<<

Electoral Studies 23 (2004) 463–483 www.elsevier.com/locate/electstud

From a two-party-plus to a one-party-plus? Ideology, vote choice, and prospects for a competitive in Thomas J. Scotto a,∗, Laura B. Stephenson b, Allan Kornberg a a Duke University, Durham, NC 27706, USA b University of Western , London, Ontario N6A 5C2, Canada

Abstract

Conventional wisdom, buttressed by numerous empirical studies, has questioned the impor- tance of both class and ideological factors on voting behavior in Canada. However, two recent studies of the 1997 Canadian national election [Gidengil et al., Canadian J. Political Sci. 32 (1999) 247; Nevitte et al., Unsteady State: The 1997 Canadian Federal Election, Oxford Uni- versity Press, Ontario, 2000] indicate that ideological factors played an important role in determining the outcome of the election, since they affected the direction of the vote for the several parties in both and the rest of Canada. In this paper, we clarify and extend these analyses with data from the 2000 election, as well as the 1997 election. We find that a number of ideological dimensions underlie the issue positions of voters in Quebec and the rest of Canada in both the elections and that these have important implications for the mainte- nance of the multiparty system that seemed to emerge in the last decade. More specifically, we find that ideological factors affect partisan volatility in Canada, and that the inclusion of these factors improves the explanatory power of a standard vote choice model in both election years. We infer from our findings that ideological differences among Canadian voters can help sustain a multiparty system in the foreseeable future but that it is a system that currently favors the —so much so, in fact, that the ability of any other party to successfully displace the as the government in the future is problematic. However, there is a curious periodicity to the electoral fortunes of Canadian parties at both the federal and provin- cial levels. Political parties can govern for years and then experience an electoral disaster of a magnitude sufficient to make them a distinctly minority party.  2003 Elsevier Ltd. All rights reserved.

∗ Corresponding author. E-mail addresses: [email protected] (T.J. Scotto); [email protected] (L.B. Stephenson); kornberg@- duke.edu (A. Kornberg).

0261-3794/$ - see front matter  2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0261-3794(03)00054-4 464 T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483

1. Introduction

For political parties and their leaders, democratic elections are usually risky busi- ness. Perhaps never was this demonstrated more forcefully in recent years than in the 1993 Canadian national election in which the governing Progressive Conservative Party (PC) was nearly annihilated, losing 167 of the 169 parliamentary seats it had won in 1988. The party’s share of the vote dropped from 43% to 16%. The left-of- center New (NDP) suffered almost as great a calamity with its seat total dropping from 43 to 7 and its vote total falling from 20% to 7%. In sharp contrast, two new parties (less than a decade old) emerged as big winners: the Que- bec-only Bloc Quebecois (BQ) elected 54 MPs and received 14% of the vote while the western-based Reform Party garnered 19% of the vote and elected 52 MPs. In their paper analyzing the meaning of the 1993 election, Clarke and Kornberg (1996) question whether the striking changes in party support are likely to endure or whether they are yet another illustration of the “plus c¸a change” adage. They note that Epstein (1964) characterized Canada as a “two-party-plus” system. At the time, only the Liberal and Progressive Conservative parties had ever formed a national government, yet historically there had been a number of occasions when a “third” party seemed poised either to transform the national party system into a genuine multiparty system or to displace one of the two old-line Liberal or parties. However, in the quarter century since Epstein wrote, neither event transpired; the system he described remained essentially unchanged. Nonetheless, the results of the 1997 and 2000 Canadian elections seem to indicate that the party system has under- gone a transformation into a genuine multiparty system. Despite the fact that the Liberals won both the elections, their current dominance rests largely on their ability to win seats in Ontario—98 out of a total of 177 seats the party won in 1993 were in Ontario, as were 101 out of 155 in 1997, and 100 out of 172 in 2000. The Con- servatives and New Democrats made modest comebacks in 1997 (the Conservatives winning 20 seats in Parliament while the NDP captured 21 seats) but lost steam in 2000 (the Conservatives won only 12 seats and the NDP 13). The Reform Party (which in 2000 “morphed” into the Canadian Alliance) improved its showing across the two elections, suggesting that it was not merely a “flash party”, a label coined by Converse and Dupeux (1966) to describe short-lived contenders in French elec- tions. Rather, Reform/Canadian Alliance appeared to have “legs”. It won 19% of the 1997 vote but no seats in Ontario, Canada’s largest province, increased its vote share to 24%, and won two seats in the province in the 2000 election. The party also managed to supplant the BQ as the official by electing 60 MPs to the Bloc’s 44 in 1997 and 66 to the Bloc’s 38 in 2000. A further indication that the old two-party-plus system might have developed into a multiparty one was the publication of two studies of the 1997 election. Gidengil et al. (1999) argued that in some instances the ideological positions of supporters of the several parties and the salience of ideological considerations more generally, although they varied by region, were important in structuring vote choice. Nevitte et al.’s (2000) study of the 1997 election went further. Nevitte and his colleagues argued that outside of Quebec, Canadian voters appear to have a basic outlook on T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483 465 a variety of social and financial issues and that—in addition to partisan identification and candidate affect—electoral support for each of the parties in part was grounded in such issues.1 The findings are especially interesting and potentially of genuine importance because they are in contrast to what has long been the conventional wisdom—that Canadian are not structured in terms of class-based ideological considerations (e.g. Alford, 1963; Pammett, 1987). This is usually attributed to the importance of federalism and the salience and duration over time of ethno-linguistic (Francophone/Anglophone) differences.2 There is also a generation of empirical stud- ies of national elections which conclude that because of the relative weakness (for example, in comparison to Britain or the United States) of partisan identifications, national elections are heavily influenced by short-term factors, notably party and leader affect and issues—even, at times, (such as the 1988 election) by single issues. These studies include those by Alford (1963), Clarke et al. (1979, 1984, 1991, 1996, 2000), Johnston et al. (1992), Kornberg and Clarke (1992), and Pammett and Dor- nan (2001). As just noted, the finding by Nevitte et al. (2000) suggests that the emerging multiparty system could be sustained because voter support for the several parties is partially grounded in sets of attitudes and values that cohere sufficiently to be labeled an ideology. Given this possibility, in this paper, we will elaborate on and extend their findings by conducting analyses of both the 1997 and 2000 national elections. These analyses will show that their views have merit in the 2000 as well as the 1997 election. By way of illustration, distinct ideological factors (four in Quebec and the rest of Canada (ROC) in 1997, four in Quebec and three in the rest of Canada in 2000), underlie the positions of Canadian voters on a variety of issues. In both elections, these factors contribute significantly to explaining the vote outside of Quebec. In Quebec, although four dimensions underlie the issue positions of Que-

1 Interestingly, an earlier comparative study of ideology in Canada and the United States by Gibbins and Nevitte (1985) found that left–right labels reflected weaker attitudinal coherence in Canada than in the United States and concluded that “there has been neither the need nor the opportunity for Canadians to structure their political world along conventional ideological lines” (598). 2 The theory that historically the relative absence of class-based, ideologically grounded electoral poli- tics in Canada is no accident was best articulated by John Porter (1965) in his landmark study of social stratification and the distribution of power. Porter argued that over time, Canadian political and economic leaders have combined pious exhortations to unite with themes (Anglophone/Francophone differences) and practices (federalism, constitutional revision) that divide in order to maintain their own positions as elites. For example, he concluded that “the major themes in Canadian political thought emphasize those characteristics, mainly regional and provincial loyalties, which divide the Canadian population. Conse- quently, integration and national must be constantly reiterated to counter such divisive sentiments. The dialogue is between unity and discord rather than between progressive and conservative forces…Can- ada must be one of the few major industrial societies in which the right and left polarization has become deflected into disputes over regionalism and national unity” (Porter, 1965: 368–369). On the conservative and elitist tendencies inherent in the operation of the Canadian federal system, also see Garth Stevenson (1989). Somewhat differently, in a series of essays party scholar John Meisel (1975) indicted Canada’s federal political parties, principally the Liberals, for failing to generate or campaign on coherent packages of policies that transcend regional differences. 466 T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483 bec voters, those pertinent to the province’s relations to the rest of Canada (and which load on the first dimension) overshadow all others. As an additional demonstration of the importance of these factors, below we will show that they influence partisan volatility in both Quebec and the ROC and that the inclusion of these variables in a standard vote model improves our ability to predict vote choice better than a model in which they are not included. Given these findings, we contend that the federal multiparty system that emerged during the past decade can be sustained in the current decade. However, such a multiparty system greatly favors the Liberal Party, so much so, in fact, that the prospects of other parties overcoming the Liberal dominance in federal elections in this decade are, at best, slim.

2. A note on data and methods

To estimate whether ideological considerations influence Canadian vote choice, one must determine whether there is an underlying basis to issue preferences. Factor analysis of voter responses to policy questions in the 1997 and 2000 Canadian national election studies (CNES) is an appropriate technique to employ for this analy- sis. Nevitte et al. (2000) and Gidengil et al. (1999) used this analytic tool with satisfactory results. We were concerned, however, that the methods each group employed may have excluded some important information in the CNES datasets. More specifically, Nevitte et al. (2000) ran two separate factor analyses with only a select set of variables. But since a primary purpose of exploratory factor analysis is to search for underlying constructs or dimensions (Kline, 1993), and since very few theories exist to guide the exercise, in selecting variables to include in an analysis “the guiding principle should be to obtain wide coverage of the domain, so as not to miss any important common factors” (Tucker and MacCallum, 1997: 132). There- fore, the starting point of our analysis is a tabula rasa.3 We use two techniques to elaborate and, hopefully, to improve upon Nevitte et al.’s analysis. First, to prevent biased results, we use all of the Likert-scaled issue questions in the 1997 and 2000 Canadian election survey, pre- and post-tests, stan- dardizing the issue variables to run from 0 to 1.4 In each survey, we also collapse the responses to domestic spending questions into a “Social Spending Cut Index” to avoid erroneous results due to high correlations. We also deal differently with the problem of missing data than do Nevitte et al.

3 Since the research of Gidengil et al. and Nevitte et al. stand in contrast to most of the previous scholarship on Canadian voting behavior, we provide additional confirmation for their work, by including as many issues as possible and running analyses on both the 1997 and 2000 data. 4 Complete results of the factor analyses are available from the authors. Likert scales involve the use of a standardized set of responses that can be used to answer a variety of questions or statements. Jaccard and Wan (1996) suggest that assuming these questions create interval level variables does not significantly increase the probability of either type I or type II errors. Variables that load on our identified dimensions are listed in Appendix A. T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483 467

With such a large set of variables, it is not surprising that a large portion of our samples did not respond to at least one of the questions included in the analysis. The standard technique to correct for this—list-wise deletion—eliminates a substan- tial portion of our dataset, making any analysis weaker. Nevitte et al. (2000) addressed the missing data problem with the use of pair-wise deletion. However, we believe that using this technique in a factor analysis might produce erroneous results.5 To avoid the possible pitfalls of pair-wise deletion and still retain as many cases as possible, we employ mean substitution to deal with missing values. Additionally, to assess the validity of the factor analysis generated with mean substitution, we used a multiple imputation technique with a simplified algorithm developed by King et al. (2001). The factor analyses generated using the imputation technique produced similar results to those generated by mean substitution.6 Accordingly, we are confi- dent that our findings are reliable. As did Nevitte et al., we divided the respondents in both elections into those domiciled in Quebec and those domiciled in the ROC. We then conducted four factor analyses: 1997 ROC, 1997 Quebec, 2000 ROC, and 2000 Quebec. In the 1997 elec- tion, for the respondents in the ROC, four factors were extracted that had eigenvalues greater than 1; after orthogonal rotation, we were able to label each dimension. The first dimension, which we label the “liberal–conservative” factor, encompasses most of the specific questions dealing with spending priorities, social issues, Canadian , and the role of business and unions. But it also includes feelings about the consequences of Quebec separation. The other three dimensions encompass fewer issues. The second focuses on policy questions dealing with Quebec and aboriginals, which we label a “minority issues” dimension. The third deals with views on issues that pertain to people’s feelings of political competence in affecting public policy and government’s accountability to them in these matters (e.g. “government doesn’t care what people like me think”). For convenience, we simply label this factor as “alienation”. The fourth dimension encompasses issues pertinent to feminism and the government’s role in the economy. Again, for convenience, we label this dimen- sion the “cultural ” factor. As indicated above, for the respondents from Quebec, the 1997 factor analysis also reveals four dimensions, albeit with slightly different content. The first dimen- sion focuses mostly on issues important to Quebec (e.g. positions on sovereignty, government’s treatment of Quebec, relations between Quebec and the ROC if Quebec separates). It also includes satisfaction with democracy in Canada, unsurprising given that Quebec separatism is rooted in the desire to exist as a nation in a separate state. The second dimension contains issues that loaded on the fourth factor for the ROC, as well as a question about the prospects of Canada and the US uniting into a single state. The third dimension for the Quebec subset is the same as for the ROC (political alienation). The fourth dimension is about support for minorities, encompassing ques-

5 With pair-wise deletion, matrices may not be “positive-definite” since each element of the covariance matrix used in the factor analysis is computed from differing subsets of cases (Arbuckle, 1996). A non- positive-definite matrix creates negative variances that produce erroneous results. 6 Analysis is available to interested readers from the authors upon request. 468 T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483 tions that are included in the ROC’s second dimension. What is evident from these results is that although there are real differences on a variety of issues between voters in Quebec and those in the rest of the country, opinions load on the same ideological dimensions, albeit in a different order. For the 2000 election, we again divided the respondents into two groups, those in Quebec and in the ROC. For the latter, three dimensions were delineated by the factor analysis. The first dimension reflects a general left–right ordering and includes opinions about Quebec and spending for minorities. This dimension seems to encompass elements from the first and second dimensions found in the 1997 ROC dataset. The second dimension corresponds to the third in 1997—responses to polit- ical alienation questions. The third dimension is similar to the second dimension from 1997, focusing on minority issues, coupled with views about abortion. The 2000 Quebec dataset is similar to that for 2000 ROC in that there is a political alienation dimension (#3), a general left–right dimension (#4), and one about min- ority issues (#2).7 As was the case in 1997, the most prominent ideological dimension encapsulates opinions on issues pertinent to Quebec’s national aspirations. Our findings regarding the multidimensionality of issue positions of voters are in accord with those of Campbell and Christian (1996:1) who observed that beliefs among Canadians are couched in ideologies dealing with “liberty, individualism, social coherence, hierarchy, and equality”. Also, in an earlier study Kornberg et al. (1975) found that Canadian MPs and an elite sample of the public as well as a cross- sectional one arrayed their own issue positions and those they ascribed to the four national political parties along several dimensions. Although in the current study the items that cohere along each dimension are interrelated, they do not always fit “logi- cally”. In this regard, they differ from the conventional view of an ideology as a set of interrelated beliefs, one or more of which occupies a position of centrality so that if it (they) change(s), logically so should the other issues (Converse, 1964). The closest to such a conception of ideology is the first factor extracted from the views of Quebec voters in both 1997 and 2000.8 That said, in the remainder of the paper, we periodically use the terms “ideology” and “dimensions” interchangeably.

3. Ideology matters: some empirical evidence

Table 1 reports the median placement of voters on each dimension by their partisan identification in ROC and Quebec for the 1997 and 2000 elections. For 10 of the 15 dimensions, the party averages complement one another with the more liberal and more conservative parties being on opposite ends. The political alienation dimen- sion is somewhat surprising. In the ROC, as would be expected, Liberal supporters

7 The minority issues dimension also includes some questions that load on the left–right dimension for the ROC, such as spending on foreign aid and whether the government should do more for minorities. 8 Hinich et al. (1998) come to the same view of Quebec voters’ positions on policy issues in their analysis of the 1993 Canadian national election. T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483 469 Conservative – 0.142 0.256 0.133 0.099 0.146 0.126 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ – – – – – 0.169 0.031 0.340 0.129 0.220 0.073 0.099 0.171 0.216 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ 0.1300.5500.391 0.046 0.347 0.510 0.357 0.139 0.093 0.068 0.165 0.180 0.024 0.175 0.271 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Conservative Minority issues Alienation Cultural conservatism Conservative Alienation Minority issues – – 0.249 0.105 0.070 0.492 0.752 0.329 0.010 0.272 0.045 0.862 0.236 0.081 0.323 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Liberal Liberal Dimension 1 Dimension 2 Dimension 3 Dimension 4 Quebec issues Cultural conservatism Alienation Minority issues Quebec issues Minority issues Alienation Liberal cation 0.272 0.023 0.237 0.094 cation 0.082 cation cation fi fi fi fi Liberal 0.361 0.173 PCPC NDP 0.362 0.691 NDPReformNo identi Liberal 0.441 0.294 0.211 0.494 0.273 0.176 0.098 0.563 Alliance 1997 ROC 2000 ROC No identi 1997 Quebec Liberal PC BQNo identi 0.819 0.236 2000 Quebec Liberal BQNo identi 0.951 Table 1 Median factor scores in Quebec and the ROC (rest of Canada) for 1997 and 2000 Canadian national elections 470 T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483 show the lowest median levels of alienation.9 However, the Reform/Alliance mem- bers tend to be less efficacious than their PC counterparts; we expected the opposite given the rapid rise of the former to its “Official Opposition” status and the monu- mental decline of the latter at the federal level. These data show that despite their party’s successes, Reform/Alliance partisans remain distrustful of government and seemingly doubt their abilities to bring about change in Ottawa. On the liberal– conservative dimension in 1997, the median factor scores for the dimension were bi-modally distributed, with the majority of cases clustering either between factor scores of 0 and 1 or between factor scores of Ϫ1.5 and Ϫ2.0. To support our view that the ideological positions of voters in Quebec and the ROC can provide a modi- cum of support for the maintenance of the multiparty system that emerged in the 1990s, first we will demonstrate that they affect partisan volatility and hence the direction of the vote. Although Canadians change their partisan identifications more frequently than do Americans or Britons, the intensity and direction of partisanship nonetheless is a powerful predictor of voting in Canada. We first ran a series of logistic regressions of partisan identification onto several control variables together with the four ideological dimensions.10 With regard to partisan volatility, a number of studies (Clarke et al., 2000:193) have indicated that partisan identification in Canada is less stable than it is in the United States, for example. There are a variety of reasons that have been given for why Canadians change their partisan identity, such as leader affect. Table 2 reports the extent to which ideological factors add to the propensity of Canadians to change their partisan identification. More specifically, we may note the change in the probability of ident- ifying with a particular party when a respondent’s factor score moves from a con- servative (low factor score) to a liberal (high factor score) position.11 Regarding the 1997 election, we find that in Quebec, Quebec issues (Dimension 1) had a negative effect on the probability of changing to a Liberal identification and an enormously positive effect on the probability of switching to a BQ identification. Dimension 2 (cultural conservatism) also had a significant positive impact on prospects of switch-

9 Due to the negative factor loadings on the alienation dimension for 1997 in the ROC and 2000 in Quebec, the least alienated have negative factor scores while those most alienated have high median factor scores. 10 The full analyses of the logistic regressions are available from the authors upon request. For each dataset (1997 ROC, 1997 Quebec, 2000 ROC, 2000 Quebec), several models were run to reveal the impact of the ideological dimensions upon the likelihood of accepting different partisan identifications or refusing any identification. Each dependent variable was coded 1 for accepting the identification (or refusing any) and 0 otherwise. Control variables include the party leaders’ thermometer scores, region, short-term issues, socioeconomic forces, age, and gender. In all the models that analyze behavior in Quebec, primary language and religion controls were added. 11 For ease of interpretation, a “conservative” position for a dimension is considered to be a factor score at the thirty-third percentile while the “liberal” position is considered to lie at the sixty-sixth percentile. For the analyses where partisan identification is the dependent variable, the socioeconomic variables are set to their median values and the respondent is assumed to reside in Ontario and be without a short-term issue. The values for the party leaders are set to their medians and the factors not being analyzed are set at their mean scores (zero). For the Quebec analyses, the respondent is assumed to speak French and be a Catholic. All analyses were carried out using the “CLARIFY” software described in King et al. (2000). T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483 471 0.5 level. Ͻ p Conservative (%) – 5.5 7.2 5.4 13.8 33.3 19.6 35.3 27.6 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ – – – – – cation in Quebec and the ROC for 1997 and fi cant in the logistic regression at the fi 0.5 1.4 6.8 9.3 5.3 6.7 7.1 11.3 17.5 14.5 11.0 11.8 14.5 14.9 15.7 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ 1.2 7.3 5.3 1.1 7.5 1.1 61.8 26.4 15.8 23.3 14.3 1.5 15.9 23.3 43.5 13.5 15.3 8.5 13.6 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Conservative (%) Minority issues (%) Alienation (%) Cultural conservatism (%) Conservative (%) Alienation (%) Minority issues (%) – – 3.4 1.1 5.4 3.8 15.9 30.5 22.6 57.5 36.3 75.8 37.5 35.9 59.3 Ϫ Ϫ Ϫ 350.0 30.6 580.3 Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Dimension 1 Dimension 2 Dimension 3 Dimension 4 Liberal Liberal Quebec issues (%) Cultural conservatism (%) Alienation (%) Minority issues (%) Quebec issues (%) Minority issues (%) Alienation (%) Liberal cation cation 2.8 cation cation fi fi fi fi 1997 ROC PC Liberal 2.6 4.3 NDP 12.1 Reform No identi 2000 ROC Liberal 7.3 PC Alliance NDP No identi 1997 Quebec Liberal PC No identi BQ Liberal 2000 Quebec BQ No identi Table 2 Impact of ideological dimensions (% increase in likelihood of change) on propensity to change party identi Percentages in italics indicate that the factor score variable for the given dimension is statistically signi 2000 Canadian national elections 472 T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483 ing to the a BQ, whereas minority issues (Dimension 4) had a significant positive effect on moving to the Liberals and diminished the probability of becoming a non- identifier. In the 2000 election, Quebec issues had an even greater effect on the prospects of movement towards BQ identification and again had a constraining effect on movement to the Liberals or to non-identification. Minority issues (Dimension 2) had significant negative effects on non-identification. Both Dimension 3 (alienation) and Dimension 4 (liberal–conservative) had no significant effects on party identification. In short, issues pertinent to Quebec (and to a lesser extent minority issues) had a profound effect on party identification and, indirectly, on the outcome of the vote in that province in both elections.12 As for the rest of Canada with the exception of Reform, liberal–conservative issues (Dimension 1) had no significant impact on partisan identification in 1997 but had significant effects on the probabilities of identifying with all parties except the Lib- erals in the year 2000—negative in the case of the Alliance and the and positive on movement to the NDP and the Liberals. In 1997, Dimension 2 (minority issues) inhibited the movement toward identifying with Reform and the PCs but had a positive effect on the probability of identifying with the NDP. Dimension 3 (alienation) had a significant positive impact on the movement toward Reform, and non-identification in 1997 and a positive effect on movement toward the Liberals in the 2000 election.13 Finally, in 1997, the fourth dimension (cultural conservatism) not surprisingly had a highly significant positive effect on a switch to the NDP and a negative effect on changing to identification with Reform or the Tories. Our next step is to show that the inclusion of the ideological dimensions in a standard model of the vote matters by demonstrating that they improve the explana- tory power of Clarke et al.’s (1996) standard vote choice model. More specifically, we want to ascertain whether one or a combination of the ideological dimensions increases or decreases the probability of voting for a party other than the incumbent Liberals. Accordingly, we estimate multinomial logit models comparing the Liberals with the other federal parties—Progressive Conservatives, New Democrats Reform (1997), Alliance (2000), and, in Quebec only, the Bloc Quebecois. We first add a number of control variables to our models consistent with the model proposed by Clarke et al., thereby including in our analyses controls for party identification, affect for all party leaders, and other standard controls.14 Table 3 depicts the results of significance tests on the four dimensions. Even with region, the importance of short-term issues, partisan identifications, and affect for the party leaders controlled, the three or four ideological dimensions helped explain

12 Note that in 2000, we exclude the Progressive Conservatives from the analysis because there were so few Conservative respondents in Quebec. 13 Recall that due to negative factor loadings, substantially greater levels of “alienation” in 1997 increase the probability of identifying with Reform. The opposite is the case for 2000. 14 Due to space considerations, results are available upon request from the authors. Because of the ability of NDP identification to perfectly predict a vote against the Alliance in 2000, both NDP and non- identifiers were consigned to the base category. Alternative specifications of the model, which included dropping NDP identifiers altogether, were undertaken and yielded no substantive differences. T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483 473 Conservative – 0.05 0.001 Ͻ Ͻ p – – – – p 0.10 n.s. 0.10 n.s. 0.05 0.005 0.05 0.000 n.s. 0.01 0.05 n.s. Ͻ Ͻ Ͻ Ͻ Ͻ Ͻ Ͻ Ͻ p p p p p p p p 0.10 n.s. n.s. 0.01 0.01 0.05 n.s. 0.001 n.s. n.s. 0.001 0.10 Ͻ Ͻ Ͻ Ͻ Ͻ Ͻ Ͻ p p p p p p p Conservative Alienation Minority issues Conservative Minority issues Alienation Cultural conservatism – – 0.001 0.001 n.s. 0.001 n.s. n.s. 0.05 n.s. 0.05 0.10 Ͻ Ͻ Ͻ Ͻ Ͻ Ͻ Quebec issues Minority issues Alienation Liberal p p p Quebec issuesp Cultural conservatism Alienation Minority issues p Liberal p Liberal Dimension 1 Dimension 2 Dimension 3 Dimension 4 2000 Quebec Liberal vs. BQ PC vs. BQ Liberal vs. BQ PC vs. Alliance1997 Quebec Liberal vs. PC n.s. n.s. Liberal vs. Alliance Liberal vs. NDP n.s. 2000 ROC Liberal vs. PC n.s. PC vs. Reform n.s. Liberal vs. Reform 1997 ROC Liberal vs. PC n.s. n.s. Liberal vs. NDP n.s. Table 3 Impact of ideological dimensions on vote choice in Quebec and the ROC for 1997 and 2000 Canadian national elections 474 T.J. Scotto et al. / Electoral Studies 23 (2004) 463–483 the vote choices of Canadians. That is, for each of the four models, log-likelihood tests were performed and all four models that included the additional ideological dimensions both fit the data better and improved our ability to explain vote choice than did the model without them. More specifically, political alienation in 2000 significantly influenced a voter’s choice between the old-line Liberals and all other parties, whereas the other dimen- sions played a significant role in the decision to vote for a particular party in either the 1997 or the 2000 elections. In Quebec, not surprisingly, issues germane to the province weighed very heavily on a decision to vote for the BQ over the Liberals. Feelings towards minorities (Dimension 4 in 1997 and Dimension 2 in 2000) also played a significant part in the voting decision, although the impact of these issues was dwarfed by the importance of the Quebec-specific issues. In another illustration of the strength of ideological issues, we constructed scen- arios in which we varied the values of certain significant issue dimensions while holding other socioeconomic control variables in the model at their median values.15 These scenarios indicate that, ceteris paribus, the voter’s ideological positions on issues had a significant impact on the probability of their casting a vote for a parti- cular party. For example, in the ROC in 1997, as one moves from a conservative position on minority issues to a more liberal one, the probability of voting Reform decreases by 10%. In the 2000 election, becoming more liberal on the liberal–con- servative, alienation, and minority dimensions combined decreases the probability of Alliance identification by 17%. For Quebec in 1997, as one moves to a more favor- able position on issues pertinent to , the probability of a Liberal vote declines from 48% to 17%, the probability of voting PC declines from 39% to 37%, while in contrast, the probability of voting BQ increases from 12% to 46%. In the 2000 election in which the Conservatives were not really a factor, a more favorable position on Quebec nationalism issues increases the probability of a non- identifier voting for the BQ by 51%! To recapitulate, we have demonstrated that in two successive national elections, the positions of Canadians in Quebec and in the rest of Canada can be arrayed along several ideological dimensions. We have shown that the dimensions have a signifi- cant impact on both the propensity of Canadians to change their partisan identifi- cations and their vote preferences. Regarding the latter, a standard model of the vote that included the several ideological dimensions adds appreciably to its explanatory power and several scenarios we constructed provided additional evidence of the influence of ideology on vote choice in both election years. There is reason, therefore, to assume that the ideological positions of Canadians have facilitated the emergence of a genuine multiparty system in the 1990s and that they may help to sustain such

15 In our analyses, factor scores were varied from the thirty-third percentile to the sixty-sixth percentile while other factor scores were held at their means (zero). Substantively, this is akin to moving from a center-right to a center-left position on the dimension. For the “alienation” dimensions, as an individual moves along a dimension, she becomes more efficacious. We also assumed that a typical voter in the two elections was not identified with any of the parties and affect levels for all four party leaders were held at their respective medians. T.J. Scotto et al. / Electoral Studies 23 (2004) 463Ð483 475 a system in the future. However, now we will show that it is also a system that disproportionately benefits the governing Liberal Party. Moreover, to the extent that current conditions can be extrapolated, they also enhance the probability that the Canadian party system may have changed from the “two party plus” system, in which the NDP was the perennial “plus,” to one in which all of the federal parties occupy that position vis a` vis the Liberals.

4. From a two-party-plus to a one-party-plus-plus-plus-plus?

Table 4 shows the distribution of strong, moderate, and weak identifiers by party. In the 1997 data, NDP and Reform include a larger proportion of strong identifiers than do the old-line Liberals and Progressive Conservatives, and, other than for the New Democrats, this condition also obtained in the year 2000. In Quebec, the pro- portions of strong identifiers in both the BQ and Liberal parties were relatively simi- lar in both election years. Note, however, that the intensity of identification with the two parties increased dramatically in the 2000 election. Indeed, in 2000, the pro- portion of strong identifiers with the Bloc and Liberal parties almost doubled. Given the importance ascribed to the intensity of identification for both the propen- sity and direction of voting, we would anticipate that in the ROC larger proportions of Reform/Alliance and New Democrat identifiers voted for their parties in both 1997 and 2000 than did identifiers with the two old-line parties, and that in Quebec,

Table 4 Distribution of strong, medium, and weak Identifiers (% of identifiers in each category) by party in Quebec and the ROC for 1997 and 2000 Canadian national elections

Strong (%) Medium (%) Weak (%)

1997 ROC Liberal (515) 22.3 53.0 23.9 PC (312) 25.3 52.6 21.2 NDP (167) 33.5 49.1 13.9 Reform (205) 37.1 52.7 9.8 1997 Quebec Liberal (187) 13.9 55.1 29.4 BQ (48) 14.4 56.9 28.7 PC (188) 10.4 70.8 16.7 2000 ROC Liberal (395) 29.1 45.8 24.1 PC (145) 15.2 53.8 31.0 NDP (96) 22.9 61.5 15.6 Alliance (225) 38.7 52.4 8.4 2000 Quebec Liberals (201) 26.9 49.3 22.9 BQ (213) 33.3 52.6 13.6

Figures in parentheses are each party’s number of identifiers in the 1997 and 2000 elections. 476 T.J. Scotto et al. / Electoral Studies 23 (2004) 463Ð483 this was also true of Bloc identifiers. Inspection of Table 5 indicates instead that in the ROC there is something of a progression with Reform/Alliance partisans the most loyal, New Democratic and Liberal identifiers almost equally loyal and Tory identifiers the least inclined to support their party in both elections. In 2000, loyalty among Alliance, NDP and Liberal identifiers is almost equal while PC identifiers were apt to defect to either the Alliance or the Liberals. In Quebec, BQ identifiers are the most loyal in 1997 and 2000, but the differences between them and the Liberals are relatively modest. Particularly striking is that in the 2000 election, vir- tually 100% of each party’s cohort of identifiers voted for their respective parties, suggesting a “politics of commitment” in Quebec that does not obtain outside that province.

5. Summary and conclusions

We noted at the beginning of this paper that the multiparty system that emerged in the 1990s, and that seemingly displaced the long time “two party plus” system, works to the advantage of the governing Liberals. Indeed, it makes the possibility of a party other than the Liberals winning a national election in the relatively near future somewhat problematic. That is because the Liberal Party has by far the largest proportion of identifiers in the electorate, and despite the frequency with which Can- adians change their partisanship, the latter variable continues to be a powerful pre- dictor of Canadian voting behavior.16 Although both Reform and the New Democrats had a larger proportion of strong identifiers than did the Liberals in 1997, the actual number of NDP and Reform identifiers was much smaller than the number of Liberal identifiers. Specifically, 10% of the 1997 national sample of voters identified them- selves with the New Democrats, 12% accepted a Reform identification, but fully 41% identified with the Liberal Party while 19% identified with the other old-line party, the Tories. In the 2000 election, the number of NDP identifiers declined from 10% to 8%, and the number of Tory identifiers declined from 19% to 12%. It is true that the percentage of Alliance identifiers rose from 12% to 18%, but they were still dwarfed by the 32% of the sample who professed a Liberal identification. In Quebec, the proportions of BQ and Liberal identifiers were almost similar, with each party’s identifiers constituting slightly over one-third of the sample, and less than 10% of the sample identifying with the Conservatives. Given these numbers, it is hard to envisage a scenario in which any of the opposition parties could displace the Liberals as the Government. Of course, the continuing hope of Alliance Party leaders is that the Conservatives will join them in a united right-of-center alternative to the Liberals. To date, their hopes have not been realized. Nor, we contend, are they likely to. The Tory party, despite its low standing currently in federal elections,

16 In their panel study of individual changes in partisanship, Clarke et al. (2000:193) estimate that 40% of the electorate changed partisan identities between 1979 and 1984, that the number of “movers” increased to 48% between 1983 and 1988, and declined to 42% between 1988 and 1993. T.J. Scotto et al. / Electoral Studies 23 (2004) 463Ð483 477 7.0 95.6 45.7 Liberal PC NDP Reform/Alliance BQ cation in Quebec and the ROC for 1997 and 2000 Canadian national elections fi cation 3.6 5.3 0.4 80.0 cation 1.5 1.5 0.5 96.6 cation 71.8 9.1 9.1 9.9 cation 80.2 13.9 5.9 cation 80.0 7.6 4.8 7.6 cation 93.0 fi fi fi fi fi fi cation 13.8 4.2 74.3 7.8 cation 15.6 4.2 80.2 0 fi fi cation 2.1 6.9 91.0 cation 4.4 cation 15.9 52.5 4.1 30.6 cation 33.2 14.3 16.0 36.5 cation 16.7 75.0 8.3 cation 32.2 34.3 33.6 cation 11.0 55.2 5.5 28.3 cation 37.9 17.1 12.0 33.0 cation 54.3 fi fi fi fi fi fi fi fi fi 1997 ROC Liberal identi PC identi NDP identi Reform identi No identi 1997 Quebec Liberal identi PC identi BQ identi No identi 2000 ROC Liberal identi PC identi NDP identi Alliance identi No identi BQ identi 2000 Quebec Liberal identi No identi Table 5 Vote choice by party identi 478 T.J. Scotto et al. / Electoral Studies 23 (2004) 463Ð483 continues to be a power in provincial politics. The well-named “Big Blue Machine” has by far the strongest organization in Ontario, having governed that province almost continuously since 1943. It remains a force in the Atlantic Provinces and in and Manitoba. In contrast, the Alliance has virtually no provincial organization, and its current prospects of winning a provincial election seem remote, even in its Alberta and British Columbia heartland. However, given the gains the party made in the 2000 federal election and the several significant ideological differ- ences between the Alliance and Tory supporters delineated in our analyses, it is unlikely that by the next election, any great number of Alliance supporters will want to return to the Tory fold. Consequently, the prospect of any party displacing the Liberals in the next election, or even in the one following, is problematic. That is on one hand. On the other, despite a tendency towards one-partyism at both the federal and provincial level, there is a curious periodicity to party fortunes in Canada. Parties may enjoy years of electoral success and then rapidly fall from grace. For example, the Liberals captured 65% of all parliamentary seats in the 1953 federal election and only 19% in 1958—a decline of more than 40% in a 5-year period. The Conservatives, for their part, won 80% of the seats in 1958, but only 36% 5 years later. Additionally, the Liberals, other than for a brief hiatus, governed Canada continuously between 1963 and 1984. In the latter election, however, they suffered a defeat in which they were left holding only 13% of the parliamentary seats, as opposed to the 52% that they had held only 4 years earlier (Kornberg et al., 1992). Finally, as we observed at the beginning of this paper, after governing for 9 years, in the 1993 election, the Conservatives were virtually shut out of Parlia- ment, losing 167 of the 169 seats they had won 5 years earlier. In short, there is reason to believe that the phoenix-like ability of the Conservatives and other Canad- ian political parties to emerge from the ashes of electoral defeat to election victory may again assert itself. The continued good health of one of the world’s oldest demands it.

Acknowledgements

The authors would like to thank John Aldrich, Jeffrey Grynaviski, Jon Pammett, Richard Potthoff, and Elizabeth Zechmeister for helpful comments. However, the authors bear responsibility for any errors of fact or interpretation that may be con- tained in this paper.

Appendix A. Issue questions that load on above dimensions

Eigenvalues are given next to description of dimension, and factor loadings as well as CNES variable names are listed next to issue variable descriptions. Variables with loadings above 0.300 are listed. Full results of the factor analyses are available from the authors upon request. T.J. Scotto et al. / Electoral Studies 23 (2004) 463Ð483 479

1997 Canada excluding Quebec

A Dimension 1: Liberal–Conservative (5.27) i Should more be done for women—pese110 (0.792) ii Canada’s ties with the United States—pese4: (0.724) iii Satisfaction with democracy in Canada—pesa5b: (0.706) iv High employment should lead to government defeat—pese16: (0.679) v Standard of living in wake of Quebec separation—pese10a-c: (0.626) vi Economic unity should still exist if Quebec separates—pese27: (0.598) vii Spending on defense—pese6a: (0.581) viii Quebec and Canada economic ties if separation—pese11d: (0.566) ix Aid given to unions—pese3: (0.555) x Spending on foreign aid—pese6g: (0.553) xi Aid given to business—pese2: (0.541) xii Gun control—pese12: (0.485) xiii Spending index on social program—pese6b-f: (0.438) xiv Trust ordinary people or the government—pese21: (0.431)

B Dimension 2: Minority issues (2.72) i Spending on aboriginals—cpsj10: (0.569) ii Should the government do more for minorities—cpsf1: (0.498) iii More should be done for Quebec-cpse3a-c: (0.483) iv Aboriginals are treated better than other Canadians—cpsj9: (0.405) v Position on the death penalty—pese13: (0.342) vi Right of Quebec to be treated as distinct—cpsj3: (0.330)

C Dimension 3: Alienation (1.56) i Government does not care about what people think—cpsb10: (Ϫ0.673) ii Ordinary people have a say in government—cpsb10b: (Ϫ0.592) iii Politicians lie to get elected—pese15: (Ϫ0.421) iv Politics is too complicated—cpsb10c: (Ϫ0.410) v Parties keeping their promises—cpsj13: (Ϫ0.397) vi Elected officials out of touch—cpsb10a: (Ϫ0.379) vii Canadian immigration policy—cpsj18 (Ϫ0.300)

D Dimension 4: Cultural conservatism (1.12) i Only married people should be having children—cpsf2: (0.542) ii A woman’s place is in the home—cpsf3: (0.476) iii Government should leave it to private sector to create jobs—cpsf6: (0.432) iv Government cannot solve unemployment problem—cpsf4: (0.389) v To maintain social programs, eliminate deficit—cpsf5: (0.345) 480 T.J. Scotto et al. / Electoral Studies 23 (2004) 463Ð483 vi Abortion position-pese5a-c: (0.338)

1997 Quebec

A Dimension 1: Quebec issues (2.95) i Position on Quebec sovereignty—cpsj3a: (0.739) ii Standard of living in wake of Quebec separation—pese10a-c: (0.680) iii Federal government’s treatment of Quebec—cpsj12: (Ϫ0.556) iv More should be done for Quebec—cpse3a-c: (0.519) v Important to defend Quebec’s interests—cpsa2h: (0.510) vi Economic unity should still exist if Quebec separates—pese27: (0.454) vii Satisfaction with democracy in Canada—pesa5b: (Ϫ0.336)

B Dimension 2: Cultural conservatism (1.96) i A woman’s place is in the home—cpsf3: (0.601) ii Only married people should be having children—cpsf2: (0.521) iii Government should leave it to private sector to create jobs—cpsf6: (0.378) iv Abortion position—pese5a-c: (0.324) v Good thing for Canada and US to be one nation—pese25: (-0.318) vi To maintain social programs, eliminate deficit—cpsf5: (0.303)

C Dimension 3: Alienation (1.29) i Government does not care about what people think—cpsb10: (0.617) ii Ordinary people have a say in government—cpsb10b: (0.433) iii Politicians lie to get elected—pese15: (0.409) iv Politics is too complicated—cpsb10c: (0.389) v Elected officials out of touch—cpsb10a: (0.360) vi Parties keeping their promises—cpsj13: (0.338)

D Minority issues (1.03) i Spending on aboriginals—cpsj10: (0.600) ii Aboriginals are treated better than other Canadians—cpsj9: (0.492) iii Should the government do more for minorities—cpsf1: (0.412)

2000 Canada excluding Quebec

A Dimension 1: Liberal–Conservative (3.749) i Should the government do more for minorities—cpsc11: (0.626) ii Should more be done for women—cpsc10 (0.609) iii Spending index on social programs—pese1b-f: (0.546) T.J. Scotto et al. / Electoral Studies 23 (2004) 463Ð483 481 iv More done to reduce gap between rich and poor—cpsc13: (0.492) v Should surplus be spent on federal programs—cpspla24: (0.429) vi More should be done for Quebec—cpsc12: (0.366) vii Power of unions—pesd2: (0.351) viii Sympathies towards feminism—pesg20: (0.324) ix Spending on foreign aid—pesd1g: (0.303)

B Dimension 2: Alienation (2.207) i Parties foster solutions to important problems—pesk8b: (0.609) ii Parties do a good job presenting issues—pesk8a: (0.598) iii Parties represent the concerns of ordinary people—pesk8d: (0.557) iv Satisfaction with democracy in Canada—cpsa8: (0.476) v Government does not care about what people think—cpsb10d: (0.460) vi Parties keeping their promises—cpsj13: (0.458)

C Dimension 3: Minority issues and social matters (1.23) i Position on the death penalty—cpsc15: (0.460) ii Canadian immigration policy—cpsj18: (0.456) iii Gay and lesbian marriage—cpsf18: (0.454) iv Aboriginals are treated better than other Canadians—cpsj9: (0.417) v Ways to deal with violent offenders—cpsj51: (0.412) vi Abortion position—pesg8: (0.33)

2000 Quebec

A Dimension 1: Quebec issues (3.70) i Position on Quebec sovereignty—pesc6: (Ϫ0.865) ii Quebec’s right to separate—pesg18: (0.775) iii Standard of living in wake of Quebec separation—pesc9: (Ϫ0.657) iv Power of province vs. federal power—pese1a, pese1b: (Ϫ0.539) v Status of French language if Quebec separates—pesc8: (Ϫ0.531) vi Is French language threatened in Quebec—pesc7: (Ϫ0.449)

B Dimension 2: Minority issues (1.95) i Canadian immigration policy—cpsj18: (0.454) ii Position on the death penalty—cpsc15: (0.454) iii Aboriginals are treated better than other Canadians—cpsj9: (0.430) iv Government should leave it to private sector to create jobs—cpsf6: (0.394) v Ways to deal with violent offenders—cpsj51: (0.364) vi Should the government do more for minorities—cpsc11: (0.323) vii Spending on foreign aid—pesd1g: (0.314) 482 T.J. Scotto et al. / Electoral Studies 23 (2004) 463Ð483

C Dimension 3: Alienation (1.35) i Parties do a good job presenting issues—pesk8a (-0.620) ii Parties foster solutions to important problems—pesk8b: (-0.579) iii Parties represent the concerns of ordinary people—pesk8d: (-0.570) iv Satisfaction with democracy in Canada—cpsa8: (-0.357) v Parties keeping their promises—cpsj13: (-0.351) vi Government does not care what people think—cpsb10d: (Ϫ0.311)

D Dimension 4: Liberal–Conservative (1.13) i Spending index on social programs—pese1b-f: (0.518) ii Should more be done for women—cpsc10: (0.446) iii More done to reduce gap between rich and poor—cpsc13: (0.407) iv Should surplus be spent on federal programs—cpspla24: (0.310)

References

Alford, R.R., 1963. Party and Society: The Anglo–American Democracies. Rand McNally, Chicago. Arbuckle, J.L., 1996. Full information estimation in the presence of incomplete data. In: Marcoulides, G.A., Schumacker, R.E. (Eds.), Advanced Structural Equation Modeling: Issues and Techniques. Lawr- ence Erlbaum Associates, Mahwah, NJ. Campbell, C., Christian, W., 1996. Parties, Leaders, and Ideologies in Canada. McGraw-Hill Ryerson, Toronto. Clarke, H.D., Kornberg, A., 1996. Partisan dealignment, electoral choice and party-system change in Canada. Party Politics 6, 75–94. Clarke, H.D., Jenson, J., LeDuc, L., Pammett, J.H., 1979. Political Choice in Canada. McGraw-Hill Ryer- son, Toronto. Clarke, H.D., Jenson, J., LeDuc, L., Pammett, J.H., 1984. Absent Mandate: The Politics of Discontent in Canada, first ed. Gage, Toronto. Clarke, H.D., Jenson, J., LeDuc, L., Pammett, J.H., 1991. Absent Mandate: Interpreting Change in Canad- ian Elections, second ed. Gage, Toronto. Clarke, H.D., Jenson, J., LeDuc, L., Pammett, J.H., 1996. Absent Mandate: Canadian Electoral Politics in an Era of Restructuring, third ed. Gage, Toronto. Clarke, H.D., Kornberg, A., Wearing, P., 2000. A Polity on the Edge: Canada and the Politics of Fragmen- tation. Broadview Press, Peterborough. Converse, P.E., 1964. The nature of belief systems in mass publics. In: Apter, D.E. (Ed.), Ideology and Discontent. Free Press, New York. Converse, P.E., Dupeux, G., 1966. Politicization of the electorate in France and the United States. In: Campbell, A., Converse, P.E., Miller, W.E., Stokes, D.E. (Eds.), Elections and the Political Order. Wiley, New York. Epstein, L.D., 1964. A comparative study of Canadian parties. American Political Science Review 58, 46–60. Gibbins, R., Nevitte, N., 1985. Canadian political ideology: a comparative analysis. Canadian Journal of Political Science 28 (3), 577–598. Gidengil, E., Blais, A., Nadeau, R., Nevitte, N., 1999. Making sense of regional voting in the 1997 Canadian federal election: Liberal and Reform support outside Quebec. Canadian Journal of Political Science 32 (2), 247–272. T.J. Scotto et al. / Electoral Studies 23 (2004) 463Ð483 483

Hinich, M.J., Munger, M.C., de Marchi, S., 1998. Ideology and the construction of nationality: the Canad- ian elections of 1993. Public Choice 97 (3), 401–428. Jaccard, J., Wan, C.K., 1996. LISREL Approaches to Interaction Effects in Multiple Regression. Sage Publications, Thousand Oaks, CA. Johnston, R., Blais, A., Brady, H.E., Creˆte, J., 1992. Letting the People Decide: Dynamics of a Canadian Election. Stanford University Press, Stanford, CA. King, G., Tomz, M., Wittenberg, J., 2000. Making the most of statistical analyses: improving interpretation and presentation. American Journal of Political Science 44, 341–355. King, G., Honaker, J., Joseph, A., Scheve, K., 2001. Analyzing incomplete political science data: an alternative algorithm for multiple imputation. American Political Science Review 95, 49–69. Kline, P., 1993. An Easy Guide to Factor Analysis. Routledge, New York. Kornberg, A., Clarke, H.D., 1992. Citizens and Community: Political Support in a Representative Democ- racy. Cambridge University Press, New York. Kornberg, A., Mishler, W., Smith, J., 1975. Political elite and mass perceptions of party locations in issue space: some tests of two positions. British Journal of Political Science 5, 161–185. Kornberg, A., Clarke, H.D., Stewart, M.C., 1992. Canada. In: Ameringer, C.D. (Ed.), Political Parties of the Americas, 1980s to 1990s. Greenwood Press, Westport, CT. Meisel, J., 1975. Working Papers on Canadian Politics, second ed. McGill-Queen’s University Press, Montreal and Kingston. Nevitte, N., Blais, A., Gidengil, E., Nadeau, R., 2000. Unsteady State: The 1997 Canadian Federal Elec- tion. Oxford University Press, Ontario. Pammett, J.H., 1987. Class voting and class consciousness in Canada. Canadian Review of Sociology and Anthropology 24, 269–289. Pammett, J.H., Dornan, C. (Eds.), 2001. The Canadian General Election of 2000. Dundurn Press, Toronto. Porter, J., 1965. The Vertical Mosaic: An Analysis of Social Class and Power in Canada. University of Toronto Press, Toronto. Stevenson, G., 1989. Unfulfilled Union: and National Unity, third ed. Gage Pub- lishing, Agincourt, ON. Tucker, L., R. MacCallum.,1997. Exploratory Factor Analysis. The Ohio State University. Unpublished.