The People's Republic of China has witnessed rapid growth in income inequality over the past three decades following the introduction of market reforms in 1978. According to World Bank estimates, the share of income held by the bottom 20 per cent of income earners fell from 9 to 5 per cent between 1981 and 2010, while the share of income held by the top 20 per cent of income earners rose from 38 to 47 per cent. Over the same period, China's Gini coefficient rose from .29 to .42.Footnote 1 Some studies suggest that income inequality may be even higher than the level reported by the World Bank.Footnote 2
The scale and trend in inequality have been major catalysts for empirical research and scholarly and policy debate, but with few exceptions, neither the implications of high levels of inequality for children's welfare in general nor the challenges presented by high levels of inequality for equality of educational opportunity have received much direct empirical scrutiny. Research elsewhere suggests that the implications could be profound. For example, the United States has experienced rising inequality since the 1970s. Sara McLanahan's Reference McLanahan2004 presidential address to the Population Association of America raised alarms about the divergent welfare destinies of children in recent decades: access to resources such as monetary investments and parental involvement has increased over time for those born to the most-educated women in America, while those born to the least-educated women have experienced a decline in access to resources at home, in part owing to changing family formation patterns among this group.Footnote 3 Educational research raised similar concerns: expanding income gaps between the rich and poor since the 1970s have coincided with a dramatic rise in the achievement gap between children at the 10th and 90th percentiles of the income distribution.Footnote 4 During this period, not only were poor children falling behind but the highest-income children were pulling far ahead, in outcomes that include achievement, household enrichment expenditures, college attendance, and selective college attendance.Footnote 5
In China, despite levels of inequality and migration-related family disruption that are unprecedented in recent history, few studies have investigated the divergent destinies of children in affluent and poor families. In this paper, we draw on two recent nationally representative surveys – the China Family Panel Studies (2012 wave) and the China Education Panel Survey (2013–2014 baseline wave) – to investigate the implications of childhood inequality in China for child welfare. We compare family environments and welfare outcomes of children in the least socio-economically advantaged households, in middle groups of households, and in the most socio-economically advantaged households. For each dimension of family environment, and for each welfare outcome, we address the following research questions: are there significant differences among the three groups? Do children with low socio-economic status fall behind children with middle and high socio-economic status on certain measures? Do children with high socio-economic status pull ahead of the rest on other measures?
Framework
Child development scholars have variously conceptualized the impact of family socio-economic status on child development,Footnote 6 but many emphasize two key mechanisms: family investments and family stress (see Figure 1).Footnote 7 Investments include monetary expenditures on tutoring, learning materials at home, parental involvement in the child's studies, and aspects of material well-being such as shelter, food, medical treatment and a safe and secure environment.Footnote 8 Examples of family stress include parental illness, parental substance abuse, strained relationships, domestic abuse, and family disruptions such as marital dissolution or the death of a parent. Related to stress and investments is the concept of family social capital, which suggests that children benefit more from the “human capital” of their parents when their parents are present; parents are close to and interact with children; parents supervise children and convey high expectations to them; parents help children with educational problems; and parents are networked into the institutions that matter for child welfare outcomes in the local context.Footnote 9 In the US, changing family formation patterns have led to the rise of single parent households and “fragile” families, and this trend is pronounced among children of mothers with lower levels of education.Footnote 10 Consequently, families with lower socio-economic status may have fewer adults available for supervision, competent investment and the kind of “concerted cultivation” of children that is practised in middle-class American homes.Footnote 11 Social capital available to a child can also be reduced by stressors such as family migration, transferring schools or living away from home.Footnote 12 In the US, Elizabeth Gershoff and colleagues found a positive relationship between income and family investments, and a negative relationship between income and exposure to family stress and material hardship.Footnote 13 The authors also identified family investments as mediators in the relationship between household socio-economic status and educational outcomes. Finally, poverty had negative implications for behavioural outcomes.Footnote 14 Drawing on these ideas, we developed the conceptual framework in Figure 1 to guide an investigation of socio-economic differences in family environments and children's welfare outcomes in China.
Figure 1: Conceptual Framework: Family Influences on Child Outcomes
The China Context
Despite recent attention to the issue of rising inequality in China, implications for the educational outcomes of children and child welfare in general have not yet been fully explored. However, there are several related lines of work. Regarding physical health, a few studies have explored the implications of socio-economic inequality in China for the nutritional status of children. Scholars have observed, for example, high child obesity rates among the affluent and a high incidence of underweight children among the poor.Footnote 15 Health disparities may be associated with differential access to health-related resources at home. Jennifer Adams and Emily Hannum, for example, found a positive relationship between household socio-economic status and children's access to health insurance.Footnote 16 To our knowledge, no sociological or economic studies have investigated disparities in psychosocial health associated with economic inequality, although studiesFootnote 17 and a recent government White Paper have investigated behavioural and mental health issues affecting children facing family disruption through migration.Footnote 18
Regarding education, many studies have investigated how poverty and economic disadvantage affect educational outcomes, with particular attention to rural poverty, which continues to be a major risk factor for school dropout.Footnote 19 Children in poor rural areas can face deprivations at a very basic level, including chronic undernourishment, food insecurity, a lack of access to vision correction, and parental absence owing to migration.Footnote 20 These deprivations have been linked empirically to various educational disadvantages.Footnote 21 Interviews in Gansu province indicated that children in poor rural communities lack both essential school materials and adults at home who have educational experience to assist with homework or school problems; they express generalized distress about money and the burden of school costs for their parents.Footnote 22 Poor rural children may be “behind before they begin,” as their parents can afford fewer expenditures on early childhood education prior to beginning compulsory basic education.Footnote 23 A study analysing the multi-province China Health and Nutrition Survey revealed that even after controlling for long-term income, children who experienced poverty in early childhood had an elevated probability of dropping out of school.Footnote 24
While there has been much scholarly attention to the education-related disadvantages associated with poverty, a smaller body of literature has explored differences between Chinese children from different socio-economic groups in educational resources and outcomes. Regarding educational attainment, Adams and Hannum found that school enrolment gaps between poor and affluent children persisted between 1989 and 1993 despite educational expansion,Footnote 25 and Elisabetta Magnani and Rong Zhu observed that the correlation between the educational attainment of parents and their children increased between 1990 and 2000.Footnote 26 A handful of studies have also investigated whether gaps exist between children from different social classes in access to educational resources, such as monetary investments, parental involvement, and access to cultural capital at home. Wei Chi and Xiaoye Qian found that both highly educated and high-income parents in urban areas spend more on out-of-school educational expenditures than do less educated and low-income parents.Footnote 27 Drawing from a dataset of 3,087 urban residents, Yuxiao Wu identified cultural capital at home as one mediator of the relationship between household socio-economic status and educational outcomes in urban China.Footnote 28 Finally, Airan Liu and Yu Xie documented a statistically significant, positive relationship between parenting practices and educational performance, although they found almost no association between income and parenting practices.Footnote 29 Nevertheless, the previous literature on socio-economic disparities in educational resources is limited. Both Wu and Chi and Qian focused exclusively on urban areas in their analyses, and research on socio-economic disparities in parenting has, until now, been limited by the lack of nationally representative data with extensive information about parenting practices.Footnote 30
It is important to add to the literature on childhood inequality for two reasons. First, while rural poverty remains an important problem in China, issues of urban poverty and inequality have become increasingly pressing with the rise of migration and the emergence of a disadvantaged urban migrant class. One recent estimate indicates that the migrant population numbered 253 million by the end of 2014 – about one-sixth of China's total population – and is expected to reach 291 million in 2020. This projected number includes 220 million rural-to-urban migrants.Footnote 31 Of particular relevance to this paper are the implications migration has had for family environments. According to 2010 census figures, over 61 million children aged 0 to 17 were “left-behind” (liushou ertong 留守儿童).Footnote 32 Close to half were left behind by both parents, 36 per cent experienced the absence of a migrant father, and 17 per cent had a migrant mother. It is reported that 38 per cent of all rural children and 22 per cent of the general population of children are left-behind.Footnote 33 In addition, a smaller but growing fraction of children are brought along by their migrant parentsFootnote 34 and experience uncertain access to schools and services in urban areas and may be part of an emerging vulnerable and poor urban population.Footnote 35 Migration can also affect family structure and family ties, which may have implications for childhood inequality.Footnote 36
A second reason to focus on inequality in childhood is the relatively recent appearance of an affluent class in China. This change highlights the need for a shift in focus to consider the advantages of the elite, as well as the disadvantages of the poor, as critical elements of inequality in childhood. As an illustration of this point, a McKinsey report highlighted a projected emerging middle and affluent class in China whose consumption is expected to grow substantially in the future.Footnote 37 This consumption extends to education for children.Footnote 38 The advantages of the wealthy are just beginning to be studied in the context of large-scale survey-based studies in China. Beyond looking at parental education and income effects on education, few studies have considered the advantages of children from affluent families.
To summarize, prior research on childhood inequality in China suggests that socio-economic inequality may translate into inequalities in children's nutritional status, access to healthcare and educational resources at home, and educational outcomes. A number of studies have focused on poverty and socio-economic deprivation in education and child welfare, and many of these studies have focused on issues of rural poverty. Although rural poverty remains a serious challenge to children's welfare in China, urbanization, migration and the emergence of both marginalized and affluent urban populations have changed the scale and nature of inequality in childhood and call for further study. While a number of studies have begun to investigate the implications of migration for children's schooling and welfare, these studies have not placed migration in the context of a broader investigation of family socio-economic status. Further, few studies have considered the advantages of children from China's nascent affluent classes. Drawing on a framework of family investments and family stress and comparing children in “middle” socio-economic groups with those of high and low socio-economic status, the current paper begins to address these gaps in the literature.
Data and Methods
Data
Because no single data source is ideal for our purposes, we make use of two national-scale datasets that are different in sample and focus but mutually complementary: the China Family Panel Studies (CFPS) and the China Education Panel Survey (CEPS). The CFPS is a household survey that includes detailed information on children aged up to 15, while the CEPS focuses on seventh and ninth graders (ages 12–16) currently enrolled in school. Critically for our purposes, the CFPS contains detailed income and expenditure data, while the CEPS contains extensive information on parenting practices. By drawing on both datasets, we are able to capitalize on the strengths of each.
The 2012 wave of the CFPS is a nationally representative sample of 13,315 households and 35,719 adults (aged 16 and above) and 8,620 children (aged between 0 and 15) living within the sampled households.Footnote 39 We use data collected during the 2012 wave, since the data collection period of this wave corresponds more closely to that of our second dataset. When using family income as our measure of socio-economic status, we restrict analysis to the 8,264 children (aged between 0 and 15) not missing data on net family income per capita in 2012. We restrict analysis to the 8,219 children not missing data on either father's education or mother's education when using parental education as our socio-economic status (SES) measure.Footnote 40
The baseline (2013–2014) wave of the CEPS is a nationally representative sample of 112 schools, 438 classrooms within schools, and 19,487 seventh and ninth graders (aged from 12 to 16) within classrooms.Footnote 41 In addition to the students, school administrators, teachers and a parent or guardian of each sampled student completed questionnaires. As with the CFPS, in producing descriptive statistics we restrict analysis to children not missing data on the given measure of socio-economic status (Table 1). The analysis covers 19,007 children when using parental education as our measure of socio-economic status, while analysis is restricted to 19,475 children when using a household assets scale as our measure.Footnote 42
Table 1: Descriptive Statistics for Two Samples
Measures of socio-economic status
The two datasets do not contain the same information about family socio-economic status. We were able to construct one common measure of family SES based on parental education, which is reported in both datasets, and one unique measure based on income for the CFPS and based on household assets for the CEPS. We use this measure because the CEPS dataset does not contain information on family income.
We use data on parental education in the CEPS and the CFPS to divide children into three groups – high parental education (high SES), low parental education (low SES), and middle parental education (middle SES). High parental education is defined as having at least one parent with post-secondary education. Low parental education is defined as lacking a parent with education beyond elementary school. Middle parental education is defined as not falling into either of the other categories. Within the CEPS, 13.65 per cent of the nationally representative sample is classified as low parental education, 74.71 per cent as middle parental education, and 11.63 per cent as high parental education. Within the CFPS, 21.08 per cent of the nationally representative sample is classified as low parental education, 67.25 per cent as middle parental education, and 11.67 per cent as high parental education.
Family income (available only in the CFPS) is operationalized as net family income per capita in 2012. We generate income quintiles to compare high SES families and low SES families to middle families. We define high SES families as those within the top quintile of the income distribution, middle families as those in the three middle quintiles (the middle 60 per cent), and low SES families as those within the bottom quintile.
Owing to the lack of direct measurement of income in the CEPS dataset, a scale of household assets is employed as a second measure of SES. Asking children or adolescents to report family income often results in high levels of missingness and/or bias.Footnote 43 Instead, scholars have proposed the adoption of household assets scales to measure socio-economic status in child or adolescent surveys.Footnote 44 Some researchers use an assets index as a linear measure of wealth,Footnote 45 although others have expressed concern that these indices are often lengthy, which can pose a problem, as children may not have full information about their household possessions.
An alternative approach is the Family Affluence Scale (FAS) developed by the research team of the Health Behaviour in School-aged Children: WHO Collaborative Cross-National Study (HBSC).Footnote 46 The questions that constitute the cross-national FAS are quick and easy for children to answer. The second version of the scale, FAS II, for example, is based on whether the family owns a car, the number of computers at home, whether the child has their own bedroom, and the frequency of family vacations. Yang Liu and colleagues found FAS II to be a valid measure of SES within certain Chinese contexts.Footnote 47
Although not strictly a household assets scale, the Economic, Social and Cultural Status (ESCS) index employed by the OECD's Programme for International Student Assessment (PISA) also includes a set of questions about home possessions. Students are asked if they possess a desk, their own bedroom, a study space, a computer, internet access, educational software, a calculator, classic works of art or literature, books, a dictionary, and a dishwasher at home. Children are also asked to list the number of books their family possesses. Other researchers argue that simply asking children about the number of books at home is a valid and useful measure of socio-economic status.Footnote 48
We use a household assets scale based on the reviewed scales as our second measure of SES status in the CEPS. Although we would have liked to replicate FAS II, the student questionnaire did not include information about family vacations or whether the child had their own bedroom. Instead, the following assets are included in our scale: access to a computer and/or internet at home, the relative number of books in the household (Coded 1 = very few; 2 = relatively few; 3 = a normal amount; 4 = many; 5 = a lot), and whether the child has access to their own desk at home. Our scale has a Cronbach's alpha of 0.66 and factor analysis suggests that the scale is unidimensional. Higher values on the scale indicate higher SES status. We generate quintiles for the household assets scale to compare high SES families and low SES families to middle families. We define high SES families as those within the top quintile of the scale, middle families as those in the three middle quintiles (the middle 60 per cent), and low SES families as those within the bottom quintile.
Family environment measures
We draw on the CFPS parent questionnaire, which includes questions about annual expenditures on education-related items for the child, to consider socio-economic disparities in monetary investments. We use both the CFPS and the CEPS to examine SES differences in enrolment in tutoring (the CFPS provides data on tutorial enrolment among 0- to 15-year-olds, while the CEPS provides data for seventh and ninth graders).
We also use the CEPS to investigate measures of parental involvement in education. We look at student reports of how frequently their parents checked or provided guidance on their homework in the past week; parent reports of attendance at parent–teacher meetings; and student reports of how frequently they read with their parents or accompanied them to museums or other cultural institutions in the past year.
Variables related to family stress come from the CEPS and include measures of parental absence, the child's migrant status, whether the child transferred primary schools, and whether the child boards at school. Parental absence is measured via two dummy variables, one for an absent mother (coded 1 if the student does not select “mother” when asked, “in your current home, with whom do you live?”), and one for an absent father (coded 1 if the student does not select “father” when asked, “in your current home, with whom do you live?”). Finally, the parent who filled out the parent questionnaire reported his or her (own) current health status.
Measures of welfare outcomes
Physical health-related variables include a measure of self-rated health in the CEPS (“how is your overall health currently?” Coded 1 = very poor; 2 = poor; 3 = average; 4 = relatively healthy; 5 = very good), and measures of illness in early childhood and disability in the CEPS dataset (visual impairment other than near-sightedness, hearing impairment, physical disability, speech impairment, autism or other mental disorder, ADHD, other).
Our first measure of psychosocial health is a CES-D scale administered by the CFPS research team to all sampled children aged 10 to 15 years old. The Center for Epidemiological Studies Depression Scale (CES-D) is one of the most widely used screening tests for depression in the world; previous research suggests that it is a valid and reliable measure of depression not only in the US and other Western societiesFootnote 49 but also in China.Footnote 50 The instrument has high reliability in the CFPS sample of children (Cronbach's alpha = 0.809). The CEPS uses a much shorter, five-item scale to measure unhappiness or depression among seventh and ninth graders. We use this as a secondary measure of psychosocial health. Although we do not have information on validity or psychometric properties of this scale in other populations in China, the reliability in the CEPS sample is high (Cronbach's alpha = 0.80).Footnote 51
Our measure of educational performance is a standardized test of logical reasoning administered by the CEPS research team.Footnote 52 The test consists of 20 items for seventh graders and 22 items for ninth graders. Similar to other psychometric tests used in international education research, the CEPS test measures student reasoning across three dimensions: language, mathematics, and graphical forms.Footnote 53 The CEPS research team used the three-parameter logistic (3PL) Item Response Theory (IRT) model to derive final test scores.Footnote 54 Item Response Theory is a well-regarded, common approach to psychometric test design and evaluation in the education field.Footnote 55 To derive final test scores, the 3PL IRT model takes into consideration the difficulty of each item, the ability of each item to discriminate among respondents with different ability levels, and the probability that a respondent correctly guesses the answer to the item.Footnote 56 The test has high reliability in the CEPS sample (Cronbach's alpha = 0.69 for seventh graders; 0.72 for ninth graders). We use standardized scores on this test to measure educational performance.
Methodological approach
To produce figures depicting family inequalities, we regress investment, stress and welfare outcome measures on each measure of SES status.Footnote 57 For the CFPS dataset, we add controls for gender, age and the month during which the household completed the survey. For the CEPS dataset, we adjust for gender, grade and the semester during which the student completed the survey. We estimate OLS regression models for continuous dependent variables. In the case of categorical dependent variables, we first dichotomize the variable and then estimate a logistic regression model of the log odds of falling into one category of the dichotomized variable.Footnote 58 Using these specifications and setting covariates at mean values, we calculate predicted values (for continuous outcomes) or predicted probabilities of falling into a given category (for categorical outcomes) for high SES children, middle SES children, and low SES children.
We test whether the group differences are statistically significant. First, we test the null hypothesis that there are no differences among any of the three socio-economic groups in the dependent variable of interest. We use an F-test for continuous dependent variables and a chi-square test for dichotomous dependent variables (alpha = 0.05). If the evidence is sufficient to reject the null hypothesis, we perform a series of post-hoc, pair-wise comparisons of the three SES groups on the dependent variable. We use the Bonferroni correction for multiple comparisons to set the family-wise error rate at 5 per cent.
Analysis and Results
Family investments
We first consider whether high, middle, and low SES families differ in monetary investments in children's education. Figure 2.1 shows the predicted educational expenditures of families within each of the three socio-economic groups. These groups are defined by parental education in the top part of Figure 2.1 and by income quintiles in the bottom part of Figure 2.1. Regardless of the measure of SES used, the predicted total educational expenditures of high SES families are very high relative to the other socio-economic groups. The predicted total educational expenditures of high SES families are twice the predicted total educational expenditures of middle SES families, and between 2.6 times (if we define SES by family income) and 4.6 times (if we define SES by parental education) the predicted total educational expenditures of low SES families. Hypothesis tests reveal that the differences in total educational expenditures between high and middle SES families and between high and low SES families are statistically significant. The gaps between middle and low SES families, however, are only statistically significant if we define SES by parental education.
Figure 2: Education-related Monetary Investments
Spending on tutoring is a major contributor to the overall gap in educational expenditures. The predicted expenditure on tutoring among high SES families is about four times that of families with middle socio-economic status. Even more drastic are the gaps in predicted tutoring expenditures between high and low SES families: the ratio is 8:1 for high income families relative to low income families, and the predicted expenditure of highly educated parents is 1,256 yuan, compared to close to 0 yuan for parents with low levels of education.Footnote 59 Differences between high, middle, and low SES families on tutoring expenditures are all statistically significant.
While the predicted values produced using the CFPS are for children aged 0 to 15, in Figure 2.2 we use the CEPS to compare the predicted expenditure on tutoring among 12- to 16-year-olds enrolled in school. Although the predicted values are different from those produced using the CFPS, the observed socio-economic gaps are comparably large.Footnote 60 Further, as in the CFPS, the differences between the three SES groups are all statistically significant.
A key reason that expenditures on tutoring are higher among high SES families is that high SES children are more likely than other children to enrol in tutoring. The top half of Figure 3 shows the predicted probability of 0- to 15-year-olds enrolling in tutoring in the month during which the family was interviewed, while the bottom half of Figure 3 shows the predicted probability of seventh and ninth graders enrolling in tutoring in the semester in which the survey was conducted. Again, we see stark disparities between low, middle, and high SES children, all of which are statistically significant.
Figure 3: Participation in Tutoring
In contrast to Liu and Xie, we also observe fairly large socio-economic disparities in parenting practices (Figure 4).Footnote 61 Notably, the dataset we use (CEPS) contains more extensive data on parenting practices than the CFPS dataset, which Liu and Xie used in their analysis. First, we observe socio-economic differences in parental homework assistance and attendance at parent–teacher meetings.Footnote 62 These differences look quite similar across both measures of socio-economic status and are statistically significant. High SES children are also more likely to read with their parents or accompany them to cultural institutions than middle or low SES children, activities that not only require cultural capital but may also generate additional cultural capital. Hypothesis tests indicate that these differences are statistically significant.
Figure 4: Parenting Practices
Family stress
In addition to disparities in family investments, the three socio-economic groups differ in exposure to family stressors (Figure 5). Low SES children are more likely than other children to live with a sick parent: the predicted probability of having a parent in poor or very poor health is about 15 per cent for low SES children, compared to 8 per cent for middle SES children and 3 per cent for high SES children. Parental absence is also more common among the socio-economically disadvantaged. Although children with highly educated parents have a 12 per cent predicted probability of living without a father, the predicted probability of paternal absence is almost three times as high for children with low educated parents. Similar patterns are observed if we use assets quintiles to measure socio-economic status or if we compare predicted probabilities of living without a mother. These observed differences between low, middle, and high SES children in parental health and parental absence are all statistically significant.
Figure 5: Family Stress
Disparities are observed in at least three other types of family stressors. First, we observe disparities in the likelihood of living at school during the week, a living situation that may weaken family ties and deplete social capital. The predicted probability of boarding at school is 63 per cent for children with low educated parents, 47 per cent for children with middle educated parents, but only 12 per cent for children with highly educated parents. The patterns are similar if we measure SES with the household assets scale, and all socio-economic differences in exposure to this stressor are statistically significant. A second stressor is whether children transferred primary schools, an event scholars have associated with the depletion of social capital.Footnote 63 The predicted probability of transferring primary schools is about 41 per cent for low SES children and 33 per cent for middle SES children, compared to 20 per cent for high SES children. These differences are all statistically significant. There does appear to be one exception to the trend of low SES children having a higher probability of experiencing family stress than other children. Children with low educated parents are no more likely to be migrants (8 per cent) than are children with highly educated parents (7 per cent), while children in the bottom assets quintile are less likely to be migrants (5 per cent) than those in the top assets quintile (11 per cent).
Welfare outcomes
We have revealed stark disparities between children from different socio-economic groups in access to resources related to child development and education. These disparities lead us to ask whether children with different socio-economic status experience different welfare outcomes. We first consider socio-economic differences in physical health (Figure 6) using three measures: self-rated health, serious illness in childhood, and physical or mental impairments. Within the CEPS, the predicted probability of reporting poor self-rated health (Figure 6.1) is higher among low SES children than among either middle or high SES children. There is no statistically significant difference between middle and high SES children on this measure.Footnote 64
Figure 6: Welfare Outcomes – Physical Health
Low SES children are also more likely to have had a serious childhood illness than other children (Figure 6.2). The predicted probability of having experienced serious illness before elementary school is about 18 per cent for low SES children, compared to 10 per cent for middle SES children and 6 per cent for high SES children. In addition, low SES children are more than twice as likely to have an impairment compared to high SES children (Figure 6.3). Hypothesis tests indicate that all of the pair-wise differences between the SES groups on these two measures of physical health are statistically significant.
Low SES children are more likely than other children to experience not only physical health problems but also psychosocial health problems (Figure 7). The predicted CES-D score for children aged between 10 and 15 years of age with low educated parents is about half a standard deviation higher than children with highly educated parents (Figure 7.1A). This difference is significant, as is the difference between children in the bottom and top income quintiles. We next use scores on the CES-D scale to calculate predicted probabilities of exhibiting evidence of depression for each socio-economic group (Figure 7.2). We find that low SES children are about two to three times more likely than high SES children to exhibit some evidence of depression (as indicated by a CES-D score of 16 or higher); this difference is statistically significant. In addition to the CES-D scale in the CFPS, we investigate differences on the CEPS psychosocial problems module (note: higher scores indicate higher frequencies of experiencing psychosocial problems). As shown in Figure 7.1B, low SES children again have a higher predicted score on this module than middle SES children, who have a higher predicted score than high SES children. Additional analyses indicate that all pair-wise differences are statistically significant.
Figure 7: Welfare Outcomes – Psychosocial Health
Finally, to investigate disparities in educational performance, we look at a standardized test of logical reasoning (CEPS). Predicted scores on the test for seventh graders with high, middle, and low SES are presented in the top half of Figure 8, while predicted scores for ninth graders are shown in the bottom half of the figure. We observe socio-economic differences in test scores for both grades. Seventh grade students with highly educated parents, for example, have a predicted score of 0.34, compared to -0.17 for seventh graders with middle educated parents and -0.44 for seventh graders with low educated parents. All pairwise differences in test scores are statistically significant, regardless of SES measure.
Figure 8: Welfare Outcomes – Educational Performance
Conclusions and Implications
In this paper, we have traced socio-economic disparities in the family environments and welfare outcomes of children in China. The overall pattern of inequality that we observe suggests that, for most measures of family environment and welfare outcomes, high SES children fare best, low SES children fare worst, and middle SES children fall about equidistant between the other two groups. But there are other domains where the key distinction appears to be high SES versus other children, and where the key distinction seems to be low SES versus other children. For example, high SES children are well ahead of the rest in terms of education-related monetary investments, including expenditure on extracurricular tutoring. For these measures, the differences between the predicted expenditures of high and middle SES children are larger than the differences between middle and low SES children. In some cases, there is no statistically significant difference between middle and low SES children. In addition to educational expenditures, high SES children “pull ahead” of the rest on a measure of educational performance. For example, the gap in predicted scores on a test of logical reasoning between ninth graders with highly educated parents and ninth graders with middle educated parents is more than twice the size of the test score gap between ninth graders with low educated parents and ninth graders with middle educated parents. In contrast, reflecting the ongoing child welfare challenges among families in poverty, children from low SES families fall far behind the rest on two measures of physical health: self-rated health and serious illness in childhood.
Our findings suggest that both emerging affluence and emerging economic inequalities in China are reflected in the family environments and welfare outcomes of children. To contextualize these findings, it is important to note that socio-economic status intersects with other domains of stratification in ways that may reinforce opportunity gaps. For example, in the United States, poverty and affluence intersect in important ways with race, ethnicity and immigration status. In China, socio-economic status may intersect with household registration type (hukou 户口), region of residence, ethnic minority status, whether one is a native speaker of Mandarin, and whether one lives in an urban area, and these other factors may carry implications for children's life chances. For example, the CEPS data show that ethnic minorities and non-native Mandarin speakers are overrepresented among low SES students. Low SES students are also less likely than middle and high SES students to live in the most developed area of China – China's eastern region; and Han Chinese (hanzu 汉族), non-agricultural hukou holders, and residents of China's eastern region are overrepresented among high SES students. An important caveat to our findings is that family socio-economic inequalities are shaped by and intersect with the broader contexts in which families operate.
Our findings raise two important considerations for policymakers. First, these results speak to the continuing challenges to child welfare facing poor families. In recent years, the Chinese government has introduced new initiatives to lift poor families out of their disadvantaged social position. In 2016, Xi Jinping 习近平 stated that a variety of programmes would be implemented to raise ten million people out of poverty each year, with the goal of eradicating poverty by 2020.Footnote 65 Poverty alleviation efforts include programmes to expand low-income families’ access to medical treatment, services and insurance. Given our finding that low SES children fall far behind other children on measures of physical health, this particular initiative is quite promising.
Second, although poverty alleviation efforts constitute an important component of addressing childhood inequality, another element of inequality that they will not address is the advantages of children in an emerging affluent class. Our findings indicate that inequality is driven by the advantages of affluence as well as by the continuing disadvantages of poverty, and the advantages that distinguish high SES children from those in the middle are in some cases different from the disadvantages that distinguish low SES children from those in the middle. As our study has shown, high SES families in China are heavily investing in their children's education by purchasing education-related goods and services outside of the school system. This investment in education may in part explain why high SES children “pull ahead” of the rest in educational performance and will likely have higher levels of educational attainment, better job opportunities and higher salaries. Similar to the situation in the United States and other countries facing high levels of economic inequality, family inequalities outside of the purview of China's educational system present a complex challenge to ideals of equality of educational opportunity.
Acknowledgements
Earlier versions of this paper were presented at the Weatherhead East Asian Affairs Institute at Columbia University and the China Education Panel Survey Users’ Meeting at Renmin University of China. The authors gratefully acknowledge comments provided by Hyunjoon Park, Annette Lareau and other seminar participants at the University of Pennsylvania's Education and Inequality Workshop. Natalie A.E. Young was supported while working on this project by a Benjamin Franklin Fellowship from the University of Pennsylvania.
Biographical notes
Natalie A.E. Young is a PhD candidate in the department of sociology at the University of Pennsylvania. Her research focuses on education and social stratification in China, with particular attention to China's emerging affluent class. Her dissertation investigates the strategies employed by high socio-economic status parents in China to maintain privilege for their children. She recently published “Departing from the beaten path: international schools in China as a response to discrimination and academic failure in the Chinese educational system” in Comparative Education (2017).
Emily C. Hannum is professor of sociology at the University of Pennsylvania. Her current work focuses on poverty and child welfare in China. Recent publications include “Early poverty exposure predicts young adult educational outcomes in China” (with Xiaoying Liu, 2017, China Economic Review) and “Chronic undernutrition, short-term hunger, and student functioning in rural northwest China” (with Li-Chung Hu, 2017, International Journal of Educational Development).
