Polit Behav (2012) 34:249–276 DOI 10.1007/s11109-011-9157-x ORIGINAL PAPER A Two-edged Sword: The Differential Effect of Religious Belief and Religious Social Context on Attitudes towards Democracy Pazit Ben-Nun Bloom • Gizem Arikan Published online: 11 April 2011 Ó Springer Science+Business Media, LLC 2011 Abstract Different components of the religious experience have differing effects on attitudes towards democracy. Using heteroskedastic maximum likelihood models and data from the fourth wave of the World Values Survey for 45 democratic countries, we show that as a personal belief system, religiosity contrasts with democratic principles, generating opposition to democracy while increasing ambivalence towards democratic principles among religious people. Nevertheless, at the group level, religion also serves as a social institution which increases the homogeneity of one’s social network, leading to lower ambivalence, and makes for an active minority group which benefits from the democratic framework, consequently increasing support overall for a democratic regime. This double-edged sword effect explains the mixed results currently found in the literature on religiosity and democracy, and clearly illustrates the multidimensionality of religiosity. Keywords Religious belief  Religious behavior  Democratic attitudes  Ambivalence  World Values Survey  Heteroskedastic maximum likelihood models Introduction Attitudes the public holds about democracy are critical for the stability of democratic regimes (Bratton and Mattes 2001), and are shaped by many factors, An earlier version of this article was presented at the Annual Meeting of MPSA, 2–5 Apr 2009. P. Ben-Nun Bloom (&) Department of Political Science, The Hebrew University of Jerusalem, Mount Scopus, Jerusalem, Israel e-mail: Pazit.Bennun@mail.huji.ac.il G. Arikan Department of Sociology, Yeditepe University, Istanbul, Turkey e-mail: garikan@ic.sunysb.edu 123 250 Polit Behav (2012) 34:249–276 with religiosity being one of the most infamous. Starting with Adorno et al.’s (1950) theory of the authoritarian personality, which suggested a psychoanalytical connection between religious and undemocratic attitudes, many scholars have argued that religion might pose an obstacle to the development of democratic values (Lipset 1981). Indeed, religiosity is associated with prejudice, political intolerance, and non-democratic norms (Altemeyer 1996; Gibson 1992; Hunsberger 1995; Karpov 2002). However, some suggest that religiosity may have a positive impact on democratic values through participation in religious social activities. In fact, studies have usually found a positive association between church attendance and electoral turnout, party membership, protest activism, and involvement in other civic organizations (Norris and Inglehart 2004; Norris 2002; Rosema 2007). We argue that the debate in the current literature reflects the varying effects that different dimensions of religiosity have on attitudes towards democracy. We build upon the framework proposed by Kellstedt et al. (1997), who regard religiosity as a multidimensional phenomenon consisting of three dimensions: belief, behavior, and belonging. We argue that religious belief, due to its association with conservativetraditional values, engenders opposition to change, desire for order, and intolerance, which are incompatible with democratic norms and values (Adorno et al. 1950; Feldman 2003; Gibson 1992; Karpov 2002). On the other hand, religiosity is also a social phenomenon. Involvement in social religious networks has a positive influence on civic skills, leading individuals to actively engage in politics and hold more positive attitudes towards democracy. Our second point is that different dimensions of religiosity also have differing effects on the ambivalence individuals experience regarding their support for and evaluation of democratic systems. We argue that religious belief induces a value conflict with democratic norms, which results in value-driven ambivalence (e.g. Alvarez and Brehm 1995, 1997, 2002; Feldman and Zaller 1992). Involvement in religious networks, and interaction with like-minded others, on the other hand, decrease one’s ambivalence towards democracy (Mutz 2002; Huckfeldt et al. 2004). There is also much controversy over the effects of religious identification on democratic support and evaluations (e.g. Anderson 2004; Fukuyama 1992; Kedourie 1994; Huntington 1996; Radu 1998; Shah 2004). Controlling for the other two dimensions of religiosity and a variety of country specific measures of religious freedoms and government regulation of religion, we find that religious affiliation by itself is not a decisive factor affecting individual attitudes towards democracy. With the exception of Hindu identifiers, identifying with major religious traditions does not, in and of itself, lead to anti-democratic attitudes. This article, thus, makes two important contributions to the literature. Firstly, although some evidence has been published pointing to the differential effects of religiosity on democratic values, most studies have focused on a single dimension of religiosity, are restricted to one or a few case studies, and usually do not control for country level contextual effects. Unlike these studies, we test our hypotheses using a pooled dataset that includes 45 of the world’s democracies, and show that religious belief and social religious activities have contrasting effects on support for democracy. Secondly, we demonstrate the differential effects of the dimensions of 123 Polit Behav (2012) 34:249–276 251 religiosity not only on the strength of democratic attitudes, but also on ambivalence in attitudes towards democracy. While the effect of attitudinal ambivalence was found consequential in reducing the predictability and stability of key political attitudes, decreasing attitudinal extremity, and lowering confidence in judgments (Alvarez and Brehm 1995, 1997, 2002; Lavine 2001; Schnell 1993; Sniderman et al. 1996), its effect on attitudes towards democracy has rarely been investigated (however, see Ben-Nun Bloom et al. 2011). We find that all else being equal, religious belief increases value-driven ambivalence towards democracy, while involvement in religious social networks decreases it. Compared to the belief and behavior dimension, religious traditions have less consistent effects, although this may be due to the utilization of a crude measure of religious identification. The Dimensions of Religiosity As mentioned above, Kellstedt et al. (1997) conceptualize religiosity as a multidimensional experience consisting of belief, behavior and belonging (also see Wald and Smidt 1993 for a similar conceptualization). The belief component encompasses theology, ‘‘an understanding of the divine and humanity’s relationship to it,’’ and ‘‘social theology,’’ which ‘‘connects the individual and the institutional church to the world,’’ and may refer to belief in God, heaven, hell, life after death, or tendencies of people to characterize themselves as religious. The behavior component consists of private and social practice. While the former refers to the private practice of religion, such as prayer or reading of the holy texts, the latter describes participation in organized religious communities and attendance at places of worship. The third dimension, belonging, consists of denominational affiliation, that is, identification as a member of a particular organized denomination, movement, or trends within a denomination. Different dimensions of religiosity have been found to have different and even conflicting effects on political behavior. For example, while religious belief influences other beliefs that enhance economic growth, church attendance leads to reductions in economic performance (Barro and McCleary 2003a). Whereas church attendance fosters volunteering, religious belief does not have any significant impact on volunteerism or participation in charitable activities (Dekker and Halman 2003). The belonging dimension is also central to understanding political behavior, yet its conceptualization varies. While Kellstedt et al. (1997) define this dimension as denominational affiliation, they also suggest that aggregating denominational preference into appropriate religious traditions could be useful on theoretical, practical, and empirical grounds. Indeed, cross-national research usually conceptualizes this dimension as identification with major religious traditions, although some scholars focus on congregation as an important unit of political socialization that affects political attitudes (Djupe and Gilbert 2006; Gilbert 1993; Jelen 1992; Jelen and Chandler 1996; Wald et al. 1988, 1990). Our position in this debate is determined by our data, as the WVS does not include data on congregations for the 123 252 Polit Behav (2012) 34:249–276 majority of countries. We therefore aggregate religious identification into nine major religious traditions and test for their effect on attitudes towards democracy. Religiosity and Ambivalence In addition to hypothesizing an effect of religiosity on democratic attitudes, we argue that the dimensions of religiosity also affect the ambivalence of these opinions. Current attitudes models agree that responses to survey questions are constructed on the spot from a distribution of relevant considerations (e.g., feelings, beliefs, and values) attached in an individual’s memory to the political concept at hand (e.g., Zaller 1992; Zaller and Feldman 1992). People typically hold several— conflicting—considerations on a subject, and build on those accessible at the time of questioning to compute an on-the-spot response. This probabilistic memory search can be conceptualized as ‘‘sampling’’ of the distribution of one’s considerations, whereas the generated attitude is in fact one’s estimation of the central tendency measure of the distribution of associations on the subject (Feldman 1995). As a result, people’s evaluations are generally expected to be unstable over time. Accordingly, individual differences in attitudes may manifest themselves in two ways. Firstly, people may come up with different estimations of their distribution’s mean, which will translate into different responses on the issue (e.g., strongly support vs. mildly oppose). Secondly, people may differ in their attitude’s consistency, which will be governed by the variance of their distributions of considerations on an issue (Zaller and Feldman 1992; Alvarez and Brehm 2002). Two people may have the same mean for their considerations distribution, but differ on the variance, such that their answer could be similar on average but the larger distribution variance leads to more volatile response across repeated sampling. The type of ‘‘considerations’’ political scientists usually stress as responsible for ambivalence in judgments involve internalized conflict among core values (e.g., Feldman and Zaller 1992; Alvarez and Brehm 1995, 2002; see Lavine 2001). When two values, such as religious tradition and gender equality, are relevant to a political issue yet ‘‘push’’ the opinion in different directions, the person internalizing both values is expected to show ambivalence in his attitude on this issue. Indeed, such value-driven ambivalence was detected on a range of political attitudes (e.g., Alvarez and Brehm 1995, 1997, 2002; Rudolph 2005; Basinger and Lavine 2005; but see Steenbergen and Brewer 2004). In addition, people may hold several and often conflicting thoughts on an issue at the same time, rather than a unidimensional bipolar attitude. This can happen, for example, when a person strongly likes several aspects of a political party’s platform and at the same time strongly dislikes other aspects, or views a candidate as liberal on some issues and conservative on others (see Lavine 2001), which manifests in greater attitudinal volatility. In terms of support for democracy, we argue that religiosity can affect both the mean of one’s distribution of considerations and its variance. We further suggest that different dimensions of religiosity have differing effects on the mean and variance of the distribution of one’s considerations concerning democracy, i.e., one’s attitude and ambivalence towards democracy. 123 Polit Behav (2012) 34:249–276 253 Religiosity and Attitudes Towards Democracy: Conflicting Values We argue that religious belief has a negative effect on democratic attitudes because of an inherent value conflict between the two. We define value conflict following Shalom Schwartz, who views values as desirable, trans-situational motivational goals that vary in importance and serve as guiding principles (Schwartz 1992). Universal human values form two bipolar conceptual dimensions representing two conflicts: a conflict between concern for the self (self-enhancement) and concern for others (self-transcendence) and between conservatism and openness to change. Religiosity is correlated positively with ‘‘conservatism’’ and negatively with ‘‘openness to change,’’ as it stresses transcendence, belief, humility, preserving the social order, protecting individuals against uncertainty, self-restriction, opposition to change, and striving for order (Schwartz and Huismans 1995). The systematic difference in value systems between the non-religious and the religious from various denominations and cultures is also verified by a number of studies (e.g., Burris and Tarpley 1998; Lau 1989; Roccas and Schwartz 1997; Rokeach 1969; Saroglou et al. 2004). At the same time, democratic principles encourage the opposite value system, stressing openness to change and suppressing conservatism (Schwartz and Sagie 2000). This well-entrenched value conflict between religiosity and democratic norms is expected to increase ambivalence towards democracy among the devout. For these individuals, opposing considerations of democracy are accessible upon appraisal, pulling them in contrasting directions, and leading to an overall lower certainty and stability of the constructed attitude. Thus, our first hypothesis suggests that: H1: An increase in the belief dimension of religiosity increases ambivalence in attitudes toward democracy. The value conflict between religiosity and democracy is expected to affect the mean of the considerations’ distribution as well, such that religious belief is associated with less support for democracy. Adorno et al.’s (1950) authoritarian personality study suggested a connection between personal inclination to religion and undemocratic leanings, and empirical studies report that groups that vary in their level of religiosity also differ on a range of attitudes towards democracy (e.g., Beit-Hallahmi and Argyle 1997; Barnea and Schwartz 1998). Thus, we hypothesize that endorsing religious beliefs will lead to lower support for democracy: H2: An increase in the belief dimension of religiosity decreases support for democracy. Religiosity and Attitudes on Democracy: Social Networks Religiosity is not just a mental concept but also a social institution, supplying abundant opportunities for social interaction and political discussion (Djupe and Gilbert 2006; Wald et al. 1988, 1990). Yet, this dialogue typically occurs between politically like-minded people, making the religious social network relatively attitudinally homogeneous (Mutz and Mondak 2006). 123 254 Polit Behav (2012) 34:249–276 As people tend to assess the correctness of many of their views by comparing their own attitudes to those held by the people around them (Festinger 1954), relative attitudinal heterogeneity in one’s social network decreases one’s confidence in their correctness and augments their ambivalence by decreasing the confidence that people have in the accuracy of their attitudes and increasing their motivation to re-examine and change their attitudes to conform to with those of others in their network. Indeed, research demonstrates that network heterogeneity reduces attitude stability, decreases resistance to persuasion (Visser and Mirabile 2004), increases political ambivalence (Mutz 2002; Huckfeldt et al. 2004), and raise consumption of external information such as news (Scheufele et al. 2004). On the other hand, as social networks increase in homogeneity, likely in the case of religious groups, individuals may be more confident and less ambivalent about their attitudes. Accordingly, we hypothesize that: H3: An increase in the social dimension of religiosity leads to less ambivalence in attitudes on democracy. Participation in religious social networks may affect the mean of one’s distribution of considerations on democracy as well. Being a minority in democracies, the religious enjoy the fruits of religious freedom and the rights and resources granted to minorities. More often than not, the religious elites and socialization agents politicize the religious group identity, and mobilize the corresponding constituency within the democratic game. For that, they need to encourage their public to practice their democratic rights as citizens. Indeed, religious institutions are often considered powerful mobilization forces, promoting political involvement and participation in democracies (Jelen and Wilcox 2002; Putnam 2000; Verba et al. 1995). Studies also point to the fact that active participation in churches positively affects political involvement (Leege 1988; Wald 2003; Rosema 2007). With religious leaders advocating the democratic game, and encouraging their publics to take an active role in it, places of worship can be hypothesized to enhance overall support for democracy. Therefore, we hypothesize that: H4: An increase in the social dimension of religiosity increases support for democracy and political participation. Religiosity and Attitudes Towards Democracy: Identification with Religious Traditions The belonging dimension can be conceptualized as identification with major religious traditions, i.e. a group of local churches and denominations that share a set of beliefs, practices, and a common heritage (Kellstedt et al. 1997), as well as the more narrow congregational and denominational affiliation within a general tradition (Djupe and Gilbert 2006; Gilbert 1993; Jelen 1992; Jelen and Chandler 1996; Wald et al. 1988, 1990). Due to the limitations of the WVS dataset, we conceptualize this dimension at the former, more general, level. 123 Polit Behav (2012) 34:249–276 255 There are conflicting theoretical arguments and empirical evidence concerning the effect of identification with religious traditions. Some scholars perceive the negative effects of identification on support for democracy as associated with the teachings of certain religious traditions such as Islam, Hinduism, Orthodoxy, or Evangelical Protestantism due to inherent conflicts between these religious beliefs and democracy (Fukuyama 1992; Kedourie 1994; Huntington 1996; Radu 1998; Huntington 1996). Others suggest that religious identification is not of itself utterly contradictory to democratic values because religions often preach compassion and ‘the golden rule’ (e.g. Canetti-Nisim 2004; Eisenstein 2006), or they cite present evidence for a positive effect of religiosity (Aaron 2009; Bratton 2003; Chaibong 2004; Fernandes 2009; Gifford 2008; Marsh 2005; Mehta 2004; Tessler 2002). Yet, empirical evidence in this debate typically relies on a small number of case studies, making the results ungeneralizable, and on a weak operationalization of religiosity that neglects the multidimensionality of the religious experience. We argue that the belief and behavior dimensions are more central to explaining attitudes towards democracy than is identification with a major religious tradition, since there are many diverse viewpoints, customs and political schools within each tradition (Weightman 1997; Filali-Ansary 1999). In addition, all religious teachings require interpretation to give them meaning in specific contexts (Bromley 1997), as religious belief systems contain within them resources that can be used to promote different visions of the appropriate political order (Anderson 2004). These variations within each religious tradition may not be captured by the crude measure of belonging at the center of contemporary debates on the relationship between religion and democracy. We therefore hypothesize that: H5: Identification with a major religious tradition will have less consistent effects on attitudes towards democracy as well as ambivalence in democratic attitudes. Model, Data, and Operationalization of Variables The Model The model employed for explaining attitudes towards democracy has two main parts. Firstly, religious belief and religious social activity are both expected to affect attitudes towards democracy, with religious belief decreasing support for democracy and religious social activity increasing it, holding religious identification constant. Secondly, holding all else equal, religious belief and religious social activity are both expected to affect ambivalence towards democracy, with religious belief increasing ambivalence and religious social activity decreasing it. While the mean attitude can be easily estimated in regression, ambivalence will be inferred from an estimation of error variance in a Heteroskedastic Maximum Likelihood (ML) Model (e.g., Franklin 1991; Alvarez and Brehm 1995, 2002). By defining a variance equation, we test hypotheses about variables that increase or decrease the error variance—i.e., the ambivalence in the dependent variable. A positive coefficient in the variance equation suggests higher error variance (and thus 123 256 Polit Behav (2012) 34:249–276 higher ambivalence) and vice versa. We define a normal function for the model, and since the variance in the dependent variables is non-negative, we define an exponential equation to constrain it accordingly. We use two different dependent variables to capture attitudes towards the democratic system. One is endorsing democracy in one’s country as opposed to support for authoritarian forms of government. The other is perception of democracy as an efficient system. We include a number of individual-level control variables in the mean equation, such as income, education, age, political ideology, authoritarianism, discussing politics, trust, nationalism, and gender. In addition, we incorporate a number of country-level variables, as contextual factors may significantly affect attitudes towards democracy. Socio-economic development is usually associated with stronger support for democratic systems, while communist legacy is negatively related to endorsing democracy (Inglehart and Welzel 2005). The quality of democracy in a country is believed to create an environment suitable for the internalization of democratic values (Muller and Seligson 1994). Religious freedoms and the absence of state regulation of religion may be expected to generate stronger support for democracy. In addition, religious tradition is regarded as one of the important conditions for democratization and the development of democratic ideals (Lipset 1959; Norris and Inglehart 2004). Historically, Protestantism and democracy have been positively correlated (Bollen and Jackman 1985), but the active role Catholic churches have played during the democratization movements in authoritarian regimes (Philpott 2004) suggest that individuals in Catholic nations could be just as supportive of democracy. Despite the arguments according to which Muslim and Orthodox religious traditions are antithetical to democratic development (Huntington 1996), recent research has found strong support for democracy in predominantly Muslim countries (Norris and Inglehart 2004), although this is not the case for predominantly Orthodox nations (Meyer et al. 2008). We therefore expect support for democracy to increase with the percentage of Muslims, whereas we do not have a clear expectation for the effect of Orthodox tradition on attitudes towards democracy. Some further factors may also be expected to affect the error variance of an attitude, since they may affect cognitive capacity or the existence of relevant contextual knowledge. Young adults may have less crystallized attitudes, since they have less experience with institutions. Next, we control for the effect of authoritarianism, to test the alternative explanation that authoritarianism, which usually correlates with religiosity, rather than religion itself is responsible for decreasing ambivalence. Next, education may reduce stochastic variance since it implies greater contextual knowledge and also may be a proxy for cognitive capacity. Finally, discussing politics may increase ambivalence, as it adds relevant considerations to one’s distribution, or may decrease ambivalence if it facilitates crystallization of the concept of democracy. Data and Countries We use data from the fourth wave of the World Values Survey (WVS), which includes data from 75 countries, collected between 1999 and 2001. Since we believe 123 Polit Behav (2012) 34:249–276 257 that individuals who have no experience with democratic forms of governance could have unreliable evaluations of democracy, we drop countries that are classified as undemocratic by Freedom House. Among the remaining countries, a number were not included in the final analysis because some of the items or country-level data were missing, leaving us with a final total of 45 countries.1 Operationalization of Variables To facilitate interpretation, all individual level variables with the exception of age were coded to vary between 0 and 1. Attitudes towards Democracy are measured by two variables: endorsing democracy in one’s country as a form of government (support for democracy as opposed to support for authoritarian forms of government), which is a summative index of the items:2 ‘‘Having a strong leader who does not have to bother with parliament and elections’’; ‘‘Having a democratic political system,’’ and perceiving democracy as an efficient system, which was constructed from the items: ‘‘In democracy, the economic system runs badly’’; ‘‘Democracies are indecisive and have too much squabbling’’; and ‘‘Democracies aren’t good at maintaining order.’’ Both measures were coded so that high values indicate higher support for democracy. Religious Belief: We use items that Kellstedt et al. (1997) and Wald and Smidt (1993) suggest are vital components of religious belief. The following are also used by many other scholars in constructing religious belief scales (Norris and Inglehart 2004): ‘‘Belief in God’’; ‘‘Belief in life after death’’; ‘‘Belief in heaven’’; ‘‘Importance of God in respondent’s life’’; and ‘‘Getting comfort and strength from religion.’’ Before constructing this measure, we controlled for differential item functioning (DIF),3 which might obstruct the comparability of the construct across different religious groups (Van de Vijver 2003), by using a multi-group (MG) structural equation modeling (SEM; Reise et al. 1993; Van de Vijver and Leung 1997) using MPLUS software. The CFA of the pooled items has good fit (CFI = .988; TLI = .975; RMSEA = .138), while the multi-group, full-invariance model, which fixes the loadings of all items across all religious groups, has a slightly poorer fit (CFI = .937; TLI = .950; RMSEA = .178). Since belief in God is generally the first and most fundamental aspect of religiosity in all major religious traditions (Jelen and Wilcox 2002; Widgery 1936), we expect this item not to have any bias across different contexts. In fact, modification indices in the full invariance 1 The final countries included in the analysis and their Freedom House scores are: Austria (1), Canada (1), Denmark (1), Finland (1), Iceland (1), Ireland (1), Luxembourg (1), Malta (1), Netherlands (1), United States (1), Belgium (1.5), Czech Republic (1.5), Estonia (1.5), France (1.5), Germany (1.5), Great Britain (1.5), Hungary (1.5), Italy (1.5), Japan (1.5), Latvia (1.5), Lithuania (1.5), Poland (1.5), Slovakia (1.5), South Africa (1.5), Spain (1.5), Bulgaria (2), Chile (2), Croatia (2), Greece (2), Peru (2), Romania (2), India (2.5), Mexico (2.5), Philippines (2.5), Argentina (3), Republic of Moldova (3), Serbia (3), Montenegro (3), Albania (3.5), Bangladesh (3.5), Macedonia (4), Ukraine (4), Tanzania (4), Russian Federation (5), Uganda (5). 2 Since the measure is composed of two items, which does not allow for the utilization of CFA, we use a summary index. 3 Unfortunately, the WVS questionnaire does not include many items that could be used as anchors for other scales. 123 258 Polit Behav (2012) 34:249–276 model do not show a statistically significant expected change in chi-square value when this item is unconstrained. Therefore, we fixed the loading of the ‘‘belief in God’’ item and allowed the rest of the items to vary freely across countries. The partial invariance model gives fit indicators better than the full-invariance model: CFA = .974, TLI = .963, RMSEA = .154, which in fact is the best model fit obtained across a number of alternative partial invariance models. We saved the factor scores, which were corrected for DIF, using the MPLUS software’s missing option.4 Social Religious Behavior: A summative index of two variables belonging to a religious organization, and the frequency of attending religious services (an 8-category variable) in which higher values represent more frequent social religious behavior. Religious Identification: Respondents who do not identify with any religious denomination form the baseline against which the effect of belonging dimension will be assessed. We made use of Fahlbusch et al. (2005), Hinnells (1997), Layman (1997), Lumsdaine (2009), and Ranger (2008) to recode denominational subcategories to nine main religious traditions: Catholic (Aglipayan, Barangay Sang Birhen, Charismatic, Christian,5 Greek Catholic, Israelita Nuevo Pacto UniversalFrepap, Roman Catholic), Protestant (Anglican, Lutheran, Methodist, Unitarian, Protestant), Evangelical (Alliance, Assembly of God, Baptist, Born-again, Christian Fellowship, Christian Reform, Church of Christ, Evangelical, Faith in God, Filipinista, Iglesia Ni Cristo (INC), Independent African Church, Jehovah’s Witnesses, Jesus is Lord (JIL), Jesus Miracle Crusade, Mennonite, Mormon, New Testament Christ/Biblist, Pentecostal, Presbyterian, Salvation Army, Seven Day Adventist, The Worldwide Church of God, United, United Church of Christ in the Philippines), Orthodox (Orthodox), Jewish (El Shaddai, Jewish), Muslim (Muslim), Buddhist (Buddhist, Self-Realization Fellowship), Hindu (Hindu, Sikh), Independent/Other (Bahai, Free Church/Non-denominational Church, Jain, Ka-a Elica, Native, Other, Paganism, Sisewiss, Spiritista, Spiritualists, Theosophists, Wicca). Trust: Coded 1 for those who chose the ‘‘most people can be trusted’’ option for the question, ‘‘Generally speaking, would you say most people can be trusted or that you need to be very careful in dealing with people?’’ and 0 for those who chose ‘‘need to be very careful.’’ Political Interest: ‘‘When you get together with your friends, would you say you discuss political matters frequently, occasionally, or never?’’. The highest value represents the highest frequency of engaging in political discussions. 4 Missing values in the dataset lead to a serious decline in the number of observations available due to listwise deletion. To overcome such problems, for measures derived from confirmatory factor analysis, we use the MPLUS software’s missing command, which uses full information of maximum likelihood from all available data unless all items in the factor model are missing. 5 Independent Christianity may mean different things in different contexts depending on the broader institutional arrangements. Although we coded ‘‘Christians’’ (14 observations from Canada only) as belonging to the Catholic tradition, Manza and Brooks (1997) code unaffiliated Christians in the ‘‘Protestant’’ category, while Mueller (2009) codes them into the ‘‘other’’ category (coding scheme obtained via personal communication, December, 2010). However, because the number of observations for ‘‘Christian’’ identifiers in our dataset is small, recoding them into an ‘‘other/independent’’ category or dropping them from the analysis leads to no substantive changes in our results. 123 Polit Behav (2012) 34:249–276 259 Ideology: A 10 point scale for left–right ideological self-placement. The ideology variable suffers from a lot of missing observations (about 15,000). While it is tempting to use multiple imputation techniques to fill in the missing values for ideology, no combination of potential predictors gives a reasonably high R-squared value (The highest R-squared obtained is .004). Besides, it is reasonable to assume that respondents who do not give an answer to this question do not feel strong about the left or the right, and thus could be considered moderates. Therefore, we gave the mean ideology score of .499 to the missing values for the left–right selfidentification variable. We also replicated results using the non-imputed measure (see the robust analysis subsection). National Pride: 1 if the respondent indicated pride in holding his/her nationality, 0 if not. Authoritarianism: We coded the item 1 if the respondent mentioned that obedience is a quality that children should be encouraged to learn, and 0 if not. Demographics: Gender: 1 for males, 0 for females; Income: A 10 point scale, where low is the lowest level of income; Education: 10-category variable coded by the WVS so that those who completed their education at an age less than 12 are assigned the minimal score and those who completed their education at an age greater than 21 are assigned the maximum score, with those who completed their education between these ages labeled accordingly; Age: Years of age. Country-Level Control Variables Freedom House Rating of Non-Democracy: Freedom House’s rating of political rights and civil liberties for the year 1999–2000. The scores vary between 1 and 7, where countries with a rating of 1–2.5 are considered ‘‘free,’’ 3–5 ‘‘partly free,’’ and 5.5–7 ‘‘not free.’’ Since this variable is coded so that more open societies score lower in the index, we call this variable the ‘‘Freedom House rating of nondemocracy.’’6 GDP Per Capita (PPP): Logged GDP per capita, adjusted for purchasing power parity for the year 2000 to capture the effect of economic development on attitudes towards democracy (Source: World Bank). Religious Tradition: Country’s religious tradition, measured by the percentage of Protestant, Catholic, Muslim and Orthodox adherents in country for the year 2000, gathered from Barro’s Religious Adherents Dataset. The reason for the exclusion of other religious traditions is the relatively smaller number of adherents of other religions in our dataset.7 Communist Legacy: Dummy variable for countries with a Communist legacy, which should be negatively related to support for and evaluations of democracy (Inglehart and Welzel 2005). 6 The average of the score in our dataset is 1.22, and the range is 1–5.5. Since non-democratic countries were not considered in our analysis, the average is lower than the world average of 3.4. 7 We reran the models using binary measures that specify a country’s dominant religious tradition (as in Norris and Inglehart 2004) rather than the reported percentage of each denomination (as in Barro and McCleary 2003a, b, and Durlauf et al. 2005). The results are robust to this alteration in measure. 123 260 Polit Behav (2012) 34:249–276 Religious Freedom: Norris and Inglehart’s (2004) Freedom of Religion Scale, which focuses on church-state relations. This index is based on 20 criteria, including constitutional recognition of religious freedom, the existence of a single official state church, restrictions on certain denominations and/or religious activities, and state subsidy of religions. The index ranges from 0 to 100, with 100 representing the highest level of religious freedom. Religious Regulation: This instrument, which measures a composite variable that addresses regulation and restriction of religious activity, is based on data from the Religion and State (RAS) Project Dataset for the year 2000. This index varies between 0 and 21, with higher values representing higher state regulation and restriction of minority and majority religions. It includes dimensions such as the presence of arrests, continued detention, or severe official harassment of religious figures, officials and/or members of religious parties; the presence of restrictions on formal religious organizations other than political parties, on public observance of religious practices, on public religious speech, on access to places of worship, on the public display of religious symbols, and so forth. Results Endorsement of democracy in one’s country and agreeing that democracy is an efficient regime are each regressed along three dimensions of religiosity and a set of control variables, both in the mean and in the variance equation. The estimates for each of the two models are presented in Table 1 (mean equation) and Table 2 (variance equation). Our key hypotheses regarding the mean equation state that levels of religious belief and religious social activity both affect attitudes towards democracy, with religious belief decreasing support for democracy and religious social activity increasing it. Indeed, frequent social activity within religious organizations significantly increases both support for democracy in one’s country (b = .023, p = .000) and the tendency to view it as an efficient system (b = .012, p = .005) (see Table 1). In contrast, religious belief significantly reduces support for democracy in one’s country as well as one’s evaluation of the democratic system as efficient. Holding all else equal, moving from the least devout to the most reduces the support for democracy in one’s country by about 1/13 of its range (b = -.076, p = .000), and belief that democracy is an efficient system by about 1/18 of its range (b = -.054, p = .000). Both these effects are larger than the effect of religious social activity, and in the opposite direction. In fact, religious belief has the strongest effect on support for democracy compared to all other individual-level variables, and has the third strongest effect on appraising the efficiency of the democratic polity.8 Holding constant for religious belief, religious social behavior, and control variables, religious denominations have inconsistent effects on attitudes towards 8 Effect sizes are comparable, as all individual level variables are coded to vary 0–1, with the exception of age, measured in years (see frequencies table in the appendix). 123 Polit Behav (2012) 34:249–276 261 Table 1 Religiosity and support for democracy: mean equation Mean equation Supports democracy in one’s country Coef. Std. err. Z Democracy is efficient Coef. Std. err. Z 28.26 Religiosity dimensions Religious belief 2.076 (.007) 210.72 2.054 (.007) Social religious behavior .023 (.005) 5.14 .012 (.004) 2.83 Catholic .008 (.004) 1.91 .016 (.004) 4.25 Protestant .008 (.005) 1.53 Evangelical -.014 (.009) -1.63 Orthodox .012 (.006) Jewish .057 (.021) Muslim .036 Buddhist .008 (.005) 1.70 -.015 (.008) -1.86 1.96 .016 (.005) 3.04 2.75 .061 (.021) 2.96 (.008) 4.75 .041 (.007) 5.52 -.015 (.012) -1.24 .043 (.010) 4.36 Hindu 2.068 (.009) 27.86 2.021 (.008) 22.56 Independent/other -.005 (.008) -.64 .024 (.007) 3.42 Individual level controls Discuss politics .054 (.004) 14.95 Authoritarianism 2.033 (.003) 212.72 .039 (.003) 11.60 2.025 (.002) 210.57 Trust people .026 (.003) 9.84 National pride .015 (.005) 3.23 .044 (.002) 17.74 -.004 (.004) Political right 2.017 (.005) 23.49 .018 (.005) -.88 3.91 Education .057 (.004) 13.83 .039 (.004) 10.12 Income .049 (.004) 11.00 Age .000 (.000) 5.32 Male -.004 (.002) Percent Catholics .000 Percent Protestants .001 Percent Orthodox Percent Muslims .064 (.004) 15.79 2.000 (.000) 22.03 -1.81 .003 (.002) 1.20 (.000) 2.51 .000 (.000) 6.27 (.000) 5.00 .002 (.000) 16.60 2.001 (.000) 24.65 .000 (.000) 4.61 .002 (.000) 17.65 .002 (.000) 18.82 Countries fixed effects Religious regulation -.001 (.001) -1.03 .002 (.001) 4.10 Religious freedom scale 2.002 (.000) 212.55 .002 (.000) 11.49 Freedom House rating of non-democracy 1999 2.039 (.002) 222.11 2.013 (.002) 28.32 GDP PPP 2000 (logged) 2.006 (.002) 22.54 -.001 (.002) -.37 Post-Communist 2.068 (.004) 219.47 2.053 (.003) 216.54 Log likelihood hetero/homo 4110.7837/4012.9575 6258.1216/6133.8159 N 36047 38286 Table entries are maximum likelihood coefficients, std. errors and Z values; significance in two-tail 95% confidence level in bold, one-tail 95% confidence level in italics democracy. Identifying as Catholic, Orthodox, Jewish, or Muslim increases support for democracy (b = .008, p = .056; b = .012, p = .044; b = .050, p = .006; b = .036, p = .000, respectively) and endorsement of the democratic system 123 262 Polit Behav (2012) 34:249–276 Table 2 Religiosity and support for democracy: variance equation Supports democracy in one’s country Coef. Std. err. Z Democracy is efficient Coef. Std. err. Z Variance equation Religious belief .097 (.021) 4.66 Social religious behavior 2.041 (.015) 22.81 .093 (.020) 4.54 2.081 (.014) 25.64 2.01 Catholic -.006 (.012) -.49 .023 (.012) Protestant 2.030 (.014) 22.16 .011 (.014) .79 Evangelical .029 (.027) 1.07 .094 (.026) 3.55 Orthodox .046 (.015) 3.05 .048 (.014) 3.44 Jewish .083 (.063) 1.31 .185 (.062) 2.99 Muslim 2.074 (.020) 23.78 .062 (.019) 3.31 Buddhist .045 (.038) 1.17 2.155 (.039) 24.00 Hindu 2.051 (.025) 22.03 -.018 (.024) -.73 Independent/other -.001 (.026) -.04 -.043 (.025) -1.68 Discuss politics 2.050 (.011) 24.33 .014 (.011) 1.24 Authoritarianism .049 (.008) 5.92 .053 (.008) 6.70 Education 2.078 (.013) 25.93 2.067 (.013) 25.23 Age .000 (.000) .75 .001 (.000) 4.91 Log likelihood hetero/homo 4110.7837/4012.9575 6258.1216/6133.8159 N 36047 38286 Table entries are maximum likelihood coefficients, std. errors and Z values; significance in two-tail 95% confidence level in bold, one-tail 95% confidence level in italics (b = .016, p = .000; b = .016, p = .002; b = .061, p = .003; b = .041, p = .000, respectively), compared to not identifying with any religious tradition. Being a Buddhist or belonging to an independent religion also increase endorsement of the democratic system (b = .043, p = .000; b = .024, p = .001 respectively), but does not affect support for democracy. In contrast, Evangelical and Hindu identifiers are less supportive of democracy, and less likely to believe democracy is an efficient system. Although Protestantism is cited as the religious tradition most compatible with democratic norms and ideals, we are not able to provide empirical evidence for Protestants being more supportive of democracy; this could be due to the diversity of Protestant denominations with different perspectives that the aggregate religious tradition measure does not express. Therefore, it seems that, when other dimensions of religiosity are controlled for, no religious tradition except for Hinduism is associated with anti-democratic attitudes. Like the religiosity variables, the individual-level and country-level control variables usually exhibit the expected effects. Engaging in more political discussion, interpersonal trust, higher education, and higher income increase both support for democracy in one’s country and endorsement of the democratic system. On the other hand, authoritarianism lessens support for and positive evaluations of democracy. Other individual level variables do not have systematic effects on the two types of democratic attitudes. The elderly support democracy more, but they do 123 Polit Behav (2012) 34:249–276 263 not necessarily evaluate democracy as an efficient system. Identification with the political right has a negative effect on democratic support, but a positive effect on democratic evaluations. While being male does not have a statistically significant effect on democratic evaluations, it is negatively related to support for democracy. National religious tradition has significant effects on support for and evaluations of democracy. Ceteris paribus, as the percentage of Catholics, Protestants or Muslims increase so does support for democracy, and viewing it as an efficient system. A higher percentage of Orthodox identifiers in a country is associated with more favorable evaluations of democracy, although not with higher support for democracy. We expected extensive religious freedoms to lead to higher support for democracy, and higher government regulation to correlate with less favorable evaluations of democracy. However, both variables have unexpected effects on the dependent variables. The religious freedom scale is negatively related to support for democracy, but has a positive effect on democratic evaluations. This suggests that individuals could be more supportive of democracy in nations with more restricted religious freedoms, perhaps due to a belief that a more open regime would also allow them to practice their faith more freely. On the other hand, in such nations, individuals tend to express more negative evaluations of the democratic system. This could be because of their negative experiences with the current democratic regime. State regulation of religion, on the other hand, does not have a statistically significant effect on individual support for democracy, but a positive effect on endorsement of the democratic system. Next, individuals in countries with a higher Freedom House rating (i.e. those that are less free) and post-Communist nations are more critical of democracy. Contrary to expectations, we find that higher GDP per capita is negatively related to support for democracy, and has no significant effect on viewing democracy as an efficient system, implying that, as far as democratic countries are concerned, socioeconomic development may not be crucial to the formation of democratic values. We also hypothesized that different dimensions of religiosity would have differing effects on ambivalence in democratic attitudes. Note that likelihood ratio tests suggest that the null of homoskedastic models can be rejected at the 95% confidence level in both models.9 As Table 2 shows, both religious belief and religious social activity affect ambivalence towards democracy in the expected direction. Social religious behavior significantly decreases ambivalence about both support for democracy (b = -.41, p = .000) and perceiving democracy as an efficient system in general (b = -.081, p = .000). In contrast, and in accord with our expectations, religious belief significantly increases ambivalence about support for (b = .097, p = .000) and endorsement of (b = .93, p = .001) democracy. Although we did not have any particular expectation for the effect of religious traditions on ambivalence, we also included them in our variance equations. We find that identification as Catholic, Evangelical, Orthodox, Jewish or Muslim increases 9 Note that we took the logs in order to constrain the variance to positive values; thus changes in the independent variables predict changes in the logs of the variance. 123 264 Polit Behav (2012) 34:249–276 ambivalence in viewing democracy as an efficient system compared to not identifying with any religious tradition, while Buddhist and independent identifiers experience less ambivalence. On the other hand, religious identification typically has no or negative effects on ambivalence concerning support for democracy in one’s country. These results suggest that it is rather the strength of one’s belief and the extent of one’s involvement in religious social networks that influence ambivalence, rather than identifying with a specific religious denomination. Other individual-level control variables usually exhibit anticipated effects on the error variance. Authoritarianism increases ambivalence towards democracy, while education and discussing politics reduce ambivalence (with the exception of a null effect of discussing politics on viewing the system as efficient). Unexpectedly, increasing age leads to higher ambivalence in perceiving democracy as an efficient system, and has no effect on support for democracy. Robust Analyses and Moderating Effects of Religious Variables To further challenge our findings, we conducted a series of robust analyses. Firstly, the private and social dimensions of religiosity retained their effects in the mean and variance equations in reduced models, where only individual level variables were specified, and where religious identification was omitted. Results were also replicated when running a non imputed measure for political ideology. Next, whereas ML estimation is sensitive to the definitions of the dependent variable and error term functions, OLS is much more robust under different distributional assumptions and does not rely on the normal error assumption when large samples are involved. Thus, we ran OLS on the mean equation, in which results replicated the differential effect of the private and social dimensions of religiosity on attitudes towards democracy. We also estimated the stochastic component of the MLE model by obtaining the squared residuals from the OLS regression, and we regressed them on the variables from the variance equation (see Franklin 1991), where both dimensions of religiosity retained their expected signs and were more than twice their standard errors in both models. While we hypothesized the independent effects of religiosity dimensions, the three dimensions may moderate each other’s effects. For example, while it is true that those who are more devout tend to participate in social religious networks more, as indicated by the high correlation between religious belief and religious social behavior (r = .56), those who are not strong believers may also participate in communal religious activities due to family or friendship circle influences, and the effect of social religious involvement could be different for them. In addition, those identifying with religions that are traditionally seen as being hostile to democratic norms such as Muslims or Orthodox may become even less supportive as they become more attached to their faith. Identification with a religious tradition may also moderate the effect of social religious behavior on support for democracy, since not all places of worship may function as environments for the discussion of worldly affairs and thereby help identifiers develop their civic skills. We therefore test for some possible refinements of our results by specifying interactions between these variables in the models. 123 Polit Behav (2012) 34:249–276 265 Fig. 1 Interactions between religious belief and behavior First, we find that religious belief moderates the positive effect of social religious behavior on democratic attitudes, as indicated by significant interaction coefficients in both models (support for democracy: b = .094, p = .000; democracy efficient: b = .027, p = .095). This interaction is presented in Fig. 1, with the left hand side panel showing the effect of religious behavior on support for democracy moderated by religious belief, and the right hand side panel—the corresponding effect on evaluation of democracy as an efficient system. The relationship between social religious behavior and support for democracy is positive for those who are moderate and strong believers. However, this positive relationship is inverted as an individual becomes less devout. That is, participation in social religious networks has a negative effect on support for democracy if the individual is not a strong believer. This could be because less devout people may be passive participants in social religious activities, and therefore not very open to positive communal influences. Therefore, while religious communities may play an important role in socializing believers into democratic norms, they do not have such an effect when the participants are not very devout. We also consider whether the effects of religious belief and behavior are contingent on the major religious tradition that the individual identifies with. To test this, we run a series of Heteroskedastic ML regressions, in which religious belief and behavior are separately interacted with each one of the denominations at a time, the baseline being all other denominations. Results show that Catholics and Muslims systematically differ from all other denominations in the extent to which religious belief and religious behavior affect their democratic attitudes. The interactive effects are presented in Fig. 2. The two left hand side panels on Fig. 2 depict the effect of religious belief on support for democracy and evaluation of democracy as an efficient system moderated by religious identification. While belief is negatively related to both 123 266 Polit Behav (2012) 34:249–276 Fig. 2 Interactions among religious identification and religious belief and behavior. Figure depicts the significant interactions between religious identification and religious belief, and identification and religious social behavior. The left side panels present the effect of religious belief on democratic attitudes for Catholics, Muslims and all other religious traditions, while the right side panels present the effect of religious social behavior on democratic attitudes for these three groups dependent variables for all religious traditions, this negative effect is more moderate for Catholics, and significantly sharper for Muslims. As Muslim identifiers become more devout, they become more opposed to democracy, while the level of support among Catholic identifiers does not depend much on their level of belief. The effect of religious behavior on support for democracy and evaluations of the democratic system moderated by religious tradition are presented in the upper and lower right panels of Fig. 2 respectively. The positive effect of involvement in social religious networks is the same for all other denominations, and is stronger for Catholics. That is, religious social behavior increases democratic attitudes among Catholics to a greater extent compared to other religious identifiers. On the other hand, Muslims’ support for democracy is less affected by participation in collective religious activities and in fact, higher participation leads to less favorable evaluations of democracy. That is, attending mosques, more than other places of worship, seem to be associated with anti-democratic attitudes. Overall, these results suggest the robustness of the differential effects of religious belief and behavior among specific denominations, with the exception of Catholics, who show a milder negative effect of belief and a steeper positive effect of behavior, 123 Polit Behav (2012) 34:249–276 267 and Muslims, who show a steeper negative effect of belief and a negative, rather than positive, effect of behavior. We also consider whether religious freedom or state regulation of religiosity moderate the relationship between individual religiosity and democratic attitudes. Since this requires a cross-level interaction between contextual (level-2) and individual (level-1) variables, we test these arguments by running multilevel models (see Steenbergen and Jones 2001). First, our main analysis does not reveal consistent effects of religious freedom or state regulation of religion on support for democracy in the mean equations, raising suspicion that their effects may be contingent on the dimensions of religiosity. We therefore specify cross-level interactions of religious belief, behavior, and denomination with the religious freedom scale and religious regulation index. However, all resulting interactive terms have high standard errors, which forbid rejecting any of the null hypotheses. Thus, as far as the nations in our dataset are concerned, religious identifiers’ support for democracy is not conditional on the level of religious freedoms or government regulation of religion. We also consider the possibility of identifiers of minority religions being more critical of democracy. In order to test for this, we specify cross-level interactions between the percentage of identifiers with a major religion in the country and identifying with this religion (e.g. Protestant identification and percentage of Protestants in country, etc.). These interactions allow us to test whether, as their religious tradition becomes dominant in the country, identifiers differ in their support for democracy compared to identifiers of all other religions.10 Of all possible interactions, the only significant effects emerged for the percentage of Orthodox nations, such that as non-Orthodox identifiers become the minority, they become less supportive of democracy. This is also in line with the finding of Marsh (2005), who finds that devout Orthodox Christians in Russia are more civic minded and committed to democracy than many of their non-Orthodox fellow citizens. However, the reason why non-Orthodox minorities become less supportive of democracy in these nations could be due to high regulation of religion and the lack of religious freedoms in predominantly Orthodox nations.11 In order to test this 10 Each model included one religious tradition at a time such that the baseline is all others. For example, the multilevel model that tests for the effect of being a Protestant on support for democracy conditional on the dominance of Protestant tradition in a country P is: The multilevel model for attitudes towards democracy is Yij ¼ b0j þ b1j  X1ij þ b2j  X2j þ pq¼3 bqj þ eij ; where Yij is support for democracy for individual i country j, b0j is the country-level intercept, X1ij Protestant dummy for individual i in country j, X2j is percent Protestant in country j, Since we control for the effect of Protestant identification on support for democracy as Protestantism becomes more dominant, the slope of the Protestant identification (b1j) variable is defined by the equation: b1j = c10 ? c11 9 X2j ? d1j. In alternative models, we also included the ‘‘no denomination’’ dummy, so that the baseline is ‘‘all other identifiers’’, which did not lead to substantive changes in the results. 11 In fact, a number of predominantly Orthodox countries are also characterized by restricted religious freedoms and by high levels of regulation of religious activity. 123 268 Polit Behav (2012) 34:249–276 possibility, we run three-way interactions between percent Orthodox, Orthodox identification, and the religious freedom index.12 The three-way interaction is statistically significant, which implies that non-Orthodox minorities are more supportive of democracy in Orthodox nations with higher levels of religious freedoms. On the other hand, we are not able to replicate this finding for religious regulation. In addition, we run the same models for other religious traditions and find no significant evidence for an effect of religious freedoms and regulation conditional on being a minority in nations. Conclusions We build on current theories of political psychology and on the multidimensional conceptualization of religiosity to settle the debate about the effect of religiosity on attitudes towards democracy. The results confirm our hypotheses about the differential effect of religiosity on attitudes towards democracy and on ambivalence in these attitudes. The social-behavioral aspect of religiosity leads to greater endorsement of the democratic system and generates stronger support for democracy; furthermore this effect is stronger for believers. The positive effect of involvement is stronger for Catholics and weaker for Muslims. The variance component of the Heteroskedastic ML model shows that social religious behavior also decreases the ambivalence and value conflict individuals may experience, as involvement in homogeneous networks and interaction with like-minded people leads to more stable and less ambivalent attitudes. On the other hand, religiosity is a double-edged sword, since religious belief has a negative impact on attitudes towards and evaluations of democracy. These results are robust when considering the strength of beliefs for identifiers of different religious traditions, except for Catholics, whose democratic support and evaluations are less affected by their level of devoutness. In addition, we find that the negative effect of religious belief on democratic support is stronger for Muslim identifiers. Religious belief also leads to more ambivalence towards democracy and democratic norms due to its inherent conflict with democratic values. Another important finding is that, once controlled for the belief and social behavior dimensions, identification with a major religious tradition does not seem to have a decisive impact on the strength of attitudes towards democracy or on ambivalence in these attitudes. Although our results consistently confirm more 12 The multilevel model for attitudes towards democracy is Yij ¼ b0j þ b1j  X1ij þ b2j  X2j þ b3j  P X3j pq¼4 bqj  Xqij þ eij ; where Yij is support for democracy for individual i country j, b0j is the country-level intercept, X1ij Orthodox dummy for individual i in country j, X2j is percent Orthodox in country j, X3j is religious freedom index value for country j. Since we are interested in the level of support of Orthodox identifiers conditional on the level of religious freedoms in countries as percent orthodox increases, the slope of Orthodox identification b1j is random and is defined by the equation: b1j = c10 ? c11 9 X2j ? c12 9 X3j ? c13 9 X2j 9 X3j ? d1j. 123 Polit Behav (2012) 34:249–276 269 anti-democratic attitudes for Hindu identifiers, the Orthodox, Jews, Catholics and Muslims hold more pro-democratic attitudes compared to those who do not identify with a major religious tradition, and there are typically no statistically significant effects for Protestants, Evangelicals, or Buddhists. These results support previous research finding that Muslim and Orthodox identifiers have positive attitudes towards democracy. In addition, we do not find any consistent results for the effect of religious identification on attitudinal ambivalence. We believe that, while the debate in literature on the effect of major religious tradition identification on support for democracy still has merit and is relevant, future research should concentrate on the effect of congregational, denominational or sectarian identification on democratic norms and attitudes. Our results also reveal that while the national context plays an important role in individual attitudes towards democracy, it does not necessarily condition the effect of religious variables. Our analysis has certain limitations. First, we did not include all aspects of democratic political culture (such as civic norms, political participation, etc.) but concentrated on assessing the effect of religiosity on support for democracy and democratic evaluations. Thus, our results are yet to be expanded to other democratic norms. In addition, future studies could use a more fine tuned measure of religious belonging, which was not available in this project due to the limitations of the WVS dataset. But above all, this study contributes to the literature by settling the contrasting evidence regarding the effects of religiosity on democratic attitudes. While some argue that religiosity poses a challenge to democracy, others suggest that religiosity may have a positive effect on democratic practices. These results are typically incomparable, as they employ different measures and different case studies, and fail to control for all the dimensions of religiosity or for country level religious characteristics. Using pooled representative samples from 45 democracies, we show that religiosity, depending on whether the belief or social behavior dimensions are measured, produces divergent consequences as far as its effects on democratic norms and behaviors are concerned. Moreover, we show that religiosity not only impacts democratic attitudes, but also affects ambivalence towards democracy and democratic behavior. These results are robust across a variety of religious traditions, as well as different contextual effects, making a persuasive case that the effects of religiosity is at the same time complex and multidimensional, but also coherent beyond a specific case study. Acknowledgment We would like to thank two anonymous reviewers and the editors for their insightful readings and constructive suggestions. All remaining errors are our own. Appendix See Appendix Tables 3, 4. 123 270 123 Table 3 Summary statistics by religious affiliation and country Total Catholic Protestant Evangelical Orthodox Jewish Muslim Buddhist Hindu Indep/other No denom. Endorses democracy in one’s country .526 .528 .574 .484 .476 .580 .593 .551 .485 .548 .514 Democracy is efficient .719 .726 .762 .678 .654 .789 .779 .711 .629 .715 .711 Religious belief .603 .682 .623 .856 .615 .638 .832 .411 .749 .692 .356 Social religious behavior .359 .423 .479 .736 .309 .481 .503 .376 .425 .594 .099 Discuss politics .434 .426 .427 .341 .481 .495 .460 .429 .357 .397 .445 Authoritarianism .351 .389 .348 .559 .268 .369 .424 .100 .571 .415 .269 Trust people .262 .238 .348 .163 .186 .279 .211 .394 .376 .309 .286 National pride .780 .818 .824 .897 .722 .773 .839 .692 .867 .776 .683 Political right .501 .514 .508 .501 .485 .475 .563 .543 .515 .519 .451 Education .649 .608 .663 .642 .685 .780 .631 .710 .499 .694 .696 Income .403 .400 .418 .299 .408 .448 .372 .396 .323 .382 .430 Age 43.12 44.47 44.68 38.42 40.01 42.43 37.25 51.46 39.45 42.99 41.59 Male .471 .444 .454 .461 .404 .561 .534 .486 .558 .409 .528 Percent Catholic 40.39 – – – – – – – – – – 13.20 – – – – – – – – – – Percent Orthodox Kas.23 – – – – – – – – – – Percent Muslim Tem.55 – – – – – – – – – – Religious regulation 1.269 – – – – – – – – – – Religious freedom scale 80.81 – – – – – – – – – – Freedom House rating of non-democracy 1999 2.089 – – – – – – – – – – GDP PPP 2000 (logged) 8.788 – – – – – – – – – – Post-Communist .326 – – – – – – – – – – N 57599 21,647 7,783 1,239 6,754 214 4,026 580 1,802 13,554 Polit Behav (2012) 34:249–276 Percent Protestants Endorses democracy Democracy efficient Religious belief Social religious behavior Albania .617 .803 .670 .342 8.8 Argentina .483 .680 .693 .307 78.3 Austria .642 .834 .572 .380 Bangladesh .640 .853 .964 Belgium .512 .742 Bulgaria .535 .661 Catholics (%) Protestant (%) Evangelical (%) Orthodox (%) Jewish (%) Muslim (%) Buddhist (%) Hindu (%) Independent/ other (%) No denom. affil. N .0 1.3 18.1 .0 58.2 .0 .0 .4 4.9 .1 .8 .0 .1 .0 .6 13.0 1000 2.3 13.1 80.0 5.1 .0 .7 .0 .2 .0 .0 .9 12.4 1279 1522 .566 .1 .1 .0 .1 .1 91.9 .5 7.1 .0 .1 1500 .468 .235 54.9 1.2 .0 .4 .1 3.0 .0 .0 4.3 35.4 1912 .441 .219 .3 .7 .0 60.3 .0 8.4 .0 .0 .2 29.8 1000 Canada .521 .756 .630 .392 42.8 21.8 4.9 .7 .3 .8 .4 .6 .8 26.5 1931 Chile .544 .648 .734 .358 53.9 2.0 7.6 .0 .2 .0 .0 .1 2.3 33.9 1200 Croatia .618 .816 .622 .366 84.7 .0 .0 .1 .0 .0 .0 .0 1.8 13.2 1003 Czech Republic .485 .778 .250 .151 29.7 3.8 .0 .1 .1 .0 .0 .0 1.3 64.5 1908 1023 Denmark .611 .859 .424 .197 .8 86.1 .0 .0 .0 .5 .1 .1 1.4 10.0 Estonia .529 .689 .326 .191 .4 13.1 .0 9.8 .1 .1 .0 .0 .7 75.7 1005 Finland .567 .714 .500 .392 .1 83.7 .0 1.1 .0 .0 .0 .0 2.9 11.7 1038 France .428 .720 .393 .127 52.6 1.3 .0 1.2 1.4 .1 .2 .1 .4 42.5 1615 Germany .641 .801 .338 .230 21.5 34.5 .0 .4 .0 1.1 .0 .0 2.4 39.6 2030 Great Britain .595 .744 .444 .153 13.8 56.6 .0 .2 .6 .9 .6 .5 11.0 15.0 1000 Greece .513 .907 .604 .338 1.5 .0 .0 93.6 .4 .0 .0 .0 .3 4.0 1142 1000 Hungary .487 .733 .448 .215 39.0 16.1 .0 .2 .1 .0 .0 .1 1.8 42.1 Iceland .669 .852 .519 .517 .4 90.6 .0 .0 .0 .0 .0 .0 4.2 4.3 968 India .462 .613 .722 .409 3.0 .0 .0 .0 .3 11.0 1.4 76.3 1.4 6.7 2002 .577 .756 .686 .472 88.8 2.0 .0 .2 .0 .0 .0 .1 1.8 6.9 1012 Italy .535 .816 .625 .365 81.2 .3 .0 .1 .0 .0 .0 .0 .3 17.8 2000 Japan .566 .731 .320 .276 .9 1.0 .0 .6 .1 .0 37.6 .0 4.2 55.6 1362 Latvia .454 .568 .499 .195 19.6 16.9 .0 16.7 .0 .1 .0 .0 5.7 40.7 1013 271 123 Ireland Polit Behav (2012) 34:249–276 Table 4 Key summary statistics by country 272 123 Table 4 continued Endorses Democracy Religious Social Catholics Protestant Evangelical Orthodox Jewish Muslim Buddhist Hindu Independent/ No N democracy efficient belief religious (%) (%) (%) (%) (%) (%) (%) (%) other (%) denom. behavior affil. Lithuania .462 .594 .555 .261 74.6 1.3 .0 2.9 .0 .0 .1 .0 1.1 19.4 1018 Luxembourg .597 .692 .463 .252 64.1 .2 .0 .4 .6 .6 .0 .0 2.7 29.9 1211 Macedonia .494 .563 .618 .346 .3 .4 .0 59.7 .1 25.3 .0 .0 .2 14.1 1051 Malta .631 .811 .850 .497 97.7 .9 .0 .0 .0 .0 .0 .0 .1 1.3 1022 Mexico .473 .603 .815 .482 72.7 1.4 3.2 .0 .1 .0 .1 .1 1.7 20.6 1535 Montenegro .539 .692 .639 .375 5.7 .2 .0 .0 69.1 20.9 .0 .0 .1 4.1 1060 Netherlands .586 .747 .388 .331 22.1 10.1 .0 .0 .4 1.1 .2 .2 11.0 55.0 1003 Peru .458 .679 .793 .486 84.0 .0 10.7 .0 .2 .0 .2 .0 .0 4.7 1501 Philippines .452 .557 .881 .543 73.7 1.9 8.9 .0 .1 3.2 .0 .0 1.2 11.0 1200 Poland .388 .685 .738 .405 94.0 .3 .0 .3 .0 .1 .0 .0 .7 4.6 1095 Republic of Moldova .505 .537 .649 .345 2.9 1.2 .0 94.1 1.3 .0 .0 .0 .4 .0 1008 Romania .424 .567 .723 .334 7.4 2.0 .0 84.7 .0 .0 .0 .0 2.4 2.4 1146 Russian Federation .394 .521 .464 .136 .2 .3 .0 47.5 .0 3.0 .1 .0 .4 48.1 2500 Serbia .512 .650 .569 .318 1.3 .3 .0 61.4 .3 4.9 .0 .0 .3 31.7 400 Slovakia .481 .711 .564 .360 64.2 11.2 .0 .8 .2 .0 .0 .0 .5 23.1 1331 .498 .694 .851 .644 9.8 42.4 21.2 .4 .3 4.2 .1 5.4 3.2 11.0 3000 Spain .627 .754 .509 .257 82.1 .6 .0 .1 .1 .1 .2 .1 1.1 15.7 1209 Tanzania .587 .897 .891 .716 28.4 18.7 .0 5.0 3.6 40.4 .0 .1 2.0 1.7 1171 Uganda .592 .749 .876 .642 36.6 43.0 1.1 .4 .0 17.0 .0 .1 .7 1.1 1002 Ukraine .478 .551 .534 .219 .9 2.2 .0 42.4 .2 .3 .1 .0 8.8 42.1 1195 United States .571 .715 .755 .608 24.5 26.2 .0 .8 4.5 .3 .7 .2 20.6 20.9 1200 The entries in columns 1–4 are means by country, those in columns 5–14 are the percentage affiliated with a certain religious denomination in any one country, those in column 15 are the percountry sample size Polit Behav (2012) 34:249–276 South Africa Polit Behav (2012) 34:249–276 273 References Aaron, S. J. (2009). Emulating Azariah: Evangelicals and social change in the Dangs. In D. H. Lumsdaine (Ed.), Evangelical christianity and democracy in Asia (pp. 83–131). New York, NY: Oxford University Press. Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, N. R. (1950). The authoritarian personality. New York: Harper and Row. Altemeyer, R. (1996). The authoritarian specter. Cambridge: Harvard University Press. Alvarez, M. R., & Brehm, J. (1995). American ambivalence toward abortion policy: Development of a heteroskedastic probit model of competing values. American Journal of Political Science, 39(4), 1055–1082. Alvarez, M. R., & Brehm, J. (1997). Are Americans ambivalent towards racial policies? American Journal of Political Science, 92(2), 345–374. Alvarez, M. R., & Brehm, J. (2002). Hard choices. Easy Answers. Princeton, NJ: Princeton University Press. Anderson, J. (2004). Does God matter, and if so, whose God? Religion and democratization. Democratization, 11(4), 197–217. Barnea, M. F., & Schwartz, S. H. (1998). Values and voting. Political Psychology, 19(1), 17–40. Barro, R. J., & McCleary, R. M. (2003a). Religion and economic growth across countries. American Sociological Review, 68(5), 760–781. Barro, R. J., & McCleary, R. M. (2003b). International determinants of religiosity. NBER Working Paper No. w10147. Basinger, S. J., & Lavine, H. (2005). Ambivalence, information, and electoral choice. American Political Science Review, 99(2), 169–184. Beit-Hallahmi, B., & Argyle, M. (1997). The psychology of religious behavior. Belief, and experience. London and New York: Routledge. Ben-Nun Bloom, P., Zemach, M., & Arian, A. (2011). Religiosity and democratic performance evaluation. Democratization, 18(1), 25–51. Bollen, K. A., & Jackman, R.W. (1985). Economic and noneconomic determinants of political democracy in the 1960s. Research in Political Sociology, 1(1), 27–48. Bratton, M. (2003). Briefing: Islam, democracy and public opinion in Africa. African Affairs, 102(408), 493–501. Bratton, M., & Mattes, R. (2001). Support for democracy in Africa: Intrinsic or instrumental? British Journal of Political Science, 31(3), 447–474. Bromley, S. (1997). Middle East exceptionalism: Myth or reality? In D. Potter, D. Goldblatt, M. Kiloh, & P. Lewis (Eds.), Democratization (pp. 321–344). Cambridge: Polity. Burris, C. T., & Tarpley, W. R. (1998). Religion as being: Preliminary validation of the Immanence Scale. Journal of Research in Personality, 32(1), 55–79. Canetti-Nisim, D. (2004). The effect of religiosity on endorsement of democratic values: The mediating influence of authoritarianism. Political Behavior, 26(4), 377–398. Chaibong, H. (2004). The ironies of Confucianism. Journal of Democracy, 15(3), 93–107. Dekker, P., & Halman, L. (2003). The values of volunteering: Cross-cultural perspectives. New York: Kluwer Academic Plenum Publishers. Djupe, P. A., & Gilbert, C. G. (2006). The resourceful believer: Generating civic skills in church. The Journal of Politics, 68(1), 116–127. Durlauf, S. N., Kourtellos, A., & Tan, C. M. (2005). How robust are the linkages between religiosity and economic growth? Tufts University Discussion Paper Series. Eisenstein, M. A. (2006). Rethinking the relationship between religion and political tolerance in the US. Political Behavior, 28(4), 327–348. Fahlbusch, E., Lochman, M., Jan, B., William, G., Barret, D. B., & Mbiti, J. (2005). The encyclopedia of Christianity. Winona Lake, IN: Eerdmans. Feldman, S. (1995). Answering survey questions: The measurement and meaning of public opinion. In M. Lodge & K. M. McGraw (Eds.), Political judgment: Structure and process (pp. 249–271). Ann Arbor, MI: University of Michigan Press. Feldman, S. (2003). Enforcing social conformity. Political Psychology, 24(1), 41–74. Feldman, S., & Zaller, J. R. (1992). The political culture of ambivalence: Ideological responses to the welfare state. American Journal of Political Science, 36(1), 268–307. 123 274 Polit Behav (2012) 34:249–276 Fernandes, S. (2009). Ethnicity, civil society, and the church: The politics of evangelical Christianity in northeast India. In D. H. Lumsdaine (Ed.), Evangelical Christianity and democracy in Asia (pp. 131–155). New York and Oxford: Oxford University Press. Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7(2), 117–140. Filali-Ansary, A. (1999). Muslims and democracy. Journal of Democracy, 10(3), 18–32. Franklin, C. (1991). Eschewing obfuscation? Campaigns and the perception of U.S. Senate candidates. American Political Science Review, 85(4), 1193–1214. Fukuyama, F. (1992). The end of history and the last man. New York: Avon. Gibson, J. L. (1992). The political consequences of intolerance: cultural conformity and political freedom. American Political Science Review, 86(2), 338–356. Gifford, P. (2008). Evangelical Christianity and democracy in Africa: A response. In T. O. Ranger (Ed.), Evangelical Christianity and democracy in Africa (pp. 225–230). New York and Oxford: Oxford University Press. Gilbert, C. P. (1993). The impact of Churches on political behavior: An empirical study. Westport, CT: Greenwood Press. Hinnells, J. R. (1997). A new handbook of living religions. Malden, MA: Wiley-Blackwell. Huckfeldt, R., Mendez, J. M., & Osborn, T. (2004). Disagreement, ambivalence, and engagement: The political consequences of heterogeneous networks. Political Psychology, 25(1), 65–95. Hunsberger, B. (1995). Religion and prejudice: the role of religious fundamentalism, quest and right-wing authoritarianism. Journal of Social Issues, 51(2), 113–129. Huntington, S. P. (1996). The clash of civilizations and the remaking of world order. New York: Simon and Schuster. Inglehart, R., & Welzel, C. (2005). Modernization, cultural change, and democracy: The human development sequence. Cambridge, UK and New York: Cambridge University Press. Jelen, T. G. (1992). Political Christianity: A contextual analysis. American Journal of Political Science, 36(3), 692–714. Jelen, T. G., & Chandler, M. A. (1996). Communalism, associationalism, and the politics of lifestyle. Review of Religious Research, 38(2), 142–158. Jelen, T. G., & Wilcox, C. (2002). Religion and politics in comparative perspective: The one the few, and the many. Cambridge, UK and New York: Cambridge University Press. Karpov, V. (2002). Tolerance in the United States and Poland. Journal for the Scientific Study of Religion, 41(2), 267–288. Kedourie, E. (1994). Democracy and Arab political culture. London: Frank Cass. Kellstedt, L. A., Green, J. C., Guth, J. L., & Smidt, C. E. (1997). Is there a culture war? Religion and the 1996 election. Paper presented at the Annual Meeting of the American Political Science Association, August 28–31. DC: Washington. Lau, S. (1989). Religious schema and values. International Journal of Psychology, 24(2), 137–156. Lavine, H. (2001). The electoral consequences of ambivalence toward presidential candidates. American Journal of Political Science, 45(4), 915–929. Layman, G. C. (1997). Religion, political behavior in the United States: The impact of beliefs, affiliations, commitment from 1980 to 1994. Public Opinion Quarterly, 61(2), 288–316. Leege, D. C. (1988). Catholics and the civic order: parish participation, politics, and civic participation. Review of Politics, 50(4), 704–736. Lipset, S. M. (1959). Some social requisites of democracy: Economic development and political legitimacy. American Political Science Review, 53(1), 69–105. Lipset, S. M. (1981). Political man: The social basis of politics. Baltimore: Johns Hopkins University Press. Lumsdaine, D. H. (2009). Evangelical Christianity and democracy in Asia. New York and Oxford: Oxford University Press. Manza, J., & Brooks, C. (1997). The religious factor in U.S. presidential elections, 1960–1992. American Journal of Sociology, 103, 38–81. Marsh, C. (2005). Orthodox Christianity, civil society, and Russian democracy. Demokratizatsiya, 13(3), 449–462. Mehta, P. B. (2004). Hinduism and self-rule. Journal of Democracy, 15(3), 108–121. Meyer, K., Tope, D., & Price, A. M. (2008). Religion and support for democracy: A cross-national examination. Sociological Spectrum, 28(5), 625–653. Mueller, T. (2009). Religiosity and attitudes towards the involvement of religious leaders in politics: A multilevel analysis of 55 societies. World Values Research, 2(1), 1–29. 123 Polit Behav (2012) 34:249–276 275 Muller, E. N., & Seligson, M. A. (1994). Civic culture and democracy: The question of causal relationships. The American Political Science Review, 88(3), 635–652. Mutz, D. C. (2002). The consequences of cross-cutting networks for political participation. American Journal of Political Science, 46(4), 838–855. Mutz, D. C., & Mondak, J. J. (2006). The workplace as a context for cross-cutting political discourse. The Journal of Politics, 68(1), 140–155. Norris, P. (2002). Democratic phoenix: Reinventing political activism. Cambridge, UK and New York: Cambridge University Press. Norris, P., & Inglehart, R. (2004). Sacred and secular: Religion and politics worldwide. Cambridge, UK and New York: Cambridge University Press. Philpott, D. (2004). The Catholic wave. Journal of Democracy, 15(2), 32–46. Putnam, R. D. (2000). Bowling alone. New York: Simon and Schuster. Radu, M. (1998). The burden of Eastern Orthodoxy. Orbis, 42(2), 283–300. Ranger, T. O. (2008). Evangelical Christianity and Democracy in Africa. New York and Oxford: Oxford University Press. Reise, S. P., Widaman, K. F., & Pugh, R. H. (1993). Confirmatory factor analysis and item response theory: Two approaches for exploring measurement invariance. Psychological Bulletin, 114(3), 552–566. Roccas, S., & Schwartz, S. S. (1997). Church-state relations and the associations of religiosity with values: A study of Catholics in six countries. Cross-Cultural Research, 31, 356–375. Rokeach, M. (1969). Value systems and religion. Review of Religious Research, 11(1), 2–23. Rosema, M. (2007). Low turnout: Threat to democracy or blessing in disguise? Consequences of citizen’s varying tendencies to vote. Electoral Studies, 26(3), 612–623. Rudolph, T. J. (2005). Group attachment and the reduction of value-driven ambivalence. Political Psychology, 26(6), 905–928. Saroglou, V., Delpierre, V., & Dernelle, R. (2004). Values and religiosity: A meta-analysis of studies using Schwartz’s model. Personality and Individual Differences, 37(4), 721–734. Scheufele, D. A., Nisbet, M. C., Brossard, D., & Nisbet, E. C. (2004). Social structure and citizenship: Examining the impacts of social setting, network heterogeneity, and informational variables on political participation. Political Communication, 21(3), 315–338. Schnell, F. (1993). The foundation of abortion attitudes: The role of values and value conflict. In M. L. Coggin (Ed.), Understanding the new politics of abortion. Newbury Park: Sage. Schwartz, S. H. (1992). Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries. In M. Zanna (Ed.), Advances in experimental social psychology (Vol. 25). New York: Academic Press. Schwartz, S. H., & Huismans, S. (1995). Value priorities and religiosity in four Western religions. Social Psychology Quarterly, 58(1), 88–107. Schwartz, S. H., & Sagie, G. (2000). Value consensus and importance: A cross-national study. Journal of Cross-Cultural Psychology, 31(4), 465–497. Shah, T. S. (2004). The Bible and the ballot box: Evangelicals and democracy in the ‘‘Global South’’. SAIS Review, 24(2), 117–132. Sniderman, P. M., Fletcher, J. F., Russell, P. H., & Tetlock, P. E. (1996). The clash of rights: Liberty, equality, and legitimacy in a pluralist democracy. New Haven: Yale University Press. Steenbergen, M. R., & Brewer, P. R. (2004). The not-so-ambivalent public: Policy attitudes in the political culture of ambivalence. In W. E. Saris & P. M. Sniderman (Eds.), Studies in public opinion: Attitudes, non-attitudes, measurement error, and change (pp. 93–129). Princeton, NJ: Princeton University Press. Steenbergen, M. R., & Jones, B. (2001). Modeling multilevel data structures. American Journal of Political Science, 46(1), 218–237. Tessler, M. (2002). Do Islamic orientations influence attitudes toward democracy in the Arab world? Evidence from Egypt, Jordan, Morocco, and Algeria. International Journal of Comparative Sociology, 43(3–5), 229–249. Van de Vijver, F. J. R. (2003). Bias and equivalence: Cross-cultural perspectives. In J. A. Harkness, F. J. R. Van de Vijver, & P. P. Mohler (Eds.), Cross-cultural survey methods (pp. 143–155). Hoboken, NJ: John Wiley & Sons, Inc. Van de Vijver, F., & Leung, K. (1997). Methods and data analysis for cross-cultural research. Newbury Park, CA: Sage Publications. 123 276 Polit Behav (2012) 34:249–276 Verba, S., Schlozman, K. L., & Brady, H. (1995). Voice and equality: Civic voluntarism in American politics. Cambridge: Harvard University Press. Visser, P. S., & Mirabile, R. R. (2004). Attitudes in the social context: The impact of social network composition on individual-level attitude strength. Journal of Personality and Social Psychology, 87(6), 779–795. Wald, K. D. (2003). Religion and politics in the United States. Lanham, MD: Rowman and Littlefield. Wald, K. D., Owen, D. E., & Hill, S. S. (1988). Churches as political communities. American Political Science Review, 82(3), 531–548. Wald, K. D., Owen, D. E., & Hill, S. S. (1990). Political cohesion in churches. The Journal of Politics, 52(2), 197–215. Wald, K. D., & Smidt, C. E. (1993). Measurement strategies in the study of religion and politics. In D. C. Leege & L. A. Kellstedt (Eds.), Rediscovering the religious factor in American politics (pp. 26–52). New York: M.E. Sharpe. Weightman, S. (1997). Hinduism. In J. R. Hinnels (Ed.), A new handbook of living religions (pp. 191–255). Malden, MA: Wiley-Blackwell. Widgery, A. G. (1936). Living religions and modern thought. New York: Round Table Press. Zaller, J. R. (1992). The nature and origins of mass opinion. Cambridge, UK and New York: Cambridge University Press. Zaller, J. R., & Feldman, S. (1992). A simple theory of the survey response: Answering questions versus revealing preferences. American Journal of Political Science, 36(3), 579–616. 123