Social Capital and Mental Health in Rural and Urban China: A Composite Hypothesis Approach

We used the 2015 sociological survey dataset provided by China’s National Survey Research Center (the dataset can be found at http://cnsda.ruc.edu.cn/↗). Data were collected using a stratified clustered sampling design and were representative for the non-institutionalized adult (i.e., 18 and above) population in China. The primary survey unit (PSU) of this survey is at the county level or district level. Using China’s latest census as the sampling frame, the survey project included 969 of the most crowded PSUs across China. One hundred and thirty PSUs were randomly chosen and each chosen PSU was assigned a quota of households based on the population of that neighborhood as a proportion of the total population of all 969 PSUs. The researchers then selected households randomly using the residence records in the sampling frame. One adult person was randomly selected from each sampled household to serve as a respondent. Survey administrators visited each household after 18:00 h on weekdays or after 14:00 h during weekends and holidays to maximize the participation rate. The dataset consists of 10,968 face-to-face interviews (6470 respondents from rural and 4498 from urban areas, respectively). The sample includes 5834 female and 5134 male individuals from 130 PSUs (hereafter, “communities”) in 28 provinces. Most of the respondents were aged 41–60 (40%). Our samples were nested in two levels: the individual level and the community (i.e., county or district) level. The detailed distributions of sample are summarized in Table 1.

Self-rated mental health status was based on five statements that are similar to the WHO-Five Well-Being Index (WHO-5): “I often feel calm and peaceful; I often feel active and vigorous; I often feel depressive and unhappy; I often feel exhausted and tired; and I often feel that I cannot stand my life”. For the first two statements, the responses were given a value from (1) “not at all” to (5) “very true”. As for the last three statements, the responses were given a value from (1) “very true” to (5) “not at all”. The Cronbach’s alpha was 0.86. The eigenvalue of one principal component of this scale is 3.29, which represents the significant amount of variance in each item that can be explained by the principal component.

The items were evaluated by a confirmatory factor analysis (CFA) model. A weighted average mental health score was generated, which was resulting from the multiplication of each statement by its corresponding CFA standardized regression weight. The mental health score was standardized to have a mean of zero and a standard deviation of one. That is, if an individual is in good mental health status, then his or her mental health score is positive. Otherwise, it is close the zero or negative.

Social capital is regarded as “connections among individuals in social networks and norms of reciprocity and trustworthiness that arise from them” [37] (p. 19). The literature has identified four dimensions of social capital:

Structural social capital refers to an individual’s social network and various forms of civic engagement while cognitive social capital means an individual’s subjective perception of level of trust and reciprocity, which can be regarded as the result of structural social capital [38,39,40,41,42]. Horizontal social capital (or bonding/bridging social capital) refers to the relations developed between individuals or groups at the same social-economic hierarchical level, while vertical social capital (or linking social capital) includes the vertical relations between individuals or groups at different levels of formal power or authority [39,43,44,45]. Finally, social capital is not just an individual’s resource [46,47,48] but also a contextual capital [39]. Community or neighborhood social capital is usually based on day-to-day interaction between individuals [49]. According to [50,51] and [52,53], an individual in a particular area is also exposed to the community-level context, which may have an effect on an individual’s health. Therefore, we included contextual social capital variables, which can be constructed by grouping individuals within the same county/district and by aggregating an individual’s answers.

In this study, we have adopted the four-dimension framework of [39] to measure both individual and community-level social capital in the Chinese context.

The first dimension is civic participation (CP), a type of structural and horizontal social capital, here assessed by the frequency of an individual’s participation in nine different civic activities over the past 12 months (i.e., festivals, kinship networking, sports, entertainments, time with fellow workers, skill development, philanthropy, educational and religious activities). The responses were given a value from (1) “not at all” to (5) “very frequently”. The Cronbach’s alpha was 0.74. The eigenvalue of one principal component of this scale is 3.37, indicating a substantially large amount of variance in each item that can be explained by the principal component. The items were standardized to have a mean of zero and a standard deviation of one, and were then evaluated by a CFA model. A weighted average CP score was generated, which was resulting from the multiplication of each statement by its corresponding CFA standardized regression weight.

The second dimension is civic trust (CT), a type of cognitive and horizontal social capital, here measured by the degree of a respondent’s trust on people, based on 13 different types of relations (i.e., close neighbors, distant neighbors, people living in the same town/village with the same family name, people living in the same town/village with different family name, relatives, acquaintances, townsmen, members of gymnasium club, members of non-political associations, members of volunteer organizations, classmates, colleagues, and total strangers). For each statement, the responses were given a value from (1) “not true at all” to (5) “very true”. The Cronbach’s alpha was 0.81. The eigenvalue of one principal component of this scale is 4.03. The items were standardized to have a mean of zero and a standard deviation of one, and then evaluated by a confirmatory factor analysis (CFA) model.

The third level is political participation (PP), a type of structural and vertical social capital, here assessed using four binary items—whether the respondent was a member of the Chinese Communist Party (CCP) or any other political party, or a government official or military officer, or a member of People’s Congress (PC) or of People’s Political Consultative Conference (PPCC), or had participated in any political protests or petitions. The sum of these four items was used as the PP score, which was then standardized to have a mean of zero and a standard deviation of one.

Fourthly, political trust (PT), a type of cognitive and vertical social capital, referred to the extent to which respondents trusted in China’s political institutions (i.e., CCP, PC, PPCC, court system, People’s Procuratorate, military, police, and local as well as central governments). PT was measured by nine statements in terms of a respondent’s satisfaction of political institution’s service performance on transparency and accountability, social inequality, healthcare disparity, social services for the elderly/poor households, education service, judiciary’s efficiency and fairness, crime crackdown, defense and policing, and environment protection. For each statement, the responses were given a value from (1) “not satisfied at all” to (5) “very satisfied”. The Cronbach’s alpha was 0.80. The eigenvalue of one principal component of this scale is 4.71. The items were standardized to have a mean of zero and a standard deviation of one, and then evaluated by a confirmatory factor analysis (CFA) model.

Confirmatory factor analysis was performed on the constructs of CP, CT, PT, and mental health using open source R statistics package (see www.r-project.org↗) to evaluate the factor structures. The results indicated a good fit of each construct yielding a root mean square error of approximation (RMSEA) smaller than 0.05 and goodness-of-fit index (GFI) as well as an adjusted GFI superior to 0.90. All items were significantly loaded on their latent variables. The internal consistency of the scales (i.e., Cronbach’s alpha) was greater than 0.70, indicating high reliability of our measures.

Thereafter, we aggregated these four social capital subcomponents into one general social capital index (SCI). Finally, five contextual, district-level social capital variables (CP-Dist, CT-Dist, PP-Dist, PP-Dist, PT-Dist, and SCI-Dist) at the community level were measured on the basis of aggregated individual answers for each county/district.

Following [54] proposition, we have included a series of socioeconomic-demographic variables to control the confounding effects.

Gender was represented by a dummy variable for female. Age was categorized into four categories (1 = 18–25; 2 = 26–40; 3 = 41–60, and 4 = 61 and above). Education was measured as the highest grade completed (0 = no formal education; 1 = primary school; 2 = junior high school; 3 = senior high school; 4 = college; 5 = postgraduate). Marriage was categorized into three groups (1 = never married; 2 = married; 3 = divorced or widows/widowers). Unemployment was a dummy variable for whether one had been unemployed for the last three months. Self-reported social-economic status was assessed by a five-point likelihood scale ranging from 1 (very low) to 5 (very high). The respondents were asked to compare their social-economic status with peers within the same community.

Household income was measured based on the household annual expenditure (see [21]). Wealth index was constructed from a factor analysis of 15 terms related to household dwelling characteristics and ownership of consumer durables (such as, cars, TV sets, telephones, and so on. See [55] for more details. Finally, household size referred to the total number of family members that were living together.

Previous study dichotomized output variable and used binominal logistic regression to estimate the effect of social capital. The benefit of the approach was to facilitate the interpretation of regression coefficient as the likelihood to report better or worse health condition. However, [39,56,57] argued that such a dichotomization of a continuous variable was often arbitrary, which could fail to provide complete information about the actual distribution of the dependent variable or even cause misleading results. Therefore, in this study, we have kept the outcome variable, i.e., self-rated mental health as continuous and used a series of two-level linear regressions to examine the effect of individual and community social capital on an individual’s mental health. Following the approaches of [58,59], and [21], we have specified the following basic model:

where MH is the mental health status for individual i (level 1) in community j (level 2); SC is the set of social capital variables measured at both the individual and community levels; and X is a vector of socioeconomic-demographic variables. This model estimated the fixed β parameters, which represented the overall relationships between individual social capital variables, covariates and mental health. There were also two random parameters μ0j (community level) and εij (individual level), which were assumed to follow a normal distribution and represent the differences from the corresponding means at both individual and community levels. The variance between-community and the variance between-individual within a given community were used to calculate the variance partition coefficient (VPC), which represented the deviation in mental health across individual and community levels [60]. That is to say, the proportion of the total residual variation that is due to differences between communities. As the community-level social capital variables were aggregated from those at the individual level, the value of individual social capital was group-centered in order to alleviate the multicollinearity problem (i.e., SCij−SCj) [21,61,62].

Our research model is illustrated in Figure 1.

We have followed the nested modeling strategy of previous multilevel health studies (e.g., [16,21,42,56,63]) to account for the hierarchical structure of our dataset. We first estimated two variance components models (Model 0-Rural and 0-Urban) that include only random intercepts. Then we estimated Model 1-Rural and 1-Urban to include random intercepts and human capital variables. Finally, we estimated Models 2a–2e that include random intercepts, human capital variables, the five individual-level social capital variables (i.e., the aggregated SCI, CP, CT, PP, and PT) and the five community-level social capital variables (i.e., the aggregated SCI-Dist, CP-Dist, CT-Dist, PP-Dist, and PT-Dist), respectively.

The study revealed that higher civic trust helped to improve the mental health at the individual (rural areas) and community level (both rural and urban areas). This finding further corroborated prior findings between positive health outcomes and trust [15,63,64,65,66,67,68], which suggested that higher civic trust may enhance social networking among individuals, thereby reducing the mental stress and then lead to better mental health status.

Our study revealed that higher political trust helped to improve mental health at the individual and community level in rural areas. This finding was consistent with a prior study [69], which indicated that higher political trust could, one the one hand, led to an enhanced feeling of security [69] and created stronger affiliations with the legal system as identified through questions on government functioning; and on the other hand, higher political trust also promoted a better sense of belongingness and sense of responsibility [69].

We found that higher political participation also tended to improve mental health at the individual level in both rural and urban areas. That is probably that political participation may promote a sense of gratification from the social welfare activities and aroused deeper feelings of social and moral responsibility among individuals. It may be also instrumental in warding off socially deviant behavior among individuals, inculcating socially beneficial norms and promoting mental health.

Finally, we revealed that higher civic participation helped to improve the mental health at the individual level in urban areas. That is probably that higher civic participation could be manifested through progressive coaching mechanism or knowledge dissemination process to fellow individuals, thereby leading to enhanced self-esteem among individuals. This could help in promoting social interactions, reducing stress, improving security networks, and thereby improving mental health. However, at the community level our study suggested that enhanced civic participation gave rise to reduced mental health in urban areas, ceteris paribus. An explanation to this could be drawn from the “network resources” approach [70]. As predicted in [71], subjective social status could be considered as a partly mediating factor in explaining the negative association between network resources and psychological stress. We found a positive correlation (0.36, p < 0.01, in Table 4) between subjective social-economic status and civic participation in our urban sample, suggesting that a higher level of community civic participation could create extra responsibility or overwhelming burden on an individual with relatively high social-economic status. This is especially true in China in where loss of “face” has been considered as a key deterrent in an individual’s access to social capital [20,72]. Individuals will stretch beyond their economic horizon to meet moral obligations and reciprocity norms in order to save “face” in their communities. Non-adherence to these norms would result in a disastrous solitary state leading to a lower mental health. However, as found in a prior study [20], over participation in social events could involve detrimental social norms like alcohol consumption, physical stress and inevitably reduced social family interactions, leading to increased mental stress and worsening mental health.

Overall our findings were in line with those previously reported in developing countries that human capital factors were important predictors of self-rated mental health. The compositional effects remained statistically significant even after taking social capital variables into account.

In particular, we found that an increase in age led to a lower self-reported mental health, which is of paramount importance when people reach their middle-age. That is probably because that responsibilities and senses of isolation or worthlessness may be bound to increase with age. We believe that this is especially true for the elderly in rural areas, when their adult children are away as migrant workers.

Female gender was also found to be related to poorer self-rated mental health in rural areas than in urban areas. That is probably because, on one hand, rural women may suffer from lower social status in their villages and on the other hand, it may be accepted that when husbands are away as migrant workers, the responsibility to take care of younger family members and elderly parents always falls on the women without anyone to share the load.

Secondary education was also found to improve self-rated mental health in both rural and urban areas. This is probably because education may enable individuals to better understand the benefits of a healthy lifestyle and healthcare utilization [56]. We noticed that when both individual and community-level social capital variables were present, college education exhibited a marginally significant, negative effect on self-rated mental health in rural areas. We believed that when compared to their urban counterparts, the college students from rural areas may have lesser chances to find decent jobs in China because their families possessed fewer social networks. This probably led to low economic returns to their college tuition fees and made them anxious and frustrated.

Our study also found that wealth mattered for mental health. Being rich or having relatively high social-economic status was strongly associated with superior self-rated mental health. We argued that this was probably because richer households or individuals may be able to access more medical care resources and experience less stress when compared to the poor, who had to work harder to earn their living.

An increase of household size in rural area was associated with a decrease in self-rated mental health status. Given family size in rural China is usually larger (mean = 4.47, in Table 4), this association may have had its origins in increased competition between siblings for parental attention, resources and extra burden to take care of elderly parents or even grandparents. However, an opposite effect was found in the urban sample. A possible interpretation of the positive family size–mental health relationship in urban China may be due to the strict one-child policy. With the family size being small in urban areas (mean = 3.47, in Table 4), children may enjoy greater parental attention in their early life and as a result, developed greater resilience to maladaptive responses and stressful events in adulthood. Likewise, a larger family size may increase the social contact of urban elderly, reducing their sense of loneliness, a general cause of many mental illnesses.

In addition, our study indicated that, the married individuals portrayed better mental health when compared to unmarried individuals in rural areas. This finding was not supported in urban areas. We found that the mental health of the urban individual was found to deteriorate following a divorce or loss of spouse.

We believe that there is a stark distinction in the levels of aspirations between the individuals hailing from rural and urban areas. In rural areas, the concept of close families, strong bonding, and well-knit relations may exist that foster easy solving of mutual discord and problems. However, in urban areas, individuals live with more autonomy in smaller nuclear families, which could be beneficial to the extent of lesser burdening an individual with monetary and social responsibilities. At the same time, smaller nuclear families could be detrimental since the individuals may face increased mental stress when handling problems since they lack the emotional support of spouse, leading to poor mental health and estranged family ties. Therefore, with the resulting non-successful matrimony or death of beloved partner, an individual in urban area may develop a sense of personal loneliness and dissatisfaction, lack of emotional stability, and a dearth of social interaction develops, resulting in reduced mental health.

Finally, the employment status had no significant impact in rural areas. However, the effect of unemployment tended to be significantly negative in urban areas. We believe that the societal set up in rural and urban areas is starkly different. In rural areas, there are more closely-knit families characterized by large joint families. These families live together and most probably the family earnings are shared for the entire family’s sustenance. This involves collective earning and collective spending to supplement the entire family’s needs. This could be in contrast to the urban areas, where the concept of nuclear families with fewer members predominates. Financially independent individuals most probably prefer to be socially independent as well. Hence for individuals in urban areas, professional upheavals coupled with lack of emotional support may lead to mental dissatisfaction and reduced mental health.

Like most cross-sectional analyses, this study may be subject to common method bias as information regarding both social capital and mental-health is self-reported. Since both social capital and mental health can be expressed as an individual’s general well-being, there may be a recall bias [16,39]. In addition, the cross-sectional design usually leads to reversed causality problem—poor mental health may reduce one’s social capital. For example, participation in political activities requires inherent enthusiasm and good mental situation as a pre-requisite. A longitudinal study is desired to check the robustness of our findings. However, the aggregated measures of social capital and the use of mean-centered individual social capital can mitigate this problem as individual co-variance can be diluted [39]. On the other hand, the problem of reverse causality at the community level may not be an issue. Since the Chinese government restricts people’s freedom to choose which district to live in under the Household Registration System, the community-level measures of social capital may be exogenous and random.

Our results suggest that social capital is reflective of mental health both at the individual and community level. We have employed four cognitive-structural and horizontal-vertical constructs to comprehensively measure social capital in Chinese contexts and empirically test for observations across samples from rural and urban Chinese population segments. The results support the argument that in general, individual social capital holds a positive association with self-rated mental health in China. This study also finds that a significant proportion of the total variation in individual’s self-rated mental health status can be explained by differences at the community level. Those living in areas with low civic trust social capital have high excess risk of poor self-rated mental health. On the contrary, urban areas offer disagreement on the structural dimension measured through community-civic participation index, offering a negative association. An explanation for this has been based on the network resources approach. The likelihood ratio tests further support our conclusions that the impact of both individual and contextual social capital is not an artifact. These results offer further dimensions for future research. A more detailed and vigorous investigation may be called upon to study the impact of social capital variation, involving development of more suitable instruments to unearth the pattern of variations more systematically.

Finally, we suggest that policy mechanisms should be set in place to strengthen social networks and create economic upliftment of individuals. The programs should be aimed at inculcating social parity among the community members and at reducing the disparity in the access of medical benefits. Community health and social wellbeing for individuals should to be given attention with special emphasis to women, middle-aged, and elderly citizens. In rural areas, schemes should be set in place to encourage employment and business opportunities for youth through promotion of cottage industries. These schemes will be able to create employment avenues for the large number of college-educated students hailing from rural areas and hence avoid their rapid transition to larger cities. It will also be helpful in curtailing dissatisfaction among their elderly family members who are plagued with issues of staying alone as their children and even grand-children migrate to urban areas for better job prospects.