a

Intergenerational Transmission of Income Gap for Chinese Urban and Rural Households

Xu Xiaohong (徐晓红)

School of Economics, Anhui University, Hefei, China

　　Abstract: An accurate understanding of the intergenerational transmission of income gap is the foundation for theoretical research and policy formulationto address this issue. This paper has employed the method of twosample instrumental variables to effectively integrate CHIPdata and CFPSdata and correct the temporal income bias, life-cycle biasand coresidence bias, which are common problems in existing studies,and investigated the tendencies of intergenerational transmission of income gap for China’s urban and rural households between 2002 and 2012. Results of empirical study indicate that the intergenerational transmission of income gap for China’s urban and rural households has been on the decline yet the level of intergenerational transmission is greater for urban residents than for rural residents. This level of intergenerational transmission of income gap in China is at a medium international level lower than that of countries like the United States, Brazil and Japan and higher than that of Sweden and Chinese Taiwan. Further analysis of the intergenerational mobility of various income groups suggests the following: the intergenerational solidification of the bottom and top income groups of urban residents has significantly improved, which is the source for the reduction of intergenerational transmission of income gap. Rural residents of bottom income group are vulnerable to falling into the trap of intergenerational transmission of low income. In order to mitigate the intergenerational transmission of income gap, efforts must be made to improve educational allowance policy and increase the opportunities for children from poor and underprivileged families to receive education and to eliminate the divide of labor markets to create equal job opportunities for each and every worker.

　　Keywords:intergenerational income elasticity, method oftwosample instrumental variables,measurement biases,income gap

　　JEL Classification: D10, D31, J62

　　1. Introduction

　　China’s rising income gap over recent years have drawn a great deal of attention from the academia, Among numerous studies, a common approach is to measure the income gap of the same generation of people using Gini coefficient. The perspective of intergenerationaltransmission is seldom followed in the analysis of income gap yetthis approach provides a dynamic explanation of factors other than income gap. In addition, intergenerationaltransmission of income gap is also an important symbol of inequality of opportunities. Compared with the inequality of results, the inequality of opportunities is less acceptable to the public whilean important objective of public policy is to provide equal opportunities for each and every social member. An accurate reflection of intergenerationaltransmission of income gap between China’s urban and rural residents helps provide a comprehensive understanding on the widening income gap in China and test whether policy deviations exist in the adjustment of income gaps, which is of great significance to narrowing income gap and promoting social fairness, justice and harmony.

　　A major indicator for measuring the intergenerational transmission of income gapis intergenerational income elasticity. How to measure inter-generational income elasticity has always been a key problem of academic research. Early economists used the least square method to create a linear model for the estimation of intergenerationalincome elasticity. Due to the lack of data, studies before the 1990s are generally based on single-yearincome data. Becker and Tome (1979) considered that intergenerational income elasticity of the United States is roughly 0.2, which indicates high social mobility. Later studies have generally adopted large representative databases. Using the USPSID database and five-year average of father’s generation’s income, Solon (1992) introduced square termof age to arrive at intergenerational income elasticity of the United States to be 0.4, which indicates very low social mobility. Simple OLS estimation has also been transcended. Methods for the correction of estimation errors such as instrumental variable (IV) method and generalized method of moments (GMM) have been wideapplied. Currently, quantile regression and kernel density technology have been applied for the analysis of intergenerational income elasticity among variousincome groups to arrive at more accurate and detailed estimation results.

　　Some Chinese scholars measured the intergenerational income elasticity of Chinese households with varying results of estimation. Certain problems and controversies still exist in their studies. First, deviations in the measurement of intergenerational income elasticity, which includes: current-phase single-year income is used as the proxyvariable of permanent income, which caused temporal income biaswith downward elasticity; income characteristics of different stages of the entire lifecycle failed to be taken into account while observable income at a randomly selected agepoint led to lifecycle bias; coresidence bias, i.e.,systemic errors caused by insufficient representation of samples due to the omission of grown-up children who do not live with their parents. Second, lack of variety in the dimensions of measurement. Most scholars have employed the data of China Health and Nutrition Survey (CHNS),butgiven the high proportion of rural households in the data and that parent-childrenmatching samples of urban households arrived at using such data are insufficient,existing studies have mainly estimated the intergenerational income elasticity of rural residents or the nation’shouseholdsas a whole. Based on China’s reality of urban and rural divide, it is highly necessary to conduct a separate estimation of countryside and cities for comparative study. What is the extent of intergenerational transmission of income gap between China’s urban and rural households? Results of major biaseswill lead to a misinterpretation of the reality of intergenerational transmission of income gap in China. Based on the measurement of intergenerational income elasticity under the method of two sampleinstrumentalvariables (TSIV), this not only effectively overcomes multiple measurement biasescaused by data deficiencies but provides sample basis for the analysis of cities and countryside from different dimensions. Due to methodological comparability, the resultsof studyhave acquired the basic data for international comparability.

　　Measurement of intergenerational income elasticity using flawed data is vulnerable to a series of biases. (1) Temporal income bias, i.e.,using current-phase income as a proxy variable of permanent income will lead to underestimation of intergenerational income elasticity. One of the methods for the correction of downward deviation of elasticity is to adopt a few years average of the current-phase income of father’s generation as a proxy variable of permanent income. However, due to the consistency of short-term fluctuations of current-phase income, a few years average of income cannot reflect income fluctuations throughout one’s life, and estimated value ofstill has downward bias(Mazumder, 2005). Instrumental variable (IV) is another approach for the correction of downward biasof elasticity, where variablesrelated to father’s income such as the level of father’s education is a common instrumental variable. The level of education usually has a positive effect on income, which means that the instrumental variable method will lead to the absolute biasof elasticity. If the level of father’s education not only affects father’s income but is also likely to affect children’s income, i.e.,related to the explained variable of the model, upward biasof elasticity also become likely. Therefore, what IV estimation provides is the upper limit of intergenerational income elasticity. (2) Life-cycle bias. Thisbiasis inevitable when cross-section data are used, given that father’sand children’s incomes are at different stages of life-cycle. When both fatherand childrenare employed, fatheris atthe late stage of his career while his children are in the early stage of their careers. Grawe(2003)discovered through empirical analysis that using father’s income at an advanced age will lead to underestimation of intergenerational income elasticity. (3) Coresidence bias. Becker and Tomes(1986)developed a model of utility maximization based on the consumption of parents and their investments onchildren, which led to an assumption that family culture increasesintergenerational income elasticity. This effect is in positive correlation with the time children live under the same roof with their parents. Therefore, using the samples that live under the same roof will cause an overestimation of intergenerational income elasticity.

　　2.2Method of TwoSample Instrumental Variablesand Correction of Measurement Biases

　　Method of twosample instrumental variablesis a recently developed approach for the estimation of intergenerational income elasticity with data deficiencies and high goodness of fit. The basic idea of this approach is as follows: information of two independent samples is utilized through instrumental variablesto arrive at “synthesized” parent and child matching data and estimate father’spermanent income and intergenerational income elasticity in two stages.

　　Judging by this process, the method of two sample instrumental variableshas the following characteristics: first, father’s permanent income can be estimated using instrumental variables related to father’s income through cross-generational independent samples of two interval years, i.e.,father’s permanent income is an estimate; second, in two independent samples, father’s and children’s incomes are in their respective lifecycles of income; third, “synthesized” father’s and children’s data do not have the problem of coresidence bias. Obviously, the method of two sample instrumental variables can overcome the above-mentioned biases, greatly increase the size of parent and children’s matching samples and avoid measurement biasesarising from homogeneous small samples. An inadequacy is that the method of twosample instrumental variables also contains biasesof output elasticity, whose value is in between the OLS estimator and IV estimator (Björklund,1997).

　　3. Data and Variables

　　3.1Data Explanation

　　This paper has employed the data of the Chinese Household Income Projects Survey (CHIP) and the Chinese Family Panel Studies (CFPS). CHIPdata have been collected from four rounds of survey conducted in 1988, 1985, 2002 and 2007 respectively andCFPS data have been collected from the test surveys of 2008 and 2009 and official surveys of 2010 and 2012. This paper has adopted the two years of official survey data, extending the period of study from 2007 to 2012.

　　(2) Level of education and occupation. The level of father’s education and occupation observable from children’s samples is aninstrumental variable of this paper. Given that the variableof father’s occupation in 2000 is lacking in rural household survey data, great differences exist in the classification of occupation for2007 and other years. Thus, this paper can only estimate the intergenerational income elasticity of rural households in 2010 and 2012. Measurement of intergenerational income elasticity of urban households is relatively complete and includes the yearsof 2002, 2007 and 2012.

　　(3) Income. Individual annual salary income is adopted, including cash income and in-kind income, which are obtained through aggregation of individual items from the original data. Individual salary income is the most comparable income indicator among urban and rural samples of all years, while other non-salary incomes accounted by the unit of households can hardly be to be allocated to individuals. Abnormal samples with negative values of individual annual incomes have been excluded. Personal annual income for years has been deflated by consumer price index (CPI).

　　4. Measurement Results and Analysis

　　4.1Measurement Results

　　The method of twosample instrumental variables is employed to estimate the intergenerational income elasticity of urban and rural households, with results shown in Table 3 and 4. With 1988 as father’ssample, the intergenerational income elasticity of urban households for 2002, 2007 and 2012 is 0.4720, 0.3708 and 0.3272 respectively and the intergenerational income elasticity of rural households is 0.4348 for 2010 and 0.2872 for 2012.

　　UsingCHIP data andCGSS data, Chen Lin and Yuan Zhigang (2012) estimated the intergenerational income elasticity of urban and rural households in 1988, 1995, 2002and 2005;of which, the intergenerational income elasticity is 0.51, 0.42, 0.33 and 0.30 for urban households and 0.42, 0.28, 0.22 and 0.24 for rural householdsrespectively. As pointed out by the author in this paper, as single-year incomes of fathersand children have been employed, results of calculation havethe biasof downwardelasticity. Using CHNSdata, He Shijun and Huang Guitian (2013) estimated the intergenerational income elasticity for Chinese households to be 0.66, 0.49, 0.35 and 0.46 respectively for 2000, 2004, 2006 and 2009. As noted by the author, such measurement results have the biasof upward elasticity arising from coresidence bias. The study ofWang Haigang (2005) containsnot only temporal income biasand coresidence biasbut life-cycle biasas well. Considering that this study employs cross-section data, father’s age in the sample is relatively high with average age of various samples around 52 while children’s age is relatively young, i.e.,around 22 on average. Among them, the maximum values of father’s and children’s age even reach 79 and 76. As mentioned above, either father’s age being too old or children’s age being too young will lead to underestimation of intergenerational income elasticity. It can be known from the comparison of results of existing domestic studies that the intergenerational income elasticity estimated in this paper has effectively overcome multiple measurement biases, thus making the results more accurate and reliable.

　　4.2 Analysis of Intergenerational Transmission of Income Gaps between Urban and Rural Households

　　Accurate estimation of intergenerational income elasticity is the foundation for the analysis of the intergenerational transmission of income gap. It is not difficult to see from the measurement results of Table 3 and Table 4 that the intergenerational transmission of income gap among China’s urban and rural households is characterized by the following features: first, the intergenerationalincome elasticity of urban and rural households has been on the decline; second, intergenerational transmission of income gap is greater for urban households than for rural households. On the whole, the intergenerational income elasticity of urban households is 0.4720 for 2002, 0.4130 for 2007 and 0.3590 for 2012. The intergenerational income elasticity of rural households is 0.3558 for 2010 and 0.2703 for 2012. The following section provides an analysis on the implications of the above-mentioned results and reasons for such variation from three aspects.

　　(1) Variations of intergenerational income elasticity of China’s urban and rural households share the same tendency with those of Ginicoefficient. According to the National Bureau of Statistics (NBS), China’s Ginicoefficient increased between 2003 and 2012 before gradually declining since 2009 and Gini coefficient is greater for urban households than forrural households. As shown by empirical studies, the higher the level of income inequalities is for a country, the lower intergenerational income mobility will be. This reverse correlation is referred to as the “great Gatsby curve”(Corak, 2013). Stiglitz (2013) described the dynamic process in which greater income inequalities will lead to less equality of opportunities which in turn causes more income inequalities as a “reverse dynamic” and “vicious cycle.” Variations of China’s intergenerational income elasticity indicate that with the slow decline of Gini coefficient over the recent decade, the intergenerational transmission of income gap for Chinese households has also demonstrated the tendency of reductionand intergenerational income mobility has somewhat increased. This process of change is closely related to China’s educational reform, labor market reform and reform of income distribution system. Judging by educational reform, resuming college entrance examination system and constantly improvingcompulsory education system have generally enhancedthe level of education for children from all backgrounds and particularly provided opportunities for children from low-income households to increasetheir income. Although Solon (1992) considers that rising return to education will lead to an increase of intergenerational income elasticity, rising return to education has significantly positive effects onincentivizing educational investment and increasing the enthusiasm of workforce. Theoretical and empirical studiesbothsuggest that increasing the level of children’s education is akey pathway for the promotion of intergenerational income mobility. Judging by labor market reform, the abolishment of hereditaryemploymentand internal recruitment, the implementation of labor contract system and the discontinuation of job assignment uponcollege graduation since the 1980s have created a more flexible and open labor market for the workforce (including rural migrant workers) after the mid-1980s. Between 2002 and 2012, the years of birth that correspond to intergenerational income elasticity are between 1962 and 1972, between 1967 and 1977 and between 1972 and 1982. By the average employment age of 20 years, these groups of people had not been employed until after the mid-1980s. Hence, China’s educational reform, labor market reform and income distribution reform introduced in recent years to narrow down income gap have to some extent reduced income gaps and mitigated the intergenerational transmission of income gap.

　　(2) Children of urban households are affected by their parents’ income to a greater extent compared with their rural counterparts and have a higher degree of intergenerational transmission of income. For this phenomenon, existing studies have mainly offered an explanation based on human capital theory and the perspective of sociology. First, relative to rural households, urban households face smaller capital constraints for human capital investment fortheir offspring and the return to education is higher in urban areas, which enables urban households to influence the income of their children through human capital investment; second, the father’s generation of urban households possessed greater social capital and easier access to relevant jobinformation, which directly or indirectly influence the jobchoice and employment status of their children and thus influence their children’s income. However, one question is left unanswered: since the father’s generation of rural households has been limited in the credit capacity and social capital and has asmallerpositive effecton their children’s income, how can such upward mobilitybe justified? It can be easily discovered through observations on the reality that for the children of rural households and especially those born after the 1970s, a major channel for them to earn income is to seek jobs in cities. According to the 2010 survey of the NBS, more than half of rural workforce born in the 1970s entered cities for employment. Hence, this paper arrives at the deduction that children of rural households have probably reduced the level of intergenerational transmission of income gap by entering cities. In order to test such a deduction, this paper has utilized the “household registration (hukou) status (of respondents) upon12 years of age” provided by CFPS data to separate those whose hukoustatus was agriculturalhukouupon the age of 12, current hukoustatus is nonagricultural hukouand father’s household hukouis agricultural hukoufor the measurement of intergenerational income elasticity for people with urban household registration converted from rural household registration. Results of Table 5 indicate that the intergenerational income elasticity is relatively low for this group of people. For 2012, the average estimation of the three samples is only 0.1348. In light of China’s reality, there are threepathways for rural hukouto be converted into urban hukou: first, urbanization, i.e.,the migration of agricultural population into cities through industrialization; second, start with lowly-paid jobs before entrepreneurship and settlement in the cities; third, receive higher education to change one’s destinies. This indicates that by entering cities for jobsand higher education, children of rural households have escaped the impact of their low-income father’s generation and reduced the level of intergenerational transmission of income gap.

　　(3) China’s intergenerational income mobility is examined through international comparison foranalysisof the determinants of country differences of intergenerational income elasticity. Table 6 identifies the intergenerational income of some countries and regions measured also using the method of dual sample instrumental variables. Compared with developed countries, China’s intergenerational income elasticity is lower than that of countries with high degrees of inequality such as the United States and Italy buthigher than that of countries with high mobility including Australia and Sweden. Compared with emerging economies, the Chinese mainland’s intergenerational income elasticity is far below that of Brazil, acountry mired in the “middle income trap”, but higher than that of Chinese Taiwan, which has successfully transcended the “middle income trap.” Compared with Japan with similar cultural traditions, the level of China’s intergenerational income transmission is even lower. Despite differences across various economic development stages of different countries, most studies have used the generationsborn in the 1960s and 1970s as subjects of research in their samples, which are consistent with the research subjects of this paper.

　　All countries strive to pursue the social objective to reduce the level of intergenerational transmission of income gap and provide equal opportunities of education and employment for each and every social member in order for them not to be affected by the income levels of their father’s generation. However, judging by the reality of various countries, intergenerational income elasticity remains high even for developed countries and varies greatly across countries. After the 1960s, the United States has enhanced public education investment in an attempt to alleviate the capital constraints of poor households to invest in the human capital of their children. Meanwhile, a series of legislation has been introduced to safeguard equality of employment. Yetintergenerational income inequalitiesstill persist andunderprivileged groups such as the blacksand women remain at the bottom of income distribution. For Canada, also a developed country with demographic diversity and shared values and social customs with the United States, its intergenerational income elasticity is only half that of the United States. The reason is that despite high US government expenditure on the promotion of social mobility, which is even higher than that of any other high-income countries, only a quarter of such expenditure has benefited mid- and low-income people (Carasso,2008). The OECD noted in its annual report of 2012 that in the United States, educational expenditure for children from poor households is insufficient and talented teachers will not teach at schools with poor teaching conditions. In comparison, Canada attaches great importance to the infrastructure development of primary schools and provided more flexibility in terms of labor market institutional arrangements for childbearing parents. According to an opinion poll conducted in the United States, the public policy introduced by the US government to achieve the “American dream” has more disadvantages than benefits and is actually more favorable for those people with an advantageous position in the economy, while upward mobility remains difficult for mid- and low-income people(Corak, 2013). Latin American countries mired in the “middleincome trap” such as Brazil and Argentina offer another story. In the 1980s, these countries vigorously increased educational investment and subsidies to narrow income gaps and increase enrolment of children from poor households. However, due to stalling growth, these measures failed to increase their income and intergenerational income elasticity and Gini coefficient remain high tillthis day. It has been demonstrated by the practice of various countries that public education policy introduced by government will not always lead to equality of opportunities and ifnotproperly implemented, will not achieve desired effects.

　　4.3 Analysis of Intergenerational Transmission of Income Gap for Different Income Groups

　　The above-mentioned intergenerational income elasticity based on ordinary regression equation reflects overall variationsin the intergenerational transmission of income gap for urban and rural households but does not reflect the direction of flow among various income groups and can still less reveal the intrinsic determinants of overall variations. Judging by the empirical studies of various countries, different levels of social mobility could be implied behind intergenerational income elasticity. For Sweden, a country of high social mobility, even children born into the poorest families are less affected by family background in theireconomic status. However,Björklund et al. (2012)also discovered that the intergenerational income elasticity stands at a staggering 0.9 for the 0.1% of the top income group, which means that the country has a “capitalist dynasty in the landof equal opportunities.” Despite high intergenerational income elasticity, income distribution in the United States is characterized by the solidification ofpolarization. In order to investigate the flow of intergenerational income among various income groupsin China, this paper divides father’sincome and children’s income into five grades to constitute an intergenerational income mobility matrix for urban and rural households. Please refer to Table 7 and Table 8 for the results. Solidification index provides an overall assessment on the mobility of different income groups and is the sum between the main diagonal line of matrix and its adjacent units. Intergenerational mobility in its ideal condition should be one where the position of children’s generation is not affected by the father’s generation, i.e.,the probability of each unit in the matrix is equal to 0.2. Therefore, for a five-order mobility matrix, the ideal solidification index should be 2.6. Table 7 and Table 8 indicate that the solidification index is high for urban households in 2002, reaching 3.0446, which indicates that the solidification of different income groups is rather serious. In 2012, this figure dropped to 2.4474 and mobility increased. Comparatively speaking, the mobility among various income groups of China’s rural households is higher than that of urban households with solidification index reaching2.5541, which is closer to 2.6.