Introduction
China’s grassroots elections (jiceng xuanju) have drawn considerable attention from China watchers around the world. In light of reform and opening-up policy launched under the leadership of Deng Xiaoping [89], the country in the 1980s began to implement a self-governing reform with the core measure to let registered residents elect administrative leaderships in rural villages and urban communities [17, 44, 57, 64, 73]. Unpreceded in Mao’s era (1949-1976), this state-led “revolution” continued into the 21st century [33, 45, 65]. It has stimulated, among other research interests, a sociological question: How do informal networks of social relationships affect Chinese people’s participation in grassroots elections?
We raise this question for good reason. In the cultural legacy of relational Confucianism, China’s grassroots elections prior to the 21st century were found to be affected by local networks of residents in significant ways. In rural areas, for instance, clan networks were used not only to organize villagers to participate in elections but also to coordinate them into collective action in favor of preferred candidates [36,37, 87, 92]. Although such networks were rarely available in city communities, neighbor networks of friendship and acquaintance were instead in place to mobilize urban residents to participate in the elections of neighborhood committee directors [33, 41]. These case studies have shown a well-established finding about the network effect in the depoliticized decades of the 1980-2000 s under the leadership of Deng and Deng’s appointed heirs [82, 93]. However, it is needed to explore whether and how informal social networks matter in grassroots elections that have become more organized and controlled by the Chinese Communist Party (CCP) under Xi Jinping’s re-politicized regime since October 2012 [78].
Advertisement
In a broader observation beyond China, the network effect on electoral politics is in fact a widespread phenomenon in social democratic countries. There, free-will citizens are under no political or ideological coercion to vote for public officials, but all of them are embedded in informal social networks through which to learn information about politics, motivate political awareness and interest, and mobilize organizing resources, which are some of the relational mechanisms to promote social movements [59, 60, 79], civic engagement [48, 70, 106], and election participation [49, 52]. Among others, Putnam [69‐71] has identified two other important relational mechanisms that increase citizens’ propensities of election participation: Whereas networks of associations, communities, and social organizations provide formal channels for civic engagement, including election participation, boundary spanning networks increase citizens’ generalized trust in people in society at large, which in turn boost their participation in local and national elections. It will be interesting to see how these relational mechanisms interact with one another to play a role in mobilizing Chinese citizens to participate in grassroots elections under Communist rule.
The organization of this article is as follows. After this introduction, we provide a review of relevant research literature on China’s grassroots elections and theoretical perspectives developed to explain the variation in election participation of Chinese citizens. Next, we develop a network-focused theoretical framework from which to derive research hypotheses about how informal social networks affect election participation through direct and indirect pathways. This will be followed by an empirical analysis of the 2017 Chinese General Social Survey (CGSS) to test the hypotheses. In the last two sections of this article, we discuss the implications of our data analysis results and draw conclusions.
Literature Review
Grassroots Elections in China
In this article, “grassroots” refers to the lowest level of the Chinese administrative hierarchy. From the highest to the lowest, this hierarchy includes six levels of jurisdiction: (1) central (zhongyang), (2) province (sheng), (3) city (shi), (4) urban district (qu)/rural county (xian), (5) urban street (jiedao)/rural township (xiang/zhen), and (6) urban community (shequ)/rural village (cun). The so-called “grassroots elections” we discuss in this study have occurred at the lowest level of the hierarchy, or the level of urban community and rural village; such elections have not been organized at higher levels of the hierarchy, and no anticipation is made about this to happen in any near future1. In China’s grassroots election operations, villagers with the residential registration (hukou) in the village have the right to vote for a village head and other members of the new village administrative committee from among nominated candidates; similarly, registered residents in an urban community have the rights to vote for a director and other members of the new community administrative committee from among nominated candidates [33, 92].
A critical issue is how these candidates are nominated. The 2010 version of China’s rural local electoral law stated that candidates for the director and other members of the new village committee must be nominated by villagers and should not be appointed by any organization or person, and those candidates who win a simple majority of votes will be elected. Similar regulations are included in the electoral law about urban neighborhood directors and committee members. These regulations were by and large implemented to guide candidate nomination and election organization in both rural and urban areas before 2012, creating opportunities for popular candidates with higher education, entrepreneurial spirit, and good track records to serve the people to win the election [33, 74, 82]. However, CCP organization at the immediately higher level of government jurisdiction continued to exercise its power and influence to recommend incumbent or new candidates, a mechanism that was widely operated to increase the probability that CCP branch secretaries or politically loyal cadres be elected for the administrative leadership positions [35, 94, 103]. Since 2012, this mechanism has been strengthened under the Central Committee’s pronounced principle of total CCP leadership (dang de quanmian lingdao), consequently, an increasing number of villages and communities have been making the CCP secretary the only candidate in village head or urban neighborhood director election [91, 97].
Advertisement
Nonrelational Perspectives
Since their inception, China’s grassroots elections in rural and urban areas have quickly attracted scholarly attention, domestically and abroad. What concrete forces have been promoting or hindering people’s participation in leadership elections at grassroots levels? Seeking theoretically-informed, empirically-rooted answers to this question, researchers have utilized three nonrelational perspectives to explain institutional, cultural, and individual-level variations in election participation.
The institutional perspective emphasizes the varying capacities and influences of CCP organizations across localities. The point of departure is that grassroots electoral reform is not officially meant to weaken but strengthen the power of the CCP. At the lowest level of the party-state hierarchy, this means that in all rural villages and urban communities, it is the CCP branch secretary, rather than the elected administrative leader, who holds the political authority and decision-making paper over local affairs of public nature. As such, the officially desired result of the election is to make either the CCP branch secretary or the politically loyal cadre the leading candidate and the ultimate choice for administrative leadership through voting, or otherwise election policies and strategies will be adjusted in time to ensure the status-quo of the ongoing governing system [6, 35, 50, 91]. Empirically, however, local CCP organizations vary in their capacities in enforcing this one-party rule. While there is a lack of national statistics, one recent study shows that before 2001, Guangdong Province had 56% of village heads being CCP secretaries, and it was 86% in the Weihai region of Shandong Province [97]. Interestingly, the same study also reveals that under Xi Jinping’s regime after 2012, CCP secretaries as village heads became a central directive that was written into China’s Strategic Plan for Rural Revitalization (2018-2022). Consequently, since 2019 the percentage of CCP secretaries as village heads has increased to 76.8% in Liaoning, 83.0% in Shaanxi, 87.1% in Beijing, 99.8% in Heilongjiang, and 100% in Tianjin [97].
The cultural perspective points to the roles of cultural traditions in legitimizing or illegitimating election participation across societies [1, 56, 98, 101]. For China, Shi (2001) argues that the Confucian cultural tradition tends to characterize the ruler-mass relationship in terms of the father-son relationship, and Confucian teachings expect rulers to cherish masses and, at the same time, expect masses to be loyal to rulers [75]. Consequently, argue Shi and Lu (2010), in contrast to the Western notion of procedural democracy that ordinary citizens elect certain candidates to represent their interests and values, the Chinese notion of “people-centered democracy” (minben minzhu) has the tendency that voters do not easily withdraw their support for certain entrusted leaders so far as they are connected and committed to the clan or community to which they belong. This Chinese notion of democracy was used to interpret higher rates of election participation in China than in Taiwan, where Western-style democracy is more deeply embedded in their political institutions [76]. Inside China, a study of economically prosperous and highly marketized southern Jiangsu shows that villagers there have adopted the Western notion of procedural democracy, and they decided to decline to participate in village elections once they felt that their votes would not generate desired results in line with the democratic principle [105].
Unlike the forementioned two perspectives, the micro-individual perspective pays the most attention to interpersonal variation in voting participation. A chief assumption behind this perspective is that election participation is voluntary and therefore varies among individual participants who differ in attributes, capacities, and interests. Empirical studies have found that voters indeed vary by gender, age, education, social class, political identity, and political efficacy. Election participation was higher for men than for women, higher for middle-aged adults than for younger and older citizens, and higher for CCP members than for non-members [39, 40, 44, 73]. Election participation is associated with voters’ education, but in a nonlinear pattern with an inverse U shape [45], and is linearly associated with voters’ evaluation of local governments [88, 99] and political efficacy [102].
Relational Perspective
Different from the nonrelational perspectives just reviewed, the relational perspective considers voters not as isolated actors but as social networkers who influence each other in election participation. In general, informal networks of personal and social relations flow nonredundant information [31], maintain interpersonal trust and reject the behavior of opportunism [32], reinforce group identity and community solidarity through generalized reciprocity [22, 70], and mobilize network-embedded scarce resources, tangible and intangible, for instrumental and expressive goal attainments [53, 63]. Specifically, Western democratic systems facilitate two kinds of relational mechanisms that promote election participation: associational engagement and institutional trust.
Multiple forms of associational engagement exist in democratic systems. Some of these forms are churches, trade unions, clubs, foundations, occupational and professional associations, and the like, which operate under the general umbrella of non-profit organizations (NPOs) or non-governmental organizations (NGOs). These NPOs and NGOs are structures of formal networks spanning group and organizational boundaries, thus providing what Putnam has called “bridging social capital” to reinforce social cohesion and integration at societal levels [42, 70]. Consequently, the higher level of associational engagement individual citizens have, the more network channels they are connected to society at large, and the stronger political awareness and interest they maintained, thus leading to greater participation in public elections [48, 71, 106].
Institutional trust is another relational mechanism through which citizens are motivated to participate in public elections under democracy. Unlike interpersonal trust that promotes cooperation, collaboration, and mutual support between people in dense networks, institutional trust goes beyond personal worlds and close-knit communities to establish citizens’ faith in the state and nonstate institutions through the professionalization of organizational agents, such as government staff, court judges, police officers, and other organizational players of societal significance [22, 28]. When these institutional agents do their jobs professionally, their behaviors are guided by the 4-A hallmarks of professionalism: Abstract knowledge of specialization, Autonomy to use the knowledge to serve clients from whom to win recognition, respect, or Authority, and Altruism in the form of common concern toward all citizens in society [38]. When a high level of institutional trust results from the professionalization of institutional agents, it helps to maintain the public sphere by decreasing organizational costs [77], minimizing operational obstacles [81], and reinforcing citizens’ civic involvement [70], thereby promoting people to take part in the election and other political activities [67].
Although China’s grassroots elections are operated under a one-party rule, Chinese scholars are encouraged to apply Western relational explanations to the interpretation of patterns of election participation in their own country. For example, Hu [39, 40] found that variables of associational engagement had a positive effect on election participation in both rural and urban areas of Fujian province, a frontier region that pushed forward market reforms and local electoral reforms. Chen and Lu [19] conducted a quantitative study of 144 communities in three cities (Beijing, Chengdu, and Xi’an), showing that residents’ participation in voluntary groups and charitable organizations was positively associated with self-governance in urban communities. Sun [83] found in his Shanghai survey that engagement in social organizations and associations played a significant role in promoting residents’ participation in community leadership elections. Finally, many studies show that institutional trust in China is oriented toward the trust in CCP organization and government at various levels of administrative hierarchy [84], and such institutional trust is positively associated with political participation [104], including participation in grassroots elections in both rural and urban areas [102].
Research Puzzles
Although the relational perspective learned from Western research tradition has shown some explanatory power in the study of China’s grassroots elections, three research puzzles have not been attended by Chinese scholars. First, no empirical studies on China’s grassroots elections have examined the extent to which informal social networks affect election participation independent of the effect of variables of associational engagement and institutional trust, especially after 2012. Prior studies in China mostly presented evidence either from case studies about the role of clan networks in rural villages and that of neighbor networks in urban communities [33, 87, 92], or from household surveys about the effect of standard variables of associational engagement and institutional trust that are borrowed from Western research in both rural and urban elections [39, 40, 83, 84]. Note that all these studies were conducted prior to 2012, the year when Xi Jinping became China’s new paramount leader. Capitalizing on the 2017 CGSS data, the present research is aimed to assess the direct impact of informal networks on people’s participation in elections after 2012.
Second, prior Chinese studies lack a theoretical framework about the interplay of three different relational mechanisms in grassroots election participation. These relational mechanisms are informal networks, associational engagement, and institutional trust. Prior studies have, by and large, separated these relational mechanisms with controversial views and findings. Theoretically, most Chinese scholars considered informal networks of personal and social ties as what Putnam has called “bonding social capital” [70], whose core role was to maintain and enhance the interests of network members from within [68], not so much relevant to the benefits of public activities beyond personal networks, such as the election of public officials [19]. Empirically, quantitative survey analysis often produced findings that informal networks had no positive impact on voting participation [39, 40, 83], even though such impact was revealed in qualitative studies based on fieldwork [33, 37, 87, 92]. In the present research, we will attempt to formulate a theoretical framework to consider the interrelationship of the three relational mechanisms and examine their joint effects on election participation.
Third, existing quantitative studies mostly use the simple linear regression method to estimate and confirm the causal effect of personal networks on voting participation. However, this kind of method cannot control the interference brought about by the potential or unobservable confounding factors (such as one’s personality, values, etc.), so the statistical results may face the estimation bias caused by omitting variables [20, 52]. Furthermore, this kind of method cannot solve the estimation bias caused by mutual causality, because not only personal networks could affect one’s associational engagement and formal networks, but also the situation of formal networks could reversely influence the establishment of personal networks, for instance, some members from a social organization becoming one’s friends during the interacting process [49, 108]. How to solve these two endogenous problems to reduce and even eliminate the bias when estimating network effect is another challenge we confront in this study.
Theoretical Analysis and Research Hypotheses
Figure 1 displays a causal diagram of our theoretical framework within which three relational mechanisms interact with one another to jointly affect election participation. We place “personal networks” in the far left of the diagram because all human actors are born into and grow up in personal networks before being able to develop the formal networks of associational engagement and institutional trust. Logically, the influence of personal networks on voting participation can be divided into two modes of path. One is the direct path in which the status-quo of personal networks directly affects individuals’ voting participation. The other is a two-route indirect path, with one route going from personal networks to formal networks of associational engagement and to voting participation, and the other route going from personal networks to institutional trust and to voting participation. Our theoretical analysis in this section is developed to elaborate on the direct path and the two indirect pathways through which Chinese citizens are facilitated to participate in grassroots elections. We will propose three sets of research hypotheses, with each set of hypotheses being about the causal effect of relevance to the corresponding pathway.
Fig. 1
A Theoretical Framework (+ for positive effect, - for negative effect)
×
![]()
The Direct Path from Personal Networks to Voting Participation
We argue that personal networks play a positive role in China’s grassroots election participation through two critical mechanisms. The first is the mechanism of transmitting information. In the actual election process, officially disclosed information about candidates is usually relatively limited, and the formal information transmission channels are often not sound or convenient. Thus, individual voters need to make use of informal ways (such as clan networks, personal contacts, etc.) to access information about election policies, voting procedures, and candidates’ backgrounds and qualifications [33, 37, 41, 63, 82, 87], which will improve voters’ knowledge about the election, then promoting the participation in elections. In social democratic countries, personal networks have also been found to be useful for people to obtain relevant information, required resources, and various supports [30, 49], all of which would enhance voters’ interest in participation, reduce uncertainty and other costs, and facilitate individuals taking part in different kinds of public activities [80, 108].
The second is the mechanism of emotional mobilization. In Chinese society, this mechanism is typically embodied in the strong personal networks [26, 82, 87, 93], because these networks are full of emotional elements and reciprocity norms, such as favor, obligation, and gratitude [8, 43]. The mobilization of these emotional resources through personal networks will encourage people to participate in elections even if they may not be willing to do so in the first place [33, 87, 92, 94]. This trend was also observed in the pioneering research of Snow et al. in the United States [79]. Later studies have confirmed that emotional elements contained in personal networks of strong ties bring in great impact on individuals, making them to change from not participating to participating in social movements [47, 59, 60], from not voting to voting in elections for public officials [85], and from disapproving to approving of social perceptions [66].
What kinds of personal networks function well to make these two mechanisms effective for voting participation in China? We pay much attention to two dimensions of personal networks. The first is the resource dimension. Inspired by Lin’s social resource theory [53], informational and non-informational resources are embedded in and can be mobilized from personal networks of contacts, and these “social” resources are beneficial for actors’ instrumental and expressive actions. We argue that the more network resources available from one’s personal networks, the greater one’s ability to diminish obstacles arising from information asymmetries and other problems during election processes, and the higher probability for one’s participation in elections. The second is the interaction dimension. Personal networks are dynamic and functional because they facilitate social interaction between and among network members on personal levels, as demonstrated in the study of discussion networks [58], daily contact networks [27], social eating networks [9, 24], and Chinese guanxi networks [43, 96]. We argue that the more interactions one engages with network members, the greater emotional density one maintains with them, the greater encouragement, pressure, and support one receives from them, and the higher probability for one being mobilized to take part in elections. Therefore, we hypothesize:
Hypothesis 1
Personal networks have positive direct effect on people’s participation in grassroots elections. To be more specific,
Hypothesis 1-1:
The richer resources embedded in personal networks, the more effective they are to promote individuals to participate in elections.
Hypothesis 1-2:
The more interactions one has with members of personal networks, the more effective they are to promote individuals to participate in elections.
The Indirect Path Involving the Mediating Role of Formal Networks
Although the network direct effect has already attracted some scholars’ attention, there is little systematic empirical analysis to investigate its indirect effects. Our theoretical framework points to two indirect pathways through which personal networks affect voting attendance, with one of them involving the mediating role of formal networks in the causal chain of “Personal networks → Formal networks → Voting participation.”
Let us focus on the latter half of the causal chain first. As indicated in our literature review, the causal effect of formal networks on voting participation has been well-established in prior studies in China and other countries. That is, in all these countries, formal networks of associational engagement are an important type of bridging social capital, which reinforces social cohesion and integration at societal levels, thus facilitating public participation [70, 71]. Even in African countries (such as Botswana, Namibia, and Zambia), a recent study reveals the same pattern about the effects of formal networks on election participation [18].
Next, let us discuss the first half of the causal chain. We argue that the positive effect of personal networks on formal networks is made possible through two underlying logics. One is the resource logic: People with larger personal networks tend to possess more information, funds, and other resources [10, 15, 53]. These resources will empower actors to contribute to formal networks of associational engagement, thus having a high likelihood to participate in and even build formal networks as evidenced in prior studies in China [15, 90, 100]. The other is the interaction logic: People who are socially active in their personal networks tend to have more emotional strong and close-knit networks, all of which will provide more spiritual supports, driving forces, and opportunities for actors to achieve instrumental and expressive goals, such as participating in formal networks of associational engagement for public interests, as shown in the study of Chinese personal networks and their roles on civic engagement in environmental movements [11, 25].
Combining our arguments about the two parts of the causal chain, we hypothesize that:
Hypothesis 2
Personal networks have positive indirect effects on people’s participation in grassroots elections. To be more specific,
Hypothesis 2-1:
The richer resources embedded in people’s personal networks, the greater their potential to contribute to and participate in formal networks, which in turn promote them to participate in elections.
Hypothesis 2-2
The more interactions people have with personal contacts, the more supports and opportunities they gain for joining formal networks, which in turn promote them to participate in elections.
The Indirect Path Involving the Mediating Role of Institutional Trust
The causal chain involving the mediating role of institutional trust is restated here: “Personal networks → Institutional trust → Voting participation.” The empirical studies we have reviewed demonstrate that the second half of this causal chain is tenable: the higher the level of people’s trust in governments and other official departments, the more conducive it is for them to take part in political public activities (like the grassroots elections). However, the first half of this causal chain contains a negative correlation between people’s personal networks and their institutional trust, and in the Chinese context we offer two explanations.
The first explanation points to the role of informal network resources to substitute for formal channels of resource mobilization. This explanation is in line with the research on the informal economy, in which the absence of formal support systems makes informal social networks the indispensable mechanism for initiating and operating the informal sector in both developed and developing countries [5]. Empirically, the informal economy is relatively larger, stronger, and more sustainable in developing countries whose citizens have lower levels of institutional trust and greater utility of personal networks than in developed countries [86]. This is also true in China, where guanxi networks of personal ties are effective informal channels of resource mobilization and acquisition before and after market-orientated reforms [12, 96], which could reduce people’s dependence on official departments and lower people’s institutional trust [34, 51].
The second explanation is that dense interactions among close confidants within personal networks tend to lower people’s institutional trust for three reasons. One reason is that informal interactions are both to enhance the emotional density within personal networks and, at the same time, to narrow the range of trust in others or official agents beyond one’s social circle [23, 28]. Another reason is more associated with Chinese guanxi culture, in which formal and institutionally based relations are inclined to be personalized through the logics of face and favor [43, 46], making it hard to establish institutional trust for people more active in personal networks [55]. The third reason is relevant to an authoritarian regime, in which deep interactions at social eating occasions or in private tend to generate conversations about political secrets, bad images, or covered scandals of state officials, significantly lowering people’s institutional trust [20, 107].
For all of these explanations, we propose the last set of hypotheses as follows:
Hypothesis 3
Personal networks have negative indirect effect on people’s participation in grassroots elections. To be more specific,
Hypothesis 3-1:
The richer resources embedded in people’s personal networks, the lower their institutional trust, which will in turn hinders them from participating in elections.
Hypothesis 3-2:
The more interactions people engage with members of their personal networks, the lower their institutional trust, which in turn hinders them from participating in elections.
Research Design
Data
The CGSS (Chinese General Social Survey) dataset collected in 2017 is analyzed to test the hypotheses. Launched in 2003, CGSS is an annual survey based on nationally representative samples of households in China, and its public data archive has since been widely used in international scholarly publications [14]. The CGSS2017 has a total sample size of 12,582, with one-third of the respondents (4,219) randomly selected to answer the questions in an International Social Survey Programme (ISSP) module about social networks and social resources [72], from which our measures of personal networks are obtained. Our other measures are obtained from other modules of the 2017 CGSS data. Excluding the respondents (299) ineligible to vote in grassroots elections, the number of respondents for the following empirical analysis is 3,9202.
Variables
The dependent variable is an event variable about whether the respondents participated in grassroots elections. The survey respondents were asked the following question: “Did you participate in the last village/community committee election?” Accordingly, a dummy variable was constructed to measure voting participation (1–Yes and 0–No).
The independent variable is respondent’s personal networks with two measures. The measure of total network resources was from the ISSP module, based on an internationally recognized battery of “position generator” [54]. Specifically, among ten occupations for each of which the respondents were asked if they had family members, relatives, friends, and acquaintances working in each of them. These kin and non-kin relations were defined in this study as the members of the respondent’s personal networks. According to the ISEI score of each occupation following the work of Ganzeboom [29], we identified for each respondent his/her personal network size (total network members), network ceiling (the highest ISEI score among network members), and total ISEI scores (the sum of ISEI scores of all network members). These three network indices were analyzed through the PCF (principal-component factor) method, and a common factor representing the total amount of network resources was automatically extracted. As shown in Table 1, our measure of network resources is a standardized Z-Score with the range from -1.48 to 1.85. Substantively, the larger the value of the measure, the richer the network resources the respondent has.
Table 1
Descriptive Statistics of All Variables (N = 3,878)
Variable | Meaning/Range | Percentage/Mean | S.D. |
|---|---|---|---|
Voting participation | No | 50.95% | |
Yes | 49.05% | ||
Total network resources | [-1.48, 1.85] | 0 | 1 |
Social eating frequency | [0, 8] | 3.48 | 1.94 |
Formal networks | [0, 12] | 1.21 | 2.07 |
Institutional trust | [0, 10] | 6.41 | 2.50 |
Gender | Female | 51.57% | |
Male | 48.43% | ||
Age | 51.16 | 16.74 | |
Marital status | Unmarried | 10.06% | |
Married | 89.94% | ||
Political identity | Non-CCP members | 88.99% | |
CCP members | 11.01% | ||
Education level | No education | 11.53% | |
Primary school | 22.46% | ||
Middle school | 28.34% | ||
High/vocation high school | 18.33% | ||
College and above | 19.34% | ||
Social class | lower class | 17.66% | |
Lower middle class | 33.06% | ||
Middle class | 42.60% | ||
Upper middle class | 6.06% | ||
Upper class | 0.62% | ||
Political efficacy | 2.79 | 1.16 | |
Residential area | Rural | 36.33% | |
Urban | 63.67% | ||
Region | East | 44.53% | |
Middle | 31.41% | ||
West | 24.06% |
The measure of social eating frequency also comes from the ISSP module. While social eating has been widely recognized as an important occasion of social interaction among kin and non-kin contacts in all civilizations around the world [24], the ISSP module included a questionnaire item from prior research of China [9]. This item asked the respondent the following question: “How often do you go out to eat or drink with three or more friends or acquaintances who are not family members?” Eight response categories were provided: (1) daily, (2) several times a week, (3) once a week, (4) two to three times a month, (5) once a month, (6) several times a year, (7) less often, and (8) never. For our measure of social eating frequency, we reversed the coded numbers to recognize the ascending order of value change in the frequency of social eating participation, making a larger value to indicate more informal interactions a respondent had with his/her network members through social eating. An alternative, continuous measure would be obtained by converting the response categories into days of social eating on a yearly basis, and this measure has generated the same results about the significance of social eating interaction [13].
Two mediating variables are formal networks and institutional trust. Formal networks refer to the networks of associational engagement or participation in various forms of social groups and social organizations. The 2017CGSS recognized three categories of social organizations and asked the respondents to check all that applied to them: (1) interests, sports, or cultural groups, (2) politically related groups or associations, and (3) charitable or religious organizations. For each category, five responses were provided: 0–Never participated, 1–Only once last year, 2–Several times last year, 3–One to three times per week, and 4–Several times per week. When all three items were added up, we obtained a Likert scale about the extensity of one’s participation in formal networks, with a higher value to indicate that one’s formal networks were more widespread.
With respect to institutional trust, this concept can be measured with a long list of institutions in which citizens have varying levels of trust [70]. In the Chinese context, however, the institutional trust that is closely related to grassroots elections is the varying levels of trust that residents have in local governments and relevant official departments having jurisdiction over their rural villages or urban communities, such as township/county/city governments, police stations, and courts. CGSS2017 only had one questionnaire item that was mostly matched to our study: the level of trust respondents had in local courts, which was evaluated on a scale from 0 to 10, with a larger value to denote a higher level of institutional trust. Existing relevant studies have demonstrated that institutional trust in local governments, police stations, procuratorates, to name just a few, is highly consistent with that in courts [20, 107], which gives us the confidence in using this single item to measure institutional trust3.
For regression analysis to be presented shortly, we include several control variables that are of influence to individual voters and therefore need to be included in the model to achieve the accuracy of network effects [20, 52]. These are gender, age and its square term, marital status, political identity, education, social class, political efficacy, residential area, and region. Among them, political efficacy is measured by the question of whether respondents think they have the right to care and discuss what the government is doing.
Methods and Models
We choose GSEM (generalized structural equation modeling) for subsequent statistical analysis and hypothesis testing. The main task is to estimate the mediating effects, which are the formal networks (positively) and institutional trust (negatively) that translate the effects of the two personal network measures on the voting participation. Since our dependent variable is a dichotomous variable, we cannot use the traditional SEM (structural equation modeling) but instead use GSEM as the chosen method to perform the model estimation [2, 62].
The specific analyzing process is as follows. The first step is to run a base model of GSEM, in which only one independent variable (total network resources or social eating frequency) and control variables are included to estimate the total effect of one personal network measure on voting participation via Logit-model (logistic regression model). The second step is to run a full model of GSEM with two mediating variables simultaneously included in. The structure of the full model is shown in Figure 2. Accordingly, the OLS (linear regression model) is used to estimate the effect of one independent variable on each of the mediating variables, while the Logit model is used to estimate the effects of the independent variable, control variables, and two mediating variables on dependent variable. As a result, the direct effect as well as the two paths of mediating effect of each personal network measure could be obtained synchronously through the full GSEM model.
Fig. 2
Variables and Their Correlations in Full GSEM Model
×
![]()
Results of GSEM and Robustness Check
Descriptive Analysis
In the following analysis, 42 samples with missing values in any of the above variables are excluded, so the number of valid samples is 3,8784. The descriptive statistical results are displayed in Table 1. Briefly, the turnout rate of election participation is 49.05%. In 2005CGSS, this was 43.8% [95], indicating that people’s participation in grassroots elections has increased a bit over the past fifteen years, and the trend has begun to stabilize in recent years. As to network resources, the common factor extracts 89.47% of the total variance of the three indices and all factor loadings are higher than 0.89, which means that the extracted common factor is very effective and representative. In the end, ten people are found to be older than 90 years, and the results are consistent even though they are removed from the valid samples.
Results of GSEM
To improve the comparability between different regression coefficients, we firstly use Z-Score to standardize the independent and mediating variables, and then linearly convert them into variables ranging from 0 to 10 using a unified formula ((specific value-minimum)/(maximum-minimum) *10), and finally take the form of Natural Logarithm on them into the GSEM5.
The results of GSEM analysis are shown in Table 2. First of all, as to the total effect, the results of Base Model 1-1 show that the odds ratio of individuals participating in the election will significantly increase by 10% if the total network resources increase by 1 unit, while the results of Base Model 2-1 show that the odds ratio will significantly increase by 31% if the social eating frequency lifts by 1 unit, indicating that personal networks as a whole could greatly promote Chinese people to participate in grassroots elections.
Table 2
Results of Base and Full GSEM Models
Base Model 1-1 | Full Model 1-2 | Base Model 2-1 | Full Model 2-2 | Model 3 | |
|---|---|---|---|---|---|
Total network resources | 0.10* | 0.059 | / | / | 0.031 |
Social eating frequency | / | / | 0.31*** | 0.23** | 0.22* |
Formal networks | / | 0.32*** | / | 0.30*** | 0.30*** |
Institutional trust | / | 0.26*** | / | 0.27*** | 0.27*** |
Total network resources◊ Formal networks | / | 0.26*** | / | / | 0.18*** |
Total network resources◊ Institutional trust | / | -0.033*** | / | / | -0.026* |
Social eating frequency◊ Formal networks | / | / | / | 0.44*** | 0.32*** |
Social eating frequency◊ Institutional trust | / | / | / | -0.043** | -0.026 |
Male | -0.038 | -0.027 | -0.065 | -0.049 | -0.051 |
Age | 0.13*** | 0.14*** | 0.14*** | 0.14*** | 0.14*** |
Age^2/100 | -0.10*** | -0.10*** | -0.10*** | -0.10*** | -0.10*** |
Married | -0.083 | -0.054 | -0.071 | -0.012 | -0.019 |
CCP members | 0.30* | 0.17 | 0.29* | 0.17 | 0.17 |
Primary school | 0.51*** | 0.52*** | 0.51*** | 0.51*** | 0.51*** |
Middle School | 0.46*** | 0.48*** | 0.44*** | 0.46*** | 0.46*** |
High/vocation high school | 0.36* | 0.35* | 0.34* | 0.33* | 0.32* |
College and above | -0.041 | -0.12 | -0.071 | -0.15 | -0.16 |
Lower class | -0.15 | -0.080 | -0.11 | -0.049 | -0.045 |
Lower middle class | -0.21* | -0.17* | -0.20* | -0.16! | -0.16! |
Upper middle class | -0.21 | -0.28! | -0.23 | -0.29! | -0.29! |
Upper class | -0.88* | -0.92* | -0.87* | -0.91* | -0.91* |
Political efficacy | 0.14*** | 0.12*** | 0.14*** | 0.12*** | 0.12*** |
Urban | -0.99*** | -1.01*** | -1.01*** | -1.02*** | -1.02*** |
Middle | -0.42*** | -0.42*** | -0.42*** | -0.42*** | -0.42*** |
West | 0.035 | 0.021 | 0.035 | 0.021 | 0.021 |
Constant | -3.91*** | -4.51*** | -4.32*** | -4.86*** | -4.88*** |
Second, in terms of direct effect, the results of Full Model 1-2 show that the odds ratio of voting participation will increase by 5.9% for 1 unit increase in the total network resources (though it fails the significance test), while the results of Full Model 2-2 show that 1 unit increase in the frequency of social eating will increase the odds ratio of voting participation significantly by 23%. These results altogether manifest that the first three hypotheses are mostly confirmed. In particular, the positive direct effect of social eating frequency is more significant and powerful than that of total network resources. What is the reason behind this? In the previous theoretical analysis, we point out that personal networks can promote the voting participation mainly through the mechanisms of information transmission and emotional mobilization, although the network resources dimension can act well to facilitate the transmission and acquisition of relevant information during the voting process, the dimension of informal interaction based on social eating will work greatly not only on transmitting and sharing information but also on emotional exchanges between network members.
With respect to the mediating effects of personal networks on voting participation, results of Full Model 1-2 show that formal networks and institutional trust both have a significant positive impact on voting attendance (coefficients are 0.32 and 0.26, both significant at 0.001 level). Furthermore, the variable of total network resource has a significant positive impact on formal networks (coefficient is 0.26 and significant at 0.001 level) and a significant negative impact on institutional trust (coefficient is -0.033 and significant at 0.001 level). Thus, using the command “nlcom” of the Stata software [2], we can get the mediating effect value of 0.084 (significant at 0.001 level) for the path “Total network resources → Formal networks → Voting participation” and that of -0.0085 (significant at 0.05 level) for the path “Total network resources → Institutional trust → Voting participation,” indicating that 1 unit increase in total network resources will raise the odds ratio by 8.4% through formal networks while decreasing the odds ratio by 0.85% through institutional trust. These results support the last six hypotheses. The results of Full Model 2-2 are the same as those in Full Model 1-2. Here, using the same command “nlcom,” we can get the mediating effect value of 0.13 (significant at 0.001 level) for the path “Social eating frequency → Formal networks → Voting participation” and that of -0.011 (significant at 0.05 level) for the path “Social eating frequency → Institutional trust → Voting participation,” once again supporting the last six hypotheses.
Considering that the network resources dimension and the informal interaction dimension are mutually independent at the theoretical level, but strongly correlated empirically (Pearson correlation coefficient is 0.43, significant at 0.001 level), we put them together into the full GSEM model to obtain the net impacts as shown in Model 3. As to direct effects, the pattern is almost the same as before, indicating that informal interaction is more powerful than network resources in promoting individuals to participate in elections. In terms of indirect effects, the positive and negative mediating effects of network resources still exist significantly. Meanwhile, the positive mediating effect of social eating frequencies also exists significantly, but the significance of its negative mediating effect has reduced a lot. These results on the whole support the tentative conclusions we have drawn out in the previous paragraph. In addition, we find that network resources have a stronger negative impact on institutional trust than that of social eating, implying that the mechanism of “informal resources substituting for formal pathways” is a more powerful explanation than that of informal social interaction as far as their negative impacts on institutional trust are concerned.
Finally, control variables also exhibit meaningful results. There is no significant difference between males and females in voting participation. The relationship between age and voting participation is an inverted “U” shape, and the inflection point is about 65-70 years old, so people around ages 65-70 are the most frequent participants in elections. CCP members have a higher probability to vote than non-CCP members. An inverted “U” shape relationship exists between people’s education level and voting participation; specifically, residents with primary school level are the most active in voting, and the second are residents with middle school level, and then are residents with high school or vocation high school level, whereas residents without formal education or with college level education and above have the lowest probability in joining the election. In terms of social class, there is also existing an inverted “U” shape on the whole, for the middle class performs best in the election, whereas the upper class has the lowest propensity to participate. People with higher political efficacy are more likely to participate in elections. Urban residents are less likely to participate in elections than rural villagers, and people in the East and West regions have higher probability to join the election than people in the Middle region.
Robustness Check
The interference brought by the two endogenous problems as mentioned above is likely to cause bias in the estimation results. To confront this challenge in a cross-sectional survey data analysis, one of the optimal solutions is to use the IV (instrumental variable) method, whose key principle is: To find out the instrumental variable for the independent variable (such as personal networks), which requires that the instrumental variable is strongly related to the independent variable, but has nothing to do with the residual term of the dependent variable (such as formal networks), making the IV completely exogenous to the dependent variable (i.e., fulfilling the exclusion condition); thus, two steps of estimation are made to get the real causal effect6[20, 61].
Since the influence of formal networks and institutional trust on voting participation has been widely proved in prior research around the world, the IV analysis below will focus on confirming the causal effect of personal networks on voting participation, formal networks, and institutional trust. We choose “the frequency of daily communication of the respondent with her/his siblings” as the IV for personal networks for two main reasons.
First, an ideal IV for personal networks is the number of respondent’s siblings. The number of siblings is closely related to personal networks, because more siblings always mean to be more conducive for an actor to establish and develop informal networks; meanwhile, the number of siblings is not directly affected by the current situation of voting participation, formal networks, or institutional trust, making it completely exogenous to these factors. In addition, existing studies in Chinese society and other countries have demonstrated that the number of siblings is a good instrumental variable for personal networks [7, 21]. Though this ideal IV isn’t measured in CGSS2017, fortunately the data includes the frequency of daily communication of the respondent with her/his siblings. According to our experiences, in Chinese society a person with more siblings tends to have more informal contacts with siblings and other relatives, which means that the ideal IV could be strongly correlated with the frequency IV, so we could further reason out that the frequency IV may have strong correlations with the independent variable.
Second, whether the frequency IV is indeed strongly correlated with independent variable can be checked out through the “weak IV test”. Specifically, voting participation is a dummy variable, so the IV-Probit model is selected, and the “Wald endogeneity test” can be used to test whether the independent variable is endogenous. Formal networks and institutional trust are continuous variables, so the two-stage least squares IV (2SLS-IV) model is chosen, and “Durbin (score) Chi2” and “Wu-Hausman F” are used to check whether the independent variable is endogenous. If the independent variable is confirmed to be a sort of endogenous variable, we should choose the IV models which could provide more accurate estimation results. In this context, the “first-stage F value test” can be used to check the problem of weak IV, which could show us the actual situation of the correlation between IV and independent variable [3, 4, 16].
The results are shown in Table 37. As to the causal effect of total network resources or social eating frequency on voting participation, neither of the IV-Probit models have passed the Wald endogeneity test (P = 0.24, P = 0.30), which demonstrates that the independent variable must be regarded as an exogenous variable, that is, it tends not to be directly affected by voting participation. In this case, we should choose and accept the results from simple regression (Logit model) rather than the IV-Probit model [3, 4]. The results of Logit model show the two personal network measures both have a significant positive influence on voting attendance (coefficient is 0.10 and 0.31, and significance level is 0.05 and 0.001, respectively), thus we have sufficient confidence that personal networks indeed have positive causal effects on people’s voting participation.
Table 3
Results of Simple Regression and IV Models
Voting participation | Formal networks | Institutional trust | ||||
|---|---|---|---|---|---|---|
Logit | IV-Probit | OLS | 2SLS-IV | OLS | 2SLS-IV | |
Total network resources | 0.10* | 0.50 | 0.16*** | 0.83*** | -0.029** | -0.15 |
(Control variables) | Yes | Yes | Yes | Yes | Yes | Yes |
Wald endogeneity test | P = 0.24 | / | / | |||
Durbin (score) Chi2 | / | P = 0.0001 | P = 0.37 | |||
Wu-Hausman F | / | P = 0.0001 | P = 0.37 | |||
The first stage F value | / | 24.71 | / | |||
Social eating frequency | 0.31*** | 0.85 | 0.29*** | 1.40*** | -0.047** | -0.25 |
(Control variables) | Yes | Yes | Yes | Yes | Yes | Yes |
Wald endogeneity test | P = 0.30 | / | / | |||
Durbin (score) Chi2 | / | P = 0.0001 | P = 0.36 | |||
Wu-Hausman F | / | P = 0.0001 | P = 0.36 | |||
The first stage F value | / | 25.42 | / | |||
The second step is to check the causal effect of personal networks on formal networks, and the results of 2SLS-IV models show that they all pass the Durbin (score) Chi2 and Wu-Hausman F endogeneity test (both significant at 0.001 level), which mean the independent variable (total network resources or social eating frequency) should be regarded as an endogenous variable (that is, it is likely to be directly affected by formal networks), in other words, there is indeed an endogenous problem resulted from mutual causality, so we must choose and accept the estimation results of 2SLS-IV rather than OLS, showing that both measures of personal networks have a significant positive impact on formal networks (coefficient is 0.83 and 1.40, and both significant at 0.001 level). Meanwhile, the first stage F value is 24.71 and 25.42 respectively, exceeding the empirical threshold value of 16, so it has passed the weak IV test [3, 4, 20], indicating that the frequency of contact with siblings is indeed strongly related to an individual’s current personal network status. Combining these two aspects, it can be concluded that the positive causal effect is indeed subsistent.
Finally, in terms of the causal effect of personal networks on institutional trust, results of 2SLS-IV models show that none of them pass Durbin (score) Chi2 or Wu-Hausman F test (P = 0.37, P = 0.36), which means that the independent variable (total network resources or social eating frequency) must be treated as an exogenous variable (that is, it is unlikely to be directly affected by institutional trust). In this case, we should choose and accept the results of OLS rather than 2SLS-IV [3, 4]. As OLS results showing both measures of personal networks have a significant negative impact on institutional trust (coefficient is -0.029 and -0.047, and both significant at 0.01 level), confirming that the negative causal effect indeed exists.
Discussion
In the 1980s, China began to implement democratic elections and self-governing reforms in rural villages and urban communities. This study focuses on the direct and indirect effects of personal networks on individuals’ voting participation. Confined to CGSS2017, we measure the resource and interaction dimensions of personal networks and simultaneously estimate their direct and indirect effects on election participation through the use of GSEM models. Furthermore, to solve the two kinds of endogeneity problem and the latent bias in the GSEM estimation, the IV method (including IV-Probit and 2SLS-IV models) is used to fulfill the analysis of robustness check.
Based on the analytical results, we now draw three main conclusions. First, personal networks have positive direct influences on Chinese people’s participation in grassroots elections around the period of 2012 to 2017. This is supported by the evidence of direct causal effects, indicating that personal networks are not a hindrance but a facilitator in the voting process, which is beneficial and effective for relevant information transmission and emotional mobilization, thus promoting people joining into elections. Second, the indirect influences of personal networks on voting participation are not monolithic, having relatively independent or even completely inverse effects through different pathways. Among them, one positive mediating path is to promote people joining into elections by improving people’s engagement in social organizations and the construction of formal networks; meanwhile, one negative mediating path is to lower down the level of institutional trust, in turn inhibiting individuals taking part in voting campaigns. Third, the degree of positive influence is more powerful than the negative one, manifesting that personal networks (especially informal interactions between individuals) are an important force which could help local residents to build and engage in social connections and organizations, to construct formal networks, and eventually to push forward people participating in voting activities and public affairs.
In view of this, to improve the participating situation of Chinese public in grassroots elections, relevant institutional arrangements and policy design must consider providing more space, platforms, and opportunities for citizens to communicate and interact, thus helping them establish more developed personal networks. Meanwhile, the barriers and costs of institutional arrangements when people seek and acquire resources (especially public resources) should be reduced or removed as much as possible. If the formal ways and official channels for resource acquisition become more convenient and sounder, the degree of people’s dependence on informal ways (like personal networks) will descend; consequently, the power of the mechanism of “informal resources substituting for formal pathways” will fall down. This sort of change will effectively decrease and even completely eliminate the negative impact of personal networks on institutional trust, driving personal networks to transform into a completely conducive force for voting participation and self-governance at grassroots levels.
Finally, three aspects are worthy to be explored in the future. The first aspect is to further analyze and confirm the causal relationship between personal networks and voting participation. Although efforts are made in this article, the casual analysis via IV method is preliminary due to the limited dataset; more accurate estimations should be obtained from further studies using ideal measurements of instrumental variables. The second aspect is to investigate the structural heterogeneity of personal network effects. China has a large population with huge internal variations in economic spheres, political arrangements, and other macro circumstances, so future research could pay more attention to investigating the heterogeneity of network effects between different groups, regions, and other structural conditions. The last aspect is to explore the change of personal network effects. A non-negligible social background is that Chinese society is in the process of all-round transformation. Just as mentioned before, the total CCP leadership has been strengthened since 2012; consequently, an increasing number of villages and communities have been making the CCP secretary the only candidate in the grassroots committee director election. This sort of change tends to make the elections more organized and closed, decreasing the strength of people’s votes in determining which candidates could win out eventually; then, more local residents would probably lose the interest and passion to take part in elections than before. Under this situation, informal networks of social relation may turn out to be more critical in elections because of their functions of information transmission and emotional mobilization. Is this theoretical judgement efficient to explain the change of network effects in different eras? How to explore and testify this judgement via an empirical approach? These questions are also worthwhile to be discussed and answered.
Conclusion
This study focuses on the influence of personal networks on Chinese people’s participation in grassroots elections. Analyzing CGSS2017 data via GSEM and fulfilling the robustness check through different IV models, we find that: Personal networks could directly promote individuals joining into the voting campaign through the mechanisms of information transmission and emotional mobilization; In terms of mediating effects, personal networks can not only promote the voting participation by improving people’s engagement in social organizations and the formation of formal networks, but also decrease people’s institutional trust level, and then hinders public taking part in the election; In comparison, the positive effects are more powerful than the negative ones, so on the whole, personal networks are an important conducive power for people’s participation in elections and the realization of effective autonomy at grassroots levels.
It is expected that research in the future could carry out more detailed empirical investigation on the theoretical mechanism and causal correlation between personal networks and voting participation, and conduct more analysis and discussion on structural heterogeneity of the network effect and its changes under Chinese society transformation. We believe that the answers to these questions will provide helpful inspiration and important reference for enriching the theoretical understanding of personal networks, deepening the reform of democratic elections, and improving the level of grassroots self-governance.
Acknowledgements
The data of CGSS2017 used in this article is from the Chinese General Social Survey collected by the China Research and Data Centre at Renmin University of China. We are grateful to the institution for providing data assistance, but are responsible for the content of this article.
Declarations
Conflict of Interest
The authors declare no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.