The impact of population aging on income inequality in developing
countries:Evidence from rural China
Hai ZHONG ⁎
School of Public Finance and Public Policy,Central University of Finance and Economics,39South College Road,Beijing 100081,China
a r t i c l e i n f o a
b s t r a
c t
Article history:Received 21January 2010Received in revised form 13September 2010Accepted 15September 2010Available online 6October 2010Population aging is an emerging issue in developing countries.In this paper,we argue that it is
largely responsible for the sharp increase in income inequality in rural China at the beginning of
this decade.As a result of the one-child policy implemented in 1979,fewer young adults have
reached working age during this period.This leads to a fall in the ratio of household members in
working age.Regression-based inequality decomposition shows that labor shortages and the
expansion of industrialization significantly increases the return of a higher ratio of household
members in working age to household income while the distribution of this ratio becomes
increasingly unequal.The interaction of two effects significantly increased income inequality in
rural China.
©2010Elsevier Inc.All rights reserved.JEL classi fication:O15J14Keywords:
Income inequality
One-child policy
Population aging
Inequality decomposition
China
1.Introduction
Income inequality in the development context has been a subject of long-standing interest among economists.Since Kuznets's (1955)seminal paper,numerous studies have examined the relationship between inequality and a number of processes associated with development.These processes include industrialization,factor-speci fic technical change,the development and prevalence of education systems,participation of women in the labor force and population aging.The experience of developed countries shows that those processes did not appear simultaneously and occurred across a very large time span.The process of population aging began unfolding only at the end of the 20th century.Compared to other processes,the effects of population aging on income inequality have received little attention.Existing studies on the relationship between population aging and income inequality have mainly explored this relationship in the context of developed countries,and many of them found that population aging accounts for only a minor fraction of the overall increase of income inequality (e.g.,Barrett,Crossley,&Worswick,2000;Bishop,Formby,&Smith,1997;Jantti,1997).
Although economically less developed countries have been slow to recognize population aging as a major public policy concern,their older population groups are growing more rapidly than those of industrialized nations as a result of rapid declines in fertility and the broad diffusion of medical knowledge.In 1975,the majority of the world's population aged over 65resided in developed countries.However,in 2000,more than 59%of persons aged over 65lived in developing countries.In the near future the distribution of the world's elderly will continue to shift considerably.A United Nations (UN)projection estimates that by 2020,67%of persons aged over 65will live in developing countries.In 2000,all developing countries except some Eastern European countries had elderly populations that were less than 7%of their total populations,which is the de finition of an “aging population ”by the UN.However,by 2020,the elderly population of China,India,Asia (excluding South –Central Asia),Latin America and the China Economic Review 22(2011)98–107
⁎Tel.:+861062288775;fax:+861062288535.
E-mail address:haizhong@cufe.edu
.
1043-951X/$–see front matter ©2010Elsevier Inc.All rights reserved.
doi:
10.1016/j.chieco.2010.09.003
Contents lists available at ScienceDirect
China Economic Review
Caribbean will increase to 11.5%,7.3%,10.5%and 8.3%of their total populations respectively (Shrestha,2000).The socio-economic and industry structures are signi ficantly different between developed and developing countries,thus the impact of population aging on income inequality in a developing country may be signi ficantly different to that in a developed country.
The reduction of income inequality is an important policy objective for the Chinese government over the next few years.The national campaign of “western development ”and the government's commitment to “building a harmonious society ”exempli fies their recognition of this problem.Furthermore,these concerns were expressly discussed during the two most important national conferences on the People's Republic of China's (PRC)political calendar,the National Party's Congress and the National People's Congress.1China is among the few developing countries that will step into an aging society first.As a result of socio-economic development and the one-child policy,the fertility rate in China has dramatically dropped from 6.0in 1957to 2.3in 1980,to 1.7since the 1990s (Cai &Wan
g,2006).At the same time,life expectancy at birth has risen continually from 35in 1949to 63in 1975,69.2in 1985,71.3in 2000and 73.2in 2008(Bergaglio,2008;CIA,2008).According to the UN's projection,11.5%of Chinese will be aged 65and older in 2020,the ratio of the working age population to the total population in China will begin to fall from 2010and the absolute number of the working population will begin to fall from 2015(UN,2003).In the 1980s and 1990s,many observers believed that China was characterized by surplus and underemployed rural labor (Bowlus &Sicular,2003;Knight &Song,1995;Taylor,1988).However,by 2003a shortage of rural migrant workers occurred in the Pearl River Delta area,a region with a high concentration of labor-intensive manufacturing enterprises.At that time,most observers believed that this was just a cyclical phenomenon.However,over time,this phenomenon continued and spread to the Yangtze River Delta area,another region dominated by labor-intensive manufacturing enterprises,and even to some central provinces such as Jiangxi,Anhui and Henan,from which migrant laborers are generally sent out (Cai &Wang,2006).One possible explanation is that as a result of the one-child policy,the growth of the working age population has slowed from the beginning of this decade.Consequently,the ratio of China's working population to total population may have fallen since then.
In this paper,we examine the evolution of income inequality in rural China from 1997to 2006,and try to i
dentify its relationship with the demographic changes.We focus our study on rural areas for several reasons.First,the majority (about 75%)of the population in the developing countries reside in rural areas (Anríquez &Stloukal,2008).Thus,such a study might be more useful to the policy makers in the developing countries.Secondly,there are compelling reasons to anticipate that the impact of population aging may be more pronounced in a rural setting.As von Weizsacker (1996)emphasizes,there is a substantial danger of underrating the distributional signi ficance of an aging population if we ignore the critical role of age-related redistributive tax-transfer systems,such as public pension schemes and health care systems,which are usually non-existent or very poorly established in rural areas compared with those for the urban population.Moreover,for those rural workers in a developing country like China,they are more likely to exit the labor force before reaching the of ficial retirement age in urban areas.This is due to a combination of factors including the nature of the work undertaken by the majority of agricultural or low skilled labor-intensive work,and the poorer physical health of workers compared to developed countries.Thirdly,aging in developing countries usually occurs earlier and proceeds more rapidly in rural areas than in the cities (Marcoux,1994,2001;Stloukal,2004).2This is mainly caused by rural-to-urban migration which comprises mainly younger adults and thus increases the proportion of older persons in the villages.Therefore,the consequences of aging are felt most by the rural population.
In this study,we also try to identify the other causes of income inequality in rural China and the ways in which they affect income inequality.While the problem of population aging cannot be solved in the short run,the information about other inequality determinants is important for the reduction of income inequality.
Our paper contributes to the existing literature in a number of ways.It is one of the first studies on the relationship between population aging and income inequality in the context of a developing economy.Since China is one of the first developing countries to experience population aging,our results may be of use not only to the policy makers in China,but also to policy makers in other developing nations.China accounts for about a quarter of the world's population,and the majority of its population resides in rural areas.As the largest developing country in the world,any advancement in the knowledge of causes and consequences of China's rural income inequality and its changes is not only important for understanding the economic development and well-being of the people in China,but also important in a global context.Secondly,as emphasized in Wan and Zhang (2006),since existing studies on income inequality in China are mostly descriptive rather than prescriptive,one area that deserves further research efforts is the cause of the inequality.We employ two different regression-based inequality decomposition methods in this paper.The Shapley value decomposition method allow
s us to identify the relative contribution of each cause of income inequality,including the measure of demographic change.In addition to that,we introduce a decomposition method commonly used in the health-related inequality literature to analyze income inequality.This approach allows us to identify not only the relative contribution of an income determinant,but also the underlying mechanism through which that income determinant affects income inequality.Finally,our study provides more updated information on the evolution of income inequality in China.One primary focus of existing literature on income inequality in China has been on estimating the levels and changes of inequality over time.Due to data limitations,the information for the period after 2002is lacking.Based on newly available cycles of the China Health and Nutrition Survey (CHNS),we find that the level of income inequality has risen sharply between 2000and 2006and a signi ficant portion of this increase can be attributed to the demographic change.
The paper is organized as follows.In the next section,we describe the methods and data used in this analysis.Section 3contains our empirical results,and in the last section we discuss the policy implications arising from these results and submit our conclusion.
1
Held in October 2007and March 2008.
2Calculations based on the data from the National Bureau of Statistics of China (2010)indicate that in 2005,the ratio of elderly people (60and older)to the working young (15–59)is 0.213in rural areas,and is 0.169in the urban areas.99
H.Zhong /China Economic Review 22(2011)98–107
2.Method and data
2.1.Method
In order to identify the impact of demographic change on income inequality,we decompose income inequality at a given time point into its causes that include a variable “ratio of household members in working age ”.The estimation focuses on household data,and income here refers to per capita household income.The three selected time points are 1997,2000and 2006.The interested variable is de fined as the ratio between the number of household members aged from 18to 60and the total number of household members.3Variation of the mean value of this ratio across the three selected time points re flects the trend of demographic change.Decomposition results reveal the overtime changes of the relative contribution of this variable to total income inequality,by which we can identify the impact of demographic change on income inequality.
The conventional approach to inequality decomposition typically decomposes the total inequality either by population groups or factor components,both of which provide limited information on the determinants of income inequality.Moreover,these approaches are unable to control other factors when trying to identify and measure the contribution of a particular variable.However,this is not the case with the regression-based approach (Fields &Yoo,2000;Morduch &Sicular,2002;Wan,2004;Wan &Zhou,2005).
Since the early 1970s,researchers have used regression-based approaches for inequality decomposition (e.g.,Oaxaca,1973;Blinder,1973;Juhn,Murphy,&Pierce,1993;Bourguignon,Fournier,&Gurgand,2001).However,most of the early studies focus on explaining the differences in income distribution between distinct groups rather than quantifying the contributions of speci fic factors to total inequality.Fields and Yoo (2000)was the first study that tried to do the latter.They employed a parametric income-generating function and decomposed the total inequality measure based on the estimated regression equation.The major limitation of this study is that their decomposition method could only be used to decompose the squared coef ficient of variation (CV 2),which is a rather crude measure of inequality and does not even satisfy the transfer axiom (Morduch &Sicular,2002).Morduch and Sicular (2002)provide a more g
eneral regression-based approach to quantify the roles of variables in total inequality.This method can be used to decompose any total inequality measure that can be written as a weighted sum of inequalities of factor incomes.However,one shortcoming of this approach is that the contribution of the constant term and residual term in the regression are not derived from the natural rule of decomposition of Shorrocks (1999)(Wan,2004).Consequently,these two terms may not relate to total inequality if we employ inequality measures such as the Gini coef ficient or CV 2.However,as shown in Podder and Chatterjee (2002),an increase (decrease)of the constant will lower (increase)the value of a Gini index or CV 2.Wan (2004)overcomes this de ficiency.In the Wan (2004)approach,the relative contributions of the variables in a regression equation to the total inequality can be quanti fied by following the decomposition procedure based on the Shapley value (Shorrocks,1999).This approach is applied by Wan and Zhou (2005)to analyze income inequality in rural China.Another advantage of this method is that it can be applied to any inequality measures.
In our analysis,we adopt the Shapley value decomposition method (Wan,2004)to identify the relative contribution of each cause to income inequality at the three selected time points.The first step of the decomposition is to specify a linear income generation equation.4The unit of analysis in our study is the household,and the dependent variable is household income per capita.We include the following ind
ependent variables that are commonly used in the income generation equation:ratio of household members in working age,males as a percentage of working adults,a set of dummy variables for the occupation of the head of the household,average education of working adults,education squared,number of individuals in the household in good health,number of individuals in the household in fair health,number of individuals in the household in poor health,a set of dummy variables for province of residence,a dummy variable for owning a home business and percentage of the community population engaged in non-agricultural activities.The de finitions of those variables and the rationale of their inclusion are presented in Table 1.
This model is supposed to capture only the short-run effects of the explanatory variables on the income generation process.It may yield biased estimates if the variations of some explanatory variables re flect feedback from the variation of income.However,this kind of bias is likely to be small as evidence indicates that income effects on variables such as health and education are usually dominated by permanent income rather than short-term income shock (Smith,1999;Liu,Dowb,Fu,Akin,&Lance,2008).Another concern is that household health variables may be endogenous due to unobservable household characteristics that determine both income and health.Liu et al.(2008)use the longitudinal component of the CHNS data to estimate the in fluence of in
dividual health status on household income.They adopt a fixed effect model to take into account the potential endogeneity problem.The essential pattern found with OLS prevails even after correction for individual-level fixed effect.5
Suppose the regression model can be written as Y =α+˜Y +u ,where ˜Y =∑βi x i =∑y i .y i is the income flow from the i th income determinant.If we denote the deterministic part of the model by ˆY =α+˜Y and a given inequality index by I ,according to Wan (2004),the contribution of the regression residual u to total income inequality can be calculated as I u =I Y ðÞ−I ˆY
.The 3
How to de fine working age is rather subjective.Sixty is the common retirement age for male workers in urban China.We also tried 55and 65,in both cases the results lead to the same conclusions.We also tried to use “labor ratio ”to measure the demographical changes.It is de fined as the ratio between number of working adults (working on either farm job or off-farm jobs)in a household and the total number of members of that household.The results are very similar to those reported in this paper.
unequal4The Shapley decomposition method can also be applied to non-linear regression models.We have tried a semi-log regression model,and the results are not qualitatively different from the standard linear
model.Since the WVW method discussed below can only be applied to the standard linear model,we use the standard linear model in this study in order to make the results comparable.
5With the cross sectional data,we run an 2SLS by using smoking behavior,drinking behavior,diet habits,health knowledge as instrument variables.The results are not very different from the OLS estimation.100H.Zhong /China Economic Review 22(2011)98–107
contribution of the constant term to I (Y )is de fined as I α=I ˆY
−I ˜Y .Replacing x i by its sample mean would eliminate any differences in x i across individuals.It is easy t o re-compute ˜Y after this replacement,and we denote the result as Y i .The contribution of the i th variable to I (Y )is de fined as I i =I ˜Y −I Y i ÀÁ.This is just the “first-round effect ”.One can obtain a higher round effect by
replacing two or more variables with their sample means.At each round,it is possible to have multiple I i ,which are averaged first and then averaged across all rounds.See Shorrocks (1999)and Wan (2004)for more details on this decomposition method.If the inequality index is Gini coef ficient,the contribution of the i th variable to I (Y )can also be calculated as E y i ðÞ=E ˜Y h i C y i ðÞj rank by ˜Y ,where E is the expectation operator and C (y i )is the concentration index of y i ranked by ˜Y .
One limitation of Wan (2004)and the other aforementioned regression-based decomposition methods is that they can only report the (percentage)share of a particular variable to the total inequality.From a policy-making perspective,we not only need to know the relative contribution of a determinant to total inequality,but we also need to identify how it affects total inequality.In other words,we need to know whether the contribution of a speci fic variable to total inequality is due to its low mean or unequal distribution,or power in the income generation process.Therefore,in our study,we also adopt the decomposition method used in the health-related inequality literature (Wagstaff,van Doorslaer,&Watanabe,2003)to analyze income inequality in China.This method allows us to further decompose the relative contribution of a particular income determinant to the total inequality into mean effect,distributional effect and income generation effect.
Wagstaff et al.(2003),hereafter WVW,propose a method of decomposing a concentration index at a given time point.The concentration index quanti fies the degree of inequality in a speci fic variable across the income distribution.It is de fined with reference to the concentration curve.On the x-axis,the concentration curve graphs the cumulative percentage of the population ranked by income,and on the y-axis,the cumulative percentage of the interested variable.If there is no inequality,the concentration curve is the 45°line.The concentration index is de fined as twice the area between the concentration cu
rve and the line of equality (the 45°line).In our analysis,the interested variable is income itself,and therefore,the concentration curve is equivalent to the Lorenz curve and the concentration index becomes the Gini index.
Table 1
Variables used in the analysis:de finition and rationale for inclusion.
Variables
De finition and rationale Dependent variables
Per capita household income In the CHNS,questions on income and time allocation probe for any possible activity each person might have
engaged in during the previous year,both in and out of the formal market.Full income from market and non-
market activities can be imputed.Inclusion of non-monetary government subsidies such as state-subsidized
housing is an especially important advance.Household income in our study is de fined as the sum of all market
earnings across the household and then the total value of all other non-market goods and services produced
within that household is added.This includes the total value of the following components:wage income,home
gardening income,farming income,livestock income,fishing income and business income.Total household
incomes are de flated by using region/year speci fic de flators.The regional de flators are not provided in 2006
data,so we de flate the incomes by 2004de flators.
Independent variables Ratio of household members in working age
This variable is de fined as the ratio between the number of household members aged from 18to 60and the
total number of household members.Due to the lack of a well-established public pension system and income
security programme,this variable is expected to have a positive relationship with household income,and
under strong in fluence of demographic changes.
Males as %of working adults Most jobs in the rural area are physically demanding.This variable therefore may be positively related to
household income.
Occupation of household head De fined by a series of four dummy variables.The first category includes of fice staff,administrator,manager,
village cadre,etc.The second and third categories are farmer and worker respectively.The fourth category
includes all the unspeci fied others.The reference category in the Shapley value decomposition is farmer.
Occupation is strongly associated with income.
Average education of working adults In years.Higher education level may increase the opportunity to obtain an off-farm job,which may positively
relate to household income.
Education squared To take into account the non-linear effect.
Number of household members with good (fair,poor)health These three variables are based on the respondent's self-assessed health status,which is one of the following four choices:excellent,good,fair,and poor.Then the number of individuals in each category in a household is
summed.To avoid potential mutilinearity problem,the variable “number of household members with
excellent health ”is not used in the regression.Self-assessed health status is a well-validated general health
measure.In the rural area,individuals with fair or poor health may need home care from other household
members,and have to reduce labor supply.Therefore,number of household members with poor (fair)health
may be negatively related to household income.
Province of residence A dummy variable for each of the provinces in the sample.These variables capture income inequality between
regions.
Owing a home business Indicator of whether the household has a small handicraft or small commercial business.
%of community population engaged in non-agricultural activities We use this variable to capture the different degrees of industrialization across regions.A higher percentage of the population in a community engaged in non-agricultural activities may imply that it is easier to find a wage
job.This variable may positively relate to household income.101H.Zhong /China Economic Review 22(2011)98–107
One way to calculate the Gini index is:
G =
2n μ∑n i =1y i R i
−1ð1Þwhere y i is income for individual i ,R i is the rank of income for that individual,μis the mean of income and n is the number of individuals.Suppose we have a linear regression model linking the income variable y to a set of k determinants,x k :y i =α+∑k βk x ki +εi
ð2Þ
if we substitute Eq.(2)into Eq.(1),it can be rearranged as 6
G =∑k
βk x k μ C k +GC εμð3Þwhere x k is the mean and C k is the concentration index for variable k ,and GC εis a generalized concentration index for εi .In this way,we are able to decompose the relative contribution of a particular income determinant to the total inequality (βk x k μ:C k )into
mean effect (x k μ),distributional effect (C k ),7and income generation effect (βk ).Furthermore,we can also consider βk x k μas the elasticity of income with respect to determinant k ,and this means that the
estimated inequality in income can be expressed as a weighted sum of the inequality in each of its determinants;the weights are the income elasticities of the determinants.GC εμis the error term.
2.2.Data
The CHNS is an ongoing international collaborative project between the Carolina Population Center at the University of North Carolina at Chapel Hill and the National Institute of Nutrition and Food Safety at the Chinese Center for Disease Control and Prevention.Although it was designed to examine the effects of health,nutrition,and family planning policies and programmes implemented by national and local governments,the CHNS contains detailed information on the household and individual economic,demographic and social factors that are necessary for our analysis.
The first round of the CHNS was collected in 1989.Six additional cycles were collected in 1991,1993,1997,2000,2004,and 2006.The survey covers nine provinces (Guangxi,Guizhou,Heilongjiang,Henan,Hubei,Hunan,Jiangsu,Liaoning and Shandong)that vary substantially in geography,economic development and public resources.A multi-stage,random cluster process was used to draw the sample surveyed in each of the provinces.Counties in the nine provinces were strati fied by income (low,middle and high)and a weighted sampling scheme was used
to randomly select four counties in each province.In addition,the provincial capital and a lower income city were selected when feasible.In two of the nine provinces,other large cities had to be selected.Villages and townships within the counties and urban and suburban neighborhoods within the cities were selected at random.For the 1989,1991and 1993surveys there were 190primary sampling units,and a new province and its sampling units were added to the 1997survey.Currently,there are about 4,400households in the overall survey which cover approximately 19,000individuals.In each cycle,about 70%of observations are drawn from rural areas.In our study,we will use the five most recent cycles to identify the trend of income inequality in the past decade and a half.After omitting unusable observations,the sample sizes are 2325(for 1993),2627(for 1997),2557(for 2000),2940(for 2004)and 2984(for 2006).We will use the 1997,2000and 2006cycles for the decomposition analysis.
3.Empirical results
In this section,we report the level of income inequality in rural China calculated from CHNS data at five selected time points:1993,1997,2000,2004and 2006,and compare them to the results in the existing studies.We then present the decomposition results for the income inequality at three given time points in the past decade:1997,2000and 2006.
3.1.Evolution of income inequality in rural China from 1993to 2006
Estimates of income inequality in rural China vary according to the different datasets used.The key reason is that almost no survey is national representative,and the coverage of most surveys is different from each other.When different combinations of provinces are used to calculate the inequality measures,results vary signi ficantly (Gustafsson &Li,2002).Moreover,income may be de fined differently in each survey.Therefore,it is not helpful to compare the estimates of inequality from different datasets at a given time point.Instead,a comparison of the estimates of inequality from a given dataset across different time points is much more helpful to measure the evolution of inequality in rural China.
Currently,most studies on rural inequality in China are based on one of the following three surveys:the rural household survey by the National Bureau of Statistics (NBS),the Chinese Household Income Project (CHIP)and the household survey collected by 6For the proof,please refer to Wagstaff et al.(2003).
7C k is the distribution of variable k across different y i .Wan (2004)shows that the contribution of a particular income determinant k to the total inequality can
be written as E y k ðÞ=E y ðÞðÞC k j rankedby ˆy
,
where ˆy is the predicted value of y i based on the deterministic part of the regression model.While this expression is similar to the term βk x k =μðÞ:C k in Eq.(3),C k j rankedby ˆy
cannot be used to measure the real distribution of variable k across income levels since ˆy is not the real income of individual i .102H.Zhong /China Economic Review 22(2011)98–107
版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。
发表评论