Social Class and Voter Turnout in China: Local Congress Elections and Citizen-Regime Relations

Transcription

1 688109PRQXXX / Political Research QuarterlyWang and Sun research-article2017 Article Social Class and Voter Turnout in China: Local Congress Elections and Citizen-Regime Relations Political Research Quarterly 2017, Vol. 70(2) University of Utah Reprints and permissions: sagepub.com/journalspermissions.nav DOI: journals.sagepub.com/home/prq Zhengxu Wang 1 and Long Sun 2 Abstract In research on China s state-controlled elections, whether pro-democratic and politically efficacious citizens vote or abstain more has dominated the debate. Analyzing an up-to-date data set, we find class status and citizens institutional environment command overwhelming power in determining voter turnout, while psychological and motivational variables show no significant impacts. Specifically, members of the middle class employed in the state sector are more likely to vote than other social groups, while their counterparts in the private and foreign sectors are far less inclined to vote. While in general, citizens show relatively high level of support for the regime, except among the state-employed middle class, this support is not translated into a higher turnout rate. Electoral behaviors of Chinese citizens, therefore, are heavily conditioned by an individual social group s particular relation with the state. Members of the Communist Party turn out to vote no more than other members of the society, revealing a critical institutional failure the system faces. Keywords voter turnout, elections in authoritarian regimes, China, state dependence, middle class Voter participation in state-controlled elections could indicate the level of public support or compliance that the regime commands. Voter abstention, however, often indicates the public s alienation from a regime. In this context, voter turnout in China s urban elections represents a highly promising subject for scholars interested in the Chinese regime. While 70 percent of China s population is expected to live in urban areas by around 2020, the field is seriously short of studies of electoral participation of urban Chinese. As a result, it has been difficult to place the China case into the rapidly growing body of literature on parties and electoral institutions under dictatorships (e.g., Gandhi and Lust-Okar 2009). A few pioneering studies have focused chiefly on subjective-psychological variables, while ignoring most of the structural and institutional variables widely considered important in determining voter turnout. This paper attempts to expand electoral studies within a postcommunist, authoritarian political framework (Chen and Zhong 2002; Gilison 1968; Karklins 1986; Roeder 1989; Shi 1999). Approximately one month following polling day in the City of Beijing s 2011 districtlevel people s congress election, we surveyed more than one thousand Beijing residents of voting-age, enabling us to analyze how structural, institutional, and psychological factors determine the election turnout or avoidance of China s urban citizens. By pointing to structural and institutional factors as the more powerful determinants of voter turnout, we help resolve a long-standing debate regarding subjective motivations for voter participation in China s local urban elections. Our findings will also show how China s political economy has conditioned the urban middle class by their economic dependence on the state, revealing some significant mechanics of the Chinese regime s rule. The paper is structured as follows. We first examine the existing studies of voter turnout and argue that the study of elections in China should include the structural and institutional factors that so far have been omitted. The Method section explains the data collection process as well as the operationalization and measurement of the main variables. The empirical results then follow, which show that first and foremost, it is the structural factors that really matter in Chinese urban voters participation or abstention in local people s congress elections. 1 Fudan University, Shanghai, China 2 Renmin University of China, Beijing, China Corresponding Author: Zhengxu Wang, School of International Relations and Public Affairs, Fudan University, 220 Handan Road, Shanghai , China. wangzhengxu@fudan.edu.cn

2 244 Political Research Quarterly 70(2) Only within a state-sector employment setting, in fact, does a person s identification with or alienation from the state matter for his or her electoral participation. We conclude with the implications of our findings for comparative studies of authoritarian regimes, political behaviors, and democratization. Voting in China: Civic Expression or State Alienation? While the studies of urban voting participation in China are relatively few, we do have the option to start with studies conducted on rural elections. One focus is on the effect on voter participation brought by the introduction of semicompetitive electoral rules. While it is reasonable to expect pro-democratic and more cognitively competent voters to abstain in a plebiscitary election, once there is a degree of competitiveness, things become more complicated. In village elections in Yunnan Province and other areas, for example, voters turn out to vote if they perceive there is a chance someone would lose (Landry, Davis, and Wang 2010; Manion 2008). Because competition, although limited in degree, means someone will receive low votes or lose the election, these elections do help link the preferences of the voters and the officials they elect (Manion 1996, 2014b). These elections can also serve an information purpose, revealing incompetent or unpopular officials (Manion 2008, 2014a; Saich and Yang 2003). Villagers would also participate because they feel that, given the power structure between the village Party Cell and the Village Committee, they have a stake in these elections as they get to choose the members of the latter institution (Oi and Rozelle 2000). This stake-holder mechanism, however, may not operate once we move up to examine elections at the county/district level, as these elections are concerned with a rather ceremonial, powerless (therefore stake-less) institution, that is, the local people s congress. Regarding urban elections, the few existing studies have debated whether the voters perceive participating in these elections as meaningful, given the potential costs of going to vote. Shi (1999) suggests that those with stronger democratic orientation tend to view these elections as an opportunity to promote China s democratization; therefore, they are more ready to participate. Similarly, an experiment showed that by framing the participation as an act to promote democracy, canvassing workers were able to mobilize more university students to turn out (Guan and Green 2006). Shi (1999) also found that citizens with a greater desire to punish corrupt officials, and those who strongly believe in their own ability to influence politics known as political efficacy in political psychology tend to vote in urban local congress elections. Chen and Zhong (2002), however, argue that voters view such elections as nondemocratic in nature and ineffective in punishing corrupt or incompetent officials; therefore, citizens with stronger democratic values and more confidence in their civic capability would tend to abstain. These two lines of thinking, in essence, represent a civic or democratic citizenship perspective and a statealienation perspective. The former sees that people with stronger democratic values, stronger internal efficacy, and a stronger civic duty to punish incompetent or corrupt officials and promote democracy will take actions and cast their votes. The latter perspective, by contrast, sees those who are democratically oriented, more efficacious, and more acutely aware of government failures as more alienated from the regime s elections. This debate (Manion 2010) has critical implications regarding how these elections are perceived by voters, and how they link up the citizens to, or alienate them from, the state. We will be able to update the field by replicating Chen and Zhong (2002) model with new data in this paper, to ascertain whether it is the democratic citizenship mechanism or the state-alienation mechanism that is at work. Our initial finding leads us to extend the inquiry into an important area untouched by existing studies. We argue that voting patterns must be understood by examining the structural and institutional factors that voters face. Structural and Institutional Factors Social class represents probably the most important structural factors in shaping voter behaviors. In a democratic setting, it is generally found that people with higher incomes and levels of education are much more likely to turn out to vote than people of lower socioeconomic and educational status (Leighley and Nagler 1992; Shields and Goidel 1997; Wolfinger and Rosenstone 1980). But in less developed countries or a less democratic environment, illiterate voters tend to vote in greater proportion, as their votes are cheaper to buy (Blaydes 2006). More educated citizens in the former Soviet Union were also found to avoid voting as their way to protest regime-controlled elections (Karklins 1986; Roeder 1989). These findings parallel the state-alienation perspective we identified above, which sees more educated and more endowed sections of the society as more likely to be alienated from the regime. This Soviet-study perspective of voter avoidance has greatly affected the pioneering studies of voter turnout in China (e.g., Chen and Zhong 2002). To view voter avoidance as protest is highly problematic, however. First, even in the Soviet Union, active political participation, in fact, is found to positively correlate with political efficacy. It appears that people of strong efficacy tend to participate more whether in a democratic or nondemocratic setting (Bahry and Silver 1990). Second, in the Chinese and many other cases, it is far from safe to assume that more

3 Wang and Sun 245 educated urbanites are more likely to reject the regime and demand for a regime change. Unpacking the relations between China s urban social groups and the state, we believe, holds the key to understanding voting participation in China s urban election. The Middle Class, the State, and Democracy Socioeconomic development, industrialization, and urbanization are generally believed to foment social changes that foster democratization (Boix and Stokes 2003; Lipset 1959). In this process, as famously represented in Barrington Moore s no bourgeois, no democracy line (Moore 1966), the middle class commands the central position as the champion of democracy. The human development theory of democratization argues that material affluence and a high level of education give rise to postmaterialist or self-expressive values, which demand democratic institutions (Inglehart 1997; Inglehart and Welzel 2005; Welzel, Inglehart, and Klingemann 2003). This emphasis on the middle class as the driving force of democratization, however, has been vigorously debated. First of all, the concept of the middle class is imprecise. If referring to capitalist or business elites, that is, business owners and top managers, which is what the term bourgeois originally means, the middle class often tends to support representative government and the protection of civil liberties, but to oppose the extension of political rights to lower classes. In many cases, it was the working class that fought for universal suffrage and brought democracy to modern societies (Bendix 1964; Marshall 1950; Rueschemeyer, Stephens, and Stephens 1992; Thompson 1963). 1 If the middle class refers to the urban, white-collar waged professions, its members view of democracy is often conditioned by their relationship to the state. In late developing countries, the state often has played a decisive role not only in promoting socioeconomic development but also in creating the waged middle class itself. Consequently, the newly created middle class often shares common interests with the state, resulting in its so-called state dependence (Gerschenkron 1962; Rueschemeyer, Stephens, and Stephens 1992). In fact, in countries such as Indonesia, Mexico, Saudi Arabia, South Korea, Thailand, and Brazil, the state-dependent middle classes tended to support the authoritarian regime because of the benefits flowing from the state. They also feared that democratization would empower the lower classes, threatening their own interests (Bellin 2000). In China s state-led economic development, the capitalist class and business elite are found to be attached to the state (Tsai 2005, 2007) and to support the regime (Chen and Dickson 2008). Regarding the mass middle class or the white-collar employees in China s urban economy (Tomba 2004), they are not showing unified pro-democratic tendencies as would have been predicted by the modernization theory. A key variable here seems to be the middle class s dependence on the state, as those socioeconomically more reliant on the state show lower support for democracy (Chen and Lu 2011). In fact, a look into China s class structure shows that its mass middle class is composed of two distinct groups. A public- or state-sector middle class includes the white-collar workers who are employed by state and party agencies, state-owned enterprises, and public (i.e., governmentsponsored non-for-profit) organizations. A private-sector middle class, by contrast, consists of white-collar workers employed by foreign firms and domestic private enterprises. Members of the public-sector middle class are more likely to be conditioned by their strong dependence on the state and therefore to be more pro-regime in attitude (Chen 2013). 2 People employed in the state or public sector, for example, benefited greatly from the privatization of statebuilt apartments, which were sold to the occupants at vastly discounted prices (Walder and He 2014). The private or non-state middle class, on the other hand, depends much less on the state for its welfare, and is less pro-regime. 3 A quick look at the voting rate distinguishes these two groups sharply our data show that 72 percent of the voters in the state-sector middle class voted in November 2011 in Beijing, while only 45 percent of those in the non-state sectors did so. One of our main theoretical concerns below, therefore, will examine the effect of state dependence on urban middle class s voting behaviors. Class Effect in China: The Institutional Layer While we expect state-sector middle class to hold stronger pro-regime attitudes, when it comes to electoral participation, there is more at play. Members of the public-sector middle class may turn out to vote not (or not only) because they identify with the regime, but because they are more subject to state mobilization during election time. The ruling Chinese Communist Party s various bodies are most actively present in state and public institutions, promoting awareness among and mobilization of voters both prior to and during an election. Plus, state employees are often registered in the precinct based on their work-unit, together with their coworkers. On voting day, it is highly likely peer pressure will play a role in moving a person to the voting booth if his or her coworkers are already going. Institutional factors and the workplace ecology can both work to increase turnout by members of the state-sector middle class. By contrast, employees in foreign or private firms are exposed to minimal, if any, formal mobilization in their workplace. Furthermore, many of them are registered in the precinct of their residential area, as non-state work-units are seldom made a precinct. Because of this, peer pressure for voting at the workplace should be much weaker.

4 246 Political Research Quarterly 70(2) Nonvoting by non-state middle-class members, therefore, may reflect their lack of information or the lower level of state mobilization or peer pressure they are exposed to. Our data reveal that, among the Beijing citizens who voted in the last local people s congress election, only 6 percent reported that they went to vote because others were going. This to a degree indicates that peer pressure is weak in both the state and the non-state sector. By contrast, non-state middle-class members were two and a half times more likely to cite no one informed me as the reason for nonvoting than members of the state-sector middle class (44% compared with 17%). This appears to show the relative effectiveness of mobilization and information efforts in the state sector state-sector employees are much more informed about the polling, while nonstate employees are not. Still, our data show a significant level of alienation among the state-sector middle class. Among nonvoters in the state-sector middle class, a significant portion (27%) reported that they just did not want to vote. To summarize, we hypothesize that two mechanisms make the state-sector middle class more likely to vote. First, state-sector middle-class members are economically and professionally dependent on the regime; therefore, they are more disposed to compliance when it comes to electoral participation. Second, at their workplace and in their experiences of daily living, they are subject to more mobilization and influence by state and Party agents than their private-sector counterparts. By contrast, members of the private-sector middle class are less likely to vote, both because they are less dependent on the state and because they are less subject to state mobilization. These structural and institutional mechanisms, we hypothesize, will account for the majority proportion of variations in urban Chinese voters tendency to participate in the election. These mechanisms, if proved, render the civic citizen vis-à-vis state alienation debate less meaningful, because it is mainly the voters socioeconomic and institutional status in the contemporary political economy of China that determines his or her turnout decision in these elections. Additional Factors: The Party and Political Interest The main objectives this paper pursues, therefore, are to separate the structural and institutional factors from the subjective motivations of voter turnout. In the following analysis, however, besides controlling the usual sociodemographic variables such as gender, age (and agesquare), and education, two additional variables will need to be present. These include an institutional variable, that is, the voter s membership in the Chinese Communist Party, and a cognitive-psychological variable, that is, the voter s political interest. As the key institution in China s political system, the Communist Party holds strong capacity in controlling the society, mobilizing political participation, and disseminating political information, among many other functions. Members of the Party are likely to identify much more strongly with the regime than nonmembers, and do so even after they have become capitalists or business managers (Chen and Dickson 2008). Shi (1997, 1999) found that Party membership greatly increases the likelihood of a person s voting. Including Party membership in our model represents a test of the viability of the Chinese political system. If, indeed, Party membership fails to show a significant effect on voter turnout, it would amount to a major institutional failure of the entire political system. A person s interest in politics is likely to have a structure-transcending effect on his or her participation in political activities. This appears to have been the situation in the former Soviet Union (Bahry and Silver 1990), that is, regardless of regime types, politically interested citizens are more active participants. In fact, this is related to the argument that education and voter turnout represent a correlation without causality: those who have a strong interest, or an inquisitive mind, more often turn out to vote, and are also more interested in pursuing a high level of education (Franklin 2004). Politically interested people are likely to engage in greater participation in elections, regardless of the institutional setting in which they find themselves. Local People s Congress Elections in China Before moving to the empirical part, it is necessary to explain the institutional settings of the elections under study here, that is, the local people s congress in China. This institution was first established in , with a hierarchical structure consisting of five different levels of People s Congress national, provincial, municipal, county, and township. Today in China, a total of 2,856 county-level people s congresses operate, including the district people s congresses of medium to large cities, with each consisting of 120 to 450 deputies, who are directly elected by the voters. 4 While the right for ordinary citizens to vote in the elections of local people s congresses was first codified by the election law of 1953, this right was not realized until much later. The 1953 law just provided for the direct election at the village, township, and city-district level people s congresses. The deputies to the National People s Congress and other levels of people s congress were elected by the deputies to the congress of the next lower level. Some reforms were undertaken as China emerged from Mao Zedong s rule. A 1979 election law for the People s Congress provided for direct election

5 Wang and Sun 247 Table 1. Socioeconomic and Institutional Backgrounds of the Sample and Voting Participation Rates across Different Groups. Party member Education level: 1: Middle school or lower 2: High school 3: College 4: University Employment groups: 1: Blue-collar workers 2: State-sector middle class 3: Private-sector middle class 4: Retired No Yes Total % in the sample % Voted Source Beijing Local Election Participation Survey (Conducted in December 2011, one month following the Election on November 8). of delegates to the county-level people s congresses, therefore, moving direct election up the hierarchy by one level. More importantly, a degree of competition is now allowed for these elections voters are now given a list of candidates that outnumbered the seats to be filled. 5 Furthermore, a more open nomination process was put in place besides state and Party organizations, voters can nominate candidates themselves if the nomination is endorsed by ten or more voters (Nathan 1985). The key institutional weakness of the People s Congress, nevertheless, remains its lack of political power. Dubbed a rubber stamp in the Chinese political system, the People s Congress obtains a largely ceremonial role as the nominal law-making institution, while the actual power of making laws resides in the Chinese Communist Party s. The People s Congress also elects or appoints state officials, but it does this by simply rubber stamping the list supplied to it by the Party Committee. While reforms since the 1980s have clearly expanded the openness and competitiveness of its elections, they have fallen far short of bringing meaningful changes to the nature of the people s congress. Therefore, how these electoral reforms have changed the nature of the local people s congress election and how citizens perceive them remain highly uncertain. It might be highly possible, for example, citizens simply abstain as they feel theirs are votes of a toothless institution. The empirical part of this paper further the study of these elections by examining how interested voters are in participating in these elections and what affect their decision to do so. Sample and Data Our data came from a post-electoral survey conducted in Beijing in December Because of political sensitivity, it was impossible to conduct the survey immediately before the polling day of November 8, Nor was it possible to conduct exit-polls on the day the ballots were cast. Therefore, we arranged face-to-face interviews within one month following the election. Our sample was designed to be representative of people aged eighteen to seventy in Beijing. We succeeded in interviewing a total of 1,318 respondents (a detailed note for the sampling process is given in Appendix A [online]). Excluding student respondents and rural-hukou holders, 6 we achieved a data set of 1,041 respondents for this paper. The first row in Table 1 reports the make-up of the sample in terms of party membership, educational level, and social class (definition below). Ours is a sample of Beijing residents and excludes the rural-hukou holders. Therefore, it is clearly more educated than the national average, with 28% holding university-level, and another 23% holding college (zhuanke)-level education. Our sample also contains a share of Party members (14%) much higher than the national average (at about 6%). In fact, given the high concentration of state/party agencies in Beijing, this percentage of Party membership (14%) is likely to be higher than that of any other Chinese cities. While these characteristics unique to Beijing make our sample unrepresentative of the whole China, our interests are mainly on the cross-sectional, inter-group patterns such as whether state-sector employees vote more than non-state employees and such patterns should not be affected by the relative structure of the sample. The bottom row in Table 1 reports the voting rates of each category of Beijing residents. The Chinese government routinely reports the turnout rate in district-level people s congress elections at 95 percent or higher (Yuan 2003). Our data show a 62 percent of self-reported turnout rate in the 2011 Beijing local people s congress election, with a nonvoting rate of 37 percent. This turnout rate is close to one recorded around 1990, reported in Shi (1999) to be 61.6 percent, and one reported for Beijing s 1993 local people s congress election, at 58.5 percent (Chen and Zhong 2002). 7 Table 2 shows that a significant portion (39%) of the nonvoters did ask someone else to vote on their behalf in the November 2011 Beijing election. 8 Table 2 reports the various reasons cited by respondents for their voting or failing to vote. Thirty percent of the nonvoting people stated that they were not informed

6 248 Political Research Quarterly 70(2) Table 2. Reasons for Voting and Nonvoting in Beijing s Local People s Congress Election. Party membership Education level: 1: Middle school or lower 2: High school 3: College 4: University Employment groups: 1: Blue-collar workers 2: State-sector middle class 3: Private-sector middle class 4: Retired No Yes Reason for not voting (for those who did not vote) No one informed me about it I was sick, so I could not go I was busy, asked others to vote on my behalf I did not want to go Other reasons Reason for participating in voting (for those voted) I genuinely hoped to elect a delegate that can represent citizens It was a formality, exercising my rights Others were going, would not look good if I did not I could not excuse myself, so I just went Source Beijing Local Election Participation Survey (Conducted in December 2011, one month following the Election on November 8). Total about the election or voting arrangements. This figure indicates that the state s system for electoral mobilization is failing to reach a large proportion of voters. A significant proportion (18%) simply refused to go to the polls, including 27 percent of the nonvoting members of the state-employed middle class. The lower half of Table 2 shows that among those who voted, 54 percent believed they did so as a formality, with even 40 percent of voting Party members feeling so. Only 39 percent felt it was genuinely meaningful to cast a ballot. These figures appear to support the state-alienation perspective, although some voters do appear to respond to their civic urge to participate. In this regard, nonetheless, only very small differences can be observed between the state- and non-state-sector middle classes, with members of the former being slightly more likely to vote for the sake of democracy (44% vis-à-vis 41%) and less likely to vote for the sake of formality (51% vis-à-vis 55%). Operationalizing Independent Variables Socioeconomic Classes Traditionally, studies of the relationship between social class and voting behavior have focused on the distinction between the working class and the middle class. Based on this tradition and Goldthorpe s seven-category class schema, some influential empirical studies operationalize social class into the following seven categories: service class, routine nonmanual class, petty bourgeoisie, farmers, skilled workers, nonskilled workers, and agricultural laborers (Evans 1999, 28). In our survey, we provided a seventeen-category occupational scale to the respondents and asked them to choose the one to which they belonged. Then, we recoded these seventeen categories into four classes: blue-collar workers, the state dominated public-sector middle class, the private-sector middle class, and retirees. 9 The first row in Table 1 shows these four classes account for 19, 29, 31, and 20 percent of the sample, respectively. Psychological variables and subjective motivations. We included four psychological-motivational variables in our analysis, that is, democratic orientation, support for the regime, perceived political efficacy, and anti-corruption sentiment. The mean and standard deviations (SD) of each is reported in Table 3, including the mean and SD within specific sections of the sample. Democratic Orientation In our analysis, democratic orientation was measured by a battery of five Likert-type scale statements. They measure the respondent s attitude or position regarding

7 Wang and Sun 249 Table 3. Mean Scores for Beijing Voting-age Residents Psychological and Motivational Attributes. Party membership Education level: 1: Middle school or lower 2: High school 3: College 4: University Employment groups: 1: Blue-collar workers 2: State-sector middle class 3: Private-sector middle class 4: Retired Voters versus nonvoters No Yes Voters Nonvoters Democratic orientation (1 5) M SD Regime support (1 5) M SD Political efficacy (1 3) M SD Anti-corruption sentiment (0 1) M SD Source Beijing Local Election Participation Survey (Conducted in December 2011, one month following the Election on November 8). Total to China s one-party political system, a multiparty political system, the relationship between government and citizens public demonstrations, and nongovernmental groups (details and actual wording provided in Appendix B). These five items were averaged to represent a respondent s democratic orientation ranging from one to five. 10 Support for the Regime (Regime Support) Support for the regime was measured with a Likert-type scale battery that included seven statements regarding belief and attitudes toward the Communist Party, China s National People s Congress, China s military, the police, courts, political system, and the values endorsed by the party and the state (details and actual wording provided in Appendix B). The battery has been administered in previous studies of regime support among private entrepreneurs (Chen and Dickson 2008). 11 These seven items were averaged to form an index of a respondent s support for the regime ranging from one to five. 12 Political Efficacy A respondent s perceived political efficacy was measured on a Likert-type scale using two statements measuring whether the respondent feels politics is too complicated and whether individual citizens should play a role in political reform, respectively (detailed wording in Appendix B). The responses to the two items were then averaged to measure a respondent s perceived political efficacy, ranging from one to three. 13 Anti-corruption Sentiment In a battery that invites the respondents to assess government performance in a range of social issues (shehui wenti), we included an item on fighting corruption (daji tanwu fubai). Respondents expressing that the government has done quite bad or very bad are coded as 1, and the rest are coded as 0 (including so-so, quite good, and very good ). This measurement of anti-corruption sentiment is similar to what was adopted in Chen and Zhong (2002). Descriptive Statistics of Psychological Variables Table 3 reports that Party members hold weaker democratic orientation, stronger support for the regime, and stronger sense of political efficacy than nonmembers. Party members and nonmembers show little difference regarding their assessment of government s corruption fighting effort. The effect of education on these psychological attributes is quite in line with what modernization theory would predict education seems to increase democratic orientation and perceived political efficacy, while decreasing support for the regime. More educated citizens also appear to be more dissatisfied with the government s corruption fighting performance, as the mean rises from 0.42 to Put it simply, more educated citizens seem to be more prodemocracy and more critical toward the regime. If state alienation is at work in these elections, we should expect more educated citizens to abstain more.

8 250 Political Research Quarterly 70(2) The data fail to show any significant trend in the various social classes subjective motivations. For example, vindicating earlier findings (Chen 2013; Chen and Lu 2011), an average middle-class member holds no stronger democratic orientation than an average blue-collar worker, while that of the state-sector middle class seems to be significantly less prodemocracy than an average blue-collar worker. The state-sector middle class, however, fails to show stronger pro-regime attitudes either, with the mean for regime support differing little between the state-sector and private-sector middle classes (4.08 vs. 4.03). This seems to contradict earlier findings that state-dependent middle class in China tend to identify with the regime and prefer status quo instead of political change (Chen and Lu 2011; Tomba 2009). Nevertheless, regime support being a 1 to 5 scale, every social group appears to hold a rather high level of regime support, all averaging higher than 4.0 as reported in Table 3. Controlling Variables Age, gender, education. We include age and gender as two demographic controls. Age square is also included in the models to capture the potential curvilinear relationship between age and turnout. The residents of urban centers such as Beijing generally have a much higher average educational level compared with residents of China s countryside. We classified respondents educational attainment into four levels: junior middle school or lower (14%), high school (35%), vocational college (dazhuan, 23%), and university and above (28%; see Table 1). Institutional Variable: Party Membership We also include as a control variable the respondent s membership in the Chinese Communist Party. In our sample, 14 percent of the respondents held party membership. The bottom row in Table 1 shows that a significantly higher portion of party members voted in the Beijing election than nonmembers 76 percent versus 60 percent, respectively. Political Interest Respondents political interest was measured with one survey question: Generally speaking, are you interested in political issues? The respondents answers are coded into a four-point scale, with 1 representing the lowest interest and 4 the highest. Multivariate Analyses At first look (see the statistics in Table 3, two far right columns), voters appear to be less prodemocracy, more pro-regime, and less efficacious than nonvoters. Plus, voters also appear to hold stronger anti-corruption sentiments. But, with the exception of anti-corruption sentiment, none of these differences are statistically significant. 14 This first look at the data led us to look for structural and institutional factors as more plausible determinants of voter turnout. We introduced our various groups of variables separately into the logistical regression models, with voter turnout as the dichotomous dependent variable (Table 4). Age and age square, gender, education, and party membership are controlled throughout, while political interest is controlled after Model 1. Model 1 replicates Chen and Zhong s (2002) analysis. Both their sample and ours were drawn from urban residents in Beijing. Since theirs was a 1995 sample while ours came 16 years later, there are reasons to believe that the time in between has produced changes in the Chinese society, so much so that we fail to repeat their results. During the 1990s, for example, the non-state economy was still relatively small, meaning the state versus nonstate division of China s middle urban residents was probably insignificant. Class structure, in other words, might be less a determinant of voter turnout in those years, leaving subjective motivations to be the main factors. In our model, subjective motivations bear no clear effect on voter turnout, with the only exception of anticorruption. Therefore, whether a person is more prodemocratic or more pro-regime, and whether he or she holds stronger or weaker political efficacy matter little when it comes to turnout decisions in China s local people s election. Anti-corruption sentiment tends to increase voting avoidance, however. This would support the state-alienation hypothesis, that is, those unhappy with corruption and/or state s inability in dealing with it are more likely to stay away from the state-sponsored election. Based on this result, we judge that neither statealienation or civic citizenship is clearly at work citizens turn out to vote not because they are prodemocracy, are pro-regime, or believe they can make a difference in politics. The factors leading a Beijing resident to vote must be found elsewhere. Model 2 simply adds one more variable, political interest, to Model 1. While political interest itself bears highly significant effect on a person s tendency to participate in voting, the inclusion of this variable generates no significant change on the other variables role in the model. Model 3 introduces the structural and institutional variables, including social class and party membership. Somewhat surprisingly, Party members fail to show a stronger tendency to vote. What really matters seems to be an individual s class status. With the blue-collar workers as the comparison group, the state-sector middle class and the private-sector middle class show opposing tendencies, with the former significantly more likely and the

9 Wang and Sun 251 Table 4. Logistic Regression of Voter Turnout. Model 1 Model 2 Model 3 Model 4 Independent variable B SE Exp (B) B SE Exp (B) B SE Exp (B) B SE Exp (B) Age (18 72).076* * * Age Gender (female = 0,.264* *** ** ** male = 1) Education level a High school College University and above 1.109**** *** ** *** Political interest (1 4).411**** **** **** Democratic orientations (1 5) Regime support (1 5) Political efficacy (1 3) Anti-corruption.331** * ** sentiment Social class b Middle-class public.364* * sector Middle-class private.493** ** sector Retired groups Party membership Constant log likelihood 1, , , ,172.1 Model χ Degree of freedom N 1,041 1,041 1,041 1,041 a Reference category is Blue-collar Workers. b Reference category is middle school and below. *p <.1. **p <.05. ***p <.01. ****p <.001. latter significantly less likely to vote. This point strikes harder when we reintroduce the psychological and motivational factors in Model 4. In this model, subjective motivations continue to show no significant effect, while the opposite directions taken by members of the two middle classes remain salient. These models get to the crux of voter participation, that is, it is China s political economy structure that is at work. Social class seems to be the actual structure and institution, with the state-sector middle class participating most actively in these elections. By contrast, the most likely abstainers are found in the private-sector middle class members of the white-collar salaried class who work in foreign or domestic private enterprises. This should help solve the earlier controversy regarding motivational variables impact on voter turnout in China (Chen and Zhong 2002; Manion 2010; Shi 1999) once the structural-institutional factors are controlled, none of the psychological variables plays a role in increasing or decreasing turnout. Education also matters greatly. University-educated citizens, 28 percent in our sample, tend to be much more likely to vote than the rest of society, even after controlling all other variables. This is in line with the argument that more educated people seem to be active participators, regardless of what institutional settings they are in (Franklin 2004). In other words, the more educated turn out to vote not because they are prodemocracy or proregime, but because they represent the more active and more inquisitive section of society. Probably to the dismay of the party, its members fail to show positive difference in participation. The State-Sector Middle Class Why do members of the state middle class participate in voting while their counterparts in the private sector fail to do so? The state s inability to penetrate the private sector appears to be a factor among nonvoters in the private

10 252 Political Research Quarterly 70(2) middle class, 41 percent stated that they were not informed about the election (compared with 16% in the state-sector middle class, see Table 2). By contrast, members of the state-sector middle class appear to be more subject to the state s mobilization. In addition, as Table 3 shows, members of the state-sector middle class appear to be more interested in politics, but our regression models have all controlled this variable. The state-sector middle class, however, does not appear to be more supportive of the regime the mean regime support score is 4.08 for the state-sector middle class but 4.03 for the private middle class (Table 3). It would appear that members of the state-sector middle class vote more simply because they are affected by a number of institutional factors that are nonexistent for those outside the state. But it does appear strange if regime support is having no effect on voter turnout for those linked directly to the state. We suspect that in this case, regime support s impact on voting behavior is conditional on the institutional environment of a certain social class. That is, the same degree of difference in the level of regime support would have a significant impact on people s tendency to vote in one social class but not the other. Following Brambor, Clark, and Golder (2006), we introduce another regression model, which includes an interaction term between regime support and the state-sector middle class (Table 5). As it turns out, the interaction term generates a positive and significant coefficiency. This appears to show that, while regime support in general has no detectable effect on voter turnout, within the state-sector middle class, it does exert a significant and positive effect. Clearly, due to some institutional factors related to the state-sector, regime support plays a role in the middleclass members turnout decisions. We have tried including interaction terms between regime support and other social classes, and no significant effect arose. According to Brambor, Clark, and Golder (2006), all constitutive terms should be included in the interaction model; therefore, we also tried to include an interaction term between the state-sector middle class and other psychological variables, including democratic orientation, political interest, and anti-corruption sentiment. None of these results in a significant coefficiency either. Therefore, only regime support works differently within the statesector than outside the state sector. Comparing Models 5 and 4 will show how regime support works to generate variation of voting turnout within the state middle class. When no interaction term was included (Model 4), state middle-class members voting rate is times of that of the working class. When the interaction term is included (Model 5), a statesector middle-class member with low level of regime support (regime support = 0) is less willing to vote than a working-class member his or her willingness to vote is Table 5. Logistic Regression of Voter Turnout with Interaction Term. Independent variable Model 5 B SE Exp (B) Age (18 72) 0.084* Age Gender (female = 0, male = 1) 0.347** Education level a High school College University and above 0.802*** Political interest (1 4) 0.387**** Democratic orientations (1 5) Regime support(1 5) Political efficacy (1 3) Anti-corruption sentiment 0.327** (0 1) Social class b Middle-class public sector 2.702* Middle-class private sector 0.530** Retired groups Party membership Regime Support Public ** Sector Middle Class Constant 2.445* log likelihood 1, Model χ Degree of freedom 16 N 1,041 a Reference category is blue-collar workers. b Reference category is middle school and below. *p <.1. **p <.05. ***p <.01. ****p <.001. only 6.7 percent of that of his or her working-class counterpart. But, when regime support rises by just one unit, the state middle-class member s willingness to vote comparing with that of the working-class member increases by times. In other words, with the increase in regime support, state middle-class members voting tendency quickly surpasses that of the working-class members. Figure 1 shows that when a person s support for the regime is the strongest, at the scale of five (5), he or she would be times more likely to vote if he or she is employed on a state middle-class job than if he or she is a blue-collar worker. From the lowest level of regime support to the highest, a state middle-class member s voting tendency has increased from 6.7% to 300.7% of that of his or her working-class counterpart. This mechanism also means that, electoral participation has a stronger effect in separating regime supporters from the nonsupporters among the state middle class. In fact, while in other classes, voters and nonvoters differ very little in their level of regime support, among the

11 Wang and Sun 253 Figure 1. Regime support and odds ratio (middle-class public vs. blue-collar workers). state middle class, that difference in regime support appears more observable (although still short of being statistically significant, see Appendix C). This finding means that state dependence alone is insufficient in commanding citizens compliance in these elections. For middle-class members employed in the state sector, it still requires a high level of regime support for them to turn out to vote. The other side of the message is support for the regime only works to increase turnout rate among the state middle class, not in other social classes. Summarizing Findings Putting all these together, we conclude that (1) the state middle class votes significantly more than other social classes in China, (2) the non-state middle class, by contrast, votes the least among all social classes, and (3) voter turnout is determined by structural and institutional factors much more than psychological-motivational ones. The explanation of the higher participation rate of the state middle class appears to be linked to a range of institutional- and social-ecological factors, which ensure the state employees are informed (and primed) about the polling and turn a person s identification with the regime into actual regime-complying behaviors, that is, turning out at the poll. Discussions and Conclusion We expected state-sector middle class to vote more than members of other social classes, while those in the nonstate sector vote less. Within the structural-institutional frameworks Chinese voters operate in, we expected to find weak or nonexistent effects generated by subjective motivations. Our findings confirmed these hypotheses. We also expected that members of the state-sector middle class vote more because they identify with the regime. The regression models find that while in general, regime support exerts no impact on voter turnout, within the state-sector middle class, those who hold stronger support for the regime do show a stronger tendency to vote. Highly educated and politically interested Chinese also seem to vote more. This is in line with one aspect of modernization theory but contradicts another more educated people are more politically active or expressive than others, but they are not explicitly evading the regime-sponsored political rituals, at least not in terms of electoral participation. The more persuasive explanation might be that more educated and more political interested people are more active participants anywhere, regardless of the nature of political system they are in (Franklin 2004). Party membership carries no premium in terms of voter participation within each social class, Party members are as willing to vote as nonmembers. China s socioeconomic modernization over the last three and a half decades, therefore, has created a large middle class, with a significant part of it dependent upon, instead of independent from, the state. Patterns of voting participation clearly manifest this state-society relationship. By understanding the trend of increasing diversification of the economic structure, deepening social stratification, and ongoing diversification of the labor market, however, one can anticipate the trends that are forthcoming. First, with the continuing growth and expansion of the private sector, the non-state middle class is likely to play more independent and important roles in the political process in the future. Its members absence from polling stations represents both the state s failure at penetrating into the ranks of the middle class and the latter s passive resistance to state efforts at mobilization. Second, Party membership generates no premium in terms of compliance to the regime in these elections, showing the Party s inability to rally its rank and file around the state on ideational grounds. Third, even among those members of the middle class who are employed by the state or state-sponsored organizations, a sense of alienation appears to be developing. Some of them explicitly reject these party-organized elections as utterly meaningless. In the end, voters participation and abstention patterns reflect the weakening power of China s political system to keep the society under the party-state s control. Appendix A Sampling Process A multistage sampling process with a probability proportional to size (PPS) measure first randomly chose fortyeight streets (jiedao) in all six urban districts (qu) of Beijing, and among twelve streets or towns (zhen) within four suburban districts (jiaoqu), totaling sixty selections. At the second sampling stage, one residential committee