Wars and Rumors of Wars: The Contexts of Cultural Conflict in American Political Behavior

Wars and Rumors of Wars: The Contexts of Cultural Conflict in American Political Behavior Geoffrey C. Layman Department of Political Science Vanderbilt University Nashville, TN 37235 geoff.layman@vanderbilt.edu and John C. Green Bliss Institute of Applied Politics The University of Akron Akron, OH 44325 green@uakron.edu

Abstract There has been a heated scholarly debate over the culture wars thesis in American politics. Drawing on the literature on mass opinion constraint and its sources, we propose a resolution to this debate: the culture wars influence mass political behavior in special religious, policy, and political contexts where logical, psychological, social, and electoral sources of opinion constraint are in effect. We test our argument with data from the 1992, 1996, and 2000 National Election Studies and find considerable support for it. We conclude that the cultural wars are waged by limited religious troops on narrow policy fronts under special leadership, and a broader cultural conflagration is largely a rumor.

And when you hear of wars and rumors of wars, do not be alarmed; this must take place, but the end is not yet. Gospel of Mark 13:7 RSV First-century Christians used these words to describe the cultural conflicts of their day. Recently, these terms have been applied to American politics, neatly encapsulating a heated scholarly debate. Some scholars argue that the nation is in the throes of increasingly bitter "culture wars" between rival religious groups, with important consequences for mass political behavior. Others argue the opposite: the country is vexed by "rumors of wars," reflecting more limited cultural tensions among religious groups, with only modest relevance to mass politics. Drawing on the literature on mass opinion constraint, we propose a resolution to this debate: the usefulness of the culture wars thesis varies by policy, religious, and political context. The culture wars strongly influence mass political behavior when religious perspectives are logically related to political issues, communal experiences encourage these connections, and electoral actors emphasize and differentiate themselves on such matters. Outside of these contexts, they have little political impact. Using data from the presidential-year National Election Studies from 1992 through 2000, we examine the culture wars thesis for different policy areas, religious communities, and electoral circumstances. The results strongly support our resolution. In sum, the culture wars are waged by limited troops on narrow battlefields under special leadership, and a broader cultural conflagration is just a rumor. The Culture Wars in Context Although the "culture wars" came to public attention with Pat Buchanan s speech at the 1992 Republican National Convention, it was James Davison Hunter s 1991 book of the same title that established the concept in scholarly discourse (see also Hunter 1994). According to Hunter, the culture wars arise from two incommensurate philosophical "impulses," rooted in different systems of moral understanding (1991:43-44), which he labels orthodoxy and 1

progressivism. 1 The former is characterized by a commitment to "external, definable, and transcendent" sources of moral authority, while the latter adheres to a relativistic view of moral authority as changing with historical circumstances and the boundaries of human knowledge. The principal troops in the culture wars are religious groups defined by the orthodoxprogressive division over religious beliefs and behaviors. The key cleavages are no longer between the historic faith traditions (e.g. Protestants versus Catholics), but rather within and across them (e.g. orthodox Protestants and Catholics versus their progressive counterparts), so that membership in such traditions is now virtually irrelevant to politics (Hunter 1991:43). The principal battlefields of the culture wars are moral issues (e.g. abortion, gay rights, family values"), where the implications of orthodox-progressive religious differences are most clear. However, Hunter argues that the rhetorical leadership (1991: 281) of orthodox and progressive elites (including religious, intellectual and political leaders) extends the influence of this religious divide to a wide range of issues, from social welfare to civil rights to environmental policy, thus creating an isomorphism between religious conservatism and political preservationism... and between religious liberalism and... and political reformism (1991:128). Scholars generally agree that there are real culture wars with broad political consequences among elites (Williams 1997:291-92; Guth et al. 1997; Green et al. 1996, Wilcox 2000; Layman 2001). Where disagreement arises is on the extent to which such elite disputes influence mass political behavior. Some researchers find little evidence of mass polarization along orthodoxprogressive religious lines or on policy issues, or of a firm connection between religious perspectives and political attitudes (Davis and Robinson 1996a, 1996b, 1997; Demerath and Yang 1997; DiMaggio, Evans, and Bryson 1996; Smith et al. 1997; Wolfe 1998). Other work, however, identifies the orthodox-progressive cleavage as central to the recent restructuring of 1 We use Hunter s terms for ease of presentation, but they are potentially misleading. Religious traditionalists (rather than the orthodox) and modernists (rather than progressives) are probably more useful terms. 2

American religion (Wuthnow 1988, 1989; Himmelfarb 1999; White 2003) and finds strong links between that cleavage, some core values and policy attitudes, and political behavior (Green et al.1996; Jelen 1997; Wilcox 2000; Green 2000; Kohut et al. 2000; Layman 2001). Thus, the dispute is between those who contend that there are real culture wars in mass politics and those who argue that they are just a rumor. The literature on mass opinion constraint and its sources can help to resolve this dispute. The culture wars thesis implies that the orthodox-progressive division produces high levels of both horizontal constraint (tying various types of issue attitudes together into something like a single ideological dimension), and vertical constraint tying the core values of the competing camps to policy attitudes and to political behavior. Our focus here is on the possibility of vertical connections between religious perspectives, political values and attitudes, and political behavior. 2 Predictions of high levels of mass constraint run counter to the accepted wisdom in the literature on public opinion and electoral behavior. Scholars searching for horizontal constraint have found that mass policy attitudes generally lack a coherent structure (Converse 1964; Campbell et al. 1960; Converse and Markus 1979; Smith 1989), while related research finds limited evidence of vertical connections between core values, policy attitudes, and political behavior (Campbell et al. 1960; Brody and Page 1972; Zaller 1992; Miller and Shanks 1996). However, the same literature identifies several sources of constraint that may tie mass political orientations together horizontally and/or vertically. In his seminal article, Converse (1964) identified three key sources of constraint: logical, psychological, and social. Other scholars denote a fourth source of constraint (which we call an electoral source of constraint) that is distinctively political (cf. Carmines and Stimson 1989; Nie, Verba, and Petrocik 1976). In keeping with the generally low levels of mass opinion constraint, these sources operate only in 2 Scholars examining horizontal constraint in mass policy attitudes find little evidence of the unidimensional structure implied by the culture wars thesis (cf. Olson 1997; Layman and Carsey 2002). 3

particular policy, social, and political contexts. Thus, there may well be connections between religious perspectives, policy attitudes, and political behavior, but only in these special contexts. Logical constraint and policy contexts. Logic can be a potent source of opinion constraint when there is an inherent rationale for the relationship between political objects, so that simply thinking about a set of idea-elements may weld them together (Converse 1964:209). However, as Converse (1964: 209-210) argues, logic can tie only a narrow range of elements together, and it is rarely a source of constraint in the mass public because most citizens think little about politics and have limited knowledge of most political objects (see also Delli Carpini and Keeter 1996). Thus, the logic of religious perspectives is unlikely to produce high levels of constraint across a wide swath of the electorate or among a wide range of political orientations. However, logic may produce vertical constraint between religious perspectives and attitudes toward a narrow range of policy concerns and, through those attitudes, to political behavior that tap directly into the moral visions proffered by the orthodox and progressive camps. For example, on moral issues such as abortion, gay rights, and family values, moral logic... preclude[s] or endorse[s] the specific proposals from the outset (Hunter 1991:127). The orthodox are likely to see a clear, logical relationship between their traditionalist religious and moral values, traditional moral values, opposition to abortion, homosexuality, and feminism, and voting for conservative candidates. Likewise, religious progressives are likely to see a clear, logical connection between their religion, moral relativism, support for abortion rights, gay rights, and women s rights, and voting for liberal candidates. Such a logic is likely to be much less clear for the linkage between these religious perspectives and non-moral core values, such as egalitarianism and individualism, and in other policy areas, such as social welfare, civil rights, and national defense. For example, it may be logical for the orthodox to have egalitarian values and support welfare programs ( as you have done to the least of these, so you have done unto me ), but it may be just as logical for them to 4

have individualistic values and oppose welfare programs ( the poor will be with you always ). Similarly, it might be perfectly logical for religious progressives to support social welfare programs, but it is just as logical for them to adopt non-progressive positions on welfare for what in the logic of moral relativism would lead to caring for the poor? All of this suggests that the impact of the orthodox-progressive divide is likely to be strongest for moral policy attitudes. The connection between religious orthodoxy and non-moral attitudes should be weaker and less consistent, appearing only where other sources of constraint compensate for the absence of a moral logic. Psychological/social constraint and religious contexts. Psychological and social factors are much more likely than pure logic to serve as sources of mass opinion constraint (Converse 1964: 209). These sources of constraint do not depend on high levels of political information or knowledge, but on personal experiences. Psychological constraints tie attitudes together when life experiences package attitudes with superordinate values or identifications; social constraints do the same when group leaders package attitudes for mass consumption and social interaction promotes and re-enforces such packaging (Converse 1964:210-212). The role of personal experiences in generating psychological and social constraints means that the nature and extent of vertical constraint varies by social context (Berelson, Lazarsfeld, and McPhee 1954; Huckfeldt and Sprague 1995). In terms of linking religious perspectives to politics, the most important social context is membership in a religious community or tradition. The strength and nature of psychological constraint should vary across religious communities because they both promulgate core values and explain the relevance of such values for politics. Religious groups are certainly a key source of the competing moral visions at the heart of the orthodox-progressive divide, but they also promulgate other core values that may influence political behavior. Indeed, it is the mix of core values that produces the distinctive ethos of the major religious traditions, including the individualism of evangelical Protestants, the civicmindedness of the Protestant mainline, and the communalism of Catholics. 5

The strength and nature of social constraint also varies across religious communities. Religious and political orientations can be packaged together by official church teachings, the cues provided by clergy, lay leaders, and interaction with fellow congregants all of which vary considerably across religious communities. Thus the connection between orthodox-progressive religious perspectives and political behavior should vary across faith traditions. For example, psychological and social sources of constraint may re-enforce the logical connection between religious orthodoxy, moral values, and moral policy attitudes to a greater degree in some traditions than in others. Such a linkage may be stronger among evangelical Protestants, where the laity is relatively uniform in its moral traditionalism and clergy often address private and public morality from the pulpit, than among mainline Protestants and Catholics, where the moral values of congregants are more heterogeneous and clergy focus more on social justice and communal good works (Guth et al. 1997; Wald, Owen, and Hill 1990). Political connections that are less clearly logical and which thus rely more heavily on other sources of constraint should vary even more across faith traditions. For example, the economic and social individualism displayed by evangelical Protestant clergy may help to connect orthodoxy with social welfare conservatism for their congregants. In contrast, the commitment of Catholic and mainline clergy to economic and social justice (Barker and Carman 2000; Guth et al. 1997; Byrnes 1991) may link orthodoxy to more-liberal social welfare attitudes for their laity. Electoral Constraint and Political Contexts. Carmines and Stimson (1989: 135-136) identify a fourth, electoral, source of constraint: the time bundling of political issues, in which political parties and candidates package together distinctly liberal or conservative stands on diverse issues, thus encouraging increased consistency in mass issue-attitudes. Similarly, other researchers show that when parties and candidates emphasize and take clearly distinct stands on a set of policy issues, the horizontal and vertical connections between attitudes toward the issues, related socio-demographic orientations, and political behavior tend to increase (Key 6

1966; Nie, Verba, and Petrocik 1976; Sniderman, Brody, and Tetlock 1991; Carsey 2000; Layman and Carsey 2002). The corresponding pattern holds at the individual level as citizens who care about policy issues and recognize party and candidate differences on them display greater horizontal and vertical constraint with regard to those issues than do other citizens (RePass 1971; Brody and Page 1972; Krosnick 1990; Layman and Carsey 2002). Thus, the electoral contexts that should encourage connections between the orthodoxprogressive religious divide and political behavior are those in which in which the political parties and their candidates emphasize the moral issues that are most clearly related to that divide and take distinct stands on them. At the individual level, there should be a connection between religious orthodoxy and political behavior only for those citizens who find moral issues to be salient and recognize party differences on them. In sum, we expect the political impact of the orthodox-progressive religious divide to vary by context, being more relevant for moral questions than for other topics; in some religious traditions than in others; and for citizens who find moral concerns to be salient and are aware of partisan differences than for other citizens. Within these contexts, the culture wars should influence mass political behavior; outside of them, they should be largely a rumor. Measuring the Orthodox-Progressive Divide To test our resolution to the culture wars debate, we use data pooled from the 1992, 1996, and 2000 National Election Studies (NES), 3 the three presidential-election-year surveys conducted since NES greatly improved and enhanced its measures of religion (Leege and 3 The 2000 NES interviewed roughly half (1,006) of its respondents in person and roughly half (801) of its respondents over the phone. Since the 1972 study, the NES has measured respondents attitudes and their perceptions of party and candidate positions on most political issues by asking them to place themselves, the two parties, and candidates on seven-point scales. The face-to-face interviews in 2000 used the traditional seven-point scale format, but the phone interviews used a branching format resulting in only five categories. To maintain continuity with the 1992 and 1996 NES surveys, our analysis includes only the face-to-face respondents in 2000. 7

Kellstedt 1993). 4 To measure the orthodox-progressive divide, we use six items: views about biblical authority, identification as a born-again Christian, religious salience (guidance), and frequency of Bible reading, worship attendance, and prayer. 5 Some of these items tap into religious beliefs and others measure religious behavior, aspects of religion that are conceptually distinct. However, it is appropriate to combine them into a single measure of religious orthodoxy for three reasons. First, religious orthodoxy within traditions is defined by both beliefs and behaviors (Kellstedt et al. 1997). For example, an orthodox Catholic is someone who follows the traditional teachings of the church, which include regular mass attendance and private devotionalism. Second, past work has shown that within the three largest American religious traditions (evangelical Protestant, mainline Protestant, and Catholic), beliefs and behaviors very rarely work at cross-purposes in creating moral, cultural, and political divisions (Layman 2001). Rather, it is the combination of belief and behavioral differences that produces the clearest divisions within these traditions. 6 Finally, consistent with past work (Kellstedt et al. 1997; Layman 2001), a principal components factor analysis of these items suggests that they all 4 We combine these three NES surveys for our analysis so that our results will not be greatly affected by the idiosyncracies of a particular election campaign and will reflect the general connection between religion and contemporary political behavior. This does have the potential to mask ongoing changes in that connection, but if the political impact of the orthodox-progressive religious divide grew over this period, then pooling the three surveys would deflate, rather than inflate, its current importance. Moreover, we conducted our analyses separately for each of the three years and did not find systematic changes over time in the political influence of religion. 5 Some of these measures for example, view of the Bible, born-again status, and Bible reading may be less than ideal because they are more central to theological and behavioral orthodoxy in evangelical Protestantism than in most other traditions. However, there is considerable evidence that these items also tap religious orthodoxy within other religious traditions (Kellstedt et al. 1997; Wilcox 1990; Smidt 1988; Jelen, Smidt, and Wilcox 1993; Kellstedt and Smidt 1993), and they are, of course, the only items available in NES data. 6 We confirmed this in our own analyses. For all of the dependent (political) variables in table 1, we estimated models that included doctrinal orthodoxy (beliefs) and religious commitment as separate variables. For each variable, and for each of the three largest traditions, increases in both doctrinal orthodoxy and commitment led to increases in political conservatism and support for the Republican party and its candidates. In fact, for the four dependent variables in table 1 and for the three largest traditions, there were only three instances (out of a possible 12) in which both doctrinal orthodoxy and religious commitment did not have statistically-significant independent effects. The results of this analysis are available from the authors by request. (Note to editors and reviewers: Appendix D shows those results. We do not plan to include this appendix if the paper is published.) 8

measure an underlying commitment to orthodox beliefs and practices. 7 Our measure of religious orthodoxy is the factor score from this analysis. 8 The Religious Contexts of the Culture Wars 9 Our measure of the religious contexts of the culture wars is membership in religious traditions. We use the NES religious affiliation battery and follow the recent literature (Kellstedt and Green 1993; Kellstedt et al. 1996; Layman 2001) to identify the five largest American religious traditions (white evangelical Protestants, white mainline Protestants, Catholics, black Protestants, and Jews), and the secular population. 10 To capture variation in the effect of the orthodox-progressive divide across traditions, our models of political behavior include dummy variables for five of the religious traditions (with evangelical Protestants serving as the comparison category), our index of religious orthodoxy, and interactions between orthodoxy and each of the tradition dummies. These interactions indicate the difference between the impact of orthodoxy for evangelical Protestants and for each of the other traditions. The dependent variables in these models are four measures of political affiliations and behavior: liberalconservative identification, party identification, comparative evaluations of the Republican and Democratic presidential candidates in each year, and the two-party presidential vote. 11 7 This analysis produced only one factor with an eigenvalue greater than one. That factor had an eigenvalue of 3.42 and explained 57 percent of the total variance in the six indicators. The factor loadings of the six indicators ranged from.66 to.81. 8 Religious orthodoxy ranges from -1 to 1, and has a reliability coefficient (alpha) of.85. We use this measure for the analyses in tables 1 and 3. In table 2, orthodoxy is measured through a confirmatory factor analysis of observed indicators, treated as having measurement error. 9 Because all of our analyses examine the political impact of the orthodox-progressive divide within particular religious traditions (i.e. take into account religious context), the most straightforward analyses are those focusing on religious contexts. Thus, although we introduced policy contexts first in developing our argument, we begin the analysis with religious contexts. 10 We discuss the assignment of religious affiliations to religious traditions in Appendix A. 11 The vote is coded zero for Democratic and one for Republican. We recoded the other three variables to range from zero (most liberal/democratic) to one (most conservative/republican). We use logit to estimate the vote model and regression to estimate the other three models. All of the models include dummy variables for 1996 and 2000 and standard demographic controls: income, education, region of residence, gender, age, and union 9

Because we are primarily concerned with the political differences between and within religious traditions, and because the coefficients in interactive models are often difficult to interpret, we use methods suggested by King, Tomz, and Wittenberg (2000) and the CLARIFY software that they developed (Tomz, Wittenberg, and King 2001). This approach allows us to focus on the predicted values and probabilities from these models for the different religious traditions and the differences in the predicted values and probabilities for the orthodox and progressive camps within traditions. 12 It also allows us to assess statistical significance by reporting the standard errors around the predicted values, probabilities, and difference scores. Table 1 shows two sets of predictions from our models. First, we capture the orthodoxprogressive divide within the four Christian traditions by showing the difference between the predicted values or probabilities of the orthodox and the progressive members of each tradition. 13 Second, we assess the political impact of religious tradition, separate from the effect of religious orthodoxy, by showing the predicted values and probabilities for traditions when orthodoxy (and each of the control variables) is held constant at its mean value for the entire sample. [Insert Table 1 About Here] The predicted orthodox-progressive differences show considerable support for the culture wars thesis. The orthodox members of the three largest religious traditions (evangelical and mainline Protestants and Catholics) are noticeably more conservative, more attached to the Republican Party, and more supportive of Republican presidential candidates than are their progressive counterparts. The small standard errors around the predicted differences indicate that membership. We do not control for race because our models include a dummy variable for black Protestants and because only non-blacks are included in our evangelical and mainline Protestant categories. 12 Appendix B presents the actual coefficients from our regression and logit models. 13 We hold all of the control variables at their sample means to compute the predicted differences between the orthodox members (defined by setting religious orthodoxy to one standard deviation above its mean for the particular tradition) and the progressive members (defined by setting orthodoxy to one standard deviation below the tradition mean) of a tradition. Because of the very small number of Jewish respondents in the NES sample and because of seculars uniformly low levels of orthodoxy, we do not show predicted differences for these groups. 10

they are statistically significant (p<.05) for all three traditions on all four political variables. Take, for example, the predicted differences (on variables ranging from zero to one) between orthodox and progressive evangelicals. The orthodox are more conservative than the progressives on the ideology scale by.15, plus or minus about.01, and are more Republican than progressives on party identification by.16, plus or minus about.02. The probability of a Republican presidential vote among the orthodox is higher than that of the progressives by.32, plus or minus about.04. Our contention that the political impact of orthodoxy varies in magnitude across religious traditions also holds true. Orthodox-progressive differences among evangelicals are twice as large as those among mainline Protestants and Catholics on party identification (.16 versus and ) and candidate evaluations, and nearly twice as large on ideology and the presidential vote. Meanwhile, the culture wars divide seems to be inconsequential for the political behavior and affiliations of black Protestants: the difference between the orthodox and progressives in this tradition is not statistically significant for any of the political variables. The second half of table 1 the predicted values and probabilities for the various religious traditions provides evidence both for and against the culture wars thesis. On the one hand, when we eliminate differences in religious orthodoxy between traditions, there are no statistically significant political differences between seculars and either white Protestant tradition. Evangelical and mainline Protestants differ significantly only in party identification, with mainliners showing their historically stronger ties to the GOP. 14 On the other hand, it is clear that, contrary to Hunter, religious tradition has not become virtually irrelevant for political behavior. Black Protestants and Jews are significantly less conservative and less Republican than the three white Christian traditions and seculars on each of 14 In a separate analysis, we allowed levels of religious orthodoxy to vary across traditions by computing predictions with orthodoxy held at its mean for each separate tradition. The differences between traditions were much sharper than those shown here. Importantly, evangelical Protestants were significantly more conservative or Republican than every other tradition (including mainline Protestants) on each political variable, and seculars were more liberal or Democratic than both white Protestant groups on each variable. 11

the political variables. There is also evidence that the Catholic-Protestant divide remains a part of American politics. Although Catholics are no less conservative than evangelical or mainline Protestants (.54 for Catholics versus.55 for both Protestant groups on ideology), they are significantly less likely than either white Protestant group to identify with the Republican party, to favor Republican candidates, and to vote Republican in presidential elections. In sum, there are clear signs of an orthodox-progressive religious divide in American politics. Religious orthodoxy creates clear political differences within religious traditions, but tradition remains a potent force, exerting an effect independent of religious orthodoxy and conditioning the impact of orthodoxy. Simply put, religious belonging is a critical context in which religious perspectives are linked to politics. The Policy Contexts of the Culture Wars To examine the policy context of the culture wars, we turn to a structural equation model that examines the extent to which the political impact of religious orthodoxy is exerted through orientations toward various key dimensions of public policy. Models of vertical constraint typically envision socio-demographic factors such as religion affecting political behavior by first affecting core political values (or policy predispositions ). 15 These values then shape attitudes on specific policy issues, which in turn influence political behavior (cf. Miller and Shanks 1996). This sort of pattern is also implicit in Hunter s (1991) argument that religious orthodoxy and progressivism lead to competing moral visions, which in turn produce different views on policy issues and, ultimately, different patterns of political behavior. Thus, we structure our model so that orthodoxy first affects core values, those values affect attitudes toward policy issues, which in turn affect political behavior. However, the model also accounts for the 15 There is a substantial literature on various types of core values and the role they play in shaping political attitudes and behavior (cf. McClosky and Zaller 1984; Feldman 1988; Feldman and Steenbergen 2001). 12

possibilities that orthodoxy may affect policy attitudes and political behavior directly and that core values may have direct effects on political behavior. As figure 1 illustrates, the model includes a direct connection between religious orthodoxy and political affiliations and behavior the combination of which we call political alignment as well as four distinct indirect links between orthodoxy and political behavior. 16 One indirect path is through moral values and attitudes, with orthodoxy first affecting the core value of moral traditionalism the degree to which individuals are committed to traditional moral values and are intolerant of those with different moral outlooks and then affecting attitudes toward moral issues such as abortion and gay rights both directly and indirectly. A second indirect path is through the values and attitudes associated with the social welfare and racial policy dimensions. Orthodoxy first affects two sets of core values egalitarianism (coded here as anti-egalitarianism) and support for a limited scope of government that are commonly associated with social welfare and racial attitudes (McClosky and Zaller 1984; Feldman 1988; Feldman and Steenbergen 2001). 17 It then affects attitudes toward issues of social welfare and economic redistribution and toward racial issues such as government assistance to blacks and affirmative action both directly and indirectly. 18 The other two indirect paths are through attitudes toward government spending on national defense and government spending on environmental 16 The observed indicators of the latent variable political alignment are the four dependent variables in table 1. We use a political alignment index rather than treating each variable as a separate dependent variable for two reasons. First, party identification, comparative candidate evaluations and the presidential vote all reflect individuals feelings of identification with or support for the two major political parties. The liberal and conservative ideologies are closely identified with the Democratic and Republican parties, respectively, and, in fact, are growing more so over time (Abramowitz and Saunders 1998). Thus the four indicators all tap into one s degree of alignment with the major combatants in American national politics. Second, we estimated our structural model for each of the four indicators separately and the results were very similar to those presented here. 17 Although orientations toward social welfare and racial issues typically fall into two distinct attitudinal dimensions (cf. Layman and Carsey 2002), we do not separate the indirect effects of orthodoxy through the two sets of attitudes because we assume that they are both influenced by egalitarianism and support for limited government. 18 Our model also includes attitude toward the death penalty as an indicator of racial attitudes. Race plays a major role in the debate over the death penalty. Also, in our exploratory factor analysis of domestic policy attitudes, death penalty attitude loaded much more strongly with racial issues than with other types of issues. 13

protection. 19 [Insert Figure 1 About Here] With the exception of the single-indicator measures of defense and environmental attitudes, all of the variables in our structural model are latent variables, measured by confirmatory factor loadings on several observed indicators, each of which is treated as having measurement error. 20 To account for our expectation that the political impact of religious orthodoxy varies across religious traditions, we estimate the measurement and structural (path) components of the model separately for evangelical Protestants, mainline Protestants, and Catholics: the three largest religious traditions and the traditions exhibiting a consistent link between orthodoxy and political behavior in table 1. 21 Table 2 shows the structural coefficients (the measurement coefficients are presented in appendix C). 22 It also shows the indirect effects of orthodoxy on policy attitudes through policy predispositions, and the amount of the indirect effect of orthodoxy on political alignment that is exerted through moral, social welfare and racial, defense, and environmental attitudes. 23 The results provide strong support for the importance of policy contexts and further highlight the importance of religious contexts. As expected, the impact of religious orthodoxy is greatest in the moral policy realm. The effect of orthodoxy on moral traditionalism is very strong 19 These spending questions are the only indicators of environmental and defense and foreign policy attitudes that are included in the 1992, 1996, and 2000 NES surveys. Also, there are no indicators of core values associated with environmental or defense and foreign policy attitudes that are included in all three surveys. 20 To provide a scale for the latent variables, we constrain the factor loading for one observed indicator of each latent variable to be equal to one. All of the observed indicators are coded to range from 0 for the least orthodox, most liberal, or most Democratic orientation to 1 for the most orthodox, most conservative, or most Republican orientations. Thus, the latent variables take on the same range of values. 21 This approach does not allow us to examine the effect of religious tradition, independent of religious orthodoxy, on political attitudes and behavior. However, we already established that independent effect in table 1. 22 Our models include the standard demographic controls and dummy variables for 1996 and 2000. 23 We do not exclude missing values from our analyses. We estimate our model using Amos 4.0, which computes full information maximum likelihood (FIML) estimates even in the presence of missing data (Andersen 1957). Wothke and Arbuckle (1996) describe this FIML procedure and show that the estimates produced by it are more consistent and efficient than those produced by pairwise or listwise deletion of missing observations. 14

and statistically significant in all three traditions. Orthodoxy has a significant direct effect on moral issue attitudes, even controlling for moral traditionalism, in each tradition. Combined with its substantial indirect effect through moral traditionalism, that produces a very large total effect of orthodoxy on moral policy attitudes in all three groups. However, even in the policy context in which the orthodox-progressive divide is manifested most fully, its effect varies considerably across religious contexts. The direct effect of orthodoxy on both moral traditionalism and moral issue attitudes is clearly greater for evangelical Protestants than for mainline Protestants or Catholics. The total effect of orthodoxy on issue attitudes among evangelicals is noticeably larger than among Catholics and is almost twice as large as that for mainliners. Not surprisingly, the linkages between religious orthodoxy and non-moral policy predispositions and issue attitudes are weaker and less consistent across religious traditions. In keeping with the individualistic tendencies of evangelicalism, orthodoxy leads to statistically significant increases in both anti-egalitarianism and support for a limited scope of government among evangelical Protestants. Thus, although, orthodoxy does not have a significant direct effect on social welfare attitudes, it does have a clearly positive indirect effect on social welfare conservatism among evangelicals. The relationship between orthodoxy and the rejection of egalitarian values is also statistically significant for mainline Protestants. However, the morecommunitarian proclivities of the Protestant mainline and of Catholics are reflected in the fact that the effect of orthodoxy on support for limited government is much smaller for them than it is for evangelicals, and that orthodoxy s affect on egalitarianism is not statistically significant for Catholics. The end result is that the total effect of orthodoxy on social welfare conservatism is clearly smaller for mainliners and Catholics than it is for evangelicals. The indirect effect of religious orthodoxy on racial issue conservatism is clearly positive for evangelical and mainline Protestants and slightly positive for Catholics. However, orthodoxy s total effect on racial conservatism is close to zero for white Protestants and negative 15

for Catholics. The reason is that religious orthodoxy has a statistically significant and negative direct effect on racial issue conservatism in all three traditions. A possible explanation is that, holding aside their support for limited government and lack of egalitarianism, orthodox white Christians regard African-Americans as fellow believers, deserving of the compassion God offers to all humankind. Religious orthodoxy leads to support for more spending on national defense and less spending on environmental protection in all three traditions. Somewhat surprising is that the effect on defense spending is actually stronger for mainline Protestants and Catholics than it is for evangelicals. The more-typical pattern holds for environmental spending, where the effect is much larger for evangelicals than for mainliners or Catholics. Lending support to the culture wars thesis is the fact that moral issue attitudes have the largest direct effect on political alignment in all three religious traditions, with conservative moral attitudes leading to greater support for the Republican Party and its candidates and greater ideological conservatism. Social welfare conservatism and support for more defense spending also have positive (pro-conservative/republican) effects on political alignment in each tradition, and support for decreased environmental spending has a positive effect for evangelical and mainline Protestants. In contrast, the effect of racial policy attitudes does not reach statistical significance in any of the traditions. The effects of core values on political alignment are mixed. The political impact of moral traditionalism is exerted entirely through moral policy attitudes in each tradition as its direct effect fails to reach statistical significance. The direct effect of anti-egalitarianism is also insignificant for both white Protestant traditions, but is positive and strong for Catholics. Support for a limited scope of government leads directly (and, of course, indirectly) to Republican and conservative political leanings in all three religious traditions. Somewhat surprising is that the direct effect of religious orthodoxy on political alignment is negative and significant for evangelical Protestants and Catholics. This simply may mean that 16

we have over-controlled for the orientations that link orthodoxy to conservative and Republican political inclinations. However, it also may reflect the traditional Democratic loyalties of orthodox evangelicals and Catholics. A large part of the reason why these individuals have abandoned the Democratic Party and, in the case of evangelicals, aligned with the GOP over the last 40 years is the growing importance of moral issues in American party politics (Layman 2001). Once we control for their conservative moral views (as well as the other policy contexts where orthodoxy is linked to conservatism), it is only natural that orthodox evangelicals and Catholics revert to their Democratic leanings. Lending support to this idea is that the negative direct effect of orthodoxy for mainline Protestants, a group with traditional Republican ties, is not statistically significant. The sources and sizes of the indirect effect of religious orthodoxy on political alignment are in keeping with our arguments about policy and religious contexts. Far and away the largest portion of that indirect effect is exerted through moral values and attitudes in each tradition. However, the size of that indirect effect is much larger for evangelical Protestants than for mainline Protestants or Catholics. There is also a noticeable indirect effect of orthodoxy through social welfare and racial attitudes, and it also is larger for evangelicals than for mainliners and Catholics. Finally, the total political effect of orthodoxy is much larger in the evangelical tradition than it is among mainline Protestants or Catholics. [Insert Table 2 About Here] The Political Contexts of the Culture Wars While religious elites and activists play a crucial role in linking religious orthodoxy to politics, their political counterparts are hardly silent partners in these relationships. The electoral activities of political parties and their candidates help structure the political choices before citizens, fostering or retarding the impact of the orthodox-progressive divide. To assess these effects, we consider the impact of issue salience and perceived issue differences between the two 17

parties. To measure salience and awareness of party differences, we turn to two sets of questions in the 1996 NES. The first simply asks respondents about the importance of a particular issue to them. Respondents who said that an issue was very or extremely important are coded as having salient opinion, while those who said that an issue is somewhat, not very, or not at all important were coded as having opinions that were not salient. The second set of questions asks respondents to place each of the major political parties on a scale for a particular issue. We classify those respondents who place both parties on the scale and view the Republican Party as more conservative than the Democratic Party as being aware of party differences on that issue. Respondents who fail to place both parties on the scale, place the parties at the same position, or view the Democrats as more conservative than the Republicans are coded as not being aware of party differences on that issue. The 1996 NES asked both of these sets of questions for three domestic policy issues: abortion, the level of government services and spending, and the importance of protecting jobs versus the environment. 24 If it is salience of and awareness of party differences on moral issues that contributes to vertical constraint between religious orthodoxy and political behavior, then we should see orthodox-progressive political divisions within religious traditions most clearly when the key moral issue of abortion is salient and citizens understand that the GOP is more pro-life than the Democrats. When these conditions do not hold, orthodox-progressive divisions should be much less evident. To see if this is true, we return to our single-equation interactive model of the impact of religious tradition and religious orthodoxy, with standard demographic controls and a 24 We use only the 1996 NES for this analysis because it contains a much more thorough set of questions on issue salience and awareness than do either the 1992 or 2000 surveys. For example, the 1996 NES was the only one of the three surveys to ask respondents both how important the abortion issue was to them and to place the two parties (and not just their candidates) on the abortion scale. 18

political alignment index as the dependent variable. 25 We estimate this model for four groups of respondents: abortion is not salient and not aware of party differences on abortion; abortion is not salient but aware of party differences; abortion is salient but not aware of party differences; and abortion is salient and aware of party differences. 26 The first section of Table 3 shows the predicted orthodox-progressive differences for evangelical Protestants, mainline Protestants, and Catholics. 27 The results are strongly supportive of our argument about political context. 28 When respondents either do not find abortion to be salient, or are unaware of party differences on abortion, or both, the difference in political alignment between orthodox and progressive members of religious traditions is almost never statistically significant. There are only two exceptions. Among Catholics who do not find abortion to be salient and are unaware of party differences, the orthodox are more liberal and Democratic than are the progressives. Among evangelicals who find abortion to be salient, but are not aware of party differences, the effect of orthodoxy is positive (pro-republican and conservative) and barely reaches statistical significance. 25 For this analysis, the political alignment index is the factor score (coded to range from zero for the most liberal/democratic orientation to one for the most conservative/republican orientation) from a principal-components factor analysis of party identification, ideological identification, comparative candidate evaluations, and the presidential vote. The analysis produced only one factor with an eigenvalue greater than one. It had an eigenvalue of 3.04 and explained 76.0 percent of the total variance in the four variables. The factor loadings for the four variables ranged from.76 to.91. The reliability coefficient (alpha) for the index is.87. 26 Respondents are coded as finding an issue to be salient if they said that it was "very" or "extremely" important. Respondents are coded as not finding an issue to be salient if they said that it was "somewhat," "not too," or "not at all" important. Respondents are coded as being aware of party differences on an issue if they placed the Republican party as more conservative than the Democratic party on the issue. Respondents who did not place one or both parties on the issue, who placed the Democrats as more conservative than the Republicans, or placed the parties at the same position were coded as not being aware of party differences on the issue. The number of Jewish respondents in some of these subgroups is too small for us to estimate meaningful coefficients on the Jewish variable and its interaction with orthodoxy. So, we exclude Jews from these analyses. 27 We do not show the predicted values of political alignment for various traditions when religious orthodoxy is at its sample mean (as in table 1) because the differences between traditions are not affected much by the salience of and awareness of party differences on abortion. 28 We performed the same analyses with party identification, ideology, candidate evaluations, and the vote as separate dependent variables. The results were all very similar to those for the political alignment index. 19

When abortion is both salient to citizens and they are aware of party differences, the impact of religious orthodoxy comes close to what the culture wars thesis expects it to be. It is significantly related to conservatism and support for the GOP in each of the three largest religious traditions, and the positive effect for each tradition is much larger than in any of the other political contexts. Importantly, even in the political context that is most conducive to an orthodox-progressive divide, the importance of religious context is evident. The effect of orthodoxy is at least twice as large among evangelicals as it is among mainliners and Catholics. Of course, there may be other political contexts in which the orthodox-progressive cleavage is linked to political behavior. Because religious orthodoxy is related to positions on non-moral issues in some religious traditions, it is possible that the orthodox-progressive divide is connected to political behavior when citizens find non-moral issues to be salient and are aware of party differences on them, and not just when the salience and awareness conditions hold for moral policy issues. Moreover, it is possible that our indicators of salience and awareness for abortion have tapped into a more general sense of political interest and awareness, factors that are always necessary for citizens to link their values and social characteristics to political behavior in an appropriate way (Campbell et al. 1960; Zaller 1992). In other words, the condition for linking religious orthodoxy to political behavior may not be moral issue salience and awareness of partisan differences on moral issues, but rather a general salience and awareness of politics. If either of these possibilities hold, then we should see an orthodox-progressive cleavage in political behavior if citizens find other issues to be salient and are aware of party differences on them, even if they do not meet the salience and awareness conditions on moral issues. To test this possibility, we first assess whether an orthodox-progressive divide exists when individuals find the issues of government services and spending and jobs versus the environment to be salient, but do not find abortion to be salient. We then examine the political impact of religious orthodoxy when individuals are aware of party differences on the two non-moral issues, but are 20