Income Inequality in China: 1995 – 2006
Master Thesis
August 2008
Author: Thesis Supervisor: Co-evaluator: Qing Zhao Dr. Gábor Péli Dr. Dirk Akkermans Faculty of Economics and Faculty of Economics and Faculty of Economics and Business Business Business University of Groningen University of Groningen University of Groningen The Netherlands The Netherlands The Netherlands
TABLE OF CONTENTS:
ABSTRACT………. 3
1. INTRODUCTON……….…………...……….4
1.1 Background………5
1.2 The Features of Inequality of Income In China………..…...6
2. LITERATU REREVIEW………...7
2.1Economy Growth and Income Inequality……….…….7
2.2 Methods for Analyzing Income Inequality……….9
2.3 Causes for Income Inequality………...11
3. HYPOTHESES..……….………...14
4. METHODOLOGY AND DATA…………..……..………..….18
4.1 Model Explanation……..……….………18
4.2 Data Sources and Sample Description...………..23
5. OVERVIEW THE INEQUALITY OF CHINA...………...25
6. RESULTS AND DISCUSSIONS………...………...26
6.1 Description of Variables……..……….26
6.2 Regression Results……….………..28
6.3 Discussions...………30
7. CONCLUSION……….……….32
7.1 The Effect of Foreign Investment……….33
7.2 The Effect of Human Resource………33
7.3 The Effect of Population...……….………..33
REFERNCES……….……….35
Income Inequality in China 1995 – 2006
Abstract
Inequality of income has been attracted attention for a long time. However, most existing literatures about the income inequality focus on developed countries, such as UK and USA. Different in economy and social institutions, the determinants of income inequality differ in different countries. The condition of post socialism nation is more complicated, since it has experienced the transition from planned economy to market economy. In this paper, I take China as the example to examine the inequality of income distribution.
Chapter 1: INTRODUCTION
It is surprising that average growth of China has remained at 8% since the policy of reform and open started 30 years ago. The income gap has attracted much attention simultaneously.
China’s economy reform has originated from a kind of distribution system of egalitarianism. Planned economy hampered the economy efficiency. The relative widening income gap, in the first instance, will bring about the changes of incentive mechanism, enhance economic efficiency and promote economic growth with positive impact. In the process of economic growth, income gap has not narrowed automatically but constantly accelerate along with economy growth.
There is an ancient saying in China “No worries about scares but inequality”, which means people prefer poverty to unequal treatment. During the past ten years, the interaction between income inequality and economic growth has been flourished in the academic study. Many economists observed that relative income equality is concomitant with persistent economic growth in many East Asian countries and regions while inequality-increasing and economy stagnancy are coexisting in the South American countries. Thus, the interaction between them is delicate. From the existing empirical studies, people have had different findings on this issue, at the same time; researchers propose a lot of questions on the income gap and economic growth to be answered. The goal of this paper on the basis of China data is to examine the trend of income inequality and find which determinants are important for the income inequality over the period from 1995 to 2006.
economy transition. China, which implemented the policy of reform and open in 1978, has transited from state-owned economy to market economy. Studying the trend of income inequality in China is useful for investigating the present condition of income inequality in developing countries which experience economy transition.
A striking feature of China’s rapid development during the past two-decade is naturally imagined with very huge changes in the distribution of income. Although some developing countries struggle to catch up with advanced countries from 1950s, they have been still getting relative needy (Lin and Liu 2003). Kuznets (1955)
suggests that during the course of secular economic growth of a country, income inequality first increases but begins to decline after reaching a peak (Ram, 1995). Scholars both in China and abroad have worried about that the phenomenon of income polarization. In short, the rich getting richer and the poor getting poorer, which indicates income distribution is so unequal that inequality of income may threaten economy and social development. It is no doubt that China is rich in labor force because it occupies the largest population in the world. The change of income distribution in labor-abundant country is dependent on the choices of economy development strategies and social policies (Wang 2006). For a developing economy, the inverted U hypothesis could be avoided if it can make full use of abundant labor force and concern the social policy in terms of income distribution as always (Lin and Liu 2003).
1.1 Background
China.
Many reasons lead to expanding income gap in China. But the equality in initial distribution has always been ignored. In the early period of reform and open policy, it was necessary for China to break the planned economy under the egalitarianism, encouraging some people to get rich first and stressing that “giving priority to efficiency with due consideration to equality”.
But in the process of developing market economy, the growing income gap has been caused by the over pursuit of efficiency and ignorance of fairness. If the equality of initial distribution is under oversight, redistribution will get more difficult to be equal and create serious social problems.
To realize perfect combination between efficiency and fairness will help increase consumption’s contribution to the economic growth rate, achieving faster and healthier development.
1.2 The Features of Inequality of Income in China
First, Gini coefficient reflecting the income distribution gap is 0.46 in China, which means that income distribution is so uneven. However, Gini coefficient of urban area is 0.37 and that in rural area is 0.34. In other words, the major income inequality is caused by the income gap between the citizens and peasantry. This situation has less destructive influence on social stability, compared to the contrast between slum and luxury house within the same residence. In short, China’s Gini coefficient is larger, but its impact is relatively small.
Second, it is critical that the starting point is not fair, compared to the distribution of income. Inherent unfairness is objective, because people possess different endowments, and come from different family. Nevertheless, acquired unfairness is related to the social policy choice. People are mainly dissatisfied with the access to education, health opportunities, the chance of movement, etc.
are different in the treatment of corporate income tax; the wages of migrant workers are not protected, and so on. Inequalities of market access are related to the set of rules and regulations. Overall, the establishment of rules is strict; the implementation is loose; the discretion varies from person to person.
Fourth, the quality of economy growth is not high. Unfair opportunities are originated from unfair start point and process. Then people’s potential and talent can not develop into creativity, and turn into unemployment. The unbalance between investment and consumption, to some extent, relates to the gap in income distribution. Employment opportunities created by per unit of GDP growth have been declining during last several years. The reduction of employment further deteriorated the income distribution.
The paper is organized as follow. Section 2 reviews some earlier researches about the inequality of income distribution. Section 3 proposes my hypotheses with motivation. Section 4 introduces the methodology used in this paper, including describing data in analysis and explaining the methodology. Section 5 takes different measures of inequality to examine the income distribution in China. Section 6 analyzes income inequality from two regression models I built. Section 7 is conclusion.
CHAPTER 2: LITERATURE REVIEW
2.1 Economy Growth and Income Inequality
inequality reduces income growth for the poor, but not for the rich. However, there is a discrepancy between Deininger and Squire’s study and Kuznets hypothesis since their data indicate that many countries which began with low level of per capital income grew rapidly without experiencing increasing income inequality, while countries that grew gradually were infected with significant swings in aggregate measures of inequality (Deininger and Squire 1998). Their explanation, which is similar to Lin and Liu (2003), is that the evolution of income and inequality is affected by initial conditions and policies which are adopted. It is usually expected that post-communist countries would experience income inequality over the period of their transformation into market economy. Kattuman (2001), whose data is on the basis of Hungarian Household Budget Survey, points that the increasing earnings inequality is accompanied with the efficacy of income taxes and the importance of state transfers. His findings do indicate that state policies play a significant role in determining the change in income distribution. Stack (1978) also finds that the level of development has no statistically significant relationship with the level of overall income inequality. Exports/GDP and debits on investment income, which are world-economy indicators, have a significant impact on income inequality within nations independent of the level of economic development (Stack 1978). When comes to political issues, the presence and intensity of democratic political institutions, the degree of equality in the distribution of political power, and the degree of working-class political participation are viewed as effective means for reducing the income inequality through state action such as minimum wage laws, full employment policy, and progressive income taxation (Stack 1978). His regression involves indicators of level of development, political organization, and world economy on the overall degree of income inequality. His results obviously predict that the greater the involvement in world economy the greater the inequality, which is in line with my hypothesis in terms of foreign investment.
unemployment and income inequality. Nevertheless, a lot of empirical studies show that this diffusion effect or radicalization effect can not realize by itself. In 1950s, the United Nations set the goal of economy development to many developing countries and these countries have reached the achievement in terms of economy. However, people’s living has not improved. To solve poverty, income inequality and unemployment is much more important than purely economic growth.
2.2 Methods for Analyzing Income Inequality
There are mainly three popular methods for analyzing income inequality: decomposition analysis, cross-section regression and time-series regression.
Decomposition method can be divided two categories:
The first one concentrates demographic characteristics. Shorrocks is representative. He focuses on the influence of population subgroups, such as age, gender, race (Shorrocks 1982). This kind of method in terms of population subgroups raises problems with regard to the proper decomposition rule and the constraints placed on the choice of inequality measures (Shorrocks 1982).
The second category of method used for examining the income inequality is to analyze the contribution of each income source to the total inequality. A number of studies have considered the disaggregation of income into different factor components and proposed methods for decomposing the overall inequality value into the corresponding component contributions (Rao 1969; Pyatt, Chen and Fei 1980). However, decomposition approach is that practical application often only control for one variable at a time (Parker 2000).
Parker (2000) thinks time-series regression has three advantages: first, time-series is appropriate for explaining the pronounced rise in income inequality observed in many economies; second, it is intended to handle the multiple explanatory variables that have been tied to inequality in the literature; third, the method has gained an apparent success in explaining a substantial portion of inequality.
Creedy (1997) applies cross-sectional method to answer whether cross-sectional comparisons between countries can provide a good prediction of lifetime inequality differences if income mobility is similar. Cross-section analysis has often criticized for restraining data at either the firm or the employee level, but rarely both (Parker 2000). So this method is not as popular as the above two.
Explaining the level of and trends in inequality is an intriguing topic but one that is often dependent on a researchers’ particular approach to inequality measurement
(Cowell and Jenkins, 1995). Sometimes the approach is simply one that accords with intuition; sometimes principles of applied welfare economies or statistical analysis are invoked. The particular approach adopted is important: the issue of the “explanation” of inequality is not just a matter of computational procedure but can significantly affect our understanding of economic inequality and can potentially guide the design of economic policy (Cowell and Jenkins, 1995).
Quantitative discussion of the extent of world income inequality presupposes a coherent picture of the world income distribution expressible as a function of inequality between nations and inequality within nations: the same point would apply on a smaller scale to the assignment of inequality components to the distribution within and between constituent subgroups of the target population in a single country (Cowell and Jenkins 1995). Therefore, it is reasonable to apply methodology of
inequality, some of the internal-development variables are important explanatory variables in my model.
2.3 Causes for Income Inequality
Jenkins (1996) find three main reasons for income inequality: (1) the greater the income growth the better-off the income group; (2) the increasing gap between the poor and the rich was accompanied by changes in the clustering of incomes in between; (3) he attributes more income changes to changes along with work status and the labor market rather than to demographic changes.
Gender earnings inequality has been a sensitive topic for a long time. Narrowing education period and gender discrimination in China widen the income gap between female and male (Zhang, Han, Liu and Zhao 2008). Women in China still experience unfair employment opportunities.
The income gap between skilled and unskilled occupies a lot in total income inequality. Hanson and Harrison (1999) suggest that the wages of more-educated, more-experienced rose more quickly, in contrast to less-educated and less-experienced when Mexico experienced a trade reform.
In addition, income gap between urban and rural area can not be ignored, which is obvious in China. Researches demonstrate that growing income gap among different regions and urban-rural income difference are the major performance of serious income distribution inequality in China (Kanbur and Zhang 1999). The income gap between urban and rural areas has been the driving factor behind the rising overall inequality in China (Yang 1999). Prices of agricultural products, agricultural tax and fees system, labor market segmented by urban and rural areas, farmer discrimination in employment are important determinants to the income gap between urban and rural regions.
Beyond the existing literature about the causes of income inequality, there are other factors which contribute to income inequality in China after the policy of reform and open:
Unfair opportunity and policy to the urban and rural residents
in rural area restricts using higher degree mechanization and modern technology; otherwise it will lead to large unemployment, thereby constraining productivity improvement. Second, existing fiscal and taxation system in rural area strengthen the income gap between urban and rural areas. There are three manifestations: a) rural public goods investments have been committed by farmers themselves. Supply of public goods is an important function of government, but they are only provided in the cities; b) farmers also have to pay for the funds for compulsory education. Compulsory education with the nature of public goods should be afforded by the state. This also results in lower level of education, higher dropout rate, hampering rural economy development; c) peasantry contributes much more than citizen in term of the ratio tax to income. Third, industrial product price is much higher than agricultural product price. Peasantry does not have a given organization on behalf of their interest, and lack of information, supervision and participation right and capacities in the process of domestic price formation. They have already been used to accepting established price which exploits their interest, rather than to bargain under the fairness principle in market economy. In addition, the sale dealers, making use of products in different circulation links, have gained high profits. Consequently, profits of peasantry have not been assigned reasonably.
The existence of improper income from the interspace between the old and new structure
There has been inevitable vacuum of the law system since the perfect market economy system has not been established and fiction has emerged in the transition period of the old and new system. Some people obtained benefits through justice vacancy and illegal operation. Lack of supervising and controlling government induces some officials, leaders of state-owned enterprises abusing power for personal benefits, resulting in corruption and bribery. There is a huge loss of state assets when state-owned enterprises began restructuring.
Policy discrimination
income in eastern regions is apparently higher than that in inland, which enables to attract abundant intellectual and other resources, aggravating the income gap between eastern and western residents. Monopoly of different sector has brought about long-term employee earnings difference, in particular such industries in long-term exclusive monopoly position as water, electricity, gas and telecommunications industries, etc. With the monopoly power, these sectors have access to the high monopoly profits in the productivity process. The monopoly benefits through various ways eventually or partly turn into staff or personal income.
Different natural resource and humanistic community
Available resources, natural condition, transport, population quality and infrastructure are closely related to economy development. Because regions differ in natural resources, transportation, infrastructure, history, population quality, economies among different regions develop at an unharmonious rate. Initial distribution of market economy is based on the resources, which increases income inequality at the very start.
Large unemployment and imperfect social security system
rich and the poor.
A detailed study of the evolution of the distribution of income in China is still much needed for several reasons although a sizable literature has developed on income distribution in China. First, the existing literature is at too much an aggregate lever to permit understanding the mechanism through which income distribution may be affected by exogenous structural changes in the economy or in the socio-demographic structure of the population. Second, it also tends to focus on a single aspect of the problem – e.g. trade, strength of competition, education – while ignoring other aspects and the way they interact to produce the observed change in the distribution of income. Third, most studies make comparisons between female and male, the skillful and the unskillful. Less focuses on the inequality among different regions.
CHAPTER 3: HYPOTHESES
My hypotheses are based on studies from Nielsen and Alderson (1999), Wan (2004)
Hypothesis 1 – Education Expansion
Secondary-school enrollment: as long ago as the middle of the 19th Century, John Stuart Mill (1848) conjectured that the spread of education in a population would decrease inequality (Williamson 1991). The view that the accumulation of human capital reduces inequality is widely shared among social scientists. Using secondary school enrollment as a measure of education expansion, there is a strong negative effect of the spread of education on income inequality (Nielsen and Alderson, 1995). Compulsory Education Law was legislated in 1986, which guarantees all school-age children have access to elementary and secondary education. People are endowed with basic skill for living by the spread of education. Unlike secondary education, higher education grants minority the privilege of getting higher income, resulting in widening income gap. Thereby, I am confident that secondary school enrollment negatively affects the income inequality within province.
Hypothesis 2
variable on inequality is expected to be positive. High population increase tends to generate greater inequality by perpetuating the phenomenon of “surplus labor”, because a large proportion of the labor force remains locked into low income employment in the backward sectors of the economy (Ahluwalia 1976). Bollen and Jackman (1985) suggest that high population increase leads to income inequality by expanding the proportion of national population in low-income groups. Within one country, different income groups grow at different rates because the lower income groups are experiencing a faster natural rate of increase in population (Bollen and Jackman 1985). It is natural to imagine the similar story would happen within province. Rapid population growth produces a large cohort of young and typically lower-paid workers. This influx is expected to increase income inequality by inflating the bottom of the income distribution and by contributing to excess labor supply, further depressing lower incomes and thereby widening wage differentials (Nielsen and Alderson 1995). One-child policy has been carried out in China since 1980s. It is well-known that China is the country with most population. The objective of this policy is to control the fast rate of population increase, because the government has realized the heavy burden from large number of population to the economic development. After two decades, the natural rate of population increase has remained stable, even decreased recently. However, natural rate of population increase in urban area and backward regions is relatively higher than that in rural area and advanced regions. This variable is associated with positive coefficient.
Hypothesis 3 – the effects of labor force shifts
fierce since the labor supply is greater than demand, which mainly impacts the low-income people a lot. They have to face the potential of being dismissed or work more studiously than before to keep their jobs because workers from rural area are willing to provide the same workload but with lower wage. Briefly, farmers have competed with citizens for unskilled jobs. This brings on widening the income gap between the high-income and the low-income people. This is consistent with Kuznets’ hypothesis which indicates the labor force shifts from agriculture sector to the modern sector is increasing income inequality in the beginning. It will terminate until the process of urbanization is completed. Agriculture industry absorbed 46.9% of population in 2004. It is still a long way to complete urbanization for China. Percentage of labor force in agriculture is expected to have positive effect on income inequality.
Hypothesis 4
Gross Dependency ratio: gross dependency ratio in Wan’s (2004) research is expected to have a negative marginal impact on household income. Ye, Ji and Lv (2007) argue that gross dependency ratio is negatively related to the family income. Therefore, disposable income per person should be lower when a large percentage of the population is very young or is very old. Coale and Hoover (1959) first coined the term “burden of dependency”, or the gross dependency ratio to describe those persons who are in a dependent status because of their age (too young or too old to work). Demographic condition is a major determinant of income, and gross dependency ratio is a statistically distinct and quantitatively important influence on the income.
ratio. Besides, social security system has not been established for farmers by now. The elder have to be dependent on the family earners. The income gap between citizen and countryman has widened although they live within province. My data shows that gross dependency ratio is higher in less developed provinces than that in developed provinces. The elder in developed provinces are less dependent because of relatively perfect social security system and children dependency ratio is lower because the natural rate of population ratio has remained stable or even decreased in developed provinces. Such different gross dependency ratio between developed and less developed provinces has increased the income gap more seriously. No matter within province or across provinces, the income inequality will increase if gross dependency ratio rises. Being identical with Wan’s (2004) hypothesis, gross dependency ratio is anticipated to increase income inequality.
The following hypotheses are related to foreign and domestic investment
Hypothesis 5
Foreign investment rate: due to the penetration of economies by investments of
MNCs located in more developed nations, high levels of economic inequality can be found in developing countries. Foreign investment exacerbates inequality because the capitalist accumulation it fosters is so strongly exclusionary and inegalitarian (Evans and Timverlake 1980). Evans and Timverlake (1980) also argue that the occupational structure is distorted by high dependence on foreign capital, bloating the tertiary sector and producing both a highly paid elite and large groups of marginalized workers. That is, foreign investment increases income inequality. The result of a cross-sectional regression analysis from Sullivan (1983) shows that foreign investment is related to uneven development, which in turn is associated with level of income inequality. Sullivan (1983) also reckons that foreign penetration affects the power distribution within nations, and that this results in a more unequal pattern of income distribution.
and knowledge-concentrated industries, most of which distribute in cities and developed regions. This kind of place, with perfect infrastructure, is rich in intellectuals, finance assistance, convenient transportation. The income gap between the highly qualified intellectual employed in MNCs and unskilled has been exacerbated. Within province, citizens enjoy more benefit than countrymen do through investment penetration.
From this view, foreign investment rate is associated with positive impact on inequality.
Hypothesis 6
Domestic investment rate:
In contrast to foreign investment, domestic investment is more likely to contribute to public revenues, more likely to encourage the development of indigenous entrepreneurship and more likely to reinvest profits in homeland (Firebaugh 1992). Moreover, domestic industries are willing to form links with other industries in the domestic economy.
Many socialists are counting on increased domestic investment to lead a moderate economic recovery when there is economic depression, because domestic investment is an efficient stimulation for increasing domestic demand. It can create more employment opportunities. Unlike foreign investment, domestic investment does not “discriminate” against traditional industry. The income gap will be declined by reducing unemployment rate. I suppose that domestic investment will play an important role in decreasing the income inequality.
CHPATER 4: METHODOLOGY AND DATA
4.1 Model Explanation
percentage of the labor force in agriculture, Marxist-Leninist regime as the measures of internal-development.
I also use all the above variables as my explanatory variables except Marxist-Leninist regime. Marxist-Leninist is a dummy variable which is coded 1 for communist countries. Since my observations are within one country, all provinces implement the same economic and political institutions. So this variable is not applicable for my case.
Alderson and Nielsen’s data set has covered 88 countries from 1967 to 1994 and has an unbalanced panel structure. They believe that the structure of data has potential heterogeneity bias and unmeasured time-invariant factors. The fixed-effects model (FEM) and random-effects (REM) model are two widely used methods designed to correct unmeasured time-invariant factors and heterogeneity (Alderson and Nielsen 1999).
Their central model is written as,
yit =α0+∑k=1 Kβkxkit+αi+εit
where α0 and αi represent the general intercept and country-specific intercepts
respectively.
There are mainly three reasons for Alderson and Nielsen (1999) to choose REM rather than FEM: (1) they think FEM may waste much information because the time series is relatively short and much of the variation appears to be between countries; (2) the FEM can not be estimated for models including variables which are time invariant because these variables are collinear with the fixed country-specific intercept specified in the FEM (Alderson and Nielsen 1999); (3) they prefer REM to FEM because they think REM is more efficient than FEM.
independent variables in my case. They are also important as the foreign investment rate. Alderson and Nielsen use the foreign investment rate to measure the foreign investment penetration. In addition, domestic investment rate is also included by
Alderson and Nielsen. Without exception, domestic investment rate is also one of my important explanatory variables. Besides, year dummies and gross dependency ratio are included. Cultivated land ratio is regarded as the control variable.
Secondary school enrollment rate (SSER) refers to the ratio of the gross students
enrolled in the secondary school, regardless of age, to the population aged 12 to 14. It is defined by the Ministry of Education of China. The equation is written as
SSER= 14 -12 aged population school secondary in the students gross *100%
Secondary school enrollment rate is available from statistics yearbook compiled by National Bureau of Statistics of China (NBSC).
Natural rate of population increase (NRPI) is defined as the crude birth rate minus
the crude death rate. The equation is written as
Natural rate of population increase = crude birth rate –crude death rate
I can also get natural rate of population increase directly from statistics yearbook.
Percentage of labor force in agriculture (PLFA) represents the population engaged in
the agriculture.
Percentage of labor force in agriculture =
persons employed Total e agricultur in Population *100%
It comes from the same source as the above data.
Gross dependency ratio (GDR)
A measure of the portion of a population which is composed of dependents (people who are too young or too old to work)
Children dependency ratio and old dependency ratio constitute gross dependency ratio.
Child dependency ratio: ratio of the population aged 0-14 to the population aged 15-64.
Old dependency ratio: ratio of the population aged above 64 to the population aged 15-64.
Statistics yearbook provides detail information of gross dependency ratio, children dependency ratio and old dependency ratio.
Foreign investment rate (FIR) is calculated as the ratio of flow of foreign investment
to stock of foreign investment by Alderson and Nielsen (1999). Foreign investment rate =
investment foreign of stock investment foreign of flow *100%
Although this variable can not be found from statistics yearbook, flow of foreign investment and stock of foreign investment are available. So I can calculate it by myself.
Domestic investment rate (DIR) is defined as the ratio of gross domestic investment to
domestic stock by Alderson and Nielsen (1999).
Domestic investment rate =
stock domestic investment domestic gross *100%
This variable is also easy to obtain because I can collect gross domestic investment and domestic stock from statistics yearbook.
Cultivated land rate (CLR) is the ratio of cultivated land of one province to the total
cultivated land of China, also expressed as a percentage.
Cultivated land rate=
land cultivated total province one of land cultivated * 100%
SSERit=Secondary-school enrollment rate of i province in t year
NRPIit =Natural rate of population increase of i province in t year
PLFAit=Percentage of the labor force in agriculture of i province in t year
GDRit= Gross Dependency ratio of i province in t year
FIRit=Foreign investment rate of i province in t year
DIRit=Domestic investment rate of i province in t year
CLRit= Cultivated land ratio of i province in t year
Year dummies: dum96, dum97…dum06
yit =αo+β1 (SSERit) +β2 (NRPIit) +β3( PLFAit)+β4 (GDRit)+β5 (FIRit)+β6 (DIRit)+β7
(CLRit)+αi+dum96+…dum06+εit
Where yit indicates the inequality measure (Gini coefficient) of i province in time t, i
denotes the province, and t denotes the time. In this regression model, αo indicates the
general intercept, and αi represents the province-specific intercepts.
There are two standard ways of dealing with panel data: random-effect model (REM) which is also known as Error Component Model, and fixed-effect model (FEM).
Alderson and Nielsen prefer REM to FEM. FEM treats the province-specific intercepts αi as fixed effects to be estimated, equivalent to the regression coefficients
of indicator variables for province, while REM treats it as a random component of the error term in the estimation equation.
To control for omitted variables that differ between cases but are constant over time, FEM is recommended. It allows for the changes in the variables over time to estimate the effects of the independent variables on dependent variable, and is the main technique used for the analysis of panel data.
gives wider confidence intervals than FEM.
The generally accepted way of choosing between FEM and REM is running a Hausman test.
The Hausman test detects the null hypothesis that the coefficients estimated by the efficient REM estimator are the same as the ones estimated by the consistent FEM estimator. If p-value is not significant (Prob larger than .05), it is safe to use REM. If gets a significant p-value, FEM should be used.
4.2 Data Sources and Sample Description
Data in my research covers 31 provinces including four autonomous municipal cities (Beijing, Tianjin, Shanghai, and Chongqing) over the period from 1995 to 2006. Chongqing is the youngest region in China; before 1997 it was part of Sichuan province (Wan 2004). Therefore, my sample consists of 30 regions in 1995 and 1996. Taiwan, Hong Kong, and Macao are excluded.
All the data are from National Bureau of Statistics of China (NBSC), mainly in charge of national statistics and national accounts work.
Dependent variable
Gini coefficient: it is a statistics indicator, measuring the degree of income inequality
(in a country or region). Gini coefficient take on values between 0 and 1 with zero interpreted as no inequality.
I apply a simply and convenient method introduced by Yao (1999) to calculate Gini coefficient1.
Most researchers use Gini coefficient as the measure of income inequality (dependent variable). Ram (1995) built a regression model that relates to a measure of income inequality with level of income and square of the income term. In his model, income
1
According to Yao, sample population should be divided into n groups before the calculation. wi, mi and pi represent per income share, average per capital income and population proportion. All the samples are ranked from the lowest average per capital (mi) to the highest average per capital income (mj). Then the equation is written as G=1－
∑
= n i pi 1 (2Qi-wi) Qi=∑
= i k 1 kw , is the accumulation from w1 to wi ,
∑
inequality is measured by Gini coefficient. Wan (2004) proposed a framework for income inequality in rural China. Without exception, he also uses the Gini coefficient as a measure of inequality. When Alderson and Nielsen (1999) studied the impact of foreign investment on income inequality, they still used Gini coefficient as dependent variable. Deininger and Squire (1998) emphasized the effect of Gini coefficient when they study the relation between economic growth and income inequality. The difference between these studies is that they focus on different factors to the inequality.
Independent variables
a) Secondary-school enrollment ratio: it refers to the education expansion.
It is calculated as the ratio of the gross students enrolled in the secondary school, regardless of age, to the population aged 12 to 14. The equation is written as
SSER= 14 -12 aged population school secondary in the students gross *100%
Its effect is expected to be negative.
b) Natural rate of population increase is calculated as the crude birth rate minus the crude death rate. Its effect is expected to be positive.
c) Percentage of the labor force in agriculture is the ratio of labor force in agriculture to the whole labor force. Its impact on income inequality has separated in two periods as Kuznets (1955) supposed. It leads to inequality-increasing in the early period, and has negative effect on inequality after reaching the peak.
d) Gross dependency ratio (GDR) is the ratio of dependent population (aged 0-14 and over 65 years) to the population aged 15-64. Higher gross dependency ratio reduces the disposable income of each member of family. Its effect is expected to be positive.
GDR= 64 -15 aged population population dependent *100%
e) Foreign direct investment rate: The foreign investment rate is calculated as the ratio of flow to stock ([flow/stock]/100). Earlier researches showed that dependence on foreign investment is to be associated with greater income inequality. I expect this ratio has positive effect on inequality in China.
f) Domestic investment rate: it is defined as the ratio of gross domestic investment to domestic stock by Alderson and Nielsen (1999). Its effect is assumed to be negative.
Control variable
Ratio of cultivated land in each province to national total
Wan (2004) finds that the coefficient of cultivated land ratio is negative when he studied rural inequality in China. He argues that regions with larger land are usually more backward and more heavily involved in farming compared to land-scarce but more affluent regions. Statistics yearbook compiled by NBSC shows that it does not change with years. For example, Beijing (capital of China) accounted for 0.26% of total cultivated land in 1995, which is the same in 2006. Instead of using provinces dummies, ratio of cultivated land is used as control variable in my case.
CHAPTER 5: OVERVIEW THE INEQUALITY OF CHINA
Before analyzing the inequality of income distribution in China in detail, I take several inequality measures to see the overall tendency of income distribution. These indexes, computed from National Bureau of Statistic of China (NBSC) income data over the period 1995 to 2006 are in table 1 and table 2 respectively.
--- Insert Table 1 about here ---
0.43 during 1997 and 2003 after the peak. And then it has been declining quickly since 2005 and got to the bottom at 0.3. From this aspect, the overall inequality in China has been relieved.
--- Insert Table 2 about here ---
Table 2 shows that the ratio among high, median and low income. P90 represents high income, and p50 ranks in the middle. P10 stands for low income. It is not difficult to find that any ratio between different levels of income has decreased. The rate between the rich to the middle has decreased from 1.57 to 1.46 after its peak at 1.6 in 1998. The gap between the middle and the poor has also been narrowed. The difference between two income poles, p90 and p10, has fallen by 40 per cent over the period from 1995 to 2006. This suggests that there is an income-increase among middle and low income people.
CHAPTER 6: RESULTS AND DISCUSSIONS
6.1 Description of Variables
--- Insert Table 3 about here
---
Nonstationarity
Table 4 illustrates the results of unit root test of each variable. Unit root test can tell us whether the variable is stationary. The null hypothesis is that there is a unit root and a unit root is an indication of nonstationarity. The principle of judgment is that the null hypothesis will be rejected if the t-statistic of ADF is lower than the given significance level. Given three significances at 1%, 5% and 10% respectively, all t-statistics of ADF are lower than 1% except CLR. However, it is a control variable in my case and it is still lower than 5%. So it is still can be considered as a stationary series. Besides, the p-values are close to zero, which proves they are stationary series again. Gini coefficient is obviously stationary, with t-statistic at -20.17. It is far lower than 1% significance. Domestic investment and foreign investment are also relatively stationary series.
--- Insert Table 4 about here ---
Correlations
Correlation among each variable presents in table 5 ---
Insert Table 5 about here ---
population increase. To avoid the multicollinearity problem, two regression models are included in this paper. One removes natural population increase rate and the other one gets rid of gross dependency ratio and population in agriculture. The correlation coefficients of other variables are lower, below 0.5. For example, the coefficient between secondary school enrollment rate and cultivated land ratio is only -0.093. Domestic investment has little correlation with other variables since the coefficients involving domestic investment are around 0.2 or even lower.
Heteroskedasticity
Table 6 shows the result of white test of OLS method.
--- Insert Table 6 about here ---
P value of F-statistic is 0.12. Although it is higher than 5%, it is not enough significant. To avoid potential heteroskedasticity, I still use WLS method.
6.2 Regression Results
Since Gini coefficient (dependent variable) is quite left-skewed, I transform Gini coefficient to a differencing form to reduce the skewness. After the transformation, the skewness of Gini coefficient is much closer to zero, from -0.68 to 0.3. And it appears much more symmetric (Figure 1).
--- Insert Figure 1 about here ---
--- Insert Table 7 about here ---
The results are shown in the third column of table 9 and are similar to the second column. Secondary education and domestic investment are still influence income inequality negatively and significantly. The coefficients and t-values of gross dependency ratio and foreign investment are both increasing. Their impact has been rising by adding AR (1). The point estimate of the first-order autocorrelation coefficient is -0.1524 and it is significant at 1 per cent level. The p-value of F-statistic is still zero, suggesting a high overall significance of the regression. The adjusted R2 has improved from 0.2492 to 0.3298. DW statistic fell from 2.2197 to 2.1822. It is more close to 2 although it is still above 2. So the whole model has improved and fits the data better.
To make DW statistic more close to 2, AR (2) is included. The results are shown in the last column of table 9. Nevertheless, some outcomes are out of my expectation: the coefficient of secondary education turns out to be positive that means it enlarges the difference of income distribution. In addition, domestic investment also presents opposite influence, in contrast with results in the first two columns. The remaining presents similar picture. The p-value of F-statistic is as same as before. The adjusted R value increased to 0.4291. DW-statistic has approached to 2 approximately (2.0082).
--- Insert Table 8 about here ---
significance. The adjusted R2 value is 0.2351, approaching the corresponding one in the second column of table 7. It is surprising to find that DW-statistic is so close to 2 without adding AR (1) and AR (2). But I still insist on including AR (1) and AR (2) to examine whether the better result can be obtained. The outcomes of including AR (1) and AR (2) are in third column and fourth column respectively. Unfortunately, they are not as close to 2 as I expect. It suggests that it is not necessary to add AR (1) and AR (2) because the first regression model has no significant autocorrelation. But the adjusted R2 value has grown gradually. Both regression models are significant. Secondary education plays an important role in increasing income inequality, which is difficult to explain.
6.3 Discussions
According to my hypotheses, domestic investment and secondary education are supposed to have a negative relationship with income inequality while the rest should be positively related to the income inequality. Due to the relatively high correlation between natural population increase and gross dependency ratio and population in agriculture, the whole analysis is divided into two parts.
Education expansion, as suggested by (Lee, Alderson and Nielsen 1994), plays an important role in decreasing income inequality. In my first regression model, it is indeed narrowing the income gap significantly. However, it behaves completely different in the second model, replacing gross dependency ratio and population in agriculture by natural population increase. The reasonable explanation is that secondary education can not provide everyone with opportunity for further study. Only the ones who can afford the tuition fees have access to the higher education. It is invisible to expel some students born in poor family. So the impact of secondary education on income inequality has two sides.
to the city. Labor force in agriculture is closely related to natural population increase. As mentioned above, natural population increase leads to high labor force in agriculture. Many migrant forces are willing to be engaged in low-income jobs, result in reducing the price of less-skilled labor. Income distribution between the high-income and low-income becomes more unequal.
Gross dependency ratio is another factor related to population. It is no doubt that high population increase can make for higher children dependency ratio. Thereby, these two variables should be analyzed separately. In my first regression model, Gini coefficient increases about 0.1 by increasing 1 unit gross dependency ratio. At present, education investment occupies most in an ordinary family. Parents have realized that the education is critical to children’s future career. The cost of raising children has not reduced although there is usually only one child per family. Improving social security system is an efficient means to relieve the income inequality caused by old dependency ratio.
Foreign investment and domestic investment are both included in the two regression models. Foreign investment is in line with my expectation, with positive coefficients in both models. Ability of attracting foreign investment is not equal among different provinces. Hence, the income inequality is caused by “different treatment”.
Domestic investment, similar to secondary education, negatively affects income inequality in the first regression model while being positive in the second model. This is possibly because that infrastructure, human resource, natural condition differ in different provinces. The return from same amount of domestic investment can not be same among different provinces.
Cultivated land ratio is opposite to my expectation. In contrast to my anticipation, this control variable is attached with negative coefficients in the two regression models. But its t-statistics are not enough significant to influence the whole models.
Although resident income grows a lot after the policy of “reform and open”, the income inequality are also increasing. The natural development of market economy, specific constitutive property in the transition period, the institution factors and other related determinants can influence the income distribution. Restricted by the limited available data, this paper can not cover all the related factors. However, there are still some findings in this paper for further study.
7.1 The Effect of Foreign Investment
China’s foreign trade centralizes in east area, and the economy growth of mid-west area relies on little foreign trade. In 1999, gross imports and exports of twelve coastal provinces occupied 90.6%, and other provinces only accounted for less than 10%. Therefore, people in east provinces enjoy more benefits which the foreign trade brings than mid-west people do. It is reasonable that the foreign investment increase the income inequality among different provinces. Its effect is proven in my both models.
7.2 The Effect of Human Resource
Along with the education development, it indeed enlarges the disparity between the skillful and less-skilled. Since the skillful have higher productivity, they will be engaged in higher income occupation. But I use secondary education, instead of higher education, as the measure of spread education. However, secondary education quality still has difference among east and mid-west provinces which is mainly caused by the different economy development level. In addition, people receiving basic education are not willing to work and stay in less developed areas. They prefer searching more opportunities in developed regions to living in less developed regions. Although secondary education is extended in less developed regions, these provinces can not keep the intellectual from leaving, not to mention attracting the talent people. This may explain why secondary education positively affects income inequality in the second regression model sometimes.
7.3 The Effect of Population
implement “one-child” policy in China.
REFERECES
Ahluwalia, M.S. (1976). Income Distribution and Development: Some Stylized Facts.
The American Economic Review, Vol. 66, No. 2, pp. 128-135
Alderson, A.S. and Nielsen, F. (1999). Income Inequality, Development, and
Dependence: A Reconsideration. American Sociological Review, Vol. 64. No. 4, pp. 606-631
Bollen, K.A. and Jackman, R.W. (1985). Political Democracy and the Size
Distribution of Income. American Sociological Review, Vol. 50, No. 4, pp. 438-457
Coale, A.J. and Hoover, E.M. (1959). Population Growth and Economic Development in Low-Income Countries: A Case Study of India’s Prospects. Annals of the American Academy of Political and Social Science, Vol. 326, pp. 173-174
Cowell, F.A. and Stephen P. Jenkins, S.P. (1995). How much Inequality Can we
Explain? A Methodology and an Application to the United State. The Economic
Journal, Vol. 105, No. 429, pp. 421-430
Creedy, J. (1997). Inequality, Mobility and Income Distribution Comparisons. Fiscal
Studies, Vol. 18, No. 3, pp. 293-302
Deininger, K. and Squire, L. (1998). New ways of looking at old issues: inequality
and growth. Journal of Development Economics, Vol. 57, pp. 259-287
the Tertiary: A Comparative Analysis of Less Developed Countries. American
Sociological Review, Vol. 45, No. 4, pp. 531-552
Firebaugh, G. (1992). Growth Effects of Foreign and Domestic Investment. The
American Journal of Sociology, Vol. 98, No. 1, pp. 105-130
Hanson, G.H. and Harrison, A. (1999). Trade Liberalization and Wage Inequality in
Mexico. Industrial and Labor Relations Review, Vol. 52, No. 2, pp. 271-288
Jenkins, S.P. (1996). Recent Trends in the UK Income Distribution: What Happened
and Why? Oxford Review of Economic Policy, Vol. 12, No. 1, pp. 29-46
Kanbur, R. and Zhang, X.B. (1999). Which Regional Inequality? The Evolution of
Rural-Urban and Inland-Coastal Inequality in China. Journal of Comparative Economics, Vol. 27, pp. 686-701
Kattuman, P. (2001). Income Inequality in Early Transition: The Case of Hungary
1987-1996. Journal of Comparative Economics, Vol. 29, pp. 40-65
Kuznets, S. (1995). Economic Growth and Income Inequality. The American
Economic Review, Vol. 45, pp. 1-28
Lee, C.S., Francois Nielsen, F. and Alderson. A.S. (1994). Income Inequality,
Global Economy and the State. Social Forces, Vol. 86, No. 1, pp. 77-111
Leff, N.H. (1984). Dependency Rates and Savings: Another Look. The American
Lin, Y.F. and Liu, M.X. (2003). Economic Growth and Income Distribution. http://old.iwep.org.cn/jwe/200308/linyifu.pdf
Nielsen, F. and Alderson, A.S. (1995). Income Inequality, Development, and
Dualism: Results from an Unbalanced Cross-National Panel American
Sociological Review, Vol. 60, No. 5, pp. 674-701
Parker, S.C. (2000). Opening a can of worms: the pitfalls of time-series regression
analyses of income inequality. Applied Economics, Vol. 32 No. 2. pp. 221-230
Pyatt, G., Chen, C.N. and Fei, J. (1980). The Distribution of Income by Factor
Components. The Quarterly Journal of Economics, Vol. 95, No. 3, pp. 451-473
Ram, R. (1995). Economic Development and Income Inequality: An Overlooked
Regression Constant. Economic Development and Cultural Change, Vol. 43, No.2, pp. 425-434
Rao, V.M. (1969). Two Decompositions of Concentration Ratio. Journal of the Royal
Statistical Society. Series A (General), Vol. 132, No. 3, pp. 418-425
Shorrocks, A.F. (1982). Inequality Decomposition by Factor Components.
Econometrica, Vol. 50, No. 1, pp. 193-211
Stack, S. (1978). Internal Political Organization and the World Economy of Income
Sullivan, G. (1983). Uneven Development and National Income Inequality in Third
World Countries: A Cross-National Study of the Effects of External Economic Dependency. Sociological Perspectives, Vol. 26, No. 2, pp. 201-231
Wakabayashi, M and Mackellar, L. (1999). Demographic Trends and Household
Saving in China. Interim Report. from http: www.iiasa.ac.at
Wan, G.H. (2004). Accounting for income inequality in rural China: a
regression-based approach. Journal of Comparative Economics, Vol. 32, pp. 348-363
Wang, R. (2006). Analysis of Income Disparity between Rural and Urban Area: The Application of Inverted “U” Theory. Statistics Research, Vol. 4, pp. 14-19
Williamson, J. (1991). Inequality, Poverty, and History. Oxford: Basil Blackwell
Yang, D.T. (1999). Urban-Biased Policies and Rising Income Inequality in China.
The American Economic Review, Vol. 89, No. 2, 306-310
Yao, S.J. (1999). On the Decomposition of Gini Coefficient by Population on Class
and Income Source: A Spreadsheet Approach and Application. Applied
Economics, No. 31, 1249-1264
Ye, H.R., Ji, X.G. and Lv, J.F. (2007). The Effects of the Economic Development
and Demographic Change on the National Savings in Taiwan. Journal of
Management and Strategy Research. pp. 189-208.
Zhang, J.S., Han, J., Liu, P.W. and Zhao, Y.H. (2008). Trends In the Gender
Relations Review, Vol. 61, No. 2, pp. 224-243
Internet
National Bureau of Statistics of China
APPENDICIES
Figure 1 Distribution of Gini coefficient
(a) Gini coefficient
Table 1 Gini Coefficient Year 1995 1996 1997 1998 1999 2000 Gini 0.4356 0.4639 0.4335 0.4125 0.4320 0.4233 Year 2001 2002 2003 2004 2005 2006 Gini 0.4268 0.4309 0.4355 0.4425 0.3994 0.3000 Table 2 1995 1996 1997 1998 1999 2000 p90/p50 1.5576 1.5517 1.5449 1.6149 1.4380 1.4735 p50/p10 1.2608 1.3366 1.2075 1.1801 1.1887 1.1486 p90/p10 1.9639 1.9257 1.8655 1.9057 1.7094 1.6924 2001 2002 2003 2004 2005 2006 p90/p50 1.5298 1.3979 1.4246 1.4422 1.4178 1.4559 p50/p10 1.0879 1.1039 1.0874 1.0859 1.1023 1.0972 p90/p10 1.6643 1.5431 1.5491 1.5661 1.5629 1.5975
Table 3 Variables and Summary Statistic Variable Mean Standard
Deviation
Table 4 Stationarity Test
Augmented Dickey-Fuller test statistic t-Statistic 1% 5% 10% Prob.* Y -20.1741 -3.4483 -2.8694 -2.5710 0.0000 CLR -3.1944 -3.4497 -2.8700 -2.5713 0.0212 DIR -9.0257 -3.9854 -3.4232 -3.1345 0.0000 FIR -6.2690 -3.4481 -2.8692 -2.5709 0.0000 GDR -3.9138 -3.7905 -3.4256 -2.8360 0.1596 NRPI -4.5850 -3.9864 -3.4236 -3.1348 0.0013 PLFA -4.7075 -3.9854 -3.4231 -3.1345 0.0008 SSER -3.5095 -3.4491 -2.8697 -2.5712 0.0083
Table 5 Correlation Matrix
CLR DIR FIR GDR NRPI PLFA SSER
CLR 1.000 DIR 0.173 1.000 FIR -0.174 -0.188 1.000 GDR -0.139 0.239 0.126 1.000 NRPI -0.107 0.110 0.384 0.785 1.000 PLFA 0.341 0.094 0.264 0.624 0.717 1.000 SSER -0.093 0.102 0.174 0.460 0.352 0.362 1.000 Table 6
White Heteroskedasticity Test:
Table 7 Regression Summary Ⅰ Variable Coefficient (t-statistic) Coefficient (t-statistic) Coefficient (t-statistic) Constant (c) 0.0583 (4.1439)*2 -0.0339 (2.2356)** -0.0920 (-4.6781)* Education -0.1945 (-5.9377)* -0.1392 (-3.3318)* 0.1393 (2.4045)** Population increase - - - Agricultural population 0.0118 (1.6151) 0.0035 (0.4837) -0.0083 (-0.9715) Foreign investment 0.0003 (5.1063)* 0.0004 ( 6.6090) * 0.0005 (8.5741)* Domestic investment -0.0312 (-4.6811)* -0.0268 (-5.1747)* 0.0113 (1.8528) Gross dependency 0.0998 (7.0391)* 0.1091 (8.7248)* 0.0813 (5.7632)* Cultivated land -0.0002 (-0.5250) -0.0003 ( -0.9074) -0.0014 (-3.5103)* Dum96 -0.0135 (-1.5440) -0.0055 (-0.1064) -0.0024 (-0.8721) Dum97 -0.0074 (3.0290)**3 -0.0310 (-4.0859) -0.1367 (1.6398) Dum98 -0.0035 (-1.5330) -0.0127 (-5.8033) -0.0266 (-3. 1057) Dum99 0.0089 (4.3087)* 0.0089 (4.9127) -0.0042 (-1.6594) Dum00 0.0038 (7.9238) 0.0027 (6.8298) 0.0017 (5.6329) Dum01 -0.0539 (3.5498) -0.0387 (1.4398) -0.01743 (0.8347) Dum02 -0.0104 (-5.4761) -0.0076 (-3.4856) -0.0095 (-1.7432) Dum03 0.0065 (3.5110) -0.0006 (-0.2931) -0.0398 (-0.5329) Dum04 -0.0157 (-8.7219) -0.0164 (-8.9213) -0.0230 (-9.6791) Dum05 0.0174 (9.6540) 0.0171 (9.3317) 0.0235 (11.9300) Dum06 -0.02124 (-7.3958) -0.0078 (-6.5292) -0.03823 ( -5.3298) AR(1) - -0.1524 (-10.771)* -0.2543 (-12.7000)* AR(2) - - -0.1319 (-8.0439)* Adjusted R-squared 0.2492 0.3298 0.4291 F-statistic(Probability) 33.0181(0.0000) 49.8022 (0.0000) 51.5420 (0.0000) Durbin-Watson stat 2.2197 2.1822 2.0082 2
* indicates significance at the 1 per cent level
3
