Social Class, Religion, and the Left-Right Political Party Divide in

Social Class, Religion, and the Left‐Right Political Party Divide in Canada and the United States: Survey Comparisons across 15 years Paper delivered to the International Sociological Association World Congress of Sociology, Political Sociology Section, Goteberg, Sweden, July 16, 2010. Douglas Baer , University of Victoria, Canada Research examining the role of religious divisions on voter behaviour/political party support suggests that religion has, far from crawling into obscurity as a social force, become increasingly salient over the past decades, especially when one distinguishes evangelical Protestant religious engagement from other forms of religious activity (Brooks, 2002; Achterberg, 2006, Pippa and Inglehart, 2004; Patrikios, 2008; van der Waal et al., 2007). As part of a wider, but highly contested thesis on the nature of political polarization in American society (Hunter, 1991; see Moutlon, 2006), this thesis can be extended by the writings of authors who distinguish between the theological orientations of Protestants who believe in an “absent deity” (a benign diety which does not intervene in human affairs) and biblical literalists whose conceptualization of a Christian God entails sternness and strong elements of obedience towards authority (see Foese and Baker, 2008; McLeod, 2007; Brint 2010). In this line of argumentation, Protestants in conservative/fundamental denominations will, as a direct result of this theological orientation, show wider policy preferences not only in response to the heavily publicized debates over abortion and gay rights, but also probably in opposition to welfare state measures as well given the affinity between this form of Christianity (which puts emphasis on the notion of individuals having responsibility for their destiny over some notion of devine inevitability) and American individualism. The writings on the affinity between Protestantism and individualism are not unambiguous, though, with respect to the nature of divisions within Protestantism; many suggest or imply that the connection pretty much applies as the “dominant form” of religious belief (see Bellah, 2004; McLeod, 2007) and as a consequence one might infer that both mainstream and fundamentalist Protestants share a common 2 individualism that leaves room for differences only on some matters of individual social/moral behavior (which, to be sure, could be mobilized to produce political division, but may not necessarily do so. Literature directed specifically at religious differences in political choice or policy preferences, based largely on data from the 1990s, suggests, to be sure, that differences between mainline and fundamentalist Protestants may have been overstated by a previous cohort of writers (see, especially, Manza and Brooks, 1997; Brooks, 2002). Still, it remains the case that evangelical church leaders in the United States have connected with a variety of right‐leaning Christian policy lobby organizations; even if the rank‐and‐file membership remains indistinguishable from the broader American public, at least the some of the major leading figures in evangelical (conservative/fundamentalist) churches anchor the political right in policy debate. This view, though, runs against the argument that all Christian religions essentially pull their adherents to the political right, a tendency which increases with higher levels of involvement in the church. This argument is in turn supported by some research that has been done by Brint (2010) and Brooks (2002); at the very least, it implies that any investigation needs to take some measure of social involvement (e.g., church attendance) into account. If the question of whether there is a large political division between rank‐and‐file adherents to mainstream versus conservative/fundamental Protestantism in the United States (and, as importantly, whether there have been increases or decreases in differences between Protestant denominations) remains as an ongoing empirical question , the issue of whether the American divisions may have been diffused elsewhere, most notable Canada, is of particular interest to the present investigation. From a variety of perspectives, authors argue that this sort of polarization cannot happen in Canada since, among other things, there has been, at least until recently, a strong political norm in Canada – adhered to by evangelical leaders – rejecting “religious politics” and threatening, in part because of the smaller proportion of Canadians presently linked to conservative congregations, to backfire against any political 3 actor seeking to employ U.S.‐style tactics in the media, in parliament or in behind‐the‐curtain lobby efforts (see Hoover et al., 2002; Farnes, 2009; Malloy, 2009). This is a sort of resource mobilization argument: in the U.S., it argues, conservative Protestants have the critical mass to make a difference, and this in turn facilitates the acquisition of political attention that may, in a variety of ways, promote the popularity of the religious orientation itself. But there is another form of historical argument suggesting that, at critical junctures from the Great Depression through the 1950s and 1960s, Canadian evangelicalism followed a different path in helping to shape, through efforts of NDP/CCF founder Tommy Douglas (himself an evangelical preacher) the Canadian welfare state (which, prior to the 1960s was considerably weaker even than the American one). In short, those Canadians who have been attracted to evangelicalism have tended to be attracted to denominations which were not necessarily hostile to “economic populism” (involving wealth redistribution) or to “social missions”. While this argument does not fully explain how it is that evangelicals in Canada moved in an almost opposite path historically, as it applies to the contemporary scene, it suggests there is less social and political space for the form of religious polarization that might have taken place in the United States. However, norms can evolve, and the strategies invoked by the religious right in the United States could easily spread (or be in the process of spreading) to Canada; the very authors, cited above, noting the distinctions between Canadian and U.S. evangelism often point out that there are emerging forms of co‐operation between conservative denominations (and across denominations) that connect Canadians and their American counterparts. These connections are undoubtedly reinforced by a small degree of media diffusion as U.S. televangelist channels reach Canadian audiences1. One form of explanation which probably can be ruled out is the argument, advanced by Lipset and others (see, for example, Malloy, 2009) that much in the way of Canada‐U.S. differences can be explained by the presence of a “state religion” in Canada and a corresponding absence in the United States. This explanation hinges in the historical claim that the Anglican church formed the de facto if not the de jure “state religion” in Canada, or at least English 4 Canada. As Grabb and Curtis (2004) point out, while it is indisputable that there was, in the 1800s, a tendency for state elites to be Anglican, English Canada exhibited the same general form of Protestant denominational competition that occurred in the United States (though, due to size differences, the total number of denominations may have been smaller, farm communities across Canada typically contained Catholic, Anglican, Methodist, Presbyterian churches as well as one or more churches from smaller denominations (e.g., Bretheren). And, on the matter of elite penetration by one particular religious group, this was hardly exceptional in North America (as members of specific religious groups dominated particular state governments in the early history of the United States). It remains, then, an empirical question as to whether there is evidence of an increasing similarity between the (English) Canadian and American cases with respect to the influence of religion on political preference, especially with respect to divisions within Protestantism and distinctions between Protestants and others (most notably Catholics). All of this presumes, of course, that religion “matters”; while the secularization thesis has not exactly found itself in a favorable climate in the sociology of religion in the United States, it is at least reasonable to suppose that, no matter how much more religious the United States is than most if not all world countries, there remains a negative slope – a secular decline – in religious practice, the rapid expansion of U.S. megachurches notwithstanding. 5 Although by no means uncontested, the claim that social class plays a diminishing role in American politics, if not politics worldwide, is now commonplace (Paklulski, 2001; Kingston, 2000; Houtman et al., 2008; Inglehart, 1997). Religion is but one of a number of “social structure factors” that can and have been investigated by prior writers. As Elff (2007) notes, there is now a literature arguing that “cognitive mobilization” is displacing “structural factors,” the most studied of which have been religion and social class. This line of reasoning bears some affinity to Inglehart’s (1997) postmaterialist thesis. Thus, while religion can, in some senses provide competition for class – the “individualistic” orientation of Protestant religiosity can be seen as undermining class solidarities – it is also in the same conceptual container as class, given that it suggests that social groupings (as with class, religion can be seen as part inherited and part chosen) are important generators of solidaristic political attitudes that differentiate one group from another. Although by no means uncontested, the claim that social class plays a diminishing role in American politics, if not politics worldwide, is now commonplace (Paklulski, 2001; Kingston, 2000; Houtman et al., 2008; Inglehart, 1997). Explanations are numerous: the blue‐collar working class is in numerical decline, and no longer has a plausible chance of achieving political dominance through the absence of sheer numbers (see also Przeworski , 1986), the social organization of neighbourhoods and suburbs no longer segregates a distinct group of “working class” individuals into geographically contiguous areas within which workplace solidarities can be reinforced, working class entertainment forms such as the pub and the folk song have been displaced by other, non‐local forms of media that cut across classes, the skills differentiation of the contemporary manual labour workplace places workers in competition and inhibits co‐operation, strong racial divisions cross‐cut class divisions and undercut the latter, capital flight for primary and secondary industry has moved the main occupational locations for 6 class definition to third world countries (where class solidarities may or may not be possible), and so on. If there has been a decline in class politics, there is reason to suspect that it might not have occurred as fully in Canada as in the United States. While studies of politics and social class have been on the wane recently in Canada, the 1980s and 1990s saw a body of research which appeared to affirm the view that social class distinctions in Canada had political consequences (see Gerber, 1986; Johnston and Ornstein, 1982, 1985; Lambert et al., 1987; Nakhaie and Arnold, 1996; Johnston and Baer, 1993). While these rarely involved cross‐national comparisons, where they did (see Johnston and Baer, 1993), the findings suggested that class divisions in political ideology were stronger in Canada than in the United States, but weaker than, say, in European Nordic countries which had followed “social democratic” paths (Esping‐Anderson, 1989, 1991). Again, from a resource mobilization perspective, the expectation that class forms a stronger basis for political differentiation in Canada arises from the presence of stronger “class” institutions in Canada (higher unionization rates, even outside the public sector, the presence of a minority, historically labour‐supported “left” party [the NDP] which has occasionally formed the government in many but not all provinces and which, federally, has never been likely to form a government but has sometimes held the balance of power). Whether the anticipated Canada‐U.S. difference in the effects of class can actually be observed, and what the trajectory of these differences are (converging? diverging?) remains an empirical question. Data The comparative analyses reported here largely compare models for respondents in English Canada (defined as respondents living outside the province of Quebec) with models for respondents in the United States. Quebeckers, large French, are excluded from the analysis not only because of expected differences related to the cultural/linguistic divide, the historical dominance of Catholicism but 7 then the rapid secularization of Quebec society (note lower levels of religious attendance in Table 1B), but also because the political choices facing voters in that province are of a different nature: the separatist Bloc Quebecois competes and often succeeds in attracting Quebec voters (despite the successive failures of two separation referenda) and has a legislative stance that is at least mildly left‐
progressive. Given the dual focus on social class and religion, individuals in the age range 21‐65 are included in the analysis, but those who were outside this age range and/or listed “unemployed,” “student,” “retired”, “disabled” or “homemaker” as their occupation were not included. For the analysis of recent trends across a period of 15 years, with an emphasis on differences in how class and religion affect voting over time, the 1990 and 2005 World Values Study data (Inglehart et al., 1991, 2006) were used. In 1990 and again in 2005, respondents were asked, “If there were a general election tomorrow, which party would you vote for?” For right party support, respondents who indicated that they would vote for the Republican party (in the U.S.) or the Conservative party (in Canada) were coded 1, with all other responses (including “don’t know”) coded as 02. In Canada only, respondents who indicated that they would vote for the New Democratic Party were coded “1” on a “left party” variable, with all other respondents coded “0”. In 1990, the English Canadian N was 959, while the US sample size was 1,254 (for respondents meeting the age and occupation criteria identified above). In 2005, the respective Ns were 788 and 567. To further investigate the effect of social class and religion on political support for right political parties (and, in the case of Canada, the left NDP party), data from the Comparative Study of Electoral Systems were used, with an N of 710 in English Canada and 781 in the US after cases not meeting the age and workforce status criteria were filtered out. The 2001‐2006 wave of the CSES dataset included a variable representing the party the respondent voted for in the most recent Presidential race (in the U.S., 2004) or the most recent Parliamentary election (in Canada, 2004). The CSES dataset was in turn 8 based on individual‐level data collected as part of post‐election waves of the American National Election Study (2004) and the Canadian National Election Study (2004). An extension of the comparison to more recent (2008) data was not possible because a dissemination version of the 2008 American National Election Study is not currently available with an occupation variable upon which class categories could be constructed. However, the 2008 Canadian National Election Study data include an extensive occupation variable with detailed 4‐digit Canadian NOC (National Occupational Classification), from which major occupational categories (and some tests of alternative categorizations) could be constructed. The inclusion of data from the Canadian 2008 Election Study is considered important because of the political transition from a scandal‐ridden centrist Liberal party administration to a right Conservative Party minority government in 2006 gave the Conservatives more political support in English Canada and, as a consequence, more statistical power in tests involving the connections between class and religion on one hand and Conservative party political support on the other. More importantly, the 2008 election occurred after two years of Conservative rule, following almost a decade during which conservative Christian organizations had been more active politically in Canada (but still with far less salience than their American counterparts) than had previously been the case. The 2008 CNES data included 860 individuals in the workforce between the ages of 21 and 65. The two main independent variables are social class and religion. For social class, the construction of class variables was limited by the nature of occupational categories supplied by the surveys used, but these categories corresponded very closely to those found in other voter studies involving class [Brooks, 2002; Brooks and Manza, 1997] and to the key categories found in Wright’s (1985) conceptualization3. The categories that were used were: 1) managers, 2) professionals, 3) non‐
manual workers, 4) skilled manual (including supervisors), 5) unskilled and semi‐skilled manual, and 6) 9 self‐employed. Since separate survey questions were available to identify those who were self‐
employed, this category took priority over others; thus, a self‐employed doctor or lawyer would fall into the “self employed” category as opposed to the “professional” category. The scheme employed here differs somewhat from the major categorization employed by Heath and his colleagues especially inasmuch as it does not collapse managers and professionals. This latter division follows Wright’s arguments in spirit, but also leaves open empirically the question of whether these two groups are consistent in political orientation. Since self employed (including small employer) professionals have been separated from other professionals, of individuals in the professional category are likely to be public sector employees such as nurses and teachers. An earlier study in one Canadian province (Bird et al., 2000) suggested that there were major differences in political orientation between public sector professionals and those in self‐employment. As an alternative measure of social status, income was used in some models to examine the extent to which the class categories outlined above did not fully capture the relationship between economic status and support for right (and, in the case of Canada, left) political parties. While variables of this sort (even, in some cases, education) are often employed as measures of “class” in areas of sociology such as health and even, occasionally, in political sociology, the problem with such measures is that it becomes difficult to imagine how, with gradational distinctions, political conscious social groupings organized around common interests might form in the same way that they might form in work cultures that put together major types of occupations in a common work environment. Still, the possibility that transformations in the occupational structure have led to a “good jobs/bad jobs” bifurcation (see Picot and Myles, 1990), or someone else along these lines which might require more refined data not available in typical surveys but which might in some way be captured by an income variable cannot be discounted entirely. Conceptually, though, one could argue that income is consequential to (and not contemporaneous with or antecedent to) social class; a “null” class finding in 10 a model that controls for income may be attributed to the misspecification (controlling for income when income should have been left out) as much as anything else. For this reason, wherever this occurs, findings in models without the income variable will be briefly reported. This having been said, a significant finding for the income variable when a class variable is present in a model may suggest that there are some aspects of occupational differentiation that lead to political outcomes but are not fully captured by existing schemes of the sort employed here. Income was measured in deciles in the World Values Surveys, in quintiles in the Comparative Study of Electoral Systems dataset, and in actual Canadian dollars in the 2008 Canadian National Election Study4. For religion, the major categories are 1) Catholic, 2) Mainstream Protestant, 3) Conservative Protestant, 4) Other religions, 5) No religion. In one set of comparisons (World Values Study, 2005), no distinction was made between the two different types of Protestants. Where simplified coding was not supplied with the survey itself, such as in the case of the Comparative Study of Electoral Systems data, Anglicans (Episcopalian), Congregational, Lutherans, Methodist and Presbyterians were coded as “mainstream” (in Canada, the United Church was included in this category)5, while such denominations as Mormons, Pentacostal, Baptists, Jehovah’s Witness and many others were coded as conservative (fundamentalist). In many American studies of the relationship between religious denomination and political attitudes or behavior (see, for example, Brooks, 2002), “black Protestant” (or even sometimes black fundamentalist Protestant) denominations are separated from other denominations. In a number of instances, the possibility that this additional religion category might be necessary in the U.S. was tested with interaction terms between religion and race or with an additional religion category for respondents who were black and Protestant. In general, the outcome of these tests was that, while Black Protestants (even Black Protestants in conservative congregations, such as Baptists) were dissimilar from their white counterparts, all of the difference could be attributed to race and not to any independent difference between the “black Protestant” category and other Protestant categories. In 11 short, black Protestant congregations are probably reflective of the black minority culture in the USA (which, to anticipate tables to be presented below, has been and remains antithetical to the Republican party there) and do not impart an additional degree of “liberalness” on adherents. Because of the presence of arguments that it is intensive of engagement in (Christian) religions that matters and not so much denomination, religious (church) attendance is included in some models. This is typically measured on 6 to 8‐point scales, ranging from “never” to “once a week or more” or “more than once a week”. Typical of these was the wording in the 1990 World Values survey: “Apart from wedding, funerals and christenings, about how often do you attend religious services these days? 1=more than once a week, 2=once a week, 3=once a month 4=Christmas/Easter 5=other specific holy days [categories 4 and 5 were collapsed], 6= once a year, 7=less often 8=never. Although not nearly as conceptually suitable for the task of assessing the impact of connections with other individuals in social institutions (churches), a variable measuring the respondent’s perceived importance of religion in his/her life was used for data based on Canadian and American election studies because these studies did not have a religious (church) attendance variable for Canada. They did, however, ask respondents, if they a) had no religious beliefs, b) not very religious, c) were somewhat religions or d) were very religious6. A similar 4‐point scale in the 2005 World Values dataset (with similar codings worded i) very important ii) rather important, iii) not very important and iv) not at all important) correlated by between .60 and .62 with the church attendance variable in each of Canada and the United States. Other variables used as control variables in the multivariate models include: a) education, b) community size (where available), c) age (in years)7, d) gender (dummy variable), e) marital status (a set of dummy variables for the categories i) single/living together ii) separated/divorced iii) widowed iv) married), f) union membership (dummy variable, yes/no). In models for Americans only, an additional dummy variable for race (black vs. not) was also used8. Education was typically measured with 6 to 8 response categories (e.g., some elementary; completed elementary; some high school; completed high 12 school; some college or university; completed university), but in the 1990 World Values Study a variable based on “age at which one completed schooling” was all that was available. Unfortunately, this variable was truncated at 21, so there was a large number of cases with a score of 21 for this variable. Methods Logistic regression models were estimated for each of the election surveys discussed above for each of Canada and the United States. In general, models were estimated for a) U.S. Republican support/ not; b) Canadian Conservative support/not ; c) Canadian NDP support / not. As detailed above, the “not” category includes non‐voters. Zero‐order models were estimated initially, followed by models which included both religion and class variables as a means of investigating the extent to which religion might suppress class effects and vice versa. Not reported in the tables to be presented here, a series of interaction tests between class and religion was constructed as well. Finally, pooled models, with Canadian and American respondents in the same model, were constructed to examine whether certain (logistic) regression coefficients differed significantly between Canada and the United States9. Rather than reducing religion to a simple “conservative Protestant versus others” dichotomy and rather than reducing class to a simple “manual working class versus others” category, the extensive categorizations described above were largely retained in the models that were estimated. Given the moderate sample sizes available in each of the datasets, though, that if there were, in fact, modest effects along the major lines anticipated, these effects might be missed by the larger number of degrees of freedom (due to a larger number of categories) in the tests that were employed. For this reason, block significance tests only reaching a probability level of p<.10 were reported, and some additional post‐hoc contrasts (limited, though to avoid too much capitalization on chance) were constructed for key comparisons (again, manual workers versus others and/or self‐employed versus others). The 13 reference category for class was unskilled/semi‐skilled manual workers, and the reference category for religion was Catholic. Missing data proportions were generally small for most variables, with the major exception of income. For models employing this variable, multiple imputation techniques (with 10 imputations, using MCMC imputation) were employed, resulting in slight increases in the variances of the parameter estimates (usually, in the order of less than 10%) and hence a reduction in the sensitivity of significance tests, but yielding improved estimates based on all selected cases (as opposed to deleting cases) and parameter estimates containing less bias, especially if the MAR10 assumption could be considered tenable. The block significance tests for models incorporating multiple imputation make use of F‐tests rather than chi‐square tests. Results Table 1 provides descriptive comparisons of the two countries. In 1990 (Table 1A), the occupational structures of Canada and the United States were fairly similar, though Canada had a higher proportion of unskilled manual workers in its workforce, and the United States had a slightly higher proportion of professionals and skilled manual workers, at least according to sample values drawn as part of the 1990 World Values Study. By 2005, the occupational structure had shifted in both countries so that the proportion of workers fitting the unskilled manual labour category had dropped substantially in both countries (from 31% to 10% in Canada, and from 23% to 5.6% in the U.S). Compared to 1990, the proportion of employed Canadian workers in the skilled manual labour category increased substantially; there was virtually no change in the United States. Self‐employment became a more important category in the United States (rising from 4% in 1990 to 19% in 2005), but changes in Canada were negligible (holding around 9‐12%). For class analysts, the two‐fold implication is that: a) the fact that the manual working class increasingly consists skilled workers, who in some occupations may 14 themselves move into and out of self‐employment more frequently than those in the unskilled and semi‐skilled category (e.g., electricians) and who possess means of production resources (“skills” in Wright’s terminology) not possessed by the declining traditional blue collar working class may make their position as the ideal bearer of anti‐capitalist attitudes more problematic, b) self employment, which usually implies that an individual is petit bourgeoise or a small employer (but not a large capitalist), rather than withering away as an employment category has become more important in Western societies, with possible political consequences. While some observers have characterized Canada as a much more Catholic country, this comparative observation applies mostly if one combines English and French Canada; English Canada, like the United States, is best characterized as mixed Protestant and Catholic in religious heritage, just as is the case in the United States. The 1990 World Values data (Table 1A) suggest that, at the time, English Canada and the United States had very similar proportions of Protestants and Catholics, though the 2005 World Values data suggest that English Canada is slightly more Catholic (at 31%) than the United States (at 20%). In both countries, there has been an increase in “Other” religions (undoubtedly associated with immigration) in the course of the 15 year period between 1990 and 2005, and the “No religion category” has expanded in the United States11. Using the Comparative Study of Electoral Systems data (not shown in Table 1B), the estimated proportion of English Canadians professing to be Catholic is 24.6%, versus 23.6% for Americans. There are, however, more Protestants in the United States (52% versus 46%) and, more importantly, many more US Protestants (32% of all respondents) list themselves in congregations that are “conservative/fundamentalist than is the case with Canadians (where only 5.5% of all respondents list themselves as members of conservative/fundamentalist Protestant congregations. Another major Canada‐U.S. difference seen in the Comparative Study for Electoral Systems data is the higher proportion of Canadians (24.6%) as opposed to Americans (15.6%) listing “No religion”. Finally, Table 1B provides an indication of the lower levels of church attendance 15 in Canada as opposed to the United States: Americans are more likely (32%) to attend church weekly than Canadians are (23%). In 1990, there are no significant class differences in the likelihood that a respondent will be a fundamentalist/conservative Protestant12. In the United States, but not in Canada13, there are class differences in the religious (church) attendance variable (F=2.88,df=5,1239, p<.014), but these are fairly small. Managers (3.5), professionals (3.4) and non‐manual workers (3.5) have scores which are lower – signifying greater church attendance – than skilled workers (4.0), unskilled workers (3.9) or those who are self employed (3.8). The first panel of Table 2A shows the extent of the relationships between religion and class on one hand and Conservative party support on the other, both with and without controls. Without controls (Models 1 and 2), there is a class effect which can largely characterized as a greater tendency for the self‐employed to support the Conservative Party. This makes sense from the standpoint of conventional class interests in relation to politically right parties, though, as noted earlier, the Conservative party of 1990 had not yet adopted the strong neo‐Conservative political stance which characterized the party in the twenty‐first century and was, if anything, largely known for its implementation of a Canada‐U.S. Free Trade Agreement (opposed by labour groups) in the 1990s. The class effect remains with controls for religion and religious engagement (Model 3), but is non‐significant with controls for education, age, community size, marital status and gender (chi‐square=.71, p<.399) or with these controls plus income (Model 4: F=1.21, p>.50), though there is some slight suggestion that self‐employed workers remain more supportive of the Conservative than other class groupings (at p<.10). The second panel of Table 2A shows parallel findings for NDP (left party) support in 1990. Here, the class effects appear to be a bit weaker; without any controls, they are significant at p<.10 but 16 not p<.05, and in any of the models with controls, they become non‐significant, although particular contrasts remain significant. Unskilled manual and skilled manual workers are most supportive of the NDP, while all other groups (but especially managers) are less supportive. This is a classic blue collar/white collar pattern of class‐based party support, although it is not a particularly strong one. With all controls except income, there is a significant difference between unskilled workers and all other class groupings combined (chi‐square=4.39, p<.036) and a significant difference between blue collar workers (skilled and unskilled together) versus all others (chi‐square=6.25, p<.0124). These results are broadly consistent with earlier findings from Canada in the 1980s (Baer, 1987): especially if one uses the blue‐collar/white collar division as the class line and left‐party (NDP) support as the criterion, there is a modest level of class voting in Canada14. In Canada in 1990, there was a strong dividing line between Catholics (who tended to support the Liberals over the Conservatives) and Protestants, and this division shows in the first panel of Table 3A, holding up even with controls. This probably reflects, among other things, a long‐standing tendency of immigrants (primarily Catholic at the time) to support the immigration‐friendly Liberal party, though at the time the percentage of Canadian residents who were immigrants was substantially lower than it is at present. Fundamentalist/conservative Protestants seem to be even more supportive of the Conservatives than other Protestants, but their numbers were very small in 1990 and the difference between mainstream and conservative Protestants is, perhaps mostly on account of low statistical power, not statistically significant (chi‐square=2.20, df=1, p<.14). For Conservative party support, religious attendance did not appear to matter in 1990. Carrying this analysis over to NDP (left party) support in 1990, religious congregation has no effect on NDP support, though at the zero‐order (without controls) level, there is a tendency for those who label themselves as non‐religious to support the NDP more (odds ratio of 1.622, p<.05). It is also the case that those who attend church more are less likely to support the NDP (it may not be clear from Table 2A, but the religious attendance variable is coded so that a high score implies less church attendance. While the 17 class effects on NDP support are fairly weak, there is an additional social status relationship involving the income variable: those with higher incomes are less supportive of the NDP (p<.05). The income effect is fairly pronounced: with income measured on a 10‐point scale, an individual who is 3 points below the mean would have an expected probability of NDP support of .28, while an individual who is 3 points above the mean would have an expected probability of only .13 of voting for the NDP. There is also a strong union membership effect. These effects do not appear to be captured with the class typology employed here; the removal of the union and income variables does lead to a finding of statistical significance for the class variable. In the United States, at the same time, Ronald Reagan had just completed 8 years in office and George Bush Sr. had been elected to replace him. The distinction between Republicans and Democrats arguably offered a starker policy contrast than the division between Liberals and Conservatives, the two dominant Canadian parties, at the time, though one cannot discount the extent to which a later U.S. President (Clinton) would go on to essentially embrace some of the key components of the Reagan era, including a vigorous attack on welfare spending, in order to form a working coalition with conservative southern Democrats. At the same time, though, Canada continued to offer a “left” choice to voters even if, in many ridings, the NDP had little hope of achieving electoral success. Table 2B suggests that there would, at the time, be little support for the notion that blue collar workers – who would soon thereafter be identified as a key component to the “red state” suburban working class said to support politically right parties (see Hunter, 1991; XXX) – disproportionately supported Reagan or his successor. In fact, unskilled blue collar workers, as was the case with Canada, were least likely to show up at the polls and vote Republican (though in part this may have been due to a pattern, not tested here, of simply not voting). With controls, Table 2B suggests that there is no difference among classes, but managers are significantly more likely to support the Republicans than unskilled workers, and with the exclusion of income from the model (not shown in Table 2B), professionals (at p<.05) and the self‐
18 employed (at p<.10) are also more likely to support Republicans than manual unskilled workers . The income effect is of fairly strong magnitude: against mean support of around 36% (this includes non‐
voters), someone who is three income deciles above the mean would have an expected probability of .46 of voting Republican. If occupational and employment status categories did not create strong “class” effects in the U.S., it was at least the case, in 1990, that there was a strong relationship between economic status and support for the politically right Republican party. In the United States, religious attendance has the same conservatizing effect (in this case, drawing support to the Republican party) as it does in pulling support away from the left NDP party in Canada (note, though, from Table 2A, religious attendance did not pull Canadian respondents into voting for the Conservative party in 1990; it just directed them away from the NDP into either not voting or voting Liberal). A one unit difference in the independent variable represented the difference between attending church weekly and monthly, or between attending church monthly and a few times a year. In Canada, someone one unit above the mean level of church attendance would have an expected probability of voting for the NDP of about .18 (down from .20), while in the U.S., someone one unit above the mean level of church attendance would have an expected probability of voting for the Republicans of .39 (39%) as opposed to .36 at the mean level of church attendance. More pronounced differences would occur between extremes (e.g., between attending once or twice a year and attending more than once a week). Moving forward fifteen years, using WVS data (see table 3A), no significant class patterns are observed for Conservative party support in Canada, while marginally significant (p<.10) class effects are observed in Canada with respect to NDP vote, with managers being considerably less likely to support the NDP than workers, but without any strong distinction between the self‐employed (with less NDP support than unskilled/semi‐skilled manual workers) and most other occupational 19 categories (e.g., professionals, non‐manual and skilled manual workers). Here, the dividing point does not unify the manual working class: rather, the division is between the lowest status jobs and all others. While there is an income effect, net of class, on Conservative Party support in 2005, there is no such effect with respect to NDP support: the NDP now seems to draw supporters from all strata (as defined by income and education), though perhaps at a higher rate from those in marginal or at least unskilled jobs. There appears to be even less of a class effect in the United States. First, no block significance tests for class approach p<.10. If there is one slight pattern – evident only without any controls – it might be for unskilled workers to be less supportive of Republicans (note how much higher than 1.0 the odds ratios for the other class categories are in Model 1 of Table 3B), but the small Ns for this group preclude any definitive findings. With this group declining in magnitude with the restructuring of the economy, even substantially large effects will be less pronounced overall, politically, because they involve fewer people. Net of class, there is an income effect in the United States (while only a p<.10, it is not significantly different from the effect observed in Canada (t=1.25, p>.20), which is itself significant at p<.01), once again suggesting that there may be something in the way of political division that has been missed by the categorization scheme using World Values Study data. In 1990, we observed in English Canada a tendency for Protestants to support the Conservative party more than Catholics. This pattern is replicated in 2005, when Protestants remains significantly more likely than Catholics to choose the Conservative Party as their first choice, even with controls (see Model 4, Table 3A, odds ratio of 1.947). This pattern does not replicate, however, when it comes to opposition to the politically left NDP: while Protestants may be more likely to support right parties in Canada according to the WVS data, they are not less likely to support the NDP (if anything, the non‐
significant coefficient is positive in log‐odds form and greater than 1.0 as a odds‐ratio multiplier). Overall, religion appears to have little effect on the likelihood that one will nominate the NDP as one’s “first choice”15. So religion matters, but not perhaps in a fully coherent manner. In the United States, 20 it is a different matter in 2005: the effects of both religion and religious attendance are strongly statistically significant, both with and without controls, with the pattern being that Protestants of all types support the Republican party much more than individuals of other religions (and, especially, those specifying “no religion”), and those with high levels of religious attendance also supporting the Republican party more. The expected probability (with all variables held constant at their means) of voting republican for someone who never goes to church is .223, while the expected probability for someone who goes to church more than weekly is .515. This is a fairly dramatic difference, and it comes in addition to differences that occur between religious groups: at average levels of church attendance (and at the means of other variables), Catholics have a .328 expected probability of voting Republican, Protestants have a .488 probability, those with “other” religions have a .304 probability and those specifying no religion have a .254 probability. Clearly, religion matters quite a bit in American politics, at least according to the World Values Study dataset. As noted earlier, the World Values Study dataset did not break down Protestants into subgroups, so a further test of the distinction between conservative and mainstream congregations cannot be undertaken with the WVS data. In both Canada and the United States, additional tests were performed to see if there was an interaction between religion and levels of church attendance, as suggested by Manza and Brooks (2002; check this). No significant interactions were found in the United States, and the interaction in Canada (chi‐square=737, df=3, p<.0609) suggested that there might be a tendency for more devout Catholics to support the Conservative party less whereas more engaged Protestants and, especially, more religiously engaged individuals from “other” religions, to support the Conservatives more. The effects here are fairly weak, though. Is social class in any way implicated in the level of church attendance? Not shown in Table 3 is a secondary analysis giving adjusted mean levels of church attendance (with controls) for each class 21 grouping. In the United States, significant between‐class differences were found (F=2.37, df=5,554, p<.038); on a 1‐7 scale, with 7 representing no attendance at all, unskilled manual workers had an expected mean of 5.131 (least religious), against 3.739 for professionals (most religious), 4.698 for managers (less religious than most) and 4.1 to 4.2 for non‐manual workers, skilled manual workers and the self‐employed. If religious engagement dampens class commitments, it does so with greater difficulty within the lower skill/income segment of the manual working class, because individuals in this group are simply less engaged religiously. The occupation effect on religiosity is not observed in English Canada (F=1.80, df=5,979, p<.110). The Comparative Study of Electoral Systems dataset provides an opportunity to examine the differences between conservative Protestants and mainstream Protestants, though its class variables do not provide an opportunity to assess the difference between unskilled manual workers and skilled manual workers and it has no church attendance variable for Canada. This dataset also provides an additional triangulation on the WVS 2005 findings, where actual vote choices rather than hypothetical preferences are modeled. The results are provided in Tables 4A and 4B. First, for English Canada, there is no evidence of a connection between social class and propensity to vote for the Conservative party, which had just formed out of a union of the former very politically right Reform party and the more centrist but weaker Conservative party prior to the previous election. The religion effect took the same pattern that had been observed with WVS data both in 1990 and in 2005, but was perhaps a bit more pronounced: Catholics had an expected probability of .238 of voting Conservative, as opposed to .414 for mainstream Protestants, .472 for conservative/fundamentalist Protestants, .235 for other religions and .276 for those with no religion. There was no additional discernible effect for variations in levels of church attendance. For New Democratic (left) party support, once again there is no discernible class effect, either with or without 22 controls, though union membership predicts significantly higher probabilities of NDP voting (odds ratio of 1.767, p<.05). Religion is significant on its own, but the biggest between‐group difference is not so much that between Catholics and Protestants (a difference which is not statistically significant) but the difference between those with “no religion” and any of the other religious groups. The “no religion” group is significantly more likely to support the NDP. If the NDP was an “old line” economic left party, it nonetheless continued to attract the very “postmodern” voter (Inglehart et al., 1997) who putatively rejects both traditional religion and traditional political parties (and the “old” left‐right alignments they represent). Across both Conservative party support and NDP support, one consistent finding is that there is virtually no difference between mainstream and fundamentalist/conservative Protestants. But, overall, the strongest finding is more or less a non‐finding: there is not much going on, from the standpoint of “structural” antecedents, systematically in Canadian federal voting patterns. The story is quite different for the United States in 2005, at least as far as election study data associated with the CSES project are concerned. First, there are significant class effects which hold even with controls such as union membership and income. The main pattern is this: manual workers are considerably less likely to vote for the Republicans than most other groups, with the exception of professionals. While this similarity in political orientation between manual workers, on one hand, and professionals, on the other, might seem odd, it is the sort of interest coalition that is anticipated by Wright (1985): inasmuch as the Republican party has persistently presented itself as the forefront of a pro‐capitalist, anti‐regulatory, anti‐redistributive agenda over the course of the past two decade in the United States, it only makes sense that those who might have the most to lose (the declining manual working class) and those whose occupations are most connected to the welfare state (non self‐
employed professionals) would be, if all other considerations are somehow kept in abeyance, most opposed to the Republican party. Expressed as expected probabilities, evaluated at the mean values of the control variables, the likelihood of voting Republican was .434 for managers, .272 for professionals, 23 .485 for non‐manual workers, .315 for manual workers and .447 for the self‐employed. To be sure, the categories are not perfect: the CSES data do not distinguish between skilled manual workers and supervisors (who, if we follow the findings from Table 3B, might be quite different from other manual working class workers), and they do not allow for the possible inclusion of certain occupations not always considered to be “manual” (e.g., check‐out clerks) in a low‐status generally underpaid category of workers. Nor did the analysis constructed here allow for the full differentiation of workers between those in the public sector and those not (due, among other things, to small Ns), though the “class” orientations of third‐sector professional workers and even private sector salaried health care workers pose interesting issues for further micro‐analyses of specific occupations or quantitative analyses using very large N surveys. But the most important story, thus far, is that class still matters – or perhaps it is more accurate to say, class matters once again – in American politics. Aside from class groupings, other, related variables also have an impact on the propensity of Americans to vote Republican. In contrast to the Canadian case where, if anything, higher educated people were less likely to vote Conservative in 2005 (though this WVS data finding in Table 3A did not replicate with the CSES data in Table 4A), better educated Americans are more likely to vote Republican, though this effect would probably disappear in a conditional logit model (not estimated here) where Republican support is only evaluated among voters (that is, lower educated people may simply be less likely to vote, rather than less likely to vote Republican when they do vote). At the same time as class mattered in the 2004 American Presidential election, religion continues to play a role as well. Without controls, fundamentalist Protestants are actually less likely to vote Republican than their mainstream Protestant counterparts (odds ratio of .858 versus 1.621), but this difference is entirely attributable to compositional factors (how the two groups differ on other demographic variables, such that in models with controls (e.g., Model 3 and Model 4), the differences 24 between the two Protestant groups disappear. Finally, it remains the case that religious attendance matters: those who attend church more frequently are, regardless of denomination, more supportive of the Republican party. The puzzle that now arises is, “why are the structural influences of religion and social class present in the United States but not in Canada?” Perhaps there was something peculiar about the 2004 Canadian national election. After all, the new Conservative party had just been formed (with a fundamentalist Protestant minister the party leader of the larger partner in the right coalition party16), and had not yet established itself in Parliament or consolidated its voter base from the two component parties. Two years later, with a major political scandal to bring down the centrist Liberal party, the Conservatives would achieve power with a plurality of votes , a vote ratio that did not change much in the subsequent 2008 election (36‐37%). Table 5, then, presents results from the 2008 Canadian National Election Study17. The main findings from this table, both for Conservative support and for NDP support, are a) that there still do not appear to be any class effects and b) both religion and religious engagement (at least as measured by the imperfect, “importance of religion” variable) matter, with those who feel that religion is important in their lives voting to the right (more likely to vote Conservative, less likely to vote NDP), and with Protestants – both fundamentalist/conservative and mainstream – being more likely to vote Conservative and with fundamentalist/conservative Protestants significantly less likely to vote NDP than other groups. As was previously observed in 2005, higher educated people are less likely to support the Conservatives, while education has no discernible impact on the likelihood that an individual will vote NDP. Overall, then, the 2005 Canadian picture has not changed much: it remains the case that differences between class groupings remain fairly small and it remains the case that religion structures voting patterns in Canadian society, with odds ratios multiplies not too dissimilar from those seen in the 25 United States. What makes the Canadian case different, of course, is that political elites rarely make pronouncements invoking religious themes18, and the fact that conservative Protestants remain a fairly small group numerically in Canadian society, with no evidence of the massive growth associated with suburban mega‐churches in the U.S. (though a few do exist in Canada). Discussion As a first step in the discussion of the findings that have been outlined above, a brief summary of key findings is in order. In 1990, religion and class both played roles in English Canada for Conservative party support, though the only clear class pattern is one in which self‐employed individuals are more to the political right than other class groupings. For religion, Protestants were more Conservative. For NDP (left) support, the major class divide was been blue collar occupations and all others, and the major religious divide was one in which, regardless of religion, those who attended church more were less supportive. The same year, in the U.S. there was a class division with the main line of bifurcation existing between unskilled manual workers and all others. Among religious groups, if anything, fundamentalist Protestants supported the (right) Republican party more (though this effect disappeared with controls), but across all religious groups, greater religious involvement (church attendance) led to greater Republican party support, giving credence to the recent claim by Brint (2010) that it is the level of religious engagement, not the denomination, that matters. Fifteen years later, in 2005, class appears to matter in the U.S. (from one of two surveys, the one which actually measures vote as opposed to hypothetical vote preference), with unskilled workers and non self‐employed professionals less likely to vote Republican . At the same time, class has disappeared as a predictor in English Canada, confined only to the greater NDP support among manual workers in the World Values dataset (but no significant between‐class differences in the election study dataset). This disappearance continues into the 2008 election study data. This combination – class 26 possibly having come to life in the U.S. but disappearing in Canada as a predictor – Is exactly the opposite to that which was anticipated earlier on the a basis of historical and institutional considerations. Across both Canada and the United States, there is no evidence that the effect of class is conditional on religion or religious engagement, or vice versa. In 2005, in Canada, Protestants were more likely to vote Conservative, consistent with the 1990 finding, with no observable difference in effect between mainstream and conservative Protestants. NDP vote was less predictable from religious denomination, with only a slight tendency for the non‐religious to be more supportive of the NDP. Religious attendance (only measured in the WVS survey) did not have an effect on vote, either for the Conservatives or the NDP. At the same time, the overall effect of religion was stronger in the United States: not only are Protestants more likely to vote Republican, but religious attendance predicts Republican support. It was also the case that the difference between mainstream Protestants and conservative/fundamentalist Protestants was negligible (indeed, without controls conservative Protestants appeared to have a lower probability of voting for the Republicans, though this difference disappeared with controls). Overall, findings in both Canada and the United States give further credence to the claim that the, at the level of the mass public, the distinction between types of Protestant religions makes little if any difference (Moutlon, 2006; Brooks, 2002). The major finding requiring further explanation is the finding of the disappearance of class differences in Canada while, at the same time, if anything, they appear to be persistent (though changing in form) in the United States. How could this be? Speculatively, there may be a number of possible explanations, none of which can be adequately explored without new data. First, it could be argued that, the election of a Conservative party with key cabinet ministers who first saw political service in the rabidly anti‐welfare Ontario Conservative provincial government of the late 1990s19 notwithstanding, the federal government has played down elements that would give rise to a 27 polarization that would take the form of “class warfare,” choosing not to heavily publicize any desire to privatize government assets, reduce expenditures (other than those which follow the widely used, “create efficiencies in the civil service” theme). It may also be the case that the very universal medicare system which arguably arose from class politics in the 1960s20 has created social cohesion in a fashion that does not apply to dispossessed individuals in the United States, where higher levels of inequality in communities has led to at least some antagonisms (though this hardly explains the tendency, in the U.S., for professionals to vote against the Republicans). In any event, at least in the U.S. case, the “decline of structure” argument can, at most, be seen as a partial and incomplete description of political reality. Politics in the United States remains structured, indeed highly structured, on the basis of class, religion and other status variables (not to mention race). That Canada has not followed the same path is most certainly worthy of further investigation. References Achterberg, P. (2006). "Class Voting in the New Political Culture: Economic, Cultural and Environmental Voting in 20 Western Countries." International Sociology 21(2): 237‐261. Bellah, R. N. (2004). American Politics and the Dissenting Protestant Tradition. One Electorate Under God? A Dialogue on Religion and American Politics. E. J. J. Dionne, J. Elshtain and K. Drogosz. Washington D.C., Brookings Institute Press: 63‐66. Brint, S. and S. Abrutyn (2010). "Who's Right About the Right? Comparing Competing Explanations of the Link Between White Evangelicals and Conservative Politics in the United States." Journal for the Scientific Study of Religion 49(2): 328‐350. Brooks, C. (2002). "Religious Influence and the Politics of Family Decline concern: Trends, Sources and U.S. Political Behavior." American Sociological Review 67(2): 191‐211. Brooks, C. (2002). "Religious Influence and the Politics of Family Decline concern: Trends, Sources and U.S. Political Behavior." American Sociological Review 67(2): 191‐211. Elff, M. (2007). "Social Structure and Electoral Behavior in Comparative Perspective: The Decline of Social Cleavages in Western Europe Revisited." Perspectives on Politics 5(02): 277‐294. 28 Esping‐Anderson, G. (1989). "The Three Political Economies of the Welfare State." Canadian Review of Sociology and Anthropology 26(1): 10‐36. Esping‐Anderson, G. (1991). Postindustrial Cleavage Structures: A Comparison of Evolving Patterns of Social Stratification in Germany, Sweden and the United States. Labour Parties in Postindustrial Societies. F. F. Piven. Cambridge, Polity Press. Grabb, E. and J. Curtis (2004). Regions Apart. Toronto, Oxford University Press. Hoover, D., M. Martinze, et al. (2002). "Evangelicalism Meets the Continental Divide: Moral and Economic Conservatism in the United States and Canada." Political Research Quarterly 55(2): 351‐374. Houtman, D., P. Achterberg, et al. (2008). Farewell to the Leftist Working Class. New Brunswick NJ, Transaction Publishers. Hunter, J. (1991). Culture Wars: The Struggle to Define America. New York, Basic Books. Inglehart, R. (1997). Modernization and Postmodernization: Cultural, Economic and Political Change in 43 Countries. Princeton, NJ, Princeton University Press. Johnston, W. and D. Baer (1993). "Class Consciousness and National Contexts: Canada, Sweden and the United States in Historical Perspective." Canadian Journal of Sociology and Anthropology 30(2): 271‐295. Johnston, W. and M. Ornstein (1982). "Class, Work and Politics." Canadian Review of Sociology and Anthropology 19(2): 196‐214. Johnston, W. and M. Ornstein (1985). "Social Class and Political Ideology in Canada." Canadian Review of Sociology and Anthropology 22(3): 369‐393. Kingston, P. (2000). The Classless Society. Stanford, Stanford University Press. Lambert, R., J. Curtis, et al. (1987). "Social Class, Left/Right Political Orientations, and Subjective Class Voting in Provincial and Federal Elections." Canadian Review of Sociology and Anthropology 24(4): 526‐549. Malloy, J. (2009). "Bush/Harper? Canadian and American Evangelical Politics Compared." American Review of Canadian Studies 39(4): 352 ‐ 363. Manza, J. and C. Brooks (1997). "The Religious Factor in U.S. Presidential Elections, 1960–1992." American Journal of Sociology 103(1): 38‐81. McLoud, S. (2007). Class in American Religion and Religious Studies. Chapel Hill, NC, University of North Carolina Press. 29 Moulton, B., T. Hill, et al. (2006). "Religion and Trends in Euthanasia Attitudes among U.S. Adults, 1977‐
2004." Sociological Forum 21(2): 249‐272. Nakhaie, M. R. and R. Arnold (1996). "Class Position, Class Ideology and Class Voting: Mobilization of Support for the New Democratic Party in the Canadian Election of 1984." Canadian Review of Sociology and Anthropology 33(2): 181‐212. Pakulski, J. (2001). Breakdown of Class Politics. The Breakdown of Class Politics. T. Clark and S. Lipset. Baltimore, Johns Hopkins University Press: 137‐160. Patrikios, S. (2008). "American Republican Religion? Disentangling the Causal Link Between Religion and Politics in the U.S." Political Behavior 30(3): 367‐389. Picot, G., J. Myles, et al. (1990). Good Jobs/Bad Jobs and the Declining Middle: 1967‐1986. A. S. B. Business and Labour Market Analysis Group, Statistics Canada. Przeworski, A. a. J. S. (1986). Paper Stones: A History of Electoral Socialism. Chicago, University of Chicago Press. van der Waal, J., P. Achterberg, et al. (2007). "Class Is Not Dead It Has Been Buried Alive: Class Voting and Cultural Voting in Postwar Western Societies (1956 1990)." Politics Society 35(3): 403‐426. Wright, E. O. (1985). Classes. London, Verso. Endnotes 1
This effect can easily be overstated, though; one merely needs to work through cable TV channels in Canada as opposed to those available on an American cable TV system to note the incredibly large number of (Christian) religion‐devoted channels available in the US in comparison to Canada. In Canada, much religious programming takes the form of programs in immigrant languages appealing to new arrivals (for example, Sikhs). 2
In 1990, the Conservative Party was called the Progressive Conservative Party. Another right party (generally more to the political right than the right‐centre Conservatives), the Reform Party, was included in the construction of this variable (thus, either support for the Reform or the Progressive Conservatives as coded “1”). In 2003, the two parties merged to form the Conservative Party. 3
The one major difference lies in the differentiation between manual and non‐manual workers here. 4
For the 2008 CNES, a logged version of the income variable was tested, but found not to improve model deviance values. For this reason, the unlogged version of this variable was retained. 5
In the case of Methodists and even Lutherans, Steensland (2000) argues that some branches should properly be labeled conservative and not mainstream, but additional information to allow for this further distinction was not available. 6
The U.S. wording was slightly different, with respondents first being asked whether they considered religion important and then, if the answer was in the affirmative, whether their religion provided a) some, b) quite a bit or c) a great deal of “guidance in your day‐to‐day living”. The lowest point on the 4‐point scale was given to respondents who answered “no” to the first question. 7
Some preliminary investigations suggested that, given the range of ages in the analysis, the use of a quadratic term for age to catch possible non‐linearities was not warranted. 8
In many American studies of the relationship between religious denomination and political attitudes or behavior (see, for example, Brooks, 2002), “black Protestant” (or even sometimes black fundamentalist Protestant) denominations are separated from other denominations. In a number of instances, the possibility that this 30 additional religion category might be necessary in the U.S. was tested with interaction terms between religion and race. In general, the outcome of these tests was that there were no substantial interactions. 9
These models excluded the race variable, which was not always available in the Canadian data. 10
The missing at random assumption assumes only that the pattern of missingness is not a function of the values of the dependent variable that have not been accounted by other variables in the model, and is less restrictive than the MCAR, missing completely at random, assumption that is made when some other approaches, such as listwise deletion, are employed. 11
As noted by Manza and Brooks (1997), this may not necessarily mean abandonment of religion altogether in the U.S. case, as many individuals listing “no religion” there still retain beliefs in God and may even attend religious services. 12
An omnibus test for class differences across Canada and the United States yielded a chi‐square value (df=4) of 1.85 (p<.764). A test for Canada‐U.S. differences was similarly non‐significant. 13
It cannot, however, be argued that Canada and the U.S. are substantively different, even though results in one country are significant and the results in the other are not: the test for differences in the effects of occupation across countries was non‐significant (F=.98, df=5,2191,p<.43). 14
In declaring the class vote in Canada to be among the lowest in the world, Alford used analyses which created a single category out of a combination of the Liberal party (whose centrist policies at the time were only slightly to the left of the Conservatives) and the NDP (which still, at the time, had “socialism” as its official policy platform objective, though in practice this tended to take the form of support for public sector initiatives and institutions within the capitalist system). 15
Though note the 1.129 odds ratio for religious attendance in Model 4, significant only at p<.10, indicating that those who attend church more are less supportive (since religious attendance is coded with higher values signifying less attendance). 16
Unlike American political candidates, though, this individual did not make strong public references to his religiosity or his belief in God in his campaigning or his public actions as a member of Parliament. 17
As noted above, comparable data for the construction of a class variable are not presently available for the American case in 2008. 18
A remark making use of the term “God” by the Canadian prime minister that would be considered a non‐event in the US. in July of 2010 drew a considerable amount of editorial reaction (not all negative) in the Canadian press inasmuch as it was fairly unusual. Malloy (2009) reports an earlier 2006 event of a similar nature. 19
The Mike Harris government was roundly defeated in 2004 after some of its radical anti‐state measures resulted in negative consequences (the most notable of which was the death of a number of individuals from a contaminated community water supply after a big reduction in regulatory oversight). Key Harris cabinet ministers, however, moved to federal politics and were able win seats in historically strongly Conservative ridings. 20
In the 1960s, the NDP played a strong role pushing the ruling (but minority government) Liberal party into adopting medicare measured the NDP party itself had pioneered in one Canadian province. Table 1A
Source: World Values Study, 1990
Percentages:
English United Canada
States
(Quebec)
Class:
Manager
Professional
Non‐Manual
Skilled Manual
Unskilled Manual
Self‐employed
7.3
17.94
22
12.2
31.28
9.28
9.97
24.16
19.94
22.81
19.32
3.75
7.38
21.23
24.62
8
30.77
8
Roman Catholic
Mainstream Protest.
Fund/conserv Prot.
Other religion
No religion
28.21
32.95
2.09
4.75
32
32.95
28.99
7.98
11.32
23.29
80
3.06
0.83
5
11.11
Attend weekly or more
Attend monthly or less
Never attend
24.62
40.68
34.7
42.54
38.81
18.65
22.75
41.83
24.88
19.28
19.75
35.75
‐‐
19.17
15.56
Religion:
Religiosity:
Conservative vote:
NDP (left) vote:
USA: Conservative vote = Republican
Canada: Conservative vote = Conservative + Reform Party
Table 1B
Source: World Values Study, 2005
Percentages:
English United Canada
States
(Quebec)
Class:
Manager
Professional
Non‐Manual
Skilled Manual
Unskilled Manual
Self‐employed
8.63
22.46
19.42
27.16
10.15
12.18
14.11
23.63
17.46
20.11
5.64
19.05
8
25.6
16.8
27.2
13.6
8.8
Roman Catholic
Protestant
31.38
18.12
20.2
28.54
66.25
2.98
Other religion
No religion
15.98
34.51
18.89
32.36
7.69
23.08
Attend weekly or more
Attend monthly or less
Never attend
23.33
45.91
30.77
32.88
38.03
29.1
12.59
41.06
46.35
Religion:
Religiosity:
Conservative vote:
30.97
31.26
16.63
NDP (left) vote:
20.43
‐‐
8.19
As a percentage of total eligible voters, incl non‐voters
Table 1B
Source: World Values Study, 2005
Percentages:
English United Canada
States
(Quebec)
Class:
Manager
Professional
Non‐Manual
Skilled Manual
Unskilled Manual
Self‐employed
8.63
22.46
19.42
27.16
10.15
12.18
14.11
23.63
17.46
20.11
5.64
19.05
8
25.6
16.8
27.2
13.6
8.8
Roman Catholic
Protestant
31.38
18.12
20.2
28.54
66.25
2.98
Other religion
No religion
15.98
34.51
18.89
32.36
7.69
23.08
Attend weekly or more
Attend monthly or less
Never attend
23.33
45.91
30.77
32.88
38.03
29.1
12.59
41.06
46.35
Religion:
Religiosity:
Conservative vote:
30.97
31.26
16.63
NDP (left) vote:
20.43
‐‐
8.19
As a percentage of total eligible voters, incl non‐voters
Table 2A
Source: WVS
Party Support, Canada, 1990
Odds ratio multipliers:
Conservative Vote
Model 1 Model 2 Model 3
Model 4
Class:
Manager
Professional
Non‐Manual
Skilled Manual
Unskilled Manual
Self‐employed
Block chi‐square
1.256
1.129
1.532
1.33
(ref)
2.132**
8.54
1.157
1.092
1.503
1.309
(ref)
2.113*
8.17
0.92
0.88
1.339
1.1788
1.712#
F=1.21
Religion:
Roman Catholic
Mainstream Protest.
Fund/conserv Prot.
Other religion
No religion
Block chi‐square
Religious Attendance*
Income
Education
Age
Community size
Union membership (not)
Gender (female)
(ref)
2.414***
5.926***
1.155
1.444
24.90***
2.476 2.381**
4.498*
4.604*
1.327
1.322
1.365
1.341
17.22** F=3.88**
1.046
1.059
1.083
1.073
1.061
0.89
1.519
0.906
Table 2A
Source: WVS
Party Support, Canada, 1990
Odds ratio multipliers:
NDP (Left) Support
Model 1 Model 2 Model 3
Model 4
Class:
Manager
Professional
Non‐Manual
Skilled Manual
Unskilled Manual
Self‐employed
Block chi‐square
.460*
.602*
.584*
0.944
(ref)
0.645
10.42#
.462*
0.647
0.617*
0.972
0.627
0.815
0.712
1.097
0.631
9.07
0.779
F=.73
1.143
1.111
1.191
1.474
2.67
1.285
0.947
1.206
1.432
F=.50
Religion:
Roman Catholic
Mainstream Protest.
Fund/conserv Prot.
Other religion
No religion
Block chi‐square
Religious Attendance*
Income
Education
Age
Community size
Union membership (not)
Gender (female)
(ref)
1.196
0.994
1.029
1.622*
5.87
1.128**
1.134**
.858*
0.975
.812**
1.03
.493**
1.082
Table 2B
Republican Party Support, USA, 1990
Odds ratio multipliers:
Model 1
Model 2
Model 3
Model 4
Class:
Manager
Professional
Non‐Manual
Skilled Manual
Unskilled Manual
Self‐employed
Block chi‐square
2.267***
2.355***
1.719**
1.554*
(ref)
2.279*
22.71***
2.165** 1.787*
2.199***
1.555
1.646*
1.482
1.451*
1.408
(ref)
2.274*
1.832
18.71** F=1.20
Religion:
Roman Catholic
Mainstream Protest.
Fund/conserv Prot.
Other religion
No religion
Block chi‐square
Religious Attendance*
Income
Education
Age
Community size
Union membership
Gender (female)
1.324
1.617*
1.04
0.833
11.63*
1.259
1.269
1.496
1.679
1.146
1.221
1.092
1.241
2.72 F=.84
.885***
.875***
1.164**
1.078#
0.974
0.896
1.480#
.719*
Table 3A
Source: WVS
Party Support, Canada, 2005
Odds ratio multipliers:
Conservative Vote
Model 1 Model 2 Model 3
Model 4
Class:
Manager
Professional
Non‐Manual
Skilled Manual
Unskilled Manual
Self‐employed
Block chi‐square
1.046
1.623
0.96
1.45
(ref)
0.956
5.73
1.069
1.525
0.953
1.48
1.1
1.959
1.169
1.536
1.034
4.53
0.986
F=1.21
1.602*
1.571#
1.947*
1.01
.630*
17.06***
0.982
0.727
6.65#
1.139
0.787
F=2.87#
Religion:
Roman Catholic
Mainstream Protest.
Other religion
No religion
Block chi‐square
Religious Attendance*
Income
Education
Age
Community size
Union membership (not)
Gender (female)
0.937 .897#
1.199***
.818***
0.986
.905*
0.907
0.781
Table 3A
Source: WVS
Party Support, Canada, 2005
Odds ratio multipliers:
NDP (left) vote
Model 1 Model 2 Model 3
Model 4
Class:
Manager
Professional
Non‐Manual
Skilled Manual
Unskilled Manual
Self‐employed
Block chi‐square
.237**
.467*
0.544
.554#
(ref)
0.587
9.01#
.232**
.457*
.519#
.519#
.275*
.468#
.474#
0.625
0.524
9.08#
0.675
F=1.27
(ref)
1.366
1.308
1.262
1.256
1.725*
5.69
1.227
1.19
0.81
1.182
1.123
F=.27
1.096
1.129#
Religion:
Roman Catholic
Protestant (main+fund)
Other religion
No religion
Block chi‐square
Religious Attendance*
Income
Education
Age
Community size
Union membership (not)
Gender (female)
0.906
1.033
0.997
1.087
1.599#
1.880**
Table 3B
Source: WVS
Party Support, USA, 2005
Odds ratio multipliers:
Republican Vote
Model 1 Model 2 Model 3
Source: WVS
Model 4
Class:
Manager
Professional
Non‐Manual
Skilled Manual
Unskilled Manual
Self‐employed
Block chi‐square
1.601
2.028
1.602
1.861
(ref)
1.978
2.3
1.707
1.368
1.266
1.627
1.331
1.139
1.172
1.359
1.882
2.37
1.375
F=.12
1.929*
2.336*
0.871
0.639
.302***
0.593
47.20*** 17.91***
0.977
0.694
4.35*
Religion:
Roman Catholic
Protestant (main+fund)
Other religion
No religion
Block chi‐square
Religious Attendance*
*high score=less attend
Income
Education
Age
Community size
Union membership (not)
Gender (female)
Black (ref)
1.714*
.781*** .779***
1.164#
1.009
0.985
0.922
0.609
0.749
.091***
Table 4A
Source: CompElSys
Party Support, Canada, 2005
Odds ratio multipliers:
Conservative Vote
Model 1 Model 2 Model 3
Model 4
Model 5
Class:
Manager
Professional
Non‐Manual
Manual
Self‐employed
Block chi‐square
Block F‐test
1.375
0.9703
1.301
1.429
1.001
1.33
1.371
1.168
1.134
1.387
1.168
1.155
1.702
3.44
1.664
3.27
1.361
1.372
0.47
0.34
Religion:
Roman Catholic
Mainstream Protest.
Conserv Prot.
Other religion
No religion
Block chi‐square
Block F‐test
Income
Education
Age
Community size
Union membership (not)
Gender (female)
1.834** 2.669*** 2.242*** 2.254***
1.882**
2.872* 3.139**
3.033*
0.82
1.038
0.976
1.025
1.001
1.399
1.166
1.335
18.65*** 16.15**
4.55**
3.86**
1.059
0.957
1.059
0.934
0.75
.640*
1.063
0.956
0.99
0.935
0.756
.637*
importance of relgion (not)
Table 4A
Source:CompElSys
1.064
Party Support, Canada, 2005
Odds ratio multipliers:
NDP (left) vote
Model 1 Model 2 Model 3
Model 4
Model 5
Class:
Manager
Professional
Non‐Manual
Manual (ref)
Self‐employed
Block chi‐square
Block F‐test
0.795
0.762
1.033
0.765
0.734
0.995
0.846
0.64
1.21
0.812
0.636
1.152
0.607
1.62
0.559
2.000
0.804
0.796
0.64
0.63
0.665
0.5225
1.453
1.727#
0.658
0.577
1.276
1.266
2.87*
0.88
1.01
0.976
0.986
1.107
1.767*
1.463
1.005
0.979
0.987
1.105
1.726*
1.486
0.866
Religion:
Roman Catholic (ref)
Mainstream Protest.
Fund Protest
Other religion
No religion
Block chi‐square
Block F‐test
Income
Education
Age
Community size
Union membership (not)
Gender (female)
importance of religion (not)
0.654
0.453
1.155
1.735*
18.94***
0.556
0.543
1.122
1.11
4.56
Table 4B
Source CompStElSys
Party Support, USA, 2005
Odds ratio multipliers:
Republican vote
Model 1
Model 2
Model 3
Model 4
Model 5
Class:
Manager
Professional
Non‐Manual
Manual
Self‐employed
Block chi‐square
Block F‐test
2.571***
1.347
2.506***
2.456***
1.336
2.517***
1.392
0.686
1.911*
1.362
0.597
1.836*
2.128*
22.69***
2.091*
21.52***
1.478
1.323
4.26**
4.84**
2.032**
2.032**
0.822
1.03
1.942**
2.038**
1.836
1.464
3.80**
3.15*
1.237*
1.224**
1.024**
0.813*
0.602
0.744
0.104***
1.262
1.212*
1.021**
0.818*
0.63
0.748
.093***
1.215**
Religion:
Roman Catholic
Mainstream Protest.
Fund. Prot
Other religion
No religion
Block chi‐square
Block F‐test
Income
Education
Age
Community size
Union membership (not)
Gender (female)
Black Religious attendance
(ref)
1.621*
0.858
0.688
.656#
15.94**
1.761*
0.947
0.784
0.743
13.35**
Table 5
Source: Cdn. Nat El Study
Party Support, Canada, 2008
Odds ratio multipliers:
Conservative Vote
Model 1 Model 2 Model 3
Model 4
Class:
Manager
Professional
Non‐Manual
Manual (ref)
Semi,unsk. Manual
Self‐employed
Block F‐test
Block F‐test
0.65
0.768
0.861
0.805
(ref)
1.01
2.45
0.939
1.408
1.215
0.916
0.934
1.478
1.228
0.930
0.931
0.950
0.58
0.660
1.328
1.783
.370*
.551*
1.360
1.504
.384*
1.117
6.16***
2.53*
1.003#
.866**
0.722
0.724
.711#
1.003
.860*
0.714
0.707
.696#
.851*
Religion:
Roman Catholic (ref)
Mainstream Protest.
Conserv Prot.
Other religion
No religion
Block Chi‐square
Block F‐test
Education
Age
Union membership Public sector employ.
Gender (female)
Importance of religion (not)
1.404*
2.006*
0.374**
0.565**
45.05***
Table 5
Importance of religion
Party Support, Canada, 2008
Odds ratio multipliers:
NDP (left) vote
Model 1 Model 2 Model 3
Model 4
Class:
Manager
Professional
Non‐Manual
Manual (ref)
Semi,unsk. Manual
Self‐employed
Block chi‐sq
Block F‐test
0.65
0.768
0.861
0.805
(ref)
1.01
2.45
1.014
0.716
0.754
0.999
1.032
0.699
0.749
0.975
0.634
0.635
0.5
0.52
1.08
0.373
0.976
1.907*
1.07
0.467
0.884
0.86
2.72*
0.46
Religion:
Roman Catholic
Mainstream Protest.
Conserv Prot.
Other religion
No religion
Block chi‐sq
Block F‐test
Education
Age
Community size
Union membership Public sector employ.
Gender (female)
Importance of religion (not)
0.921
0.403*
0.777
1.361
10.27*
0.998
0.997
1.058
1.069
0.995
0.994
2.359*** 2.459***
0.738
0.749
1.901**
1.936**
1.210*