเทียมกันโดยรวมใน GDP ต่อหัวจังหวัดปฏิรูปในอสมการ interprovincialแยกเป็นการจัดสรร โดยภาคหลัก รอง และระดับตติยภูมิ ดูเหมือนว่าซึ่งแนวโน้มของความไม่เท่าเทียมกัน interprovincial ได้รับส่วนใหญ่เกิดจากการปฏิรูปของ การภาคอุตสาหกรรมซึ่งทำให้ค่าหุ้นใหญ่ภาคสอง นอกจากนี้ แห่งชาติข้อมูลรายได้จากระบบบัญชีแห่งชาติสังคมนิยมใช้ระบุถึงผลกระทบของขั้นตอนทรัพยากร interprovincial ในความเหลื่อมล้ำทาง interprovincial ผลการศึกษาดูเหมือนจะ แนะนำการปฏิเสธที่สัมพันธ์ในบทบาทของการโอนย้ายงบประมาณ redistributiveประเภท JEL: 05 P 2 P52 R12Keywords: China; Interprovincial inequality; Inequality decomposition* E-mail: b040785@mailserv.cuhk.hk.0304-3878/96/$15.00 Copyright © 1996 Elsevier Science B.V. All rights reserved.PII S0304-3878(96)00406-3354 K.-y. Tsui / Journal of Development Economics 50 (1996) 353-3681. IntroductionThere is a widespread fear in China that market-oriented economic reforms,though improving economic efficiency, may increase interprovincial inequalities. 1The spectacular economic growth of such coastal provinces as Guangdong has allbut reinforced the above impression. Some believe that China may be inexorablyentering the initial phase of WiUiamson's inverted-U process of regional development(e.g., Zhang, 1989). In this connection, this paper is an attempt to answer thefollowing two questions:1. Have interprovincial income disparities been increasing in the post-1978 reformera?2. What are the factors behind the dynamics of interprovincial inequalities?With regard to the first question, there is now a small but expanding literaturewhich explores quantitatively the trends of interprovincial inequalities (Lyons,1991; Tsui, 1991; Denny, 1991; Tang, 1991; Yang, 1992a,b; Wei, 1992; Dong,1992; Jiang and Liu, 1992; Liu et al., 1993). The recent release of such provincialincome statistics as provincial gross domestic product (GDP) furnishes us with anopportunity to take a new look at the first question. 2 While one should bear inmind that China's GDP statistics are far from perfect, it is nevertheless a seriousattempt on the side of the Chinese to switch from the socialist national accountingsystem to the United Nation's SNA system. GDP, which includes the tertiarysector, is a more comprehensive index of the overall economic activities in aregion than such output figures as provincial net material product in the socialistnational accounting system. 3With regard to the second question, reform measures and policy shifts mayexercise significant impact on interprovincial inequalities. It is often contendedthat fiscal decentralization and the reorientation of China's regional development1 The basic units of analysis are the provinces, centrally administered municipalities and autonomousregions. The three centrally administered municipalities are Beijing, Tianjin and Shanghai. In thefollowing analysis, they are incorporated into neighbouring provinces. Guangdong does not includeHainan. Owing to data limitation, Xizang (Tibet) and Hainan are excluded. This paper does not lookinto disparities among the subprovincial units (e.g., counties).2 The provincial GDP data used in this paper are from State Statistical Bureau (1990).3 For more detailed discussion of the problems with respect to GDP data of China, see Guo and Han(1991) and Tsui (1993). One referee questions the accuracy of the data for the tertiary sector since theChinese only started collecting data on the tertiary sector for the purpose of compiling GDP data andthe historical data were reconstructed in the 1980s. On numerous occasions, I raised this question withofficials working in the state and provincial statistical bureaux. While the Chinese GDP data are farfrom perfect, it seems to me that they are not derived based on crude back-of-the-envelope calculation.Before any better data are released, the tertiary GDP data are probably the best available. Since theshare of provincial tertiary GDP varies across provinces, it makes a difference whether the tertiarysector is included, The referee further suggests that perhaps it is better to use net material product(NMP), which has a longer time series. The NMP data are, however, not necessarily better for thesimple reason that the historical NMP data were only reconstructed in the 1980s.K.-y. Tsui / Journal of Development Economics 50 (1996) 353-368 355policy in favour of the richer coastal provinces have widened the gap between therich and poor provinces (see, for example, Guo and Wang, 1988; Wang and Bai,1992; Yang, 1990). Furthermore, sectoral reform policies such as the liberalizationof the agricultural sector, the promotion of rural industries and the rapid developmentof the service sector may also induce significant reinforcing as well asoffsetting effects on regional disparities. To disentangle the various forces behindthe dynamics of regional inequalities, it is useful to decompose the overall level ofinequality into contributions by the factors mentioned above.The organization of this paper is as follows: section 2 reports the empiricalfindings with respect to overall interprovincial inequality in the reform era. Section3 attempts to explore the forces behind the changes in overall regional inequalityby decomposing the overall inequality. The relationship between reforms and thechange in overall interprovincial inequality is discussed. Finally, the prospect withrespect to interprovincial disparities for the 1990s is explored.2. Overall interprovincial inequalityTo answer question (1) above, a number of inequality indices are employed tomeasure the degree of interprovincial inequality of real per capita GDP for theperiod 1978-1989. 4 Presenting the results for different indices ensures that thetrend of interprovincial inequality is robust. 5 As we shall point out below, theindices have different degrees of sensitivity with respect to transfers at differentparts of the income distribution.Of the indices included, the Gini coefficient and the coefficient of variation(CV) are well known. Less popular is the class of Generalized Entropy (GE)measures which has many desirable properties (see, for example, Shorrocks,1984). The formulae of the GE class of indices may be found in Appendix A. Theclass of GE measures depends on a parameter c. In this paper, the values - 1, 0and 1 are consecutively assigned to c; in Table 1, they are denoted by GE(- 1),
GE(0) and GE(1) respectively. When c is less than 2, the index is more
4 All the provincial data of GDP from 1978 and 1989 are from State Statistical Bureau (1990) to
ensure comparability. Data after 1989 may be collected from provincial statistical yearbooks. However,
for many provinces, the data available are those of gross national product (GNP) and not GDP. To
ensure comparability, our analysis only includes GDP data up to 1989.
5 The indices included satisfy the Pigou-Dalton transfer principle, that is, a mean- and rank-preserving
redistribution of income from the rich to the poor reduces inequality. It is well known that their
rankings of two income distributions always coincide with that of the Lorenz dominance criterion if
and only if the corresponding Lorenz curves do not intersect (see, for example, Sen, 1973). However,
when the Lorenz curves intersect, rankings by the inequality indices may be different from each other.
356 K.-y. Tsui / Journal of Development Economics 50 (1996) 353-368
Table 1
Trends of interprovincial inequality based on real per capita GDP
Gini CV GE(-1) GE(0) GE(1)
1978 0.226 0.445 0.08 0.081 0.087
1979 0,222 0.436 0.077 0.078 0.084
1980 0.222 0.434 0.077 0.078 0.083
1981 0.22 0.433 0.075 0.077 0.083
1982 0.215 0.422 0.071 0.073 0.078
1983 0.215 0.422 0.071 0.073 0.078
1984 0.216 0.416 0.072 0.073 0.077
1985 0.216 0.422 0.073 0.074 0.079
1986 0.219 0.426 0.075 0.076 0.08
1987 0.221 0.428 0.077 0.077 0.082
1988 0.221 0.429 0.078 0.078 0.082
1989 0.218 0.422 0.077 0.076 0.08
Source: author's calculation based on SSB (1990).
Notes: Gini, Gini coefficient; CV, coefficient of variation; GE( - 1), generalized entropy measure when
c = -1; GE(0), generalized entropy measure when c = 0; GE(1), generalized entropy measure when
c=l.
See Appendix A for the formulae of the inequality indices.
transfer-sensitive in the sense that the index is more sensitive to income transfers
at the bottom end of the income distribution (Shorrocks and Foster, 1987). The GE
index is equivalent to the CV when c = 2. Thus, CV is not transfer-sensitive. It
can also be shown that Gini coefficient is not transfer sensitive (see, for example,
Sen, 1973).
Table 1 and Fig. 1 report the trends of interprovincial inequality with respect to
different inequality indices (see Appendix A for the formulae of the indices). To
facilitate comparison, the values of the indices in Fig. 1 are normalized so that the
values of the indices for 1978 are all equal to 1. Without exception, the trends
1
0.98
0.96
~ 0.94
i 0.92
0.9
-~ 0.88
0,86
0.84
0.82
1978
I I I I I I I I I I I
1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989
Year
การแปล กรุณารอสักครู่..
