The mechanisms of social network on subjective well-being from a life course perspective: the mediating role of individual resilience

Abstract

Background

Social networks are an important factor affecting Chinese residents’ subjective well-being (SWB). Dividing social networks from the functional perspective can elucidate the SWB effect more effectively. This paper aims to explore how different social network supports affect SWB, further analyze the mechanism of individual resilience between them, and observe its changing trend throughout life.

Methods

Based on the 2016 China Labor Dynamics Survey data (CLDS), this paper constructs an indicator system of individual resilience and, from the functional perspective, divides social networks into instrumental and emotional functions (SNI and SNE). According to age and the sequence of major life events, classify age groups as the operational variables of life course. The mediating effect and moderated mediating effect are the main models.

Results

First, different types of social support in social networks are significant predictors of SWB. Among them, SNI has a greater effect on SWB than SNE and remains stable throughout life. Second, individual resilience is vital to SWB, and its effect does not vary throughout life. Third, individual resilience mediates between social networks and SWB, and its mediating effect increases over life.

Conclusions

The theoretical implications of developing social network measurement from a functional perspective and policy implications of focusing on the social support function of social networks and preventing it from disrupting social resource allocation rules are proposed.

Introduction

Diener et al. ([1]:253) define SWB as: “People’s appraisals and evaluations of their own lives. It includes reflective cognitive judgments, such as life satisfaction, and emotional responses to ongoing life in terms of positive and pleasant emotions versus unpleasant and negative emotions.” The social context in which individuals are embedded is an important factor affecting SWB [2]. As a bridge connecting individuals and society, the social network is a structural context in which the individual is deeply embedded [3]. Therefore, it’s an important influence for SWB.

Sociologists have dissected the relationship between social networks and SWB from functional [2] and structural perspectives [4] based on the differences in understanding how social networks work. The structural perspective is concerned with the influence of the overall social network structure on SWB, such as the network’s density, the strength of the relationship, and the type of network [5]. In addition, scholars argue that actors initiate personal networks and form solidified or institutionalized patterns of interaction that, in turn, influence actors [6]. The functional perspective focuses on the social support function of social networks. Lin [7] divides social interaction motivation among social network members into maintenance and acquisition of valuable resources, which correspond to expressive and instrumental behavior. The former can provide SNE for actors, while the latter mainly provides SNI. Numerous empirical studies have shown that SNI and SNE benefit individuals’ SWB [8].

The division of functional and structural perspectives regarding social networks is two sides of the same coin. Usually, SNE originates from informal or strong relationship networks [7], and SNI may come from formal and informal networks simultaneously [8]. However, social networks affect SWB because they can provide social support for individuals [9]. Therefore, examining the SWB effect of the social network from the functional perspective is more straightforward and effective. However, some studies have directly equated informal or strong relation networks with SNE and formal or weak relation networks with SNI [10], blurring the boundary between structural and functional perspectives. Since social networks often provide diversified social support [11].

In addition, previous studies have paid more attention to the static characteristics of social networks, ignoring the dynamic characteristics of social networks. An individual’s social network will constantly adjust and change during life [12]. Different life course stages have different demands on social networks [13], resulting in different effects on SWB.

Finally, existing research has under-analyzed the mechanisms of the SWB effect of social networks, which can not only act directly on SWB but also increase the SWB by improving individual resilience. For individuals, resilience refers to building self-esteem and self-confidence, developing positive self-perceptions and correct coping behaviors, and reducing negative perceptions, self-denial, psychological tension, stress, and behavioral disorders in the face of normative stress and adversity [14]. It is evident that individual resilience, as an individual trait, is an important protective factor for maintaining SWB in a risk society [15, 16]. At the same time, the formation of individual resilience is also inseparable from the shaping of the surrounding environment; as a structural factor, social networks can affect individual resilience [17, 18]. So, is individual resilience a mediating mechanism between social networks and SWB? In addition, as a component of personality traits, it changes with life course [19]. However, existing studies have typically focused only on the relationship between resilience and SWB in specific groups [20], ignoring the group differences.

To sum up, the research questions of this paper are as follows: How do social networks affect the SWB of Chinese residents? Does its influence differ at different stages of the life course? We will use CLDS 2016 data to test the questions by adopting mediating and moderated mediating effects models.

The contribution of this paper is as follows: First, from the functional perspective, the social support functions of social networks are categorized into emotional and instrumental, and the effects of the two on SWB and their differences are dissected. Second, we reveal the mechanism of individual resilience between social networks and SWB. Finally, it focuses on the life course trends of social networks and individual resilience and studies the life course differences of their effects on SWB.

Core concept definition

Social network

Gamper [3] thinks, “Social networks are a link between the micro-level, or the individual action (agency), and the macro-level, or the institutions. Accordingly, networks consist of actors who build relationships with one another, and those relationships create overall social structures.” Marin and Wellman [21] define “Social networks as a set of relevant nodes connected by one or more relations.” Although scholars have not reached a consensus on the concept of social networks, they all agree that the connection between individuals is the core of social networks. According to the different research perspectives, social networks can be divided into ego-centered and whole networks [22]. An ego-centered network focuses on a focal actor and several other actors directly connected. The whole network focuses on the network composed of all actors in a bounded scope, the relationships among them, and the overall structure of the network [23]. In this paper, the social network refers to the ego-centered network.

Classification of social networks. There are different types of social networks according to different criteria. According to the nature of relationships, they can be divided into formal and informal networks [10], family and working networks [24], and, according to the intimacy of relationships, divided into strong and weak networks [25]. From a functional perspective, social networks can be divided into emotional and instrumental support networks [7] according to the types of social support relationships can provide. Since studying the SWB effect of social networks from a functional perspective can provide a glimpse into the essence of SWB, this paper divides social networks into SNI and SNE from a functional perspective and studies their SWB effect, respectively.

Individual resilience

“Resilience” originated in the natural sciences and has since extended to catastrophology, ecology, and psychology. However, disaster researchers focus on the dynamic movement process of the whole society/ecosystem [26]. Psychology focuses on individuals who have suffered the shock of a major catastrophic event. Therefore, early research on resilience in psychology aims to look for factors that ensure individuals are protected from stress and distinguish between well-adapted and those who are not [27]. Psychologists have defined many resilience concepts, such as psychological, ego, and individual resilience. However, what they have in common is an emphasis on “adversity” and “positive adaptation” [27]. Luthar et al. [28] believe that adversity mainly refers to an unfavorable environment full of risks and difficulties. This definition associates adversity with risk and is an insular understanding of adversity. Some scholars also define adversity as a difficult, unfortunate, or traumatic experience [29]. The above definitions associate negative circumstances with negative consequences. They focus on established and statistically significant maladjustment to adversity. However, their definition focuses only on people with problems rather than ordinary people’s resilience. With the advent of the risk society, adversity is a normal part of everyone’s life. Besides, for most people, adversity is not necessarily a disaster. Still, mild destruction embedded in our daily lives [30], such as work stress and investment funds, incur losses. Scholars define positive adaptation as “demonstrating social competence in behavior or completing a developmental task at a certain stage” [28]. An essential but often overlooked issue in studying positive adaptation is the sociocultural context the individual embeds. Combining the two critical concepts defined above, scholars believe resilience refers to a series of traits that enable individuals to adapt to adversity [27]. With the deepening of research, scholars have also paid attention to the dynamic characteristics of resilience, suggesting that resilience will change with time and the environment [31]. In conclusion, resilience is an important trait everyone should possess in a risk-prone modern society, not just one suffering significant trauma. Accordingly, resilience is the appropriate coping behavior when facing constant daily stress.

Scholars have developed several scales to measure individual resilience. Windle et al. [31] test the quality of 15 individual resilience scales according to content validity, internal consistency, standard validity, construction validity, and repeatability and find that the CD-RISC scale, RSA scale, and Simple Resilience scale have the highest scores. However, the scores are only a medium level and still need improvement. On the one hand, it lacks a multi-level perspective and pays insufficient attention to resources at the family and community levels; on the other hand, it lacks a dynamic perspective and ignores the changes of time and environment. On this basis, Windle et al. [31] propose a measurement framework including “optimism, self-esteem, personal ability, social skill, problem-solving ability, self-efficacy, social resources, insight, independence, creativity, humor, control, family cohesion, spiritual influence, and initiative.” But he does not give specific measurements. In addition, Tusaie et al. [32] also propose a multi-level measurement framework that divides individual resilience into individual and environmental factors. Among them, individual factors include cognitive and ability factors. These cognitive factors include optimism, intelligence, creativity, humor, belief systems (which provide existential meaning), and self-efficacy. Ability factors include coping strategies, social skills, educational ability, and above-average memory. Environmental factors include perceived social support, which is defined as the objective amount of resources at the household and community level, as well as subjective perception. Drawing on the measurement framework of individual resilience proposed by the scholars as mentioned above, this study constructs an index system on individual resilience, divides individual resilience into individual and environmental factors, tries to improve the inadequacy of psychological resilience by focusing only on mental health problems, and applies it to every ordinary individual in the risk society instead of individuals with unfortunate experiences, to expand the application scope of individual resilience.

Life course theory

The life course theory originates from the life cycle theory and the life history theory. American sociologist Elder [33] expanded the lifespan development theory, elaborating on the social meanings of age and forming the basic framework of the life course theory. This framework inherits and integrates concepts such as lifespan development, life cycle, and life history, presenting a comprehensive picture of the social attributes of time. Elder believes that biological concepts like age, growth, and death are socially constructed in the life course, and the age hierarchy also expresses a kind of social expectation. Individuals will continuously play socially prescribed roles and events throughout their lives, which are arranged in an age-graded order. Although the life course theory emphasizes the constraints of social structure and social environment on individual development, it also acknowledges the subjective initiative of individuals in the life course. It is believed that individuals can maximize the benefits or minimize the costs of life events through mechanisms such as “choice,” “optimization,” and “compensation.” Therefore, the life course is dynamic and continuous. Individuals are both constrained and able to adjust actively.

Social capital and individual resilience are important forms of capital in an individual’s life course. In combination with the dynamic characteristics of the life course, an individual’s needs for social capital and resilience will change at different life course stages. Individuals will actively adjust to meet their own developmental needs. However, at the same time, an individual’s social capital and resilience are also constrained by structural factors in the life course. This paper will discuss the changes in an individual’s social capital and resilience throughout life and examine their impact on well-being.

Research hypothesis

The subjective well-being effect of social network

Bian and Guo [34] believe that social catering networks have both emotional and instrumental functions, and the former strengthens the connection between individuals and society through social interaction. The latter enhances an individual’s ability to integrate and acquire resources through capital operation. They are all beneficial for improving individuals’ SWB. Guo and Wang [9] argue that the influence of social networks on SWB includes two paths: social support and reference groups. The social support path provides individuals with material and emotional support through interpersonal interactions, affecting their SWB. Webster et al. [2] suggest that social networks can act as an emotional pressure-reducing valve, reducing negative emotional experiences and increasing SWB through emotional communication between network members. Lin et al. [35] studied people who suffered a significant shock and concluded that the more frequent emotional interactions between network members, the fewer depressive symptoms and more positive coping with difficulties. Ji and Zhao [36] divide social networks into family-based and work-based. They find that both can reduce women’s happiness loss caused by lack of income through emotional compensation, which leads to a catch-up of women’s SWB to men.

Studies have verified the SWB effects of SNI and SNE, but the SNE has received more attention. Lin [7] divides individual actions in social networks into expressive and instrumental actions. He believes different social support produces different social consequences. Instrumental actions aim to obtain new resources, increasing individual material resources. Expressive actions aim to maintain existing resources, resulting in satisfying emotional experiences, so expressive support (SNE) affects mental and psychological health more than instrumental support. SWB is a subjective feeling at the mental and psychological level [37]. Therefore, it is also an emotional experience. Thus, the SNE affects SWB more than SNI [38]. To sum up, we propose the following research hypothesis:

• Hypothesis 1: Both social network’s instrumental and emotional functions positively affect subjective well-being, but the emotional function effect is more vital than the instrumental function.

The subjective well-being effect of individual resilience

SWB is a comprehensive evaluation of one’s life from subjective emotional and cognitive dimensions, including three aspects: feeling satisfied with life, experiencing long-term happiness, and less negative emotions [39]. Therefore, positively facing adversity will reduce negative emotional experiences, increase positive emotional experiences, and improve SWB. Optimism, humor, self-efficacy, and other positive factors are essential to resilience. Moreover, according to individual resilience, positive adaptation to adversity and success at a particular stage or event signify high resilience. Success usually accompanies high life satisfaction as well as positive emotional experiences. Thus, high resilience is an important protective factor for maintaining happiness in a risky society [40]. For example, Zhang et al. [16] investigated the transmission effect of resilience between parent–child attachment relationships and life satisfaction, verifying the life satisfaction effect of resilience. Furthermore, Paredes et al. [41] pay attention to the impact of people’s psychological advantages on SWB during COVID-19, again validating the importance of psychological resilience as a transmission mechanism between mental health and well-being under external shocks. In addition, some scholars compare group differences in the effect of individual resilience on SWB and find that the effect is more significant in males than females [42]. Finally, some scholars have focused on the effects of different levels of resilience on well-being. For example, Gentz et al. [43] find that individual, family, and school resilience contribute to children’s SWB. Still, family resilience is the most important protective factor in maintaining mental health and SWB. Chang et al. [44] compare the effects of individual and family resilience on SWB and show that although individual and family resilience are highly correlated, their effects on SWB are independent; both are effective in enhancing SWB, and individual resilience is a mediating mechanism between family resilience and SWB. In summary, individual resilience is an essential factor influencing SWB, which leads to hypothesis 2.

• Hypothesis 2: The higher the individual resilience, the stronger the subjective well-being.

The mediating role of individual resilience

Studies have shown that social surroundings and cultural factors influence individual resilience [31]. The social network is where individuals deeply embed themselves and bridge the larger social surroundings [3]. On the one hand, as a structural factor, social networks can affect individual resilience. When social networks act as a structural constraint, a social network with rich social capital and harmonious relationships among members will provide more social support for individuals to cope with risks and adversity [17]. If a social network is closed and exclusive, restricts members’ freedom and violence, it will have some negative effects [45]. For example, violence in kinship and peer networks will negatively affect the formation of individual resilience [46]. On the other hand, from a functional perspective, social networks can promote individual resilience [18]. When a social network is the source of social support, it will be the protective factor of individual resilience. The larger the social network size the individual possesses, the more social support he can obtain and the stronger the ability to cope with adversity [47]. Sippel et al. [48] believe that social networks can provide individuals with different social support: emotional support, which can give emotional comfort and make people feel they are loved, respected, and cared for by others; material support, which can provide material help and services; and information support, which offers heterogeneous information. The above social support can help individuals improve their ability and enthusiasm to cope with external shocks. In conclusion, the effect of social networks on individual resilience may have different results, but it is more likely to be a positive protective factor than a destructive factor. Therefore, social networks can enhance individual resilience, and there are differences in the effects of social networks with different functions on individual resilience. In conjunction with the above, individual resilience also positively affects SWB. Then, do social networks enhance SWB by improving individual resilience? As a result, we propose hypothesis 3:

• Hypothesis 3.1: Individual resilience mediates the positive relationship between the social network’s instrumental functions and subjective well-being.

• Hypothesis 3.2: Individual resilience mediates the positive relationship between the social network’s emotional functions and subjective well-being.

The moderating role of age

Social networks have different evolutionary logic at different stages of an individual’s life course [13]. The socioemotional selectivity theory [49] and the social convoys theory [50] describe social network trends throughout life. Both argue that social networks change throughout life but with different attributions. Based on the perspective of future time, the former believes that a lifetime is limited, and the remaining future time is different in different life course stages, so the selection strategy is also different. Information acquisition is the highest priority goal in youth and early adulthood when individuals actively seek to expand peripheral social networks and enhance network heterogeneity. As one enters late adulthood, emotion regulation goals become increasingly important, and people emphasize the emotional aspects of relationships, focusing more on intimacy [49]. In old age, individuals actively sever some social ties and focus on enhancing the strength of intimate relationships [51]. Social convoys theory attributes changes in social networks over the life course to environment transition. It argues that people’s social networks act like escorts to convoy them through the life course. The innermost members of the escort are highly stable throughout life, and the relationships at the periphery change dynamically, possibly ending with changes in the external environment [52]. Drawing on these theories, Wrzus et al. [53] have combined age and significant life events to divide the life course into six stages (11–18, 18–30, 30–45, 45–60, 60–80, and 80 +). They examine the effect of regular life events (adolescence, marriage, parenthood, entry into work, widowhood, death of relatives, divorce, relocation) on social networks. Changes in social networks synchronize with major events in the life course [54]. Other studies on the relationship between social networks and age are mainly the cumulative enhancement theory [55], which argues that an individual’s social network accumulates with age. On the other hand, the decaying theory [56] holds that social networks decay with age. However, some studies hold the optimization theory, which argues that social networks show a quantitative decrease and a qualitative increase with age [12].

Dynamic change with age is an important characteristic of individual resilience, i.e., an individual’s ability to cope with adversity is dynamic [31]. Bian and Xiao [37] believe older people can cope with life pressures and shocks more positively after experiencing hardships. Resende [57] suggests that individual resilience increases with age. Eshel et al. [58] divide the participants into youth, middle-aged, and elderly groups (18–35, 36–64, and 65 +) and conclude that in terms of individual resilience, the elderly > middle-aged > youth. Conversely, some studies suggest that resilience negatively correlates with age in older people aged 65 and older in the post-disaster environment [59]. In summary, no consensus exists on the relationship between individual resilience and age, showing dynamic and complex characteristics. So, the present study combined age and life events to divide the participants into youth, middle-aged, and elderly groups to dissect the changes in individual resilience and its predictive effect on SWB over the life course.

Combining the above studies [60], we divided participants into youth, middle-aged, and elderly groups (18–35, 36–64, and 65 + years) according to the sequence of significant life events [58]. The age of 18–35 is a critical period for individuals to enter work, start a family, become a parent, and other life transitions, during which social network and individual resilience are in a rapid generation stage. 36–64 years old is a stable period of life. After the precipitation of major life events such as marriage and settling down, the social capital in this period is relatively rich. Individual resilience is also improved and tends to be stable. 65 is the retirement age for most individuals. Their life takes another turn, leaving the workplace and returning to the family; physical function gradually declines, and at this time, the social network and resilience change again.

According to the socioemotional selectivity theory, the demand for heterogeneous resources is high in youth when SNI increases and changes rapidly. The Middle-aged group’s SNI is relatively stable and supportive. Older adults leave their jobs, and the stock of SNI decreases. As a result, we propose the following hypothesis:

• Hypothesis 4.1: The role of the social network’s instrumental functions on individual resilience differs across the life course. The middle-aged group has a more substantial predictive effect than the youth group. The youth group has a stronger predictive effect than the elderly group.

SNE primarily provides individuals with emotional support. As individuals age, their emotional attachments increase [49]. Middle-aged people have the most prosperous intimate relationships after experiencing major life events, such as starting a family and having children. As a result, they can provide the most emotional support and are the strongest predictor of individual resilience. In old age, the attachment to emotional social capital increases, but they are experiencing major life events of illness and death, the number of intimate relationships decreases, the social support they can provide decreases, and the predictive effect on individual resilience weakens. Nevertheless, the need for SNE in the youth is relatively small, and the stock of SNE in youth is also smaller. By contrast, there is no difference between young and old groups. Thus, we propose the following hypothesis.

• Hypothesis 4.2: The role of the social network’s emotional functions on individual resilience differs across the life course, with the middle-aged group having a stronger predictive effect than the youth group; no differences existed between the elderly and youth groups.

Individual resilience changes along with the life course. Therefore, we propose hypothesis 5:

• Hypothesis 5: The effect of individual resilience on subjective well-being varies across life course stages.

Combining the above assumptions, we further propose the following hypotheses 6:

• Hypothesis 6.1: The mediating effect of individual resilience between the social network’s instrumental functions and subjective well-being varies throughout life.

• Hypothesis 6.2: The mediating effect of individual resilience between the social network’s emotional functions and subjective well-being varies throughout life.

In summary, the theoretical model is in Fig. 1.

Fig. 1

The theoretical model

Research design

Data

The data in this paper came from the 2016 CLDS conducted by the Social Science Research Center of Sun Yat-sen University in 29 provinces, municipalities, and autonomous regions (except Hong Kong, Macau, Taiwan, Tibet, and Hainan) in China. The survey adopted a multi-stage, multi-level, probability sampling method proportional to the size of the labor force. The survey completed a total of 401 village questionnaires, 14,226 household questionnaires, and 21,086 individual questionnaires, which are representative. This study combined data at the community, household, and individual levels. After removing the samples with missing values, the final sample is 17,386.

Variables

Dependent variable. The dependent variable in this study was SWB, using the questionnaire item, “In general, do you think you are happy with your life?” The answers were coded from 1 to 5 as “very unhappy” to “very happy,” respectively. Overall, the mean SWB of Chinese residents is 3.78.

Independent variable. The independent variable in this paper is social network.

The social network scale is an important measurement indicator of social networks [61]. Currently, the mainstream measurement methods of social networks include name-generating and position-generating methodologies [62]. The name-generating methodology involves creating a list of participants’ networks, thus obtaining a sample of the respondents’ social relationships. It is a person-focused methodology. The method will yield quantitative and qualitative data on the respondent’s social network. The position-generating methodology systematically samples a list of jobs within a social hierarchy [63]. The respondents were asked whether they knew someone in that sampled position. The position-generating methodology is structure-focused. Empirical research has verified that the two methods can measure individuals’ social networks well. However, they can’t effectively distinguish the types of social support social networks provide. Therefore, Van Der Gaag et al. [64] propose a resource generator methodology devoted to measuring the types and quantity of available resources in social networks. Specifically, they ask whether there is anyone in their social network (acquaintances, friends, family) who could provide a certain type of support (including 37 needs), then divided the required supports into four categories: prestige and education, entrepreneurship, skills, and personal support, and then measured the scale. This method is a good attempt to measure social network capital from a functional perspective. However, its disadvantage is that it only measures whether a certain kind of support exists in different networks and cannot effectively measure the scale. Puccia et al. [65] divided social support in social networks into expressive and instrumental support according to Lin’s division, combined with Van Der Gaag’s resource generator method to classify the social support that college students obtain from social network members (classmates, teachers, parents, etc.) into instrumental (specific advice and resources) and expressive support (emotional support and encouragement). The results show that parents’ social support positively impacted social network members. In particular, parents’ expressive support had a greater impact than instrumental support. This article also measures the types of social support social network members provide from a functional perspective and compares its impact. What these articles have in common is that given social network members, then ask if they can provide a certain type of support. So, they focus on the types of social support different social network members provide. Our paper also attempts to investigate the types of social support social network members provide from a functional perspective. However, we focus on the amount of a particular type of social support an individual has access to rather than who provides it.

We draw on the idea of a resource generator to measure the size of a social support network to make up for the deficiency of the resource generator method [64]. Specifically, the former asked: “How many of these close local people can you tell to your heart? The latter asked: “How many of these local people can you borrow money ($5,000) from?” Since participants’ social networks vary widely and the mean value is small, it is taken as a logarithm in the paper to ensure the reliability of the results.

Mediating variable. We combined the measurement framework of individual resilience proposed by Windle et al. [31] and Tusaie et al. [32], selected the measurement indicators agreed upon by the two papers and tried to construct the index system with multi-level and considering the dynamic change of the environment. Table 1 shows the results. This paper used principal component analysis to process the indicators, objectively endowed weights to the indicators, and ultimately generated individual resilience. We used the Kaiser–Meyer–Olkin (KMO) and Bartlett’s sphere to test whether the above index system is suitable for principal component analysis. This study used KMO = 0.628 > 0.5 and SIG = 0.00 < 0.05 in Bartlett’s sphere test, thus making it suitable for principal component analysis [66]. According to the guideline of eigenvalues greater than one, it selected seven principal components. The cumulative variance contribution rate was 62.5% > 50%, which was within the reasonable range for social science research [60], resulting in determining the weights of each dimension (see Table 1) and generating individual resilience.

Table 1 Index system of individual resilience

Moderating and controlling variables. The participants were divided into youth (18–35), middle-aged (36–64), and elderly groups (65 and above), combined with age and sequence of life events, and used as operational indicators of the life course [58]. Other control variables included gender, marital status, education, health status, annual income, and urban–rural. Studies have tested the effects of these variables on SWB. Table 2 shows the descriptive statistics of the variables.

Table 2 Descriptive statistics of variables (N = 17,386)

Models

This paper used the mediating and moderated mediating effect model to test hypotheses. We constructed the models as follows:

$$Y=\alpha_0\;+\;\alpha_{1i}X_i\;+\;\alpha_{2j}S_j\;+\;\alpha_3W\;+\;\mu_0$$

(1)

In Eq. (1), Y is SWB. $\alpha 0$ is the constant term, X_i is the vector of control variables, where i = 1, 2, 3, 4, 5, 6, and $\alpha 1i$ is the parameter for each control variable. S_j is the social capital, where j = 1, 2, represents SNI and SNE, respectively, and $\alpha 2j$ is the parameter for the social capital. $\mu 0$ denotes the error term.

$$M\;=\;c_0\;+\;c_{1i}X_i\;+\;c_{2j}S_j\;+\;\mu_1$$

(2)

In Eq. (2), M denotes individual resilience, c₀ is the constant term, X_i is the vector of control variables, and c_1i is the parameter for each control variable. S_j is the social capital, and c_2j is the parameter for the social capital.$\mu 1$ denotes the error term.

$$Y\;=\;\alpha'_0\;+\;\alpha'_{1i}X_i\;+\;\alpha'_{2j}S_j\;+\;\alpha'_3W\;+\;\alpha_4M\;+\;\mu_2$$

(3)

In Eq. (3)_, $\alpha 4$ is the parameter of the mediating variable _M, $\mu 2$ is the error term, and other variables are the same as above.

$$M\;=\;c_0\;+\;c_{1i}X_i\;+\;c_{2j}S_j\;+\;c_3W\;+\;c_{4j}S_j\ast W\;+\;\mu_3$$

(4)

$$Y\;=\;\alpha"_0\;+\;\alpha"_{1i}X_i\;+\;\alpha"_{2j}S_j\;+\;\alpha"_3W\;+\;\alpha"_4M\;+\;\alpha_5M\;\ast W\;+\;\alpha_{6j}S_j\;\ast W\;+\;\mu_4$$

(5)

Moderated mediation effect model. In Eq. (4), M denotes individual resilience, c₃ is the parameter of age groups, S_j*W is the interaction between age groups and social capital, and c_4j is the parameter. $\mu 3$ is the error term. In Eq. (5), $\alpha^{\prime\prime}4$ is the estimated parameter of individual resilience, and M*W is the interaction term between individual resilience and age groups, $\alpha 5$ is the parameter. S_j*W is the interaction term between social capital and age group,$\alpha 6j$ is the parameter. The others are the same as above.

In this paper, we used Model 4 (a simple mediation model) and Model 58 (a moderated mediation effect model) in the PROCESS macro in SPSS, combined with the sequential test for coefficient product and the bias-corrected percentile Bootstrap CI method to test the research hypotheses.

Results

Descriptive statistics of variables

Table 3 presents the results of the group difference analysis of SWB. The mean of SWB for females is higher than that for males, but the difference between the two is not statistically significant. The SWB of the young group is higher than that of the elderly and middle-aged groups, and the differences among the three groups are statistically significant. Urban residents have a higher level of SWB than rural residents, and this difference is statistically significant. Married residents have a higher level of SWB than unmarried residents, and this difference is also statistically significant. SWB increases with education, and the differences among different educational levels are statistically significant.

Table 3 Group difference analysis of SWB

Figure 2 further compares the distribution of major variables at different life course stages and shows the results of the mean difference test. The small white dots in the figure represent the median, and the shadow indicates the distribution of the samples. The wider the shadow corresponding to the variable value, the more samples are distributed at the point. Comparing age group differences in SWB in 2(a) shows that youth > elderly > middle-aged and the mean differences are significant across all three groups. The comparison of individual resilience by age group in 2(b) shows that the elderly > the middle-aged > the youth, and the mean differences are all significant, indicating that individual resilience increased with age. The comparison of SNE in 2(c) shows that the middle-aged > the elderly > the youth group. Still, the differences are significant. Finally, the comparison of SNI among age groups in 2(d) shows that youth > middle-aged > elderly, indicating that the SNI decreases with age, and the differences between groups are significant.

Fig. 2

Distribution of main variables by life and mean difference test. Note: SNE and SNI are taken logarithmically in the figure, and samples where SNE and SNI take the value of 0 are added 0.01 to minimize sample loss when taking the logarithm

The mediation effect model

Table 4 shows the results of mediating effects under stepwise regression. Model 1 takes individual resilience as the dependent variable and SNI as the independent variable. After controlling for demographic characteristics, SNI significantly positively affects individual resilience (B = 0.226, t = 30.577, p < 0.001). Model 2 takes SWB as the dependent variable. The results show that when not adding individual resilience, the total effect of SNI on SWB is 0.092 (t = 12.139, p < 0.001). In Model 3, after adding individual resilience, SNI still plays a significant positive role (B = 0.018, t = 2.475, p < 0.05), but the coefficient decreases. It shows that individual resilience partially mediates SNI and SWB, which verifies Hypothesis 3.1. Individual resilience also has a significant and positive effect on SWB (B = 0.327, t = 44.242, p < 0.001), indicating that the higher the individual resilience, the stronger the SWB, which tests hypothesis 2. Table 5 shows the decomposition results of the mediation effect under the Bootstrap method. The direct effect of SNI accounts for 19.57%, and the mediating effect accounts for 80.43%, indicating that most of SNI’s effect on SWB transmits through individual resilience.

Table 4 The mediation effect under the stepwise regression method

Table 5 Decomposition results of mediation effect under Bootstrap method

Model 4 in Table 4 shows that, after controlling for demographic characteristics, SNE significantly and positively affects individual resilience (B = 0.214, t = 30.305, p < 0.001). In Model 5, the total effect of SNE is 0.076 and significant at the 0.001 level. In Model 6, after adding individual resilience, the effect (B = 0.005, t = 0.727, p > 0.05) of SNE on SWB is no longer significant. In contrast, individual resilience significantly positively affected SWB (B = 0.330, t = 44.151, p < 0.001), indicating that individual resilience is mediating, which tests hypothesis 3.2. In Table 5, the direct effect of SNE accounts for 7.89%, and the mediating effect accounts for 92.11%, indicating that SNE is more closely related to individual resilience, and its effect on SWB mainly transmits through individual resilience.

The above results indicate that SNI and SNE positively contribute to SWB. Moreover, the total effect (0.092 > 0.076), direct effect (0.018 > 0.005), and indirect effect (0.074 > 0.070) of SNI are greater than that of SNE. Therefore, part of hypothesis 1 is verified, and the other part is falsified.

Addressing endogenous problems

There may be a bidirectional causality between social networks and SWB/individual resilience. On the one hand, social networks can enhance SWB/individual resilience. Meanwhile, individuals with high SWB/resilience are likely to be more open to interacting with others and, therefore, have richer social networks. To overcome this endogeneity problem, we drew on the studies of Wang and Zhang [67] and Yin et al. [68] to take the mean of other residents’ (remove themselves) social network size at the village level as an instrumental variable. We conducted a series of tests to test the instrumental variable’s validity. Table 6 shows the results. Model 1 is the first stage regression result of 2SLS for SNI. First, the endogeneity test. The Hausman test was used to test the endogeneity of SNI, and the result showed that the P-value was less than 0.05, indicating that SNI was an endogenous variable. It is necessary to use the instrumental variable method to correct it. The second is the weak instrumental variable test. The Cragg-Donald Wald F statistic is 1314, much higher than the critical value of 16.38 at the 10% bias level, which rejects the null hypothesis that endogenous variables are uncorrelated with the instrumental variable and indicates that there is no weak instrumental variable problem. The Kleibergen-Paaprk Wald F statistic is 1300; in addition, the F-value of the first stage is 237.33, which proves that there is no weak instrumental variable problem. Finally, the unidentifiable test. The p-value corresponding to the Kleibergen-Paap rk LM statistic is significant at the 1% level, strongly rejecting the null hypothesis of “unrecognizable,” the instrumental variable meets the basic requirement of holding the rank condition. To sum up, the instrumental variable is valid. Model 4 shows that the instrumental variable for SNE is also valid. The coefficients of the instrumental variables on SNI and SNE in Models 1 and 4 are positive and significant, which indicates that the instrumental variables are valid.

Table 6 Test and regression results of instrumental variables

Notably, when SWB is the dependent variable and SNE is the independent variable, the regression result fails the Endogeneity test, indicating that SNE is an exogenous variable and the OLS model is more reasonable. Model 2 is the second stage regression, which takes SNI as the independent variable and SWB as the dependent variable. The results show that the effect of SNI on SWB after adding the instrumental variable is still positive and significant, with a coefficient of 0.147, which is larger than the result without adding instrumental variables (0.092), indicating that without considering the endogeneity problem underestimated the SWB effect of SNI. Model 3 is a second-stage regression when individual resilience is the dependent variable, and SNI is the independent variable. The results show that the effect of SNI on individual resilience is still positive and significant after adding the instrumental variable. Still, the coefficient is smaller than the model before adding the instrumental variable, suggesting that without considering endogeneity, it overestimates the effect of SNI on individual resilience. Model 5 is the second stage regression when SNE is the independent variable, and individual resilience is the dependent variable, and the results are consistent with model 3. Therefore, although there is an endogeneity problem, the results of adding instrumental variables are consistent with those of not adding them, indicating that the above results are robust.

The Moderated Mediating Model

Results of SNI

The role of SNI on individual resilience. Model 1 in Table 7 shows that SNI significantly and positively contributes to individual resilience, indicating that for individuals in the youth group, the richer the SNI, the stronger the individual resilience. Meanwhile, the interaction terms are all positive and significant, indicating that SNI in the middle-aged group and the elderly group had a more substantial effect on individual resilience than that in the youth group. The simple slope analysis in Fig. 3 shows this result more intuitively. The SNI of the elderly group has the most potent predictive effect on individual resilience (the highest slope) (simple slope = 0.258, p < 0.001), followed by the middle-aged group (simple slope = 0.237, p < 0.001), and the weakest in the youth group(simple slope = 0.172, p < 0.001). Thus, part of hypothesis 4.1 is verified.

Table 7 Moderated mediation model of SNI (N = 17,386)

Fig. 3

Life course differences of SNI on individual resilience

Life course differences in the SWB effect of individual resilience. Model 2 in Table 7 shows that the main effect of individual resilience is positive and significant, indicating that for youth, the higher the individual resilience, the higher the SWB. However, the interaction terms between age groups and individual resilience are insignificant. Therefore, the predictive effect of individual resilience on SWB is consistent across the life course, and hypothesis 5 is not verified.

Table 8 shows the differences in the mediating effects of individual resilience at different life course stages. The results indicate that the mediating effects of individual resilience between SNI and SWB are significant at different life course stages, increasing with age. Further comparison shows that the differences are significant between the groups, except for the elderly and middle-aged groups, which tests Hypothesis 6.1.

Table 8 Differences in mediation (N = 17,386)

Results of SNE

The effect of SNE on individual resilience. Model 1 in Table 9 shows that SNE positively contributes to individual resilience, indicating that the higher the SNE, the higher the individual resilience for individuals in youth. The interaction term between the middle-aged group and SNE is significant at the 0.05 level, indicating that SNE in the middle-aged group is a stronger predictor of individual resilience than in the youth group. The interaction term between the elderly group and SNE is insignificant, indicating no difference between the elderly and youth groups. The simple slope analysis in Fig. 4 shows this result more visually. The SNE of the middle-aged group has the most potent predictive effect on individual resilience (the largest slope) (simple slope = 0.221, p < 0.001), followed by the elderly group (simple slope = 0.219, p < 0.001). The youth group is the weakest (simple slope = 0.177, p < 0.001), verifying hypothesis 4.2.

Table 9 A moderated mediation model of SNE (N = 17,386)

Fig. 4

Life course differences of SNE on individual resilience

Table 10 demonstrates the differences in the mediating role of individual resilience across the life course. At different life course stages, individual resilience is the mediating mechanism between SNE and SWB, but the intensity of the effect is different. The mediating effect of individual resilience is most potent in the middle-aged group, followed by the elderly group, and weakest in the youth group. Further comparison shows that the differences were significant for the middle-aged and youth groups and insignificant for the other groups, which tests hypothesis 6.2.

Table 10 Differences in mediation

Robustness test

This paper used replacement models, estimation methods, and variables to test the robustness of the results.

First is the substitution model. Since SWB can be regarded as an ordered categorical variable with values of 1–5, the Ologit model can also be used to test the hypothesis. To compare the effects of SNI and SNE on SWB further, we put them in the same model. Figure 5 shows the results. In the OLS model, the effect of SNI on SWB is greater than that of SNE. The results of the Ologit model are also consistent, indicating the robustness of the above results.

Fig. 5

Comparison of SWB effects between SNI and SNE

Second, replace the estimation method. Use the KHB method to verify the robustness of the mediation effect and show the results in Table 11. The results show that when SNI is the independent variable, the total, direct, and indirect effects are significant in OLS and Ologit models. Furthermore, the total and indirect effects are significant when SNE is the independent variable. In contrast, the direct effects of the two models are all insignificant, consistent with the above results.

Table 11 Mediating effects estimated by the KHB method

Third, replace the dependent variable. SWB and life satisfaction are all measures of happiness [69]. So, we tested the moderated mediated effects model again using life satisfaction, and the results are consistent with the above. Due to space limitations, we will not show the results.

Discussion

In summary, some issues still need to be further discussed. First, the SNI has a greater impact on SWB than the SNE, contrary to existing research. This paper suggests that the sources of difference are as follows: on the one hand, the difference in research perspectives. The existing study tests social networks’ SWB effect from a structural perspective. The naming and positioning methods of social network measurement are designed based on the structural characteristics of social networks [70], ignoring the functional characteristics of social networks. Even though some structural characteristics can reflect the social functions of social networks, they cannot effectively separate specific types of social support. For example, some studies equate informal or kinship networks to SNE and formal networks to SNI [10]. This division blurs the boundary between network structure and function and could not effectively distinguish the difference in the SWB effect of the social network. A social network’s effectiveness in enhancing SWB depends on multiple social supports [51]. Therefore, we need to differentiate the types of support from social networks to effectively evaluate the role of social networks and give practical suggestions for improving SWB. From the functional perspective, this paper avoids ambiguous attribution from a structural standpoint and is more conducive to excavating the social value of social networks. The above results mean that when examining the social effects of social networks, the classification may influence the results [71]. Therefore, the researcher should select more applicable classification criteria based on the specific study.

On the other hand, the difference concerns China’s development stage. Bian and Guo [34] believe that if a country’s market economy develops sufficiently, the SNI will weaken, and its primary function will be to provide emotional support. Although China’s economic and social development has been rapid in recent years, and people’s living standards have improved significantly, China is still a developing country, and the market economy is imperfect. During the transition period, social networks still play a role in resource allocation [72]. As a developing country, resources related to material living standards remain an important influence on residents’ SWB [73]. The SNI mainly provides individuals with resources such as financial, material, and heterogeneous information, which contribute to improving living standards. The satisfaction of material life also has emotional value [74]. Therefore, the SNI has a more substantial SWB effect. Although the SNI can improve residents’ SWB and help compensate for formal institutional security deficiency. However, policymakers should also be wary of the potential social injustices of social networks as informal channels for resource allocation.

Second, the effect of individual resilience on SWB does not vary across the life course. Although descriptive statistics show that individual resilience increases with age, studies have suggested that individual resilience changes with the social environment and age [31]. However, the effect of individual resilience on SWB remains stable, i.e., individual resilience is an important protective factor for SWB throughout life. These results indicate that resilience is important for individuals to cope with risks and maintain a positive life attitude throughout the life cycle in a risk society. One of the important consequences of modernity is the increase in risk [75]. Although risks are omnipresent, there are group differences in the ability to cope with them, which may lead to greater social inequality. For example, the COVID-19 epidemic that spread to the world in 2020 has impacted everyone’s production and life, but the negative impact on high social class groups is far less than that of low social classes [76]. To cope with ubiquitous risks, everyone should strive to improve their resilience. In addition, policymakers should pay more attention to the lower class groups.

Third, the role of social networks on individual resilience differs throughout life. Specifically, SNI is a stronger predictor of individual resilience in the elderly and middle-aged groups than youth groups. The simple slope analysis shows that the elderly > middle-aged > youth group, and conversely, the descriptive statistics show the SNI in the youth > middle-aged > elderly group. The contradictory finding deserves further attention. Socioemotional selectivity theory explains changes in social networks from a future time perspective. They argue that future time appears infinite in youth, and individuals focus on cultivating instrumental social relationships. Thus, youth have more extensive peripheral social networks [77]. However, individuals in youth are in the formative stages of their careers and families and are experiencing significant life events. Thus, while the stock of SNI is high, it is less stable and has a limited effect on individual resilience. In middle age, SNI becomes more stable and supportive [51]. Individuals in old age withdraw from the workplace, and their SNI declines. Some researchers have suggested this is an active selection process in older people [78]. However, the results indicate that older people’s SNI predicts individual resilience more than the youth and middle-aged groups. On the one hand, as the physical functions of older people deteriorate and their income decreases, their demand for SNI that can provide material support has not reduced but increased. It also warns our aging society to improve the material living standard of older people. On the other hand, SNI declines with age, which is more of a passive process. The fundamental reason is that older people can no longer contribute to organizations or jobs. It also suggests that we further explore the social value of older people and extend their years of output value, which is beneficial to individual resilience and SWB and can help us better cope with the various challenges of the aging society. The role of SNE on individual resilience is more vital in the middle-aged group than the youth group, and there is no difference between older people and youth groups. Socioemotional selectivity theory suggests that individuals in youth focus on developing instrumental social capital and turn to maintaining intimate relationships after middle age. In middle age, individuals have the most SNE, which is (strongly) redundant, dense, and cohesive. It provides expressive support that can strengthen individuals’ identities, influencing their attitudes and abilities to cope with adversity [35]. Although the elderly are highly attached to the emotional support provided by intimate relationships, they face major life events of illness and death, and intimate relationships diminish [78]. Therefore, they can provide relatively less support. In conclusion, the resilience and SWB of older people need more attention.

Note that this paper still has some limitations. First, the inadequacy of social network measurement. This paper aims to analyze the SWB effect of social networks from a functional perspective. However, existing social network measurement methods are mainly from the structural perspective. Although the resource generator method attempts to measure different types of social support in a social network from a functional perspective, it can only measure whether a certain type of support exists in the network but cannot measure the quantity and quality of support [64]. This paper draws on the idea of a resource generator to measure the amount of certain social support in a social network. However, due to data limitations, the types of social support measured in this paper are relatively simple, and it is impossible to identify the quality of social support provided by social networks. However, despite this, it can explain the research problem of this paper to a certain extent. In the future, we should further explore social network measurement methods from a functional perspective to make up for the shortcomings of this study.

Second, life course measurement is inadequate. Life course theory attempts to explain the connection between individual life patterns and social changes, describing and explaining individuals’ changing social roles and positions throughout their lives [79]. Using the 2016 CLDS data, this study combines age and regular life events to divide the survey respondents into different age groups. Then, we examine the mechanisms of social capital, individual resilience, and SWB to compensate for the lack of long-term dynamic tracking data. However, the drawbacks of this approach are obvious, as the different life stages of different individuals cannot replace the life course of the same one. In the future, we should collect long-term tracking data to make up for the shortcomings of this study.

Policy suggestion

Firstly, heterogeneous social networks can be constructed through multiple channels to enrich individuals’ instrumental support and enhance their well-being. Empirical results show that instrumental social support has a greater impact on the well-being of Chinese residents than emotional support, and this impact remains stable at various stages of the life course. It indicates that at the current stage, instrumental support that can improve material living standards is still the main demand of Chinese residents. Heterogeneous social networks can provide richer heterogeneous information and offer the supported individuals more diverse development opportunities. It reminds policymakers to encourage social participation, vigorously develop social organizations, provide residents diverse participation opportunities, and expand their heterogeneous social networks.

Secondly, the administrators can cultivate individual resilience, enhance individuals’ endogenous development force, and improve their subjective well-being. Empirical research shows that individual resilience is important to subjective well-being and remains stable throughout life. Resilience refers to an individual’s ability to cope with external risks and shocks, which includes personality traits and external environmental factors. Policymakers can enhance the governance capacity at the grassroots to provide individuals with a harmonious, safe, and secure external environment. On the one hand, it is conducive to reducing risk factors, and on the other hand, it is beneficial for the cultivation of individual resilience.

Finally, the social capital and resilience of the elderly should be cultivated, and an active aging strategy should be implemented. Empirical results show that the elderly group’s social capital has the strongest impact on resilience, and the elderly group also has a high demand for social capital. As the elderly group withdraws from society and returns to the family, their stock of social capital decreases. However, their physical functions decline, and their needs for daily care and emotional support increase, which leads to a decline in the material quality of life and poor mental health status of the elderly, which contradicts the strategy of active aging advocated in China. Social administrators should provide diversified channels of social support and social participation to enrich the social capital of the elderly group. In addition, building an age-friendly community environment can enhance the resilience of the elderly group. Doing so will improve the material and spiritual living standards of the elderly group.

Conclusion

Using data from the 2016 CLDS, this paper examines the role of social networks on SWB and their changing trends over the life course and analyzes the mediating mechanism of individual resilience. The findings are as follows. First, SNI and SNE can effectively enhance individuals’ SWB. Still, SNI has a more substantial effect. Follow-up comparisons of direct and indirect effects by age groups find that the SWB effect of SNI is consistently more potent than SNE, which contradicts previous research. Second, individual resilience positively affects SWB, and there is no difference during the life course. Third, individual resilience mediates between SNI/SNE and SWB; the mediating effect differs in the life course.