The GLBT Movement: From Identity to Behaviour

Political research about the GLBT community has focused for the most part on the issue of equal rights and identity, and on the evolution of the GLBT “movement” within an urban context (see, for example, Bailey, Reference Bailey1999; Rimmerman et al., Reference Rimmerman, Wald and Wilcox2000; Smith and Haider-Markel, Reference Smith and Haider-Markel2002). As David Rayside (Reference Rayside1998) pointed out more than 10 years ago, studies that focus on GLBT voters are relatively rare, a fact that persists today.

The little work that exists on the voting behaviour of this constituency focuses on the United States and suggests these voters have a propensity to support the Democratic party (Bailey, Reference Bailey1998; Egan, Reference Egan2004; Hertzog, Reference Hertzog1996). A 2007 online poll of gay, lesbian and bisexual respondents in the US reports that between 83 and 88 per cent of them identify with the Democratic party (Egan et al., Reference Egan, Edelman and Sherrill2008). However, such support is far from absolute. Between a quarter and a third of this constituency voted Republican in presidential and mid-term elections during the 1990s (Bailey, Reference Bailey1998, Reference Bailey2000), a pattern still evident in the 2010 US midterm elections (GOProud, 2010). Some recent research suggests support in this community is leaning more heavily than previously to independent and Green candidates (Rollins and Hirsch, Reference Rollins and Hirsch2003). Nonetheless, the vast majority of gay, lesbian and bisexual voters identify as Democrats, and vote consistently with that identification.

Several possible reasons have been proposed for the community's decided tilt to the Democrats. One of these relates to the history of the movement, in particular its decision to use the political process as a means of advancing its fight for human rights, equality and social benefits (Rayside, Reference Rayside1998; Riggle and Tadlock, Reference Riggle and Tadlock1999). In US politics over the past three decades, there is strong evidence of increasing party polarization on a broad range of issue dimensions including homosexual rights (Baldassarri and Gelman, Reference Baldassarri and Gelman2008; Saunders and Abramowitz, Reference Saunders and Abramowitz2004). In this context, the Democratic party has clearly been seen as more sympathetic to the aspirations of the community (Bailey, Reference Bailey1998, Reference Bailey2000; Egan, Reference Egan2004, Reference Egan2008). It may be, then, that its distinctive partisan hue is simply a function of a single-issue constituency making an interest calculation. Supporting this thesis, Schaffner and Senic (Reference Schaffner and Senic2006) use survey data to show that group-related benefits, both economic (such as same-sex spousal benefits) and symbolic, are important for understanding lesbian, gay and bisexual support for the Democratic party.

The predisposition to vote Democrat may also result from a more general ideological perspective of the GLBT community. There is certainly evidence to suggest GLBT voters in the US tend to position themselves on the “left,” and not just on issues overtly associated with sexual orientation; they are also found to hold more progressive positions on other policy issues such as the environment, the role of government and the Iraq War (Egan, Reference Egan2008.; Egan et al., Reference Egan, Edelman and Sherrill2008; Hertzog, Reference Hertzog1996).

Three explanations have been proposed for why this is so, all deriving from the fact that publicly identifying as a member of this community is usually a deliberate choice of the individual—a choice typically made in late adolescence or adulthood.Footnote ¹ The first of these explanations holds that ideology may be a precondition for such a choice. That is, while there is no reason to think sexual orientation per se is anything but randomly distributed across the ideological spectrum, the deliberate choice to identify publicly as a member of the GLBT community may require views about societal norms, tolerance of difference and openness to societal change typically associated with the left side of the political spectrum. Egan (Reference Egan2009) reports evidence supporting such a process among lesbian, gay and bisexual respondents in the US General Social Survey. These respondents were more likely to come from families with more progressive views and from regions with more liberal voting records. If such a selection bias is at work in the identification process, then the leftist ideological hue of this constituency may arise in part as a condition for membership.

The other two explanations reverse the temporal relationship between the two, in that the choice to identify publicly is thought to initiate a resocialization of the individual. One version of this views the process as a relatively immediate and personal “conversion” experience. That is, a decision to embrace the gay, lesbian, bisexual or transgendered lifestyle is a profound life-altering event, triggering a reconsideration of assumptions about society and politics, greater sympathy for other marginalized groups, and a rethinking of political allies (Egan et al., Reference Egan, Edelman and Sherrill2008). Bjorn Lomborg, author of The Skeptical Environmentalist (Reference Lomborg1998) makes this argument very explicitly in accounting for the development of his environmentalism (Allemang, Reference Allemang2009). The other version—what Egan (Reference Egan2009) terms “embeddedness”—involves a slower process of acculturation into the political subculture that results from new primary and reference group associations (Bailey, Reference Bailey1999; Sherrill, Reference Sherrill1996). Evidence for either of these processes is largely anecdotal, but Egan's (Reference Egan2009) recent reanalysis of US exit poll data finds strong evidence for the “conversion” thesis and only mixed support for the “embeddedness” alternative. Should we expect this ideological profile to be replicated in Canada? If theories explaining the leftist ideological profile have validity, there is no reason to think the same forces are not at work here, leading us to expect a decidedly leftist Canadian GLBT community.

On the question of partisan preferences, application of the American findings is complicated by the fact that the Canadian party system offers more choices. Party polarization on most issues has not been as dramatic in Canada, and party polarization on the issue of homosexual rights has been more latent than in the United States.Footnote ² Nevertheless, sizeable proportions of the Canadian public are able to order the federal parties on a left–right political continuum (Lambert, Reference Lambert1983; Lambert et al., Reference Lambert, Curtis, Brown and Kay1986, Reference Lambert, Curtis, Brown and Kay1988); there are clearly discernable differences between the Conservative Party of Canada (CPC) and the other parties in their willingness to accommodate the homosexual rights agenda. And on the same-sex marriage issue, in particular, there is a clear ordering of partisan support. The CPC and its forerunner, the Canadian Alliance, have voted consistently to preserve the term “marriage” for unions of heterosexual couples only. The CPC opposed the 2005 Civil Marriage Act that legalized same-sex marriage in Canada, and attempted unsuccessfully to reopen the question when it gained power in 2006 as a minority government. The New Democratic party and Bloc Québécois have been consistently supportive of legalization, while the Liberal party has been more equivocal. Since 2003, Liberal party leadership has been largely supportive of same-sex marriage, but the caucus has been more divided. The minority Liberal government of Paul Martin was responsible for introducing same-sex legislation in 2005, but about one-third of the Liberal backbenchers eventually voted against that legislation in a free vote. The Liberal party division was not as much in evidence when the CPC minority government moved to reopen the question the following year. For that motion, only 13 of 98 Liberals voted to reconsider the legalization of same-sex marriage. It therefore comes as no surprise to find the NDP gaining more on this issue among those who favour same-sex marriage (Dostie-Goulet, Reference Dostie-Goulet2006).

All of this might lead us to expect the GLBT community to have a distinct partisan profile. That is, if the GLBT rights agenda affects the vote, or if the community positions itself ideologically on the left and allows that to shape their decisions, these voters should clearly favour the NDP and the Bloc Québécois, and clearly reject the Conservatives. We might expect to see this pattern most evidently in Quebec. Until recently, given the strength of the BQ in most of that province's constituencies, a BQ vote will both support a traditional partisan ally and enhance the likelihood of defeating the Conservative candidate. Outside of Quebec, however, the multiparty system in Canada complicates the picture somewhat because an NDP vote there may actually assist Conservative party candidates by splitting the centre–left vote in the many constituencies where Liberal candidates are the main opposition. So, while we would certainly expect little support for the Conservative party in the rest of Canada (ROC), strategic voting on the part of GLBT voters may weaken voting support for the NDP.

An additional question we explore here concerns possible differences in voting behaviour between GLBT males and females. For the most part, extant literature on the political behaviour of this community has not addressed this question, but there are reasons to suspect such differences. In the gender and voting literature, there is evidence women voters are more likely than their male counterparts to support parties of the left (Erickson and O'Neill, Reference Erickson and O'Neill2002). Part of this gap has been explained by a difference in value structures, in that women are found to be more concerned about social issues, and are not nearly as “sold” on market-oriented solutions to social problems (Gidengil, Reference Gidengil1995; Gidengil et al., Reference Gidengil, Hennigar, Blais and Nevitte2005; Inglehart and Norris, Reference Inglehart and Norris2003). But the gap has also been attributed in part to the feminist movement and its role in sensitizing women to issue concerns typically found on the left (Conover, Reference Conover1988). The fact that the feminist movement in Canada has close ties to the NDP and BQ and that it has traditionally championed the causes of marginalized women in society, suggests it may be an especially potent political reference group for GLBT women, with the consequence that, relative to GLBT men, women may have stronger partisan commitments to the NDP and BQ. As Blais (Reference Blais2002) has demonstrated, strength of partisanship affects one's loyalty to a preferred party, and negatively affects one's likelihood of defecting to a more competitive party for strategic reasons.

Data and Methods

Data for this analysis originate from a survey administered by Ipsos Reid's Public Affairs Division to the company's online panel on January 23, 2006, the day of Canada's 39^th federal election. Recruitment to this panel is ongoing and is conducted by a variety of means. Some panellists are invited from online sources such as websites and email, and others from offline sources such as telephone interviews. On the election day, all panel members were invited to participate once they had cast their ballot, and 35,675 of them responded to the invitation.Footnote ³

Opt-in samples of this kind provide an attractive opportunity to acquire very large samples of the public at much lower costs than the alternatives, but they also pose a number of challenges for scholarly research. Perhaps the most vexing of these challenges is the difficulty of establishing that the sample adequately represents the target population in socio-demographic, attitudinal and behavioural terms. The difficulty arises in part because parameters for the target population of voters (in our case) are themselves not firmly established. In larger part, however, it is because they are not probability samples: they consist only of those people reachable by the internet who have chosen to join the panel and to respond to the company's survey invitation. Hence, there is ample reason to suspect sample biases have crept in. While “coverage” biases would seem to be diminishing as a concernFootnote ⁴ (especially relative to the telephone alternative with the public's switch from landlines), “self-selection” and “non-response” biases remain significant issues. Extant research to assess the nature and extent of these biases suggests that reweighting to approximate socio-demographic parameters and to adjust for propensity to join a panel helps to eliminate some differences between offline probability and online non-probability samples, but the success rate here is not universal. As a consequence, researchers who have looked at this question suggest opt-in panel surveys should be used with caution, and are perhaps more useful for exploring group differences than for estimating absolute frequencies of phenomena (see, for example, Borges and Clarke, Reference Borges and Clarke2008; Chang and Krosnick, Reference Chang and Krosnick2009; Loosveldt and Sonck, Reference Loosveldt and Sonck2008; Malhorta, Reference Malhorta2008; Stephenson and Crête, Reference Stephenson and Crête2011; Vavreck and Rivers, Reference Vavreck and Rivers2008; Yeager et al., Reference Yeager, Krosnick, Chang, Javitz, Levindusky, Simpser and Wang2009).

Informed by this literature, we adopted two strategies to address potential sampling biases. First, to assess the degree to which the online sample exhibited a different pattern of relationships among variables, we compared the Ipsos Reid election-day sample to the subset of voters in the Canadian Election Study (CES) survey of 2006. For this, we employed a similar methodology used in Malhorta and Krosnick (Reference Malhorta and Krosnick2007). That is, the Ipsos Reid and the CES surveys were pooled together and tested against a series of regression models based on common variables. Each model contained an interactive term composed of the demographic variable and a separate variable that identified the survey mode (1 = on-line survey; 0 = CES). Significant interactive terms would suggest a “sample” effect, indicating that relationships in the online poll are substantially different than those found in the CES.

In our case, the dependent variable was vote choice. We ran separate models for Quebec and the rest of Canada. In both cases, we were restricted to using only demographic items, as the Ipsos Reid survey did not contain variables that measured partisanship, media exposure, leadership perceptions and a host of other variables typically found in the CES. Of the six demographic variables tested (region, age, religion, education, income and gender), few interactions were detected,Footnote ⁵ enhancing confidence that sampling biases in the Ipsos Reid survey are modest in proportion, at least relative to the benchmark CES.

Second, to attenuate self-selection biases in our sample, we employed two approaches. We reweighted the data to reflect the profile of voters found in the 2006 Canadian Election Study data. It should be noted that we were limited to reweighting on demographic variables, so other differences such as in political sophistication, information and media use remain possible biases in our data (see, for example, Hill et al., Reference Hill, Lo, Vavreck and Zaller2007). That said, the fact our sample yields reasonable estimates of other known political parameters, such as party support levels, suggests it provides a serviceable profile of the Canadian electorate. We also included the demographic variables as controls in our vote-choice regression models (see Loewen, Reference Loewen2010: 670).

About 30 questions were posed in the questionnaire. Most pertain to the respondent's attitudes and behaviour surrounding the election, but there are also questions probing the respondent's socio-demographic background. Among the latter is the following question: “Are you a member of the gay, lesbian, bisexual or transgender community?” In all, 1,270 respondents or 3.7 per cent of the sample responded in the affirmative.

Given the nature of our data, we should be clear about the population to which our results might apply. Certainly, it is not the population of gays, lesbians, bisexuals and transsexuals in the larger community, but that subgroup of this population which is prepared to identify publicly as such. The picture is further complicated by use of the term “community” in the wording of the question posed. It is conceivable that some individuals publicly identify their sexual orientation as other than exclusively heterosexual but do not see themselves as members of an associated “community.” In that event, our data will underestimate the size of the target subgroup. Moreover, because feeling a sense of community may affect the expression of political solidarity with the group, our data may overestimate the degree to which the target subgroup exhibits cohesiveness in its political choices.

That said, our estimate of the size of this subgroup population is generally consistent with what other surveys have reported for similar populations in the US (see, for example: Bailey, Reference Bailey1989, Reference Bailey1998, Reference Bailey2000; Edelman, Reference Edelman1993; Egan et al., Reference Egan, Edelman and Sherrill2008, Egan, Reference Egan2009; Schaffner and Senic, Reference Schaffner and Senic2006). The most recent and arguably the most rigorous survey of publicly identified gay, lesbian and bisexual respondents so far undertaken (transgendered were not included in the study population) is the 2007 Hunter College poll of US adults (Egan et al., Reference Egan, Edelman and Sherrill2008). It estimates the size of that community at 2.9 per cent of the US population. About half of these respondents describe themselves as “bisexual” rather than “gay,” “lesbian” or “homosexual.” More recent still, the 2008 American National Election Study included a question on sexual orientation which shows 4.1 per cent of respondents (93 out of 2,322) publicly identified as homosexual or bisexual.

Questions pertaining to the respondents' political attitudes and behaviour were not exhaustive, but they do furnish some respondent observations about the campaign, their perceptions of the parties, the immediate rationale for their party choice and the nature of that choice. In addition, the survey asked a series of items that tap respondents' ideological orientations and views on several issues current in the campaign.

In the analysis that follows, we first sketch the socio-demographic and political profile of GLBT voters, before exploring possible explanations for their political behaviour. Analysis employs a variety of techniques. Cross-tabulations are generated to provide a comparative profile of the GLBT community while regression models test expectations about the motivations of the GLBT vote.

Social and Political Profile of the GLBT Electorate

What does the Canadian community of publicly identified gays, lesbians, bisexuals and transgendered look like in socio-demographic terms? A range of such variables were recorded for each respondent, including region, community size, religion, gender, income and education.Footnote ⁶Table 1 profiles GLBT and non-GLBT subsamples for these variables. It can be seen that the GLBT community is distinct in a number of respects. First, the community tends to be disproportionately male by a factor of three to one. This is not out of line with other research. Previous studies in Canada (Canada, 2004) and elsewhere (for instance, Black et al., Reference Black, Gates, Sanders and Taylor2000; Smith, Reference Smith2008; Wellings et al., Reference Wellings, Field, Wadsworth, Johnson, Anderson and Bradshaw1990) have consistently reported a higher male-to-female ratio of respondents who publicly identify as homosexual or bisexual. Second, the GLBT community is distinctive in that it has a relatively younger profile relative to the entire sample. Slightly less than 20 per cent of the group is under the age of 35 compared to only 12 per cent of the general sample, while the proportion over 55 years is considerably less than that of the general sample (25.7% versus 41%). Again, this relatively youthful profile has been observed in other contexts (Bailey, Reference Bailey1998, Reference Bailey2000), with the pattern attributed to a greater reticence on the part of older homosexuals to publicly identify. Given this age difference, it is perhaps somewhat surprising the GLBT subsample has only a marginally lower income profile than the larger community.Footnote ⁷ The educational profile may help to explain this. As Table 1 indicates, the GLBT subsample as a group is better educated than the general sample; 74 per cent of this group (versus 63 per cent of other respondents) claims some post-secondary education.Footnote ⁸

Table 1 Profile of the GLBT Electorate

Note: Data is weighted to the 2006 Canadian Election Study subset of voters.

In terms of residency, the GLBT sample is drawn fairly proportionately from the various geographic regions of the country, although Quebec residents are somewhat overrepresented in the subsample and those from the Prairies somewhat underrepresented. However, there is a striking departure from proportionality in terms of urban–rural representation. As we might expect, the GLBT community (defined here as those who have publicly identified as members) is disproportionately drawn from urban settings—the larger the city, the greater the overrepresentation.

Turning to voting patterns within the sample, Table 2 reveals a voting profile of the GLBT community that is also distinct.Footnote ⁹ GLBT/non-GLBT vote differentials tend to be in the double digits outside of Quebec, although they are less so in Quebec where the BQ's overall strength has constrained such differences. All of these are statistically significant in the rest of Canada (p < .001), while only differences in vote proportions for the Conservatives and the BQ are significant in Quebec (p < .01).Footnote ¹⁰

Table 2 Vote Distribution

Note: Data are weighted to the 2006 Canadian Election Study subset of voters.

Closer examination of the table suggests the distinctiveness is mainly a function of this community's sound rejection of the Conservative party. For example, while 44 per cent of non-GLBT respondents cast a Conservative ballot outside Quebec, the corresponding percentage among GLBT respondents was 10. Conservative support among Quebec GLBT respondents was also in the single digits (8.2%). These are support levels normally expected for minor parties in the Canadian party system. As well, it should be noted that the rejection of the Conservative party within the Canadian GLBT community is far more decisive than rejection of the Republican party by GLB voters in recent US elections (Bailey, Reference Bailey1998, Reference Bailey2000). Further analysis (not shown) suggests that the relatively few Conservative supporters among the GLBT are found mainly in smaller communities. For instance, 6 per cent of GLBT voters in large metropolitan areas voted Conservative in 2006, while twice that level of support is found in smaller towns.

Contrary to expectations, the GLBT community is actually not very distinctive in the way it distributes its non-Conservative votes across the other parties. If the Conservative vote is removed from Table 2 (data not shown), the NDP proportion of the non-Conservative vote outside Quebec is not significantly greater for the GLBT community than it is for the larger sample, but GLBT voters in the rest of Canada are still more likely to support the Liberal than the NDP option. In Quebec, among GLBT non-Conservative voters, the NDP's support is equal to that for the Liberals, but their BQ support is not at all disproportionate to the non-GLBT vote there.

We might therefore conclude there is indeed what Hertzog (Reference Hertzog1996) refers to as a “sexuality gap” in Canadian voting, but it is a gap defined largely by the extraordinary weakness of the Conservative party within the GLBT community. However, our treatment of the GLBT community as one constituency hides dynamics within this electoral segment. While it is true males and females in this community tend to be on the same page in rejecting the Conservative option, there are some intriguing gender differences in their preferences among the remaining partisan alternatives (see Table 3). To be sure, gender differences within this constituency are not much in evidence in Quebec, with the Bloc attracting the lion's share of support from both gender groups. In the rest of Canada, however, the genders are markedly different in their choices. Males lean heavily toward the Liberal party and to a lesser extent toward the NDP; females, on the other hand, have the NDP as their modal choice and are much less supportive of the Liberals.

Table 3 Vote Distribution within GLBT Electorate

Note: Data are weighted to the 2006 Canadian Election Study subset of voters.

Explaining the Distinctiveness of the Community

What accounts for the distinctiveness of the GLBT community's voting profile? In addressing this question, we should first determine that it is not simply a function of the group's socio-demographic composition. As already noted, the GLBT community tends to reside in urban areas and tends as a group to earn slightly lower incomes. Neither of these factors favours the Conservatives. In addition, the generally higher proportion of males in the community might explain why Liberal support is as strong as it is.

To assess this possibility, we employed multinomial logistic regression, testing separate models for Quebec and the ROC. We used unweighted data for the regressions, although similar results are obtained when weights are invoked. First, to provide initial benchmarks, we ran voting models without controls, setting the Conservative party as the base category. In ROC, the GLBT variable's coefficients for the Liberal and NDP were 1.79 and 1.95, respectively, confirming the GLBT community's decided preference for non-Conservative options. A similar but more modest pattern is in evidence in Quebec, with coefficients of 1.26 for the Liberals, 1.79 for the NDP, and 1.33 for the Bloc Québécois. If socio-demographic differences account for these effects, we should expect the magnitude of the GLBT coefficients to drop substantially once appropriate controls are introduced.

To operationalize the socio-demographic variables, community size has been converted to a dummy variable, with 1 reflecting a large urban area (a population of at least 500,000). Religion is represented by three dummies: Catholic, Protestant, other religions, with “no religion” set as the reference category. Income has been collapsed into three categories: low income (under $35,000); medium income ($35,000 to under $80,000); high income (at least $80,000); in the analysis, high income was set as the reference.Footnote ¹¹ Education is represented by two dummy variables: less than high school and post-secondary; high school diploma only was set as the reference. In the ROC analyses, region has been collapsed into five dummy variables: Atlantic Canada, Ontario, Saskatchewan–Manitoba, Alberta, and British Columbia.

Results in Tables 4a and 4b suggest GLBT membership remains important. While most control variables behave in a predictable fashion,Footnote ¹² the coefficients reflecting membership in the GLBT community change very little from the benchmark model. That is, GLBT voters in Quebec and the ROC prefer all other major parties over the Conservatives. And as a predictor of vote, GLBT membership approaches the strength of region.

Table 4a Sexual Orientation and Vote Choice, Outside Quebec

Note: Multinomial logistic estimates set “Conservative party” as base category; robust standard errors;

^a p < .001;

^b p < .01;

^c p < .05.

Table 4b Sexual Orientation and Vote Choice, Quebec Only

Note: Multinomial logistic estimates set “Conservative party” as base category; robust standard errors;

^a p < .001;

^b p < .01;

^c p < .05.

If the distinctiveness of the GLBT vote is not simply a function of the constituency's socio-demographic composition, are there political variables that might help to explain it? As discussed above, the US experience suggests two possible scenarios. First is a “group interest” or “single issue constituency” account, in which votes are shaped by party differences on an issue of direct and immediate relevance to this constituency (Egan, Reference Egan2009). The second is a case of ideological congruence. American research consistently shows GLBT voters positioning themselves more to the left than the rest of the electorate and voting accordingly (Egan, Reference Egan2008; Hertzog, Reference Hertzog1996). Does the same pattern hold in Canada, and, if so, to what extent does this help to explain the distinctive voting profile of the GLBT community?

At least for the 2006 election, the conditions for a group interest effect were present. For several years leading up to that election, the “same-sex marriage” issue occupied a prominent position on the political agenda. The public's attention was sustained by a series of salient court cases and by polling evidence that the country was divided almost evenly on the issue. While the Liberal government's introduction and passage of the Civil Marriage Act in 2005 brought the issue to a head, same-sex marriage remained contentious in the subsequent 2006 federal election campaign. Conservative party leader Stephen Harper campaigned on the promise to reopen the issue if he won the election; and the Liberal party attacked the Conservative's position as evidence of intolerance towards GLBT voters in general, and opposition to same-sex marriage legislation in particular.

The 2006 election, then, clearly presented an issue of direct and unique relevance to this constituency, and there is prima facie evidence in the Ipsos Reid survey to suggest the GLBT community was duly mobilized. Not surprisingly, the vast majority (90%) of GLBT voters in the sample favoured such legalization.Footnote ¹³ Furthermore, in a separate question that asked respondents to select the one issue that “mattered most in deciding which party's candidate you voted for today,” 35 per cent of GLBT respondents selected “moral issues like abortion and same-sex marriage.”

While “group interest” on the same-sex marriage issue may provide a persuasive account of GLBT distinctiveness, there is also the possibility this community simply has a more general ideological inclination to support parties of the left. To investigate this, we first ask whether GLBT voters in Canada tend to position themselves on the left.

The Ipsos Reid survey does not include a strong measure of ideological perspective, but it does ask a number of questions with ideological import, and these should provide a tentative basis for positioning the GLBT community relative to others. The most direct question in this regard asked respondents whether they see themselves as “a liberal, a moderate, or a conservative.”Footnote ¹⁴ There is an obvious confound of party and ideological labels with this measure, so we prefer to place more emphasis on whether the policy views of GLBT respondents resemble those that typically bundle on the left. Again, the Ipsos Reid surveys devoted only limited attention to tapping respondents' policy preferences, but at least two items are potentially useful for this purpose. First, respondents were asked their views about the role of government in societyFootnote ¹⁵—an issue on which the left and right typically differ in Canadian politics. As well, respondents were asked their views on abortion,Footnote ¹⁶ a social issue which figures prominently for distinguishing moral traditionalists on the right in Canada (Lusztig and Wilson, Reference Lusztig and Wilson2005).

Table 5 displays the response distributions of GLBT and non-GLBT respondents for these questions, and suggests GLBT voters are indeed more likely to align themselves on the left side of the spectrum. Relative to non-GLBT voters in both elections, they are almost twice as likely to self-identify as liberal and much less likely to self-identify as conservative. As well, they are considerably more likely than non-GLBT voters to endorse a more active role for government in solving problems, and to endorse abortion “in all cases.”

Table 5 Ideology by Sexual Orientation and Gender

Given gender differences in voting within the GLBT community, it is worth examining whether these differences extend to political perspective as well. Table 5 shows that while there does seem to be an ideological gender gap within the non-GLBT subsample—females are significantly (p<.001) more likely than males to offer the “left” response to all indicators except for ideology—this does not consistently extend to the GLBT community. Among the GLBT segment, females and males are equally likely to give the “left” response, except on the issue of abortion, where females are significantly more liberal than their male counterparts (p < .01, t = 2.75, df = 1215, two-tailed).

While such limited measures of political perspective do not permit a strong conclusion about the leftist ideological distinctiveness of the GLBT community, the evidence here is certainly consistent with such a conclusion and consistent with US findings on the question.

Returning to the central question, then, to what extent is the distinctive voting pattern of the GLBT community explicable in terms of the community's response to the same-sex marriage issue and does the community's ideological perspective also make an independent contribution? To explore this question, we introduced these factors as control variables into models originally reported in Tables 4a and 4b. For this purpose, we recoded responses to the “same-sex marriage” question, the “abortion” question and the “role of government” question to form three scales ranging from “0” (strong opposition) to “1” (strong support). If any of these factors do play this explanatory role, we should see the impact of the GLBT variable shrink substantially—perhaps to insignificance—when they are included in the same model.

Results of that analysis (not shown here) suggest that while both dimensions of ideology and the same-sex marriage issue are relevant vote predictors, they do not fully explain the distinctiveness of the GLBT community.Footnote ¹⁷ The story, however, need not end there. It is plausible, and in fact quite congruent with the literature, that support for same-sex marriage is a more important vote driver within the GLBT community than it is for the rest of the electorate. As well, because the gay rights agenda has been championed by parties of the left, ideological propensities may also be more relevant vote predictors for this community. To test these possibilities, we include interactive terms for GLBT membership and these three controls: same-sex marriage, role of government and abortion.

Tables 6a and 6b show results of the full model with the three interactive terms. In Table 6a, it can be seen that inclusion of the interactive terms reduces the impact of the GLBT variable to insignificance, but that only one of the three interactive terms—that for same-sex marriage—is itself a significant predictor of the vote. This suggests that, outside Quebec, the GLBT vote remains distinctive, but that distinctiveness arises mostly from this community's support for same-sex marriage. While the GLBT voters tend to be on the ideological left, their partisan choices are, for the most part, not driven by this ideological proclivity any more than non-GLBT voters. That said, there is a weak (p < .05) significant “GLBT × Role of government” interaction in the case of the NDP.

Table 6a Effect of Campaign Issue and Ideology on Vote Choice, Outside Quebec

Notes: Multinomial logistic estimates set “Conservative party” as base category; robust standard errors;

^a p < .001;

^b p < .01;

^c p < .05.

Table 6b Effect of Campaign Issue and Ideology on Vote Choice, Quebec Only

Notes: Multinomial logistic estimates set “Conservative party” as base category; robust standard errors;

^a p < .001;

^b p < .01;

^c p < .05.

In Quebec, a different pattern emerges. Table 6b suggests that knowing a voter is a GLBT community member adds no predictive power once controls are introduced for the voter's position on same-sex marriage, the role of government and abortion. Unlike the ROC, all of the interactive terms here are insignificant, suggesting that neither same-sex marriage nor ideology was more important to the GLBT voter than it was for non-GLBT voters with the same views.Footnote ¹⁸

Given partisan policy debates of the period, the GLBT rejection of the Conservative party poses little in the way of a puzzle. However, the community's level of support for the NDP relative to the Liberal option is more intriguing because the NDP has traditionally been supportive of the rights agenda related to sexual orientation and the Liberal party has been less so. One possibility is that our assumption about this “natural” affinity between the GLBT community and the New Democratic party of Canada is simply wrong.

To explore this possibility in the absence of a measure of partisanship, we have derived an indicator of “party preference” that is conceptually independent of voting choice. The measure is based on responses to a battery of questions about which party “would do the best job of dealing with” each of 18 policy areas.Footnote ¹⁹ We have defined a respondent's “party preference” operationally as the party he or she adjudged “best” across the greatest number of these policy areas. Our measurement strategy here draws on a valence politics model of voting which holds that political choice is largely about which party or party leader is most competent to realize commonly desired ends (Clarke et al., Reference Clarke, Kornberg, MacLeod and Scotto2005, Reference Clarke, Sanders, Stewart and Whiteley2006; Stokes, Reference Stokes1963, Reference Stokes and Kavanagh1992; Whiteley et al., Reference Whiteley, Stewart, Sanders and Clarker2005). As constructed, the measure likely oversimplifies the voter's calculus with its assumption that all policy areas are equally salient to respondents. Bélanger and Meguid (Reference Bélanger and Meguid2008) have demonstrated that issue salience is variable across voters and conditions the influence of any particular issue-competence judgment on the vote. That said, our measure does exhibit an impressive degree of concurrent validity: for 93 per cent of respondents, our measure of “best” party correctly predicts the party judged “best overall” when explicitly asked to make that assessment.

Using our measure of “party preference,” Table 7 reports the distribution of preferences separately for Quebec and the rest of Canada, broken down by GLBT community membership. The table suggests the GLBT community outside Quebec is about equally likely to prefer the NDP or the Liberals on their handling of the issues. Had GLBT voters outside of Quebec supported their “preferred” party, we would have seen approximately equal support for the two parties from this community. However, this is largely a function of the preferences of females which tilt more than two to one toward the NDP, 61 per cent NDP versus 26 per cent Liberal (not shown in Table 7). Among males, the Liberal party is the modal choice. Outside Quebec, then, our assumption about a natural affinity between the GLBT community and the NDP is partly wrong in that it applies only to females.

Table 7 Perception of “Best Party”

Note: Data are weighted to the 2006 Canadian Election Study subset of voters.

In Quebec, where the NDP and the BQ might both be regarded as “natural” homes for the GLBT community, the BQ is preferred over the NDP. Given that the BQ is also the modal choice of all Quebecers, it cannot be said the GLBT community in that province is especially distinctive when choosing among the non-Conservative options.

This profile of GLBT party preferences helps to explain the group's electoral tilt to the BQ in Quebec and to the Liberal party in the rest of Canada, but there are clearly other forces at work because GLBT vote support levels for these parties are substantially higher in their respective regions than “preference” levels would lead us to expect. While there are several plausible reasons voters might transfer their vote to a less preferred party, strategic voting would seem to be a prime candidate. Blais and Nadeau (Reference Blais and Nadeau1996: 40) define a “strategic vote” as “a vote for a second-preferred party (candidate) rather than for the first-preferred one, motivated by the perception that the former has a better chance of winning the election.” Blais and his colleagues elsewhere suggest that “for an election to elicit substantial strategic voting there must be some generalized feeling that one specific party that is perceived to have a good chance of winning is completely unacceptable” (Blais et al., Reference Blais, Nadeau, Gidengil and Nevitte2001: 350).

For the GLBT community in 2006, the conditions for strategic voting would seem to be nicely met. That is, a case can be made that 1) a Conservative victory in 2006 was a distinct possibility; 2) a Conservative victory was decidedly unacceptable to most of the GLBT community; and 3) Liberals (outside Quebec) and BQ candidates (in Quebec) represented a better bet to defeat Conservative candidates in many if not most electoral constituencies in their respective regions. To what extent, then, is there evidence of strategic voting in this election within the GLBT community?

Scholars have adopted a number of different methodologies to identify the extent of strategic voting in an electorate. Some of these rely on analyses of aggregate voting patterns (Blais and Carty, Reference Blais and Carty1991; Johnston and Pattie, Reference Johnston and Pattie1991; Spafford, Reference Spafford1972), others on self-reported motivations (Evans and Heath, Reference Evans and Heath1993; Heath et al., Reference Heath, Jewell, Curtice, Evans, Field and Witherspoon1991; Niemi et al., Reference Niemi, Whitten and Franklin1992), and still others on a combination of the voter's party preference rankings and voter expectations about the likelihood of each party winning in a plurality election (Abramson et al., Reference Abramson, Aldrich, Paolino and Rohde1992; Alvarez et al., Reference Alvarez, Boehmke and Nagler2006; Blais and Nadeau, Reference Blais and Nadeau1996; Brady and Johnston, Reference Brady, Johnston, Orren and W1987; Cain, Reference Cain1978; Kselman and Niou, Reference Kselman and Niou2010). In this last case, which is arguably the consensus choice today, potential strategic voters are those whose second-preferred party is perceived to have a greater probability of winning than the voter's first-preferred party (Kselman and Niou, Reference Kselman and Niou2010), and a strategic vote is inferred if such a voter actually supports the second-preferred party.

Lacking information about respondents' electoral expectations or about the nature of the contest in their constituency, we developed a hybrid measure of strategic voting based on responses to two sets of questions. The first is a single direct question about the voter's motivation: “Now, would you say that you voted for this party today because you thought they would offer the best government for Canada, or because you were trying to stop another party from winning and forming the government?” This question converges on our notion of strategic voting, but has limitations for making such an inference. First, it does not distinguish between “strategic” and “protest” voting. As defined by Kselman and Niou (Reference Kselman and Niou2010), a protest vote is support for a second-preferred party that is perceived to be less competitive in the constituency than the first-preferred party. An example of this might be the possible shift of some traditional Liberal supporters to the Green party in 2006 to protest the Liberal party's role in the sponsorship scandal. Strategic voting is limited to situations where the voter is seen to shift support to a more competitive second preference. Second, the response options for this question likely pose a problem for many BQ supporters in Quebec who may well view the BQ as the best party to represent their interests, but not the one that is best, or even able, to form the government. Because neither response option clearly reflects the motivations of such Quebec voters, interpretation of their response is problematical.

To supplement this measure, we introduced a second condition for strategic voting by comparing a respondent's voting choice with her party preferences (see above), and inferring a potential strategic vote if that choice was her second preference.Footnote ²⁰ Our measure of strategic voting combined responses from these two measurement approaches. That is, we inferred a strategic vote if the respondent claimed her vote was motivated to stop another party from winning, and if she actually supported the party she liked “second best” on the issues. About 7 per cent of the entire sample satisfied both conditions.

Our measure departs from the ideal in that we are still unable to determine that the second-preferred candidate was actually more competitive (or perceived to be such) than the first-preferred candidate in the respondent's constituency; all we have is affirmation of a tactical motivation for such a vote. Nevertheless, the measure yields a pattern of voting consistent with expectations. Depending on electoral circumstances, estimates of strategic voting in recent Canadian elections have varied from a low of 3 per cent for the 1997 election (Blais et al., Reference Blais, Nadeau, Gidengil and Nevitte2001) to a high of 11 per cent for the 2004 election (Clarke et al., Reference Clarke, Kornberg, MacLeod and Scotto2005); hence our estimate of 7 per cent for this election is not inconsistent with the Canadian track record. As well, our strategic voters shifted their votes in expected ways. The NDP was the most affected, losing 14 per cent of those who believed the party was “best” overall on the issues; overwhelmingly, these strategic defectors moved to the Liberal or BQ candidate.Footnote ²¹ The Liberals lost about 8 per cent to strategic voting, again mostly to the NDP and the BQ. On the other hand, both the Conservatives and the BQ saw less than 3 per cent of strategic defections among those who rated them as “best” by our measure.

Does strategic voting help to explain the distribution of GLBT votes among the non-Conservative party options? Overall, 10 per cent of GLBT voters appeared to cast a strategic vote, compared to less than 7 percent among non-GLBT respondents; this suggests strategic considerations were relatively important to the GLBT community. Outside Quebec, the main beneficiary was the Liberal party, where two-thirds of GLBT respondents who were flagged as strategic voters cast a ballot for the Liberals. In Quebec, on the other hand, the BQ was clearly regarded as the modal choice among GLBT strategic voters, with almost 75 per cent of them defecting to that party.