What this is
- This research examines the role of in mental health inequality in China.
- It analyzes data from the China Family Panel Study (CFPS) to explore how family and community impact and .
- The findings suggest that can buffer mental health inequalities, particularly for poorer populations.
Essence
- buffers mental health inequality in China, with family-level reducing and village-level enhancing , especially for the poor.
Key takeaways
- Higher levels of correlate with lower and higher . This relationship is more pronounced among poorer individuals.
- Family-level effectively reduces depressive symptom inequality, while village-level enhances inequality. This indicates a pro-poor effect.
- Despite having less , poorer individuals can leverage it more effectively to improve their mental health compared to wealthier individuals.
Caveats
- The study's cross-sectional design limits causal interpretations of the relationship between and mental health.
- Self-reported measures of happiness and well-being may introduce bias, as individual definitions of happiness can vary significantly.
Definitions
- Social Capital: Resources linked to social networks that can improve individual health outcomes and facilitate coordinated actions.
- Subjective Well-Being (SWB): An individual's self-reported assessment of their happiness and life satisfaction.
- Depressive Symptoms: Indicators of depression measured by the CES-D scale, reflecting the frequency of negative feelings and behaviors.
Simplified
Introduction
Income-related health inequality has been one of the hottest topics in research fields of health economics and public health, and has arose the intense interest of health policymakers. Health inequality exists all over the world, especially in low- and middle-income countries such as China [1–3]. It is widely acknowledged that better health is associated with a higher level of income, educational status, socioeconomic status (SES), and more advantaged ethnic groups. A study using nationally representative data found that although average per capita income had increased a lot in the past decades in China, income-related health inequality had also doubled [4]. Another study focusing on health outcomes and SES suggested that higher SES could positively affect health indicators such as hypertension situation and life-functioning among the Chinese elderly [5]. Not only physical health but also mental health is severely unequal in China. Previous research using data from the China Health and Retirement Longitudinal Study found that for mid-aged and elderly, the individual depressive symptom was significantly related to personal educational status and expenditure [6]. Inequality caused by the hukou system also arises mental health inequality. Scholars found that urban residents possessed a lower level of depressive symptoms than rural residents and temporary migrants [7]. It is apparent that health inequality, including physical and mental health inequality, is an important issue in current China.
Since social capital (SC) was brought forward by Bourdieu [8], it has gained greatly developed by some other scholars such as Coleman [9] and Putnam [10]. The original definition of social capital is “aggregate of the actual or potential resources which are linked to possession of a durable network of more or less institutionalized”, given by Bourdieu [8]. Putnam defined SC as “features of social organization, such as trust, norms, and networks that can improve the efficiency of society by facilitating coordinated actions” [10], which was widely accepted by other scholars. Different approaches to measuring social capital are adopted in empirical studies. A common distinction is structural and cognitive dimensions. Structural social capital refers to the individual’s participation in formal associations such as religious organizations and labor unions, neighborhood activities, and social resources from these associations and activities. The cognitive dimension describes an individual’s attitude, such as trust and social cohesion of the neighborhood. Yip and colleagues measured structural social capital by how many organizations residents participated in, cognitive social capital by scores related to trust, reciprocity, and mutual help [11]. Another distinction often used is bonding, bridging, and linking components, while the bonding dimension is defined as the relationship with relatives or friends, the bridging dimension is measured by the relationship between more loosely connected people such as neighbors or colleagues, the linking dimension refers to relationship connected by authority or power such as employer and employee [12]. In health research fields, social capital has long been regarded as a useful intervention to promote an individual’s health status. For example, using nationally representative data and measuring social capital by social participation, Liu and colleagues found that SC had a significantly positive effect on the elderly’s physical health and well-being [13]. In Taiwan, scholars found that social capital could improve health limitation status [14]. Understanding the positive role social capital plays in health can help policymakers develop more useful policies to promote health.
So far, what role social capital plays in mental health inequality is still ambiguous. As an important component of individual health conditions, the relationship between mental health inequality and social capital has not been paid enough attention before, especially in China. Mental health, such as depression is gradually more severe in recent years. A national study reported that the prevalence of depressive symptoms was 37.9% and depression 4.1% for Chinese adults [22]. Understanding how social capital affects mental health and how mental health inequality will develop related policies to promote mental health and improve mental health inequality. Therefore, the aim of this study is to explore the relationship between social capital and mental health inequality in China. First, we will analyze the relationship between social capital and mental health. Second, we will conduct an analysis to explore whether social capital has a buffer effect or dependency effect with income level on health, namely, whether social capital will exacerbate or reduce income-related health inequality. Finally, we will analyze the relationship between social capital and mental health inequality in different areas separately, for example, urban and rural areas. In this way, we can better understand how social capital acts in mental health inequality in different institutions and contexts.
Buffer effect of social capital on health inequality
Methods
Data source
Data is extracted from the China Family Panel Studies (CFPS). CFPS is a nationally representative, longitudinal survey conducted every 2 years since 2010 by Peking University. All individual-, family- and community-level data is collected in CFPS. Information in economic activity, education condition, family relationship, population migration, and physical and mental health is captured in CFPS [23]. The baseline survey was conducted in 2010 and covered 14960 households, 33600 adults, and 8990 children.
Data in 2018 and 2016 of CFPS is used in this current study. 2016 database mainly provides us residents’ social capital information. The lagged variable is widely used to avoid endogeneity [24, 25]. After eliminating those who are under 18, our total sample size contains 25322 adults with 6591 urban respondents and 18731 rural respondents.
Variables
Dependent variables
Depressive symptoms
The first indicator of mental health is depressive symptoms. Depression is one of the most common mental health disorders in China [26]. Center for Epidemiologic Studies Depression Scale (CES-D) was deployed to measure depressive symptoms in CFPS. CES-D20 was found to be difficult for respondents to finish; thus, in 2018, CES-D20 was replaced by CES-D8. Therefore, in this current study, we used the CES-D8 score to indicate the respondent’s depressive symptoms. In CES-D8, respondents are asked to answer how often in the past week they felt for 8 items, including feeling depressed, feeling everything he did was an effort, sleep restless or not, happy, feeling lonely, enjoying life, feeling sad, and unable to get going. The choices are ranging from ‘none or almost none of the time’ to ‘all or almost or all of the time’, with values 0 to 3. Total scale scores are from 0 to 24, with higher scores representing a higher frequency of depressive symptoms. The validity of CES-D8 has been documented in previous studies [27, 28].
Subjective well-being
Another important indicator is well-being. In psychology, the concepts of happiness, well-being, and mental health are often used as synonyms [29]. The World Health Organization defined mental health as a state of well-being in which individuals were able to reach their potential, cope with stresses of life, able to work productively, and make contributions to society [30]. Good mental health indicates an absence of negative symptoms and positive psychological functioning [31]. Hence, it is more and more acknowledged that when measuring mental health, both mental illness (e.g. depression) and well-being should be considered. Mental health is indispensable to subjective well-being, and subjective well-being is the positive side of mental health [31]. Therefore, in this current study, subjective well-being (SWB) is used as another indicator of mental health. Subjective well-being is based on respondents’ answers, instead of pre-judgment of researchers. In CFPS, the respondent was asked “how happy do you think you are”, with values from 0 to 10.
Independent variables
Social capital
In this study, we measured both family-level and community-/village-level social capital. To avoid the endogeneity of social capital and mental health, we used lagged terms in 2016 of these two SCs. In the specific Chinese context, family SC is more related to social networks and guanxi [32, 33]. Guanxi means relation in Chinese. Unlike western society where in-kind gift is common than money gift, in China, money is also widely accepted as a gift, especially on some important occasions such as weddings and New Year Celebration [34]. Giving and receiving money gift has been one of the cores of social relationships in Chinese societies since ancient times [35]. Giving gifts including money and in-kind gifts provides access not only to show filial piety, but also to maintain social networks and acquire social capital [36]. Therefore, we set family gifts as a measurement of family-level social capital. In CFPS, the value of money and in-kind gifts in the past 12 months of all family members given is recorded. The total value of gifts is our indicator of family-level social capital.
As for community-/village-level social capital, neighbor relationship, neighbor assistance, and neighbor feeling are measured. In CFPS 2016, the neighbor relationship was recorded by the question “in general, how do you feel about your neighbor relationship” with five answers from “very good” to “very bad” and we coded them 5, 4, 3, 2, 1, respectively. The neighbor assistance was asked by “if you need help, do you think anyone in your neighborhood will offer assistance”; the answer to this question varied from “there must be” to “there must not be”, with values from 5 to 1. The neighbor sentiment was identified by the question “how much is your feeling to your neighborhood”; the answers were from “very much” to “not at all”, with scale scores from 5 to 1. The Cronbach’s alpha value of these three questions is 0.63, which is regarded as satisfactory and acceptable [37, 38]. Factor analysis of these 3 items is adopted in our study and the result shows a one-factor solution, with eigenvalue equals 1.72 and 57.34% of original information can be explained by factor.
Control variables
In our analysis, some variables related to mental health are also included. Income level is indicated by annual household income per capita. This measurement is widely used to analyze income-related health inequality [39]. Scholars recommended that a Body mass index (BMI) of 18.5 to 23.9 should be considered as a standard weight for Chinese people, and 24 to 27.9 as overweight, and >=28 as obese [40, 41]. According to this criterion, we categorize BMI into 4 groups- underweight, normal, overweight, and obese. Education level is divided into 5 groups according to respondent’s highest education qualification, including illiterate or semi-illiterate, junior school and below, high school and technical secondary school, junior college, bachelor and above. Marital status is grouped into unmarried, married, divorced, and widow. Besides, variables including gender, age, health insurance, self-rated health status, currently working or not, rural or urban area, region are all included as control variables.
Analysis
A series of OLS regression models were conducted to estimate the effects of social capital on mental health and mental health inequality. Firstly, the effects of two kinds of social capital on mental health were examined in total sample, rural sample and urban sample separately. Next, the product term of social capital variables and income level were analyzed to estimate how the effect of SC on mental health varies across different income levels. If the positive effect for the poor is higher than the rich, the mental health disparity can be reduced by social capital (as shown in Fig. 1); otherwise, the disparity will be exacerbated.
In addition, we calculated the marginal effect of social capital on mental health based on the OLS regression results, to analyze whether the effect changes due to the change in income level. The marginal effect of social capital (s) on mental health (y) is calculated by the following equation:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{\partial y}{\partial s}={\beta}_1s+{\beta}_1 Income$$\end{document}∂y∂s=β1s+β1Income
Where β1 represents the coefficient of social capital in the OLS model, and β3 is the coefficient of the interaction term of social capital and income level. In each regression model, we calculated the variance inflation factor (VIF) to analyze the collinearity. The results are shown the Table A1 in the appendix. The VIFs are always less than 10, indicating that there is no collinearity in our models.
Results
Descriptive Results
| Variable | Definition | Total sample | Urban sample | Rural sample | |||
|---|---|---|---|---|---|---|---|
| Observation | Mean (S.D.) | Observation | Mean (S.D.) | Observation | Mean (S.D.) | ||
| Depressive symptom | CES-D8 score ranges from 0 to 24 | 24,996 | 5.589 (4.031) | 6538 | 4.891 (3.764) | 18,458 | 5.837 (4.092) |
| Subjective well-being | Ranges from 0 to 10 | 25,145 | 7.490 (2.157) | 6561 | 7.702 (1.908) | 18,584 | 7.415 (2.234) |
| Family-level social capital | Total values of household money and in-kind gift in the past 12 months | 25,158 | 4496.511 (7734.711) | 6554 | 5450.018 (8489.503) | 18,604 | 4160.6 (7421.734) |
| Neighbor relationship | Ranges from 1 to 5 fromtovery badvery good | 24,371 | 2.837 (0.816) | 6447 | 2.795 (0.795) | 17,924 | 2.851 (0.823) |
| Neighbor assistance | Ranges from 1 to 5 fromtothere must not bethere must be | 24,384 | 3.477 (0.804) | 6447 | 3.407 (0.784) | 17,937 | 3.502 (0.810) |
| Neighbor feeling | Ranges from 1 to 5 fromtonot at allvery much | 24,388 | 2.946 (0.863) | 6450 | 2.794 (0.826) | 17,938 | 3.001 (0.870) |
| Income level | Annual household income per capita | 25,322 | 26,722.900 (62,339.060) | 6587 | 44,477.140 (74,310.980) | 18,731 | 20,475.600 (56,221.110) |
| BMI group | =0 if underweight;=1 if normal;*=2 if overweight;=3 if obese | 25,322 | 1.424 (0.779) | 6591 | 1.487 (0.769) | 18,731 | 1.402 (0.782) |
| Education level | =0 if illiterate or semi-illiterate;*=1 if junior school and below;=2 if high school and technical secondary school;=3 if junior college;=4 if bachelor and above | 25,322 | 1.211 (1.035) | 6591 | 1.834 (1.189) | 18,731 | 0.992 (0.876) |
| Health insurance | =0 if not join any health insurance;*=1 if join any health insurance | 24,932 | 0.922 (0.268) | 6503 | 0.904 (0.295) | 18,429 | 0.928 (0.258) |
| Self-rated health status | Ranges from 0 to 4 fromtovery badvery good | 25,080 | 1.931 (1.219) | 6555 | 1.906 (1.093) | 18,525 | 1.940 (1.261) |
| Working status | =0 if not working;*=1 if currently working | 24,437 | 0.705 (0.456) | 6366 | 0.581 (0.494) | 18,071 | 0.749 (0.434) |
| Marital status | =0 if unmarried;*=1 if married;=2 if divorce;=3 if widow | 25,322 | 1.023 (0.610) | 6591 | 1.034 (0.618) | 18,731 | 1.019 (0.607) |
| Rural | =0 if in urban area*=1 if in rural area | 25,322 | 0.740 (0.439) | ||||
| Gender | =0 if female;*=1 if male | 25,322 | 0.496 (0.500) | 6591 | 0.509 (0.500) | 18,731 | 0.492 (0.500) |
| Age | 25,322 | 48.260 (16.276) | 6591 | 49.785 (16.583) | 18,731 | 47.723 (16.132) | |
OLS Regression
Depression Symptoms
Marginal effect of family-level SC on depressive symptoms in the urban area
| Model 1 | Model 2 | Model 3 | |
|---|---|---|---|
| Family-level SC | −0.026**(0.013) | −0.191*(0.113) | −0.026**(0.013) |
| Community−/Village- level SC | − 0.497***(0.028) | − 0.497***(0.028) | − 0.753***(0.273) |
| Log of income level | − 0.370***(0.034) | − 0.491***(0.089) | − 0.369***(0.034) |
| Family-level SC × log of income level | 0.017 (0.011) | ||
| Community−/Village- level SC × log of income level | 0.027 (0.028) | ||
| BMI group | |||
| Underweight | 0.383***(0.108) | 0.381***(0.108) | 0.382***(0.108) |
| Overweight | −0.179***(0.055) | − 0.179***(0.055) | − 0.179***(0.055) |
| Obese | − 0.265***(0.089) | − 0.266***(0.089) | − 0.265***(0.089) |
| Education level | |||
| Junior school and below | −0.601***(0.075) | − 0.601***(0.075) | − 0.602***(0.075) |
| High school and technical secondary school | −0.947***(0.095) | − 0.950***(0.095) | − 0.947***(0.095) |
| Junior college | − 0.855***(0.122) | − 0.862***(0.122) | − 0.854***(0.122) |
| Bachelor and above | − 0.853***(0.132) | − 0.865***(0.132) | − 0.848***(0.132) |
| Health insurance | − 0.601***(0.104) | − 0.602***(0.104) | − 0.601***(0.104) |
| Self-rated health status | −1.013***(0.023) | −1.013***(0.023) | −1.013***(0.023) |
| Working | 0.125**(0.062) | 0.126**(0.062) | 0.125**(0.062) |
| Marital status | |||
| Married | −0.301***(0.103) | −0.301***(0.103) | − 0.305***(0.104) |
| Divorced | 1.151***(0.219) | 1.152***(0.218) | 1.147***(0.219) |
| Widowed | 1.47***(0.177) | 1.468***(0.177) | 1.466***(0.177) |
| Rural | 0.391***(0.067) | 0.391***(0.067) | 0.392***(0.067) |
| Male | −0.540***(0.048) | −0.540***(0.048) | − 0.539***(0.049) |
| Region | |||
| Middle | −0.432***(0.074) | −0.434***(0.074) | − 0.432***(0.074) |
| Eastern | −0.632***(0.069) | − 0.635***(0.069) | − 0.630***(0.069) |
| Age | − 0.021***(0.002) | − 0.022***(0.002) | − 0.021***(0.002) |
| R2 | 0.1821 | 0.1822 | 0.1821 |
| N | 22,969 | 22,969 | 22,969 |
| Urban area | Rural area | |||||
|---|---|---|---|---|---|---|
| Model 4 | Model 5 | Model 6 | Model 7 | Model 8 | Model 9 | |
| Family-level SC | −0.066***(0.022) | − 0.672***(0.255) | − 0.066***(0.022) | −0.008 (0.015) | − 0.170 (0.145) | −0.008 (0.015) |
| Community−/Village- level SC | −0.472***(0.051) | − 0.472***(0.051) | −1.216*(0.633) | − 0.510***(0.033) | − 0.510***(0.033) | − 0.671**(0.326) |
| Log of income level | − 0.400***(0.068) | − 0.812***(0.191) | − 0.392***(0.068) | − 0.332***(0.039) | − 0.457***(0.117) | − 0.332***(0.038) |
| Family-level SC × log of income level | 0.058**(0.024) | 0.017 (0.015) | ||||
| Community−/Village- level SC × log of income level | 0.072 (0.061) | 0.017 (0.034) | ||||
| BMI group | ||||||
| Underweight | 0.245 (0.218) | 0.226 (0.219) | 0.236 (0.219) | 0.412***(0.124) | 0.412***(0.124) | 0.412***(0.124) |
| Overweight | −0.217**(0.099) | −0.218**(0.099) | −0.216**(0.099) | − 0.171***(0.065) | −0.171***(0.065) | − 0.171***(0.065) |
| Obese | −0.544***(0.153) | − 0.549***(0.153) | − 0.541***(0.153) | − 0.175 (0.108) | −0.176 (0.108) | − 0.175 (0.108) |
| Education level | ||||||
| Junior school and below | −0.457**(0.202) | − 0.460**(0.203) | − 0.459**(0.202) | − 0.592***(0.082) | − 0.593***(0.082) | −0.593***(0.082) |
| High school and technical secondary school | −0.913***(0.212) | − 0.926***(0.212) | − 0.917***(0.212) | −0.867***(0.113) | − 0.870***(0.113) | −0.867***(0.113) |
| Junior college | − 0.646***(0.238) | 0.66***(0.239) | − 0.648***(0.238) | −1.044***(0.158) | − 1.050***(0.158) | − 1.044***(0.158) |
| Bachelor and above | − 0.634**(0.248) | − 0.663***(0.248) | − 0.628**(0.248) | − 1.386***(0.187) | − 1.396***(0.188) | − 1.382***(0.187) |
| Health insurance | − 0.721***(0.184) | − 0.705***(0.184) | − 0.719***(0.184) | − 0.521***(0.126) | − 0.521***(0.126) | − 0.521***(0.126) |
| Self-rated health status | − 1.040***(0.046) | − 1.037***(0.046) | − 1.040***(0.046) | −1.001***(0.026) | − 1.002***(0.026) | − 1.001***(0.026) |
| Working | 0.37***(0.114) | 0.367***(0.114) | 0.371***(0.114) | −0.049 (0.075) | − 0.047 (0.075) | − 0.049 (0.075) |
| Marital status | ||||||
| Married | −0.266 (0.210) | − 0.268 (0.209) | − 0.271 (0.210) | − 0.393***(0.120) | − 0.393***(0.120) | − 0.396***(0.120) |
| Divorced | 0.804**(0.343) | 0.808**(0.342) | 0.796**(0.343) | 1.308***(0.029) | 1.306***(0.289) | 1.306***(0.289) |
| Widowed | 1.598***(0.339) | 1.582***(0.338) | 1.594***(0.339) | 1.318***(0.208) | 1.317***(0.208) | 1.315***(0.208) |
| Male | −0.592***(0.090) | − 0.588***(0.090) | − 0.589***(0.090) | −0.519***(0.058) | − 0.519***(0.058) | − 0.519***(0.058) |
| Region | ||||||
| Middle | −0.289**(0.145) | −0.288**(0.145) | − 0.286**(0.145) | − 0.506***(0.085) | − 0.507***(0.085) | − 0.506***(0.085) |
| Eastern | −0.615***(0.138) | − 0.614***(0.138) | − 0.608***(0.138) | − 0.627***(0.079) | − 0.630***(0.079) | − 0.626***(0.079) |
| Age | −0.030***(0.005) | − 0.030***(0.005) | − 0.030***(0.005) | − 0.017***(0.003) | − 0.017***(0.003) | − 0.017***(0.003) |
| R2 | 0.1677 | 0.1689 | 0.168 | 0.1774 | 0.1775 | 0.1774 |
| N | 6099 | 6099 | 6099 | 16,870 | 16,870 | 16,870 |
Subjective well-being
Marginal effect of community-/village-level SC on subjective well-being in the rural area
| Model 1 | Model 2 | Model 3 | |
|---|---|---|---|
| Family-level SC | 0.02***(0.007) | 0.072 (0.063) | 0.02***(0.007) |
| Community−/Village- level SC | 0.339***(0.016) | 0.339***(0.016) | 0.608***(0.157) |
| Log of income level | 0.121***(0.019) | 0.159***(0.049) | 0.12***(0.019) |
| Family-level SC × log of income level | −0.005 (0.006) | ||
| Community−/Village- level SC × log of income level | −0.028*(0.016) | ||
| BMI group | |||
| Underweight | 0.095 (0.058) | 0.096*(0.058) | 0.096*(0.058) |
| Overweight | 0.053*(0.031) | 0.054*(0.031) | 0.054*(0.031) |
| Obese | 0.224***(0.048) | 0.225***(0.048) | 0.225***(0.048) |
| Education level | |||
| Junior school and below | 0.024 (0.042) | 0.024 (0.042) | 0.024 (0.042) |
| High school and technical secondary school | 0.044 (0.054) | 0.045 (0.054) | 0.045 (0.054) |
| Junior college | 0.202***(0.067) | 0.204***(0.067) | 0.201***(0.067) |
| Bachelor and above | 0.249***(0.070) | 0.253***(0.071) | 0.244***(0.071) |
| Health insurance | 0.13**(0.060) | 0.131**(0.060) | 0.13**(0.060) |
| Self-rated health status | 0.356***(0.013) | 0.356***(0.013) | 0.356***(0.013) |
| Working | −0.186***(0.035) | −0.186***(0.035) | −0.185***(0.035) |
| Marital status | |||
| Married | 0.327***(0.060) | 0.327***(0.060) | 0.332***(0.060) |
| Divorced | −0.838***(0.125) | − 0.838***(0.125) | − 0.833***(0.125) |
| Widowed | −0.107 (0.097) | − 0.107 (0.097) | − 0.103 (0.097) |
| Rural | − 0.119***(0.037) | − 0.119***(0.037) | − 0.119***(0.037) |
| Male | − 0.080***(0.027) | − 0.080***(0.027) | − 0.081***(0.027) |
| Region | |||
| Middle | 0.306***(0.041) | 0.307***(0.041) | 0.306***(0.041) |
| Eastern | 0.326***(0.039) | 0.327***(0.039) | 0.324***(0.039) |
| Age | 0.011***(0.001) | 0.011***(0.001) | 0.011***(0.001) |
| R2 | 0.0963 | 0.0963 | 0.0965 |
| N | 23,035 | 23,035 | 23,035 |
| Urban area | Rural area | |||||
|---|---|---|---|---|---|---|
| Model 4 | Model 5 | Model 6 | Model 7 | Model 8 | Model 9 | |
| Family-level SC | 0.03***(0.011) | 0.171 (0.134) | 0.03***(0.012) | 0.015*(0.009) | 0.077 (0.080) | 0.015*(0.009) |
| Community−/Village- level SC | 0.32***(0.027) | 0.32***(0.027) | 0.232 (0.347) | 0.347***(0.019) | 0.347***(0.019) | 0.686***(0.189) |
| Log of income level | 0.147***(0.035) | 0.243**(0.100) | 0.148***(0.035) | 0.106***(0.022) | 0.154**(0.064) | 0.105***(0.022) |
| Family-level SC × log of income level | −0.014 (0.013) | −0.007 (0.008) | ||||
| Community−/Village- level SC × log of income level | 0.009 (0.033) | −0.036*(0.019) | ||||
| BMI group | ||||||
| Underweight | 0.196*(0.105) | 0.201*(0.105) | 0.195*(0.105) | 0.073 (0.068) | 0.073 (0.068) | 0.073 (0.068) |
| Overweight | 0.117**(0.051) | 0.117**(0.051) | 0.117**(0.051) | 0.027 (0.038) | 0.027 (0.038) | 0.028 (0.038) |
| Obese | 0.157*(0.082) | 0.158*(0.082) | 0.157*(0.082) | 0.251***(0.058) | 0.252***(0.058) | 0.252***(0.058) |
| Education level | ||||||
| Junior school and below | −0.037 (0.104) | −0.036 (0.104) | − 0.037 (0.104) | 0.027 (0.047) | 0.027 (0.047) | 0.028 (0.047) |
| High school and technical secondary school | −0.085 (0.109) | −0.082 (0.109) | − 0.086 (0.110) | 0.091 (0.065) | 0.092 (0.065) | 0.091 (0.065) |
| Junior college | 0.148 (0.121) | 0.151 (0.121) | 0.148 (0.121) | 0.196**(0.089) | 0.198**(0.089) | 0.195**(0.089) |
| Bachelor and above | 0.12 (0.125) | 0.126 (0.126) | 0.12 (0.125) | 0.423***(0.102) | 0.427***(0.102) | 0.413***(0.102) |
| Health insurance | 0.19*(0.099) | 0.186*(0.099) | 0.19*(0.099) | 0.097 (0.076) | 0.097 (0.077) | 0.098 (0.076) |
| Self-rated health status | 0.318***(0.024) | 0.318***(0.024) | 0.319***(0.024) | 0.365***(0.015) | 0.365***(0.015) | 0.365***(0.015) |
| Working | −0.324***(0.061) | −0.323***(0.061) | − 0.324***(0.061) | − 0.111**(0.043) | −0.112***(0.043) | − 0.111**(0.043) |
| Marital status | ||||||
| Married | 0.249**(0.101) | 0.249**(0.101) | 0.248**(0.101) | 0.376***(0.074) | 0.376***(0.074) | 0.383***(0.074) |
| Divorced | −0.742***(0.178) | −0.743***(0.178) | − 0.743***(0.178) | − 0.896***(0.171) | − 0.895***(0.171) | − 0.892***(0.170) |
| Widowed | − 0.237 (0.166) | − 0.234 (0.166) | − 0.238 (0.167) | −0.027 (0.119) | − 0.027 (0.119) | −0.021 (0.119) |
| Male | 0.02 (0.047) | 0.019 (0.047) | 0.02 (0.047) | −0.118***(0.034) | −0.118***(0.034) | − 0.119***(0.034) |
| Region | ||||||
| Middle | 0.256***(0.078) | 0.256***(0.078) | 0.256***(0.078) | 0.323***(0.047) | 0.323***(0.047) | 0.322***(0.047) |
| Eastern | 0.243***(0.074) | 0.243***(0.074) | 0.244***(0.074) | 0.346***(0.045) | 0.347***(0.045) | 0.344***(0.045) |
| Age | 0.011***(0.002) | 0.012***(0.002) | 0.011***(0.002) | 0.01***(0.002) | 0.01***(0.002) | 0.01***(0.002) |
| R2 | 0.102 | 0.1022 | 0.102 | 0.0925 | 0.0925 | 0.0927 |
| N | 6113 | 6113 | 6113 | 16,922 | 16,922 | 16,922 |
Subsample analysis
From the results of model 1, model 3, model 5, model 7, model 9 and model 11 in Table 6, we can conclude that family-level SC doesn’t have any observed positive impact on family heads’ mental health, neither measured by depressive symptoms, nor by SWB. The coefficient of the interaction term in model 4 in Table 6 indicates that the family-level SC can buffer mental health inequality, which shows the same result with model 5 in Table 3.
| Depressive symptoms | Subjective well-being | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Total | Urban area | Rural area | Total | Urban area | Rural area | |||||||
| Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | Model 6 | Model 7 | Model 8 | Model 9 | Model 10 | Model 11 | Model 12 | |
| Family-level SC | −0.025 (0.016) | −0.199 (0.128) | − 0.042 (0.027) | − 0.580**(0.269) | − 0.018 (0.020) | − 0.104 (0.164) | 0.013 (0.009) | 0.089 (0.070) | 0.013 (0.014) | 0.181 (0.137) | 0.012 (0.011) | 0.072 (0.092) |
| Community−/Village- level SC | −0.513***(0.037) | −0.513***(0.037) | − 0.447***(0.068) | −0.448***(0.068) | − 0.295***(0.044) | −0.546***(0.044) | 0.345***(0.020) | 0.345***(0.020) | 0.329***(0.035) | 0.329***(0.035) | 0.355***(0.025) | 0.355***(0.025) |
| Log of income level | −0.380***(0.042) | −0.504***(0.100) | − 0.525***(0.088) | −0.877***(0.196) | − 0.547***(0.049) | − 0.361***(0.133) | 0.16***(0.023) | 0.215***(0.055) | 0.201***(0.045) | 0.311***(0.010) | 0.137***(0.028) | 0.182**(0.075) |
| Family-level SC × log of income level | 0.018 (0.013) | 0.052**(0.068) | 0.009 (0.017) | − 0.008 (0.007) | − 0.016 (0.013) | − 0.006 (0.010) | ||||||
| Control variables | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| R2 | 0.1986 | 0.1986 | 0.1876 | 0.1884 | 0.1945 | 0.1945 | 0.1072 | 0.1073 | 0.1235 | 0.1236 | 0.0983 | 0.0982 |
| Observations | 11,012 | 11,012 | 3060 | 3060 | 7952 | 7952 | 11,050 | 11,050 | 3067 | 3067 | 7983 | 7983 |
Discussion
Does social capital worsen or narrow mental health inequality? Is the mental health gap between the poor and the rich enlarged or closed by social capital? In this study, we measured both family-level and community-/village-level social capitals and investigated their associations with mental health and income-related mental health inequality. The results of this current study contribute to relevant literature on how to promote mental health and decrease mental health inequality.
Firstly, our results validate the positive effect of social capital on mental health. Our results indicate that a higher level of community/village SC is related to fewer depressive symptoms in both urban and rural areas, and family SC also has a negative impact on depressive symptoms in the urban area. These findings are consistent with previous research on social capital and depression. An existing study focusing on the effect of social capital on depression among Chinese urban elderly suggests that trust, reciprocity, and social network are all significantly associated with lower geriatric depression [44]. A study conducted in other countries also provides similar results [45]. For SWB, both family-level social capital and community-/village- level social capital play a positive role no matter in the urban or rural areas. This finding is consistent with some previous research. Several studies have reported that social capital originated from social networks and neighborhood relationships could improve residents’ happiness [46, 47]. China is a typical relational society. Scholars stated that the behaviors of giving and receiving gifts were based on the principle of reciprocity [48]. A Chinese scholar furtherly pointed out that giving and receiving gifts in Chinese society could help to preserve the partnership among intimate associations [49]. Guanxi or relation plays an essential role in socio-economic development, and is regarded to be an important way to gain social capital. In this study, we set the total value of money and in-kind gifts in the past 12 months as family-level guanxi. This approach has been adopted by other scholars when measuring social capital in the Chinese context [50]. However, when measuring individual’s SWB, money/in-kind gift as social capital hasn’t been considered while it’s very important to mental health. Our results contribute to the existing literature on the measurement of social capital and improvement of SWB.
Unlike village-level SC, family-level SC, measured by household gift in this study, is found to be unrelated to depressive symptoms in the rural area. Other scholars also have found that community-/village-level SC has a much greater effect on depression This may be possibly due to the intimate correlation within village. In rural China, a village was usually originated from a clan or tribe. Villagers have tighter relations and thus, they rely less on the relationship within family compared with urbanites [51]. Another potential reason is that in rural China, the function of giving gifts is to keep relative standing and cope with competitive mating pressure rather than maintaining social network and social capital [52]. The subsample analysis among household financial heads indicates that family-level social capital doesn’t have an observed impact on mental health. This may be because the financial head is the person who is most familiar with the household’s income and expenditure. Although social capital could bring mental health benefits to residents, it also needs economic costs for maintaining social capital. Therefore, the one who manages household’s property needs to bear the costs of giving money or in-kind gifts, which reduces the benefits brought by social capital. Previous studies have also documented that household heads experienced more mental stress due to household food burden and financial expenditure [53, 54].
Secondly, in terms of social capital and mental health inequality, this current study provides interesting findings. Although whether social capital brings buffer effect or dependency effect on health inequality has been a commonplace topic and scholars found different answers for this question, our regression results and decomposition results prove that social capital has a buffer effect on mental health inequality. For the urban residents, family-level SC has a greater effect on inhibiting depressive symptoms for the poor, and for the rural residents, village-level SC has a greater effect on improving SWB for the poor. These findings confirm the idea that social capital is the capital of the poor. Some scholars have been arguing that social capital could improve health but would also exacerbate health inequalities, because not everyone would benefit in the same way. People with a low-income level usually have less social capital; moreover, they have limited social capital to make full use of to promote health. However, the buffer effect indicates that social capital has a greater effect on health for the disadvantaged, and no effect or limited health benefits for the advantaged. A previous study conducted in America concluded that especially for American Jewish with low income, religious bonding can significantly improve self-rated health [55]. Another study conducted in Latin American also found that generalized and neighborhood trust could modify the effect of SES on health, especially for low SES groups [56]. Researches focusing on China also found similar results. Research conducted in rural China concluded that social networks could improve the well-being of rural residents, and the positive effect was much more significant for low-income groups [16]. Similarly, a study conducted in China showed that only for the poor, social capital was statistically associated with self-rated health [57]. There are several potential reasons for our results. The first potential explanation is that the poor possess much more time to manage their social network than the rich. Usually, the rich don’t have as much leisure time as the poor have, so the poor can make use of their leisure time to benefit from social participation and networks. Another potential reason is as Collier noted that the poor had a smaller amount of physical capital or economic capital; thus, they had to rely more on social capital as a substitution of physical or economic capitals [17].
Thirdly, we find that mental health in urban area is much better than in the rural area. This is consistent with previous findings. Our results also suggest that income level is strongly associated with mental health, both depressive symptoms and subjective well-being. Plenty of previous literature suggests that income-related mental health inequality exists all around the world [58–60]. Additionally, education level, health insurance, being married, and self-rated health status are positively associated with mental health. These findings are consistent with previous findings [61]. However, the regression results show some differences. We find that people with higher BMI have higher mental health status, while in another research, BMI is negatively related to better mental health [62]. This unusual result has also been reported in several other literatures [63–65]. With more and more research finding that overweight or obese patients usually live longer, scholars concluded the existence of so-called “obesity paradox”. Our results are consistent with the findings of the “obesity paradox”. In addition, while other researchers found that employed people usually had better mental health [66], our finding reports that work could decrease personal mental health. A potential explanation for this result is that the survey in CFPS measured working status but not employment status. Currently working means less leisure time and more physical burden, thus brings poorer mental health. However, previous literature usually measures the effect of being employed or unemployed on mental health. Being employed implies a stable income and work environment.
It should be noted that reducing poverty is still the most important way to improve the mental health status of the poor. However, our findings on the relationship between social capital and mental health inequality provide some new ideas to improve the poor’s mental health status and reduce mental health inequality. For example, we advocate establishing more vulnerable groups in the community or village, especially for the disadvantaged people to expand their social network. Measures should be taken to promote social participation of the poor, such as recommending them to join social organizations. In the rural area, the service quality of the village committee and the participation of villagers should be enhanced to improve their village-level social capital, especially for people with a low-income level.
We want to caution against several limitations in this study. The first limitation is that the data we used was cross-sectional and hence we couldn’t provide a causal analysis. However, we used the lagged terms of social capital to avoid the endogeneity of social capital and mental health. This is the best solution to our problem because we tried to find a perfect instrumental variable but failed. The second limitation is the self-reported happiness indicator we used. Because different people have different definitions of happiness, subjective well-being may cause some bias to measure the happiness status of residents. Despite these limitations, our study still provides important evidence on the relationship between social capital, mental health, and mental health inequality in China.
Conclusion
This study investigates whether social capital buffers or exacerbates mental health inequality. The results show that severe mental health inequality exists in China; family-level social capital can buffer depressive symptom inequality, and village-level social capital can buffer SWB inequality. Our study advocate enhancing social participation and communication of the poor.
Supplementary Information
Additional file 1.