Sage Journals: Discover world-class research

Abstract

Although there has been a longstanding curiosity about the socio-political consequences of China’s remarkable urban–rural divide, we have yet to understand the divide’s possible influence on mental health. Using data from the 2016 wave of the China Labor-force Dynamics Survey (CLDS), we find that depressive symptoms of both rural–urban migrants and rural residents are significantly higher than those of urban residents. Consistent with the fundamental-causes-of-disease and stress-exposure perspectives, results from zero-inflated negative binomial regression suggest that such differences in depressive symptoms can be attributed to socioeconomic status and proximate stressors such as unemployment, living alone, and the unaffordability of medical services. In particular, the rural–urban difference in depressive symptoms is explained away by educational attainment. A further investigation using spline Poisson regression suggests that the protective effects of the period of middle school, which vary substantially across demographic groups, are especially relevant to the rural–urban disparity in depression. We argue that hukou is a fundamental cause of disease in China and mental health is an important yet understudied area where China’s salient urban–rural inequality strikes.

Keywords

China depression fundamental causes of disease rural migrants rural–urban inequality zero-inflated negative binomial regression

Introduction

In China, hukou, the household registration system, is an institutional arrangement that categorizes people according to their place of birth and subsequently determines their (unequal) rights, privileges, and access to resources (Chan and Zhang, 1999; Fu and Ren, 2010; Lin and Bian, 1991; Tong and Piotrowski, 2012; Wu and Treiman, 2007). It is widely argued that social inequality in contemporary China results more from institutional arrangements such as hukou or danwei (one’s work unit) than from individual characteristics (Bian and Logan, 1996; Fu and George, 2015; Lin and Bian, 1991; Walder, 1992; Zhou, 2000). Social stratification, or more precisely, unequal access to resources, social support, and health care, is shown to be a key driving force in determining the mental health inequality across societies and political economies (Aneshensel et al., 2013; Miech et al., 1999; Muntaner et al., 2007; Wheaton, 2001). Therefore, it is worthwhile to explore the ramifications of China’s hukou system for mental health across social groups. Although mental health in China has attracted considerable attention in recent years from epidemiologists, demographers, and medical sociologists (Chen et al., 2010; Graham et al., 2017; Shi et al., 2016; Yip et al., 2007), this specific research question has yet to be thoroughly investigated.

To address this question, we analyze data from the most recent wave (2016) of the China Labor-force Dynamics Survey (CLDS). More specifically, we employ zero-inflated negative binomial (ZINB) models and spline Poisson regression models to address the following questions: (a) Does a rural–urban disparity in mental health exist in China? (b) If yes, what socio-economic indicators in contemporary Chinese society may account for such a disparity? and (c) If there were a key variable shaping the rural–urban disparity in mental health, how would it moderate mental health inequality? Next, we discuss the relevance of hukou to mental health in contemporary China.

Does hukou matter for mental health? Stratification and mental health in China

Students of social stratification are often motivated by two major questions: how social inequality comes into being, and what social inequality produces and reproduces (Link and Phelan, 1995; Phelan et al., 2004; Williams and Collins, 1995). Regarding the first question, inequality occurs at both the individual and group levels. For example, significant variations in income occur not only among individuals, but also across social groups defined by, for example, gender, ethnicity, immigration status, or place of origin. While theorists propose several concepts (e.g., categorical inequality, horizontal inequality, and between-group inequality) to understand how social inequality arises among individuals who have similar attributes but are positioned in segmented social groups (Jiang et al., 2012; Stewart, 2016; Tilly, 1998), the link between socio-political segmentation and inequality has attracted significant attention in recent years (Lin, 2000; Massey, 2007; Stewart, 2016).

The extensive impact of the hukou system on social inequality in China has been an important research topic in recent decades (Chan and Zhang, 1999; Fu and Ren, 2010; Wu and Treiman, 2007; Xu and Xie, 2015). Established in the 1950s, hukou divided the Chinese population into agricultural hukou holders in rural areas and non-agricultural hukou holders in cities. Prior to China’s economic reforms starting in the late 1970s, the migration of agricultural hukou holders was severely restricted to prevent a shortage of people and resources devoted to agricultural production. Given that China’s economic reforms developed a strong demand for cheap labor in urban and coastal areas, the hukou’s role in controlling rural–urban migration has gradually been lessened (Chan, 1994; Fu and Ren, 2010). One consequence of the deregulation of rural–urban migration in China is an influx of rural–urban migrants who are neither urbanites according to their de jure registration status, nor villagers based on their de facto work/residence locations. Yet, it is misleading to interpret the trend of massive rural migration as a retreat of hukou in Chinese society; due to their agricultural and nonlocal hukou status, the rural migrants—referred to by demographers as the ‘floating population’—are not entitled to public services, welfare benefits, or local resources such as education and housing in urban China (Chan and Buckingham, 2008; Whyte, 2010). The hukou system continues to play a key role in producing socioeconomic disparities in China because one’s place of birth is inextricably linked with access to social benefits, availability of state resources, and career opportunities (Chan and Zhang, 1999; Fu and Ren, 2010; Wu and Treiman, 2007).

Today, China is probably one of the most unequal societies in the world, and the rural–urban divide has been shown to be a driving force behind China’s social inequality (Wu and Treiman, 2004; Xie et al., 2012; Xie and Zhou, 2014). Despite recent reforms to the hukou system, scholars still view it as a key institutional arrangement shaping social stratification in Chinese society (Rarick et al., 2017; Xie and Zhou, 2014; Zhang and Treiman, 2013). While prior studies on hukou have tended to focus on inequalities in income, wealth, housing, life chances, and physical health (Chan and Zhang, 1999; Kanbur and Zhang, 1999; Liu, 2005; Lu and Qin, 2014; Wu and Treiman, 2007), we argue that the impact of hukou on mental health should not be under-estimated. The existing literature often approaches the socioeconomic gradient in mental health from a micro-level perspective and views the mental health disparity in China as a consequence of individual backgrounds and characteristics (Chen et al., 2010; Graham et al., 2017; Shi et al., 2016). In these studies, the socio-political nature of hukou was not sufficiently developed. For example, Yip et al. (2007) examined the association between social capital and mental health in rural China and found that, at the individual level, trust affects the subjective well-being of rural peasants. Using a survey of 210 unemployed migrant workers in Zhejiang Province, Chen et al. (2012) investigated the effects of the duration of unemployment and related coping strategies on mental health. This research suggested that long-term unemployment and under-developed coping strategies are associated with the mental health problems faced by these unemployed migrant workers. Most recently, Graham et al. (2017) studied the correlations between psychological health and individual traits such as age, education, and income. They found that urban educated Chinese are more likely to report depression than others. Again, hukou was recognized in this study as an individual characteristic rather than as the institutional backdrop against which the mental health of segmented social groups is differentiated.

The focus of these studies, which situated the link between hukou and mental health within the macro context of China’s socio-political transformation, tended to be on specific social groups, such as rural migrants, left-behind children, or rural residents. For instance, using data collected in Beijing in 2004 and 2005, Lin et al. (2011) found that both the experience of discrimination and perceived social inequity were strongly associated with mental illness among rural migrants. Drawing on a panel dataset of more than 19,000 students from 252 rural primary schools in northwestern China, Shi et al. (2016) reported a significant and negative association between parental out-migration and the mental health of left-behind children in rural China. Similarly, a different study by Lu et al. (2012) also suggested that adults in rural households with out-migrants are at a higher risk of depression than those in households without out-migrants, despite the fact that monetary remittances can compensate for the mental cost of out-migration. In a study of rural children, Xu and Xie (2015) found that migrating with their parents had no negative effect on the children’s subjective well-being. While these studies are valuable for understanding the broad impact of a particular institutional arrangement on psychological well-being, we have not thus far developed a more holistic view of how the link between hukou and mental health is differentiated among segmented social groups.

More recently, a few studies have compared differences in physical health (rather than mental health) between social groups defined by hukou, such as between rural migrants and urban residents, or between rural residents and urban residents. In order to test for selective migration in China, Tong and Piotrowski (2012) compared the health status of migrants and non-migrants using data from the 1997–2009 waves of the China Health and Nutrition Survey. They found that migrants were positively selected on the basis of health. Based on a nationally representative survey, the analysis conducted by Niu and Qi (2015) showed that rural migrants in urban areas were generally healthier than urban natives, but residents with non-agricultural hukou were healthier than those with an agricultural hukou. Yet, given this study’s focus on physical health, it remains unclear whether the rural–urban disparity also exists in mental health. In this regard, one study published in The Lancet appeared to support the link between hukou-based segmentation and mental health. Based on an epidemiological survey conducted from 2001 to 2005 in four provinces, Phillips et al. (2009) found that rural residents were more likely than urban residents to have depressive disorders and alcohol dependence. Likewise, a qualitative study conducted in eleven provinces in China suggested that suicide, especially elderly suicide, has increasingly been a critical issue in rural China since the 1980s (Liu and Fu, 2017). In the present study, drawing upon two influential frameworks explaining (mental) health disparities across different socio-cultural contexts—the fundamental-causes-of-disease and stress-exposure perspectives—we further explore the pathways through which the hukou system may affect mental-health conditions of segmented demographic groups.

Hukou and mental health: Exploring possible mechanisms

Sociological research on mental health is informed by two conceptual frameworks, one focused on the fundamental causes, and the other on the proximate determinants of diseases. As posited by the fundamental-causes-of-disease framework (Link and Phelan, 1995; Phelan and Link, 2005), social conditions such as education, income, and social support serve as underlying forces driving inequalities in health. Because the privileged always have more access to health services, better knowledge of risk factors for diseases, and earlier exposure to advances in clinical practices and medical technology, economic growth and social development will not eliminate the socio-economic gradient in health. Rather, health inequality persists in any society in which power and resources are distributed unevenly across social groups. According to this fundamental-causes perspective, policy makers’ focus on risk factors (or proximate causes) of diseases—things such as inactivity, air pollution, and neighborhood disadvantage—is likely to be misplaced. The development of effective interventions is then based on a concrete understanding of what fundamental causes subject people to these risk factors.

The concept of fundamental causes of diseases points to the significance of underlying social conditions in shaping the epidemiology of diseases; however, the stress-exposure framework focuses on proximate causes (or risk factors) and offers a different explanation of mental health inequality within a society (Gee and Payne-Sturges, 2004; George and Lynch, 2003; Lantz et al., 2005). As suggested by this latter framework, the reason why certain socio-demographic groups appear to be more susceptible to mental illness has nothing to do with their intrinsic vulnerability. Similarly, if scholars were able to measure stressors in a more comprehensive way (Lantz et al., 2005) or properly model the impact of stressors on mental illness (George and Lynch 2003), exposure to a wide range of social, economic, or environmental stressors might account for more mental health problems experienced by the disadvantaged groups.

Both conceptual frameworks are useful in exploring the link between hukou and depressive symptoms of segmented demographic groups. According to the fundamental-causes-of-disease theory, inequality in social conditions is a primary vehicle through which the health disparity comes into being and perpetuates itself. In empirical research across different societies, such inequality in social conditions is often operationalized by measures of socioeconomic status. Our first hypothesis is thus formulated as follows:

Hypothesis 1: Differences in depressive symptoms among demographic groups segmented by hukou are attributable to inequality in socioeconomic status.

Education is shown to be the key dimension of one’s socioeconomic status (Hauser and Warren, 1997), and as such, we specifically consider the role of education in shaping mental health inequalities. Since China’s current policies of human-capital investment favor urban residents over their rural counterparts and discriminate against children of rural origin, the country’s salient rural–urban divide also has a far-reaching influence on education inequality (Fu and Ren, 2010; Heckman, 2005). As it stands, one year of schooling, depending on where that schooling takes place, may not necessarily mean the same thing for everyone and tends to have different impacts on mental health. Owing to the educational inequality that exists as part of China’s rural–urban divide, we argue that among the various indicators of socioeconomic status, education is probably most relevant to the rural–urban disparity in depressive symptoms. Moreover, high schools and universities in China are largely located in urban areas; thus, it appears that the rural–urban difference in the quality of school education is particularly striking in primary and middle schools, which, taken together, are officially referred to as the ‘nine-year compulsory education system’ in China. The shortage of school teachers (especially teachers with a college education), under-investment in school buildings, inadequate maintenance of teaching facilities, and distance to school locations, all contribute to the poor quality of compulsory education in rural China (Fu and Ren, 2010; Heckman, 2005). We thus have the following hypothesis:

Hypothesis 1A: Among the various indicators of socioeconomic status, education (especially the period of compulsory education) is most relevant to the rural–urban disparity in depressive symptoms.

From a stress-exposure perspective, individuals with agricultural hukou status are exposed to or at a higher risk of (more) stressors with known consequences for depression, as compared with those with non-agricultural hukou status. For example, most rural migrants tend to work as manual laborers in private sectors (Chan and Zhang, 1999). Because these jobs provide little job security, these laborers are at a higher risk of unemployment, an important stressor for mental illness (Kessler et al., 1987). Meanwhile, rural migrants, especially those working in foreign-invested factories, have to live apart from their family members, which means a lack of social support from strong ties, and has negative implications for mental health (Dean et al., 1992). Moreover, stress exposure should be understood in a relative sense, and in the existing literature, a correspondence with opportunity hoarding has been found (Tilly, 1998). Opportunity hoarding happens through a process of social exclusion whereby the disadvantaged are deprived of access to valuable resources for upward mobility that are available to the advantaged (Tilly, 1998). Because of this opportunity hoarding, it is the combination of ‘the allocation of people to social categories, and the institutionalization of practices that allocate resources unequally across these categories’ (Massey, 2007: 5–6) that constitutes the socio-political origins of stress faced by the disadvantaged (Lin, 2000; Massey, 2007; Tilly, 1998).When rural peasants and migrant workers are deprived of key resources and supportive environments in the urban space, their relative deprivation also constitutes one form of stressor. For example, the unavailability of (free) medical services or affordable homeownership in urban areas can also have implications for the mental health of rural–urban migrants (Katon and Schulberg, 1992; Rossi and Weber, 1996). Informed by the stress-exposure perspective, we hypothesize that:

Hypothesis 2: Differences in depressive symptoms across segmented demographic groups are related to their differential exposure to proximate (social, economic, or medical) stressors.

Data and methods

Data

This study is based on the latest wave (2016) of the CLDS, which is administered by the Center for Social Surveys at Sun Yat-sen University. As a large-scale nationally representative survey of China’s labor force, the CLDS covers all provinces in mainland China except for Tibet and Hainan. A multistage stratified cluster sampling technique with probability proportionate to size (PPS) was employed by the CLDS to select households from a selected residents committee (RC, or juwei hui) or peasants committee (PC, or cunwei hui). The 2282 counties (xian), county-level cities, or county-level districts served as the primary sampling units. Street offices (jiedao ban) and townships (zhen) within a selected county and then RCs and PCs within a selected street office or township, respectively, were randomly chosen. Face-to-face interviews were conducted by trained interviewees using a Computer Assisted Personal Interviewing (CAPI) system. The response rate for the 2016 wave was 71.0% (81.6% in rural areas and 60.9 % in urban areas). Our analysis is based on 12,555 individuals from the 2016 CLDS project.

Measures

The outcome variable is the Center for Epidemiologic Studies Depression (CES-D) 20-item self-reported depressive symptom scale (Lewinsohn et al., 1997; Radloff, 1977). The CES-D scale asks respondents to assess how often over the last week they experienced any of the 20 depressive scenarios/items, including sleeping, appetite, loss of interest, sadness, and fatigue. The response categories for each item are the same: rarely or none of the time (0 days and coded as zero), some or a little of the time (1–2 days and coded as one), moderately or much of the time (3–4 days and coded as two), and most or almost all the time (5–7 days and coded as three). The Cronbach’s alpha of these 20 items reported by respondents is 0.95, suggesting high internal consistency. The CES-D score, or the sum of scores of the 20 CES-D items, is used as the outcome variable and a higher score indicates that a respondent is more depressed (Ragland et al., 2005).

Because the main thrust of this study is to investigate the mental health conditions of demographic groups segmented by hukou, the overall sample was categorized into three groups: rural residents, rural–urban migrants, and urban residents. These three groups are defined by their de jure hukou status and de facto residence. Rural residents had agricultural hukou status and resided in rural areas; rural–urban migrants also had agricultural hukou status but lived in urban areas at the time of interview; and urban residents held non-agricultural hukou status and resided in urban areas (urban-to-urban migrants are also included in this category). It should be noted, however, that the pattern of internal migration in China is more complex than that revealed by the three demographic groups. For example, the original dataset also included 201 individuals who held non-agricultural hukou but lived in rural regions. Yet, the three categories investigated here allow us to have a general picture of the Chinese population segmented by hukou.

Several socioeconomic indicators are considered. Since the link between hukou and education has been well documented in the existing literature (Chan, 2009; Fu and Ren, 2010; Heckman, 2005), educational attainment is measured by a categorical variable consisting of four categories: primary school, middle school, high school, and college or above. Income is measured by a continuous variable denoting the logarithm of annual per capita household income. For one’s occupation, we consider two dummy variables denoting administrative jobs (e.g., government officials, military officers, and local cadres) and professional jobs (e.g., professors, doctors, and lawyers), respectively, and treat all other jobs as reference. Given the documented impact of work units on one’s (mental) health status (Fu and George, 2015; Lai, 1995), a dichotomous variable, state-sector employment, is included to denote whether respondents work in state sectors (governments, institutes, or state-owned enterprises).

A series of proximate stressors related to mental illness are incorporated into this research (Baker et al., 2013; Dean et al., 1992; Kessler et al., 1987; World Health Organization, 2001). Given that unemployment, living alone, and housing tenancy are possible risk factors for mental illness (Baker et al., 2013; Dean et al., 1992; Evans et al., 2003; Kessler et al., 1987), these stressors are denoted by three dichotomous variables. Stressors pertaining to conditions of physical health and medical services are also considered due to their known consequences on mental health (Phelan et al., 2001; Richardson et al., 2005). In China, the public insurance program (gongfei yiliao) covers the majority of, if not all, medical expenses and is often provided to state employees (Shi, 1993). The absence of such coverage is denoted by a dichotomous variable, unavailability of public medical insurance. In the 2016 CLDS survey, respondents were asked whether cost was the reason they did not see a doctor or visit a clinic when they were sick. A dichotomous variable, unaffordability of medical services, is coded as one if a respondent reported such a scenario. The presence of any self-reported occupational disease is also denoted by a dichotomous variable. Finally, control variables are male (female as reference), age (in years), being married (other marital status as reference), and region (coastal, central, and western areas).

Methods

We consider the outcome variable, the sum of scores of the 20 CES-D items, as a count response. Generalized linear models from the Poisson family of distributions are often used to model count responses (Fu et al., Forthcoming; Fu et al., 2017; Long and Freese, 2006; Powers and Xie, 2008). Since mental disorders are distributed unevenly across socio-demographic groups in China (Parker et al., 2001; Phillips et al., 2009), these Poisson-based models are particularly useful in analyzing skewed counts of depressive symptoms (Cacioppo et al., 2006; Hays et al., 1998; Wade et al., 2002). The Poisson distribution requires that its mean be the same as its variable, which is seldom satisfied in empirical data. In practice, researchers find that the variance of count data tends to be larger than its mean, which is referred to as overdispersion. By and large, two methods have been proposed to account for overdispersion in count data. First, the zero-inflated Poisson approach deals with excessive zeros in count data and suggests that those who are not at risk of the count response (e.g., depressive symptoms) account for the overdispersion issue (Fu et al., 2016). Second, and in contrast, the negative-binomial approach separately models the variance and the mean of count data and thus allows the former to be greater than the latter. Despite a disproportionately large number of respondents (25.4% of the sample) reporting zero depressive symptoms, it remains unclear whether the zero-inflated Poisson approach is preferred over the negative-binomial approach before formal statistical tests are performed. As expected, a mixture of the two approaches—the ZINB model—outperforms either of the two aforementioned approaches in terms of replicating the empirical distribution (see Figure 1). We thus use the ZINB model to analyze depressive symptoms across demographic groups segmented by hukou. According to the ZINB model (Cameron and Trivedi, 1998), the probability mass function of the count variable (i.e., the CES-D depression score) is as follows

\begin{matrix} Pr (y_{i}) = {\begin{matrix} p_{i} + (1 - p_{i}) g (y_{i}) & when y_{i} = 0 \\ (1 - p_{i}) g (y_{i}) & {wheny}_{i} > 0 \end{matrix} \end{matrix}

(1)

Figure 1.

Distributions of observed and predicted depressive symptoms.

In the equation (1) above, y_i is the CES-D depression score and p_i is an exposure parameter denoting the probability that a respondent is not at any risk of depression. Here $g (y_{i})$ represents a negative binomial function that generates the CES-D depressive score, which can be written as

g (y_{i}) = \frac{Γ (y_{i} + α^{- 1})}{Γ (α^{- 1}) Γ (y_{i} + 1)} (\frac{1}{1 + α μ_{i}})^{α^{- 1}} (\frac{α μ_{i}}{1 + α μ_{i}})^{y_{i}}

where α is the dispersion parameter and

g (y_{i})

reduces to a Poisson function when α is zero. The logarithm of

μ_{i}

corresponds to the link function

\begin{matrix} \ln μ_{i} = β_{0} + \sum_{p = 2} β_{D_{p}} D_{p} + \sum_{q = 7} β_{{SES}_{q}} {SES}_{q} + \sum_{r = 6} β_{{SC}_{r}} {Stressor}_{r} + \sum_{s = 6} β_{{Control}_{s}} {Control}_{s} \end{matrix}

where D,

SES

Stressor

, and

Control

represent measures pertaining to demographic groups segmented by hukou, socioeconomic status, stressors, and control variables, respectively. The

β s

are their corresponding regression coefficients.

Equation (1) means that there are two components of the probability that a zero CES-D score (i.e., $y_{i} = 0$ ) is reported: (a) the probability that a respondent is not at risk of depression, which is denoted by p_i; and (b) the probability that a respondent is indeed at risk of depression and the zero CES-D score is instead generated by the negative binomial function $g (y_{i})$ governing CES-D depressive scores, which is denoted by $(1 - p_{i}) g (y_{i})$ . Clearly, a non-zero CES-D score (i.e., $y_{i} > 0$ ) is reported only when a respondent is at risk of depression and its probability is denoted by $(1 - p_{i}) g (y_{i})$ .

Hypothesis 1A posits that education, especially the period of compulsory education, is most relevant to the rural–urban disparity in mental health. To test this hypothesis across demographic groups, we compare the effects of different stages of educational attainment on depression using spline regression (Fu and Ren, 2010; Marsh and Cormier, 2001).¹ When dummy variables are used to measure educational attainment, the effect of each stage of educational attainment can be understood as a combined effect of the educational attainment being investigated and all lower stages of educational attainment. For example, the effect of high school corresponds to the combination of the effects of high school, middle school, and primary school. Spline regression models are thus used to tease out the separate effects of different stages of educational attainment. To focus on those who are actually at risk of depression, a ZINB model with only control variables (sex, age, marital status, and regional location) was first conducted to identify respondents who are at little risk of depression (defined as probability of zero inflation is greater than 0.9). These respondents who are at little risk of depression were subsequently excluded from the spline Poisson regression analysis. Spline negative binomial regression is also employed in an auxiliary analysis (results not shown here) to assess the association between educational attainment and depressive symptoms, and our key findings remain virtually the same.

Results

The description of the sample is shown in Table 1. First and foremost, respondents with agricultural hukou status (both rural residents and rural–urban migrants) reported higher CES-D scores than their peers with non-agricultural hukou status. In particular, rural residents on average had the highest level of CES-D scores among the three demographic groups segmented by hukou. A salient rural–urban disparity is also observed in their socio-economic status. More than 40% of urban residents had completed higher education, but the corresponding percentages for rural–urban migrants and rural residents are only 13.2% and 4.3%, respectively. As expected, rural residents had the lowest level of per capita household income and very few respondents with agricultural hukou status had administrative jobs (0.1% for rural–urban migrants and 0.6% for rural residents). More than a quarter (28.2%) of urban residents were employed by state sectors but far fewer respondents with agricultural hukou status worked in state sectors.

Table 1.

A summary of sample characteristics by demographic group (n = 12,555).

	Urban residents		Rural–urban migrants		Rural residents
	Mean	SD	Mean	SD	Mean	SD
CES-D scores	6.2	8.8	7.0	9.3	7.7	9.2
Socioeconomic status
Educational attainment
Primary school	4.0%		22.2%		42.3%
Middle school	20.0%		40.4%		39.8%
High school	35.1%		24.1%		13.7%
College or above	40.9%		13.2%		4.3%
Per capita household income (10,000 yuan)	3.598	33.557	2.128	3.729	0.883	1.429
Occupation
Administration	1.3%		0.1%		0.6%
Professional	12.1%		4.6%		1.6%
Others	86.7%		95.4%		97.9%
State-sector employment	28.2%		6.4%		2.2%
Stressors
Unemployment	11.4%		12.7%		10.3%
Living alone	4.9%		7.0%		1.1%
Housing tenancy	10.3%		50.5%		13.1%
Unavailability of public medical insurance	89.8%		97.5%		99.3%
Unaffordability of medical services	5.2%		5.3%		5.3%
Presence of any occupational disease	7.1%		7.9%		4.7%
Control variables
Male	46.7%		44.1%		47.0%
Age	40.8	12.9	37.3	11.8	42.7	12.4
Married	71.9%		77.7%		83.8%
Region
Coastal area	44.9%		56.0%		29.7%
Central area	30.6%		19.0%		9.1%
Western area	24.5%		25.1%		61.2%
Number of observations	2766		1177		8612

For stressors considered in this research, rural–urban migrants reported the highest level of unemployment (12.7%) among the three demographic groups investigated. As compared to the other two groups, rural–urban migrants were also more likely to live alone (7.0%) and be housing tenants (50.5%). The public (medical) insurance program was essentially unavailable to rural–urban migrants and rural residents (especially the latter). For the three demographic groups, there is little variation in their average proportions of medical services unaffordability. Despite the fact that rural–urban migrants reported the highest prevalence rate of occupational diseases (7.9%), one surprising finding is that rural residents reported the lowest prevalence rate of occupational diseases (4.7%). Given the self-reported nature of this variable, this surprising finding may suggest that rural residents are less aware of their history of diseases (Li et al., 2009). For control variables, rural–urban migrants tend to be female and younger as compared with the other two groups. Respondents who were married account for 83.8% and 77.7% of rural residents and rural–urban migrants, respectively, while the corresponding percentage for urban residents is 71.9%. Over one half of rural–urban migrants (56.0%) were interviewed in coastal areas and most rural residents (61.2%) were interviewed in western areas. Urban residents were distributed more evenly across the three regions.

Results from the ZINB models are presented in Table 2 and Table 3. Since the results from Vuong tests suggest that zero-inflation models should be employed to analyze the empirical count data, we first assess the characteristics associated with the probability that a respondent is not exposed to depression, or p_i in equation (1). In Table 2, when control variables are considered in Model 1, both rural–urban migrants and rural residents are less likely than urban residents to be in the group that is risk-free of depressive symptoms. In other words, individuals with agricultural hukou status tend to be more exposed to depression. For the control variables, male respondents are also more likely to be exposed to depression than their female counterparts. When socioeconomic status is considered in Model 2 and Model 3, more educated individuals are less likely to be exposed to depression (primary school as reference) but no other variables of socioeconomic status are significantly associated with exposure to depressive symptoms. Model 4 considers several major stressors and shows that those who were unemployed, failed to obtain public (medical) insurance, and found medical services unaffordable were more likely than their counterparts to be exposed to depression. When all measures are incorporated in Model 5, the significant disadvantage of rural–urban migrants’ exposure to depression disappears, while rural residents are still more likely than urban residents to be exposed. Income becomes a protective factor and living alone becomes a risk factor for exposure to depression.

Table 2.

Results from zero-inflated negative binomial regression: zero inflation of depressive symptoms.

Variables	Model 1	Model 2	Model 3	Model4	Model 5
Demographic groups (urban residents as reference)
Rural–urban migrants	−0.570***	−0.469**	−0.569***	−0.396*	−0.290
Rural residents	−0.563***	−0.371***	−0.515***	−0.542***	−0.307**
Socioeconomic status
Educational attainment (primary school as reference)
Middle school		0.464***			0.396***
High school		0.536***			0.497***
College or above		0.498**			0.444**
Log of per capita household income			0.037		0.0827*
Occupation (others as reference)
Administration			0.006		0.051
Professional			−0.142		−0.207
State-sector employment			0.000		−0.099
Stressors
Unemployment				−0.724***	−0.699***
Living alone				−0.142	−0.724*
Housing tenancy (vs. homeownership)				−0.272	−0.267
Unavailability of public medical insurance				−2.919*	−2.392**
Unaffordability of medical services				−2.306**	−2.279**
Presence of any occupational disease				−0.269	−0.268
Control variables
Male	−0.324***	−0.261***	−0.315***	−0.299***	−0.237**
Age	0.009	0.023	0.004	0.012	0.028
Age squared	0.000	0.000	0.000	0.000	0.000
Married	0.155	0.127	0.195	0.069	0.055
Region (coastal area as reference)
Central area	0.027	0.055	0.024	−0.027	0.007
Western area	−0.139	−0.106	−0.135	−0.239*	−0.186*
Constant	−1.410**	−2.257***	−1.453***	−0.873	−1.918***
Log-likelihood	−37309.6	−37239.2	−37286.9	−37102.9	−37016.7
Likelihood ratio χ²	189.63***	278.47***	228.73***	379.26***	497.55***
Vuong test (z-value)	8.58***	8.92***	8.71***	10.38***	10.76***

p < .05; ** p < .01;*** p < .001 (two-tailed tests)

Table 3.

Results from zero-inflated negative binomial regression: counts of depressive symptoms.

Variables	Model 1	Model 2	Model 3	Model 4	Model 5
Demographic groups (urban residents as reference)
Rural–urban migrants	0.098*	0.010	0.044	0.041	−0.058
Rural residents	0.117***	−0.023	0.030	0.126***	−0.065
Socioeconomic status
Educational attainment (primary school as reference)
Middle school		−0.216***			−0.213***
High school		−0.273***			−0.248***
College or above		−0.384***			−0.309***
Log of per capita household income			−0.020**		−0.037***
Occupation (others as reference)
Administration			−0.289*		−0.182
Professional			−0.236***		−0.177**
State-sector employment			−0.136**		−0.066
Stressors
Unemployment				0.167***	0.153***
Living alone				0.037	0.289**
Housing tenancy (vs. homeownership)				0.114*	0.083
Unavailability of public medical insurance				0.165*	0.105
Unaffordability of medical services				0.465***	0.447***
Presence of any occupational disease				0.198***	0.235***
Control variables
Male	−0.148***	−0.111***	−0.142***	−0.140***	−0.104***
Age	0.022**	0.012	0.027***	0.020**	0.013
Age squared	0.000	0.000	−0.000**	0.000	0.000
Married	−0.226***	−0.216***	−0.237***	−0.189***	−0.179***
Region (coastal area as reference)
Central area	0.176***	0.175***	0.166***	0.178***	0.169***
Western area	0.191***	0.170***	0.188***	0.201***	0.178***
Constant	1.428***	2.003***	1.488***	1.209***	1.897***
α (overdispersion)	1.131***	1.110***	1.121***	1.083***	1.056***

p < .05; **p < .01;***p < .001 (two-tailed tests)

Next, we investigate the depressive symptoms of those who are at risk of depression in order to test our research hypothesis, which is implemented by the negative-binomial component of the ZINB model (see Table 3). As suggested by the significant values of α across the five models, negative binomial models should be employed to address the overdispersion issue. When only control variables are included in Model 1, both rural–urban migrants and rural residents are significantly associated with higher CES-D scores than urban residents with non-agricultural hukou status. Yet, one remarkable finding is that the significantly higher CES-D scores reported by individuals with agricultural hukou status become non-significant once educational attainment is considered in the analysis (see Model 2 in Table 3). Moreover, the association between rural residents and the CES-D score even becomes negative. While other socio-economic variables (income, occupation, and state-sector employment) can also explain away the significant effects of demographic groups on depression, a crossover in the effect of rural migrants is not observed in Model 3. We interpret these findings as support for Hypothesis 1 that the inequality in socioeconomic status plays a key role in shaping different mental health conditions across demographic groups segmented by hukou. Various proximate stressors are then included in the analysis to test Hypothesis 2. In Model 4, rural–urban migrants are no longer significantly associated with higher depressive scores but rural residents are. Hypothesis 2 is thus partially supported. When all measures are incorporated in Model 5, the association between demographic groups and the CES-D score is non-significant. Among socioeconomic variables, education, income, and having a professional job appear to be protective factors in mental health; unemployment, living alone, unaffordability of medical services, and the presence of an occupational disease are risk factors for depression. For control variables, male and married respondents reported significantly lower CES-D scores, while individuals living in western or central areas (coastal areas as reference) reported significantly higher depressive scores.

Results from Model 2 in Table 3 suggest that educational attainment is relevant to the rural–urban disparity in mental health, which is in line with Hypothesis 1A. We next use spline Poisson regression to further assess the association between different stages of educational attainment and depressive symptoms in separate demographic groups (see Table 4). Since we are interested in the total effects of educational attainment on depressive symptoms, socioeconomic variables and stressors were excluded from the analysis. Meanwhile, individuals at little risk of depression were also excluded from the analysis based on their probabilities of exposure to depression (

p_{i} \geq 0.9

), which were predicted using a ZINB model with control variables. In Table 4, the effects of the control variables are similar to those presented in Table 3. In terms of the effect of one year of schooling, Table 4 suggests that the most striking difference across demographic groups happens in the middle-school period. To illustrate this finding, we multiply the effect of one year of schooling by its corresponding years of stages of education (six years for primary school, three years for middle school, three years for high school, and four years for college) and then plot the total effects of different stages of education in Figure 2. Across demographic groups segmented by hukou, their educational disparity in depression is salient. Beyond the stage of primary school, education completed by urban residents appears to be a stronger protective factor for depression than that completed by rural residents or rural–urban migrants. The most striking difference takes place in the period of middle school. The middle-school period greatly reduces depressive symptoms for urban residents, whereas the effect of the same period of education is much smaller for rural residents and even becomes positive for rural–urban migrants.

Table 4.

Spline Poisson regression analysis of depressive symptoms: the effect of educational attainment.^a

	Urban residents	Rural–urban migrants	Rural residents
Years of schooling
1 – 6 years (primary school)	0.050***	0.030**	−0.016***
7 – 9 years (middle school)	−0.229***	0.017	−0.070***
10 – 12 years (high school)	−0.049***	−0.043***	−0.019***
Above 12 years (college or above)	−0.060***	−0.030*	−0.006
Control variables
Male	−0.039*	−0.264***	−0.160***
Age	0.018***	0.001	0.019***
Age squared	−0.000***	−0.000	−0.000***
Married	−0.192***	−0.123**	−0.307***
Region (coastal region as reference)
Central region	0.156***	0.127***	0.159***
Western region	0.007	−0.077*	0.195***
Constant	2.167***	1.716***	1.749***
Log-likelihood	−12877.028	−6284.306	−48017.79
Likelihood ratio χ²	965.59***	195.17***	2913.47***
Number of observations	2211	1100	7738

Those who are at very low risk for depression (probability of exposure to depression is less than 0.10) are excluded from the analysis.

p < .05; **p < .01;***p < .001 (two-tailed tests)

Figure 2.

Partial effects of educational attainment on CES-D scores by demographic groups.

Conclusion

Although there has been a longstanding curiosity about the socio-political consequences of China’s remarkable urban–rural divide, we have yet to understand the divide’s possible influence on mental health. Using data from the 2016 wave of the CLDS, we find that depressive symptoms reported by individuals with agricultural hukou status (rural–urban migrants and rural residents) are significantly higher than those reported by urban residents. Results from ZINB regression suggest that this rural–urban difference in depressive symptoms can be attributable to socioeconomic status and proximate stressors such as unemployment, living alone, and unaffordability of medical services. In particular, the rural–urban difference in depressive symptoms is explained away by educational attainment: after controlling for educational attainment there is an urban–rural crossover in depression such that depressive scores reported by rural residents become lower than those reported by urban residents. As suggested by results from spline Poisson regression, the protective effects of the period of middle school, which vary substantially across demographic groups, are especially relevant to the rural–urban disparity in depression. We argue that hukou is a fundamental cause of disease in China and its effects on depression are mediated by various social, economic, and medical factors. Mental health is an important yet under-studied area in which China’s salient urban–rural inequality strikes.

The existing literature suggests that people in China may deny experiencing depression or express it somatically given the low prevalence rates of depression reported in China (Parker et al., 2001). Findings from this study show that a proportion of the Chinese population is actually at little risk of depression, which may contribute to the low prevalence rates of depression in China. Furthermore, socioeconomic, cultural, or cognitive factors shaping the exposure to depression in China deserve attention in future studies. The important role of education in explaining the rural–urban disparity in depression is in line with conclusions from previous studies: educational inequality is a primary mechanism through which the hukou system perpetuates and magnifies social inequality in China. For example, the period of middle school has a salient protective effect against depression among urban residents, whereas such a protective effect appears to be much smaller among rural residents and even disappears among rural–urban migrants. The life-course trajectories of these demographic groups segmented by hukou may shed light on this interesting finding. For urban residents, the middle-school years tend to be a comparatively positive and optimistic period for urban residents as they prepare for high school. In contrast, middle school is a critical period for rural adolescents to see their family hardships and limited career opportunities in relation to China's broader social context. After middle school, a large proportion of them have to leave their families, work in labor-intensive industries and endure difficulties in cities.

The key issue raised by this study is probably not what variable predicts depression across demographic groups. Rather, our findings show that mental health problems experienced by the disadvantaged with rural origins are deeply embedded in the political structuring of their everyday lives. With the persistence of power in reform-era China, scholars cannot fully understand the origin of mental illness among the Chinese unless the role of politics has been carefully addressed (Fu, 2018). Despite the prevailing public outcry against the hukou system and piecemeal hukou reforms implemented at both the national and local levels, this study demonstrates that the system still functions in China as an institutional arrangement creating inequality and injustice. This socialist policy exists, moreover, against the backdrop of institutional arrangements that have been established in other modern societies to correct for inequality and injustice. Despite the country’s rapid economic growth since the 1980s, individuals with rural origins remain at the bottom of China’s social hierarchy. With the poor quality of their education, inadequate investment in infrastructure, unavailability of state resources, and limited life chances, rural residents cannot flourish in an environment where they are institutionally discriminated against. From the social-stratification perspective, our findings do not deviate materially from Hu Feng’s vivid portrayals of Chinese peasants in the 1930s. In his epilogue to Xiao Hong’s novel The Field of Life and Death, Hu Feng wrote:

[Peasants in China] live like mosquitoes, pointlessly reproduce themselves, chaotically pass away. [They] use their own blood, sweat and lives to fertilize the land, cultivate crops, raise cattle. [They] creep beneath the power of the tyranny of nature and tyrannies with two legs (Hu, 1991 [1935]: 145).

Even if they migrate to urban areas, those with rural hukou remain strangers in the city, have no social connections, and construct houses that they cannot afford. Their need for medical services, children’s education, and housing in the city do not concern policy makers. Once the flow of rural–urban migrants is at odds with urban or regional development policies, they become dispensable in the eyes of both policy makers and urban residents. As long as China’s rural–urban divide persists, the hukou system remains a fundamental force driving the disparity in mental health, amongst other social problems in China.

Footnotes

Acknowledgment

The authors are grateful for comments from Professor Eric Fong, Luying Wang, Erin Williams and two anonymous reviewers.

Funding

The authors gratefully acknowledge financial support from The Chiang Ching-kuo Foundation for International Scholarly Exchange, Hampton Scholar Award at The University of British Columbia and the National Natural Science Foundation in China (grant number 71540031).

Note

References

Aneshensel

Phelan

Bierman

(2013) The sociology of mental health: Surveying the field. In: Aneshensel

Phelan

Bierman

(eds) Handbook of the Sociology of Mental Health, New York: Springer, pp. 1–19.

Baker

Bentley

Mason

(2013) The mental health effects of housing tenure: Causal or compositional? Urban Studies 50(2): 426–442.

Bian

Logan

(1996) Market transition and the persistence of power: The changing stratification system in urban China. American Sociological Review 61(5): 739–758.

Cacioppo

Hughes

Waite

et al. (2006) Loneliness as a specific risk factor for depressive symptoms: Cross-sectional and longitudinal analyses. Psychology and Aging 21(1): 140–151.

Cameron

Trivedi

(1998) Regression Analysis of Count Data, Cambridge: Cambridge University Press.

Chan

(1994) Urbanization and rural–urban migration in China since 1982: A new baseline. Modern China 20(3): 243–281.

Chan

(2009) The Chinese hukou system at 50. Eurasian Geography and Economics 50(2): 197–221.

Chan

Buckingham

(2008) Is China abolishing the hukou system? The China Quarterly 195: 582–606.

Chan

Zhang

(1999) The hukou system and rural–urban migration in China: Processes and changes. The China Quarterly 160: 818–855.

10.

Chen

Yang

Liu

(2010) Social change and socioeconomic disparities in health over the life course in China: A cohort analysis. American Sociological Review 75(1): 126–150.

11.

Chen

et al. (2012) Mental health, duration of unemployment, and coping strategy: a cross-sectional study of unemployed migrant workers in eastern China during the economic crisis. BMC Public Health 12(597): 1–12.

12.

Dean

Kolody

Wood

et al. (1992) The influence of living alone on depression in elderly persons. Journal of Aging and Health 4(1): 3–18.

13.

Evans

Wells

Moch

(2003) Housing and mental health: A review of the evidence and a methodological and conceptual critique. Journal of Social Issues 59(3): 475–500.

14.

(2018) Bringing urban governance back in: Neighborhood conflicts and depression. Social Science and Medicine 196: 1–9.

15.

George

(2015) Socioeconomic determinants of childhood overweight and obesity in China: The long arm of institutional power. Sociology of Health and Illness 37(6): 805–822.

16.

Fu Q, Guo X, and Land KC (Forthcoming) Optimizing count responses in surveys: A machine-learning approach. Sociological Methods and Research.

17.

Guo

Land

(2017) A Poisson-Multinomial Mixture Approach to Grouped and Right-Censored Counts . Communications in Statistics-Theory and Methods. DOI: 10.1080/03610926.2017.1303736.

18.

Land

Lamb

(2016) Violent physical bullying victimization at school: Has there been a recent increase in exposure or intensity? An age-period-cohort analysis in the United States, 1991 to 2012. Child Indicators Research 9(2): 485–513.

19.

Ren

(2010) Educational inequality under China's rural–urban divide: The hukou system and return to education. Environment and Planning A 42(3): 592–610.

20.

Gee

Payne-Sturges

(2004) Environmental health disparities: A framework integrating psychosocial and environmental concepts. Environmental Health Perspectives 112(17): 1645–1653.

21.

George

Lynch

(2003) Race differences in depressive symptoms: A dynamic perspective on stress exposure and vulnerability. Journal of Health and Social Behavior 44(3): 353–369.

22.

Graham

Zhou

Zhang

(2017) Happiness and health in China: The paradox of progress. World Development 96: 231–244.

23.

Hauser

Warren

(1997) Socioeconomic indexes for occupations: A review, update, and critique. Sociological Methodology 27(1): 177–298.

24.

Hays

Landerman

George

et al. (1998) Social correlates of the dimensions of depression in the elderly. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 53B(1): 31–39.

25.

Heckman

(2005) China's human capital investment. China Economic Review 16(1): 50–70.

26.

(1991 [1935]) Epilogue to the Field of Life and Death, Harbin, China: Harbin Press.

27.

Jiang

Sato

(2012) Identity, inequality, and happiness: Evidence from urban China. World Development 40(6): 1190–1200.

28.

Kanbur

Zhang

(1999) Which regional inequality? The evolution of rural–urban and inland–coastal inequality in China from 1983 to 1995. Journal of Comparative Economics 27(4): 686–701.

29.

Katon

Schulberg

(1992) Epidemiology of depression in primary care. General Hospital Psychiatry 14(4): 237–247.

30.

Kessler

House

Turner

(1987) Unemployment and health in a community sample. Journal of Health and Social Behavior 28(1): 51–59.

31.

Lai

(1995) Work and family roles and psychological well-being in urban China. Journal of Health and Social Behavior 36(1): 11–37.

32.

Lantz

House

Mero

et al. (2005) Stress, life events, and socioeconomic disparities in health: Results from the Americans' Changing Lives Study. Journal of Health and Social Behavior 46(3): 274–288.

33.

Lewinsohn

Seeley

Roberts

et al. (1997) Center for Epidemiologic Studies Depression Scale (CES-D) as a screening instrument for depression among community-residing older adults. Psychology and Aging 12(2): 277–287.

34.

L-K

J-F

et al. (2009) Knowledge and attitudes about human papillomavirus (HPV) and HPV vaccines among women living in metropolitan and rural regions of China. Vaccine 27(8): 1210–1215.

35.

Lin

Wang

et al. (2011) Discrimination, perceived social inequity, and mental health among rural-to-urban migrants in China. Community Mental Health Journal 47(2): 171–180.

36.

Lin

(2000) Inequality in social capital. Contemporary Sociology 29(6): 785–795.

37.

Lin

Bian

(1991) Getting ahead in urban China. American Journal of Sociology 97(3): 657–688.

38.

Link

Phelan

(1995) Social conditions as fundamental causes of disease. Journal of Health and Social Behavior (Extra Issue). 80–94.

39.

Liu Y and Fu Q (2017) Changing demographic characteristics and motives for suicide in rural China, 1980–2009. Chinese Sociological Dialogue 2(3–4): 136–148.

40.

Liu

(2005) Institution and inequality: The hukou system in China. Journal of Comparative Economics 33(1): 133–157.

41.

Long

Freese

(2006) Regression Models for Categorical Dependent Variables Using Stata, College Station, TX: Stata Press.

42.

Treiman

(2012) Migration and depressive symptoms in migrant-sending areas: Findings from the survey of internal migration and health in China. International Journal of Public Health 57(4): 691–698.

43.

Qin

(2014) Healthy migrant and salmon bias hypotheses: A study of health and internal migration in China. Social Science and Medicine 102: 41–48.

44.

Marsh

Cormier

(2001) Spline Regression Models, Thousand Oaks, CA: Sage.

45.

Massey

(2007) Categorically Unequal: The American Stratification System, New York: Russell Sage Foundation.

46.

Miech

Caspi

Moffitt

et al. (1999) Low socioeconomic status and mental disorders: A longitudinal study of selection and causation during young adulthood. American Journal of Sociology 104(4): 1096–1131.

47.

Muntaner

Borrell

Chung

(2007) Class relations, economic inequality and mental health: Why social class matters to the sociology of mental health. In: Avison

McLeod

Pescosolido

(eds) Mental Health, Social Mirror, New York: Springer, pp. 127–141.

48.

Niu

(2015) Internal migration and health stratification in urban China. Asian and Pacific Migration Journal 24(4): 432–462.

49.

Parker

Gladstone

Chee

(2001) Depression in the planet’s largest ethnic group: The Chinese. American Journal of Psychiatry 158(6): 857–864.

50.

Phelan

Link

(2005) Controlling disease and creating disparities: A fundamental cause perspective. The Journals of Gerontology: Series B 60(2): S27–S33.

51.

Phelan

Link

Diez-Roux

et al. (2004) ‘Fundamental Causes’ of social inequalities in mortality: A test of the theory. Journal of Health and Social Behavior 45(3): 265–285.

52.

Phelan

Stradins

Morrison

(2001) Physical health of people with severe mental illness: Can be improved if primary care and mental health professionals pay attention to it. BMJ: British Medical Journal 322(7284): 443–444.

53.

Phillips

Zhang

Shi

et al. (2009) Prevalence, treatment, and associated disability of mental disorders in four provinces in China during 2001–05: An epidemiological survey. The Lancet 373(9680): 2041–2053.

54.

Powers

Xie

(2008) Statistical Methods for Categorical Data Analysis, Bingley, UK: Emerald Group Publishing.

55.

Radloff

(1977) The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement 1(3): 385–401.

56.

Ragland

Satariano

MacLeod

(2005) Driving cessation and increased depressive symptoms. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences 60(3): 399–403.

57.

Rarick JR, Dolan CT, Han WJ, et al. (2017) Relations between socioeconomic status, subjective social status, and health in Shanghai, China. Social Science Quarterly, DOI: 10.1111/ssqu.12360.

58.

Richardson

Faulkner

McDevitt

et al. (2005) Integrating physical activity into mental health services for persons with serious mental illness. Psychiatric Services 56(3): 324–331.

59.

Rossi

Weber

(1996) The social benefits of homeownership: Empirical evidence from national surveys. Housing Policy Debate 7(1): 1–35.

60.

Shi

(1993) Health care in China: A rural–urban comparison after the socioeconomic reforms. Bulletin of the World Health Organization 71(6): 723.

61.

Shi

Bai

Shen

et al. (2016) Effects of parental migration on mental health of left-behind children: Evidence from northwestern China. China and World Economy 24(3): 105–122.

62.

Stewart

(2016) Horizontal Inequalities and Conflict: Understanding Group Violence in Multiethnic Societies, Basingstoke, UK: Palgrave Macmillan.

63.

Tilly

(1998) Durable Inequality, Berkeley, CA: University of California Press.

64.

Tong

Piotrowski

(2012) Migration and health selectivity in the context of internal migration in China, 1997–2009. Population Research and Policy Review 31(4): 497–543.

65.

Wade

Cairney

Pevalin

(2002) Emergence of gender differences in depression during adolescence: National panel results from three countries. Journal of the American Academy of Child and Adolescent Psychiatry 41(2): 190–198.

66.

Walder

(1992) Property rights and stratification in socialist redistributive economies. American Sociological Review 57(4): 524–539.

67.

Wheaton

(2001) The role of sociology in the study of mental health… and the role of mental health in the study of sociology. Journal of Health and Social Behavior 42(3): 221–234.

68.

Whyte

(2010) One Country, Two Societies: Rural–Urban Inequality in Contemporary China, Cambridge, MA: Harvard University Press.

69.

Williams

Collins

(1995) US socioeconomic and racial differences in health: Patterns and explanations. Annual Review of Sociology 21: 349–386.

70.

World Health Organization (2001) Mental Health: New Understanding, New Hope, Geneva, Switzerland: World Health Organization.

71.

Treiman

(2004) The household registration system and social stratification in China: 1955–1996. Demography 41(2): 363–384.

72.

Treiman

(2007) Inequality and equality under Chinese socialism: The hukou system and intergenerational occupational mobility. American Journal of Sociology 113(2): 415–445.

73.

Xie

Thornton

Wang

et al. (2012) Societal projection: Beliefs concerning the relationship between development and inequality in China. Social Science Research 41(5): 1069–1084.

74.

Xie

Zhou

(2014) Income inequality in today’s China. Proceedings of the National Academy of Sciences 111(19): 6928–6933.

75.

Xie

(2015) The causal effects of rural-to-urban migration on children’s well-being in China. European Sociological Review 31(4): 502–519.

76.

Yip

Subramanian

Mitchell

et al. (2007) Does social capital enhance health and well-being? Evidence from rural China. Social Science and Medicine 64(1): 35–49.

77.

Zhang

Treiman

(2013) Social origins, hukou conversion, and the wellbeing of urban residents in contemporary China. Social Science Research 42(1): 71–89.

78.

Zhou

(2000) Economic transformation and income inequality in urban China: Evidence from panel data. American Journal of Sociology 105(4): 1135–1174.

Live like mosquitoes: Hukou,rural–urban disparity,and depression

Abstract

Keywords

Introduction

Does hukou matter for mental health? Stratification and mental health in China

Hukou and mental health: Exploring possible mechanisms

Data and methods

Data

Measures

Methods

Results

Conclusion

Footnotes

Acknowledgment

Funding

Note

References