Negative Partisanship in a Multi-party System: The Case of Canada

Abstract

The origins and implications of partisan identification are well studied, but negative partisan attitudes – dislike for a particular party – have escaped such scrutiny. We argue that the concept of partisanship, especially in a multi-party system, is incomplete until negative sentiments are considered. In this paper, we refine the concept of negative partisanship (NPID) by providing an improved method of operationalizing it, examine its incidence and relationship with positive party identification in a multi-party system, and propose two theories of its origins. Our results, based upon data from five Canadian federal elections, indicate that studying NPID in a multi-party system requires a broader understanding of the mechanisms that lead to negative party attitudes than in a two-party system.

Notes

1. This model is compatible with Key's (1966) contention that “switchers” change their minds based on legitimate political evaluations – under this rubric, their running tally is the impetus to switch.

2. Positive partisanship is conventionally described as long-standing, and NPID may be as well. Nevertheless, we remain agnostic on this dimension of the concept, given the inability to test such a proposition with current data.

3. Of course, as in most studies of partisanship and voting behaviour, some endogeneity in the relationship between ideology, competition and NPID is possible. Negative partisanship may influence ideological positions and perceptions of competition, similar to how PPID might shape political attitudes and judgements. We do not evaluate to what extent the relationship is endogenous here; doing so would require panel data of a sort not available for the elections under consideration. Nevertheless, while we have no evidence of causation, we are encouraged that our results are consistent with the explanations we have provided for the potential pathways to NPID.

4. Negative partisanship has not been examined directly in the multi-party Canadian context. Wattenberg's (1982) study of the relationship between positive and negative attitudes among voters towards the two main political parties in Britain, Canada, Australia and the USA shows that those with positive opinions of a major party need not necessarily hold negative opinions of the other major party, and that in fact much diversity in combinations of positive, neutral and negative evaluations of parties is possible. Particularly in Canada, the idealized pattern of “my party is good and the other is bad” does not hold, which Wattenberg suggests may be a result of the parties’ own unwillingness to adopt polarized stances. Bowler, Lanoue, and Savoie (1994) argue for a phenomenon related to NPID, in that they expect that the perceived strength of an “ideologically threatening” party should positively influence the strength of a voter's positive partisan attachment, especially under a “winner-take-all” electoral system like Canada's. Finally, Medeiros and Noël (2014) consider NPID with respect to the Conservatives and Liberals exclusively.

5. Data are available from the Canadian Opinion Research archive at http://www.queensu.ca/cora/ces.html.

6. Still another potential method of operationalizing NPID involves the “propensity to vote” questions considered in some election surveys (though not in the CES). The questions provide an indication of the likelihood that respondents would consider voting for each party. NPID could be assigned to those parties that receive a score of 0 on this measure, and that also receive a low feeling thermometer score.

7. The feeling thermometer questions specify that “zero means you REALLY DISLIKE the party and one hundred means that you REALLY LIKE the party”. We thus infer that 50 is the neutral point on this scale, and anything below that point suggests some degree of dislike of a party. This is a conservative threshold to ensure that a party someone would never vote for is not rated positively.

8. We allow for NPID in cases where multiple parties receive the same feeling thermometer score, provided that our other two criteria are met (no other party can receive a lower thermometer score than the party listed in the “would not vote for” question, and that party must receive a feeling thermometer score less than 50). Overall, 20.3% of cases classified as having NPID include ties of this nature. The average feeling thermometer score for NPID cases where ties exist is 16.6, and in 37% of cases with ties, the thermometer score given to the NPID party is 0. Classifying such individuals as not holding NPID would ignore the negative attitudes of a sizable and important segment of the sample. Nonetheless, when we replicate our analyses classifying these cases as not having NPID the substantive conclusions of our analyses remain unchanged. Furthermore, these results support the notion that rating another party at the same level does not negate the notable antipathy implied by the respondent's stated intention never to vote for a particular party; as many such respondents tend to rate other parties equally poorly, it is possible that this generosity with low scores is simply a part of the negative partisan's mindset.

9. There are a small number of cases where individuals hold PPID and NPID for the same party. That is, they state that they are partisans of a party, but that same party meets our criteria for NPID. Such cases are removed here. Interviewee response error may partially account for this phenomenon. However, the existence of such cases suggests that the concept of NPID may be of use when attempting to identify disloyal partisans.

10. In 1988, 12.7% of respondents selected the “neither other” option, indicating that they would never vote for two different parties. Of that group, 82% of cases met the other two NPID criteria (i.e. the two NPID parties had a feeling thermometer score of less than 50 and lower than the non-NPID party). Of that smaller group, 30% of individuals had tied feeling thermometer scores. In order to focus upon single NPID only, these cases are dropped (this represents 3.1% of the sample).

11. In 2006 and 2008, respectively, 6.1% and 4.4% of applicable respondents selected the “multiple/other” option.

12. Comparative Manifestos Project (CMP) data reinforce our assumptions about the relative ideological positions of the parties. CMP data are based upon a content analysis of official party manifestos, and contain an absolute left–right variable (Klingemann et al. 2006). Positive values for this variable indicate the right, and negative values indicate the left. The CMP does not yet have data for the 2011 election, but based upon 1988, 1993, 2006 and 2008 ratings, the CMP assigns an average value of +17.1 to the Conservatives, –5.7 to the Liberals and –26.5 to the NDP. The Greens were only coded in 2008, receiving a score of –13.8, and data are only available for the Reform Party in 1993, when they were assigned a value of +40.4. The parties thus fall into the relative positions described above.

13. Rates of NPID using the method applied by Medeiros and Noël (2014) are much higher than with our operationalization. These rates are 77.6%, 83.3%, 86.0% and 79.9% for 1988, 1993, 2006 and 2008, respectively. Thus, our measure is more conservative.

14. The average ideology score for those with Liberal NPID is 5.1, a finding that may be surprising, given that this score is roughly the midpoint of the scale, and that the Liberal Party generally is thought to be in the centre of the Canadian political spectrum. This average value makes sense, however, if we assume that individuals opposed to the Liberals are split between those on the left and those on the right (taking the average of these extreme values produces a result near the centre of spectrum).

15. These questions tend to be placed in either the post-election or mail back segment of the CES (these portions of the CES have limited sample sizes), and this information is altogether unavailable for the 1993 and 2006 elections.

16. An analysis of available ideological scores for positive partisans suggests that PPID is a valid proxy. Pooling data from 1988, 2008 and 2011 (the elections with available ideology self-placement values), the average ideology scores of Conservative, Liberal, NDP and Green partisans, respectively, are 6.8, 4.6, 3.4 and 3.9 (this trend holds for each individual election). This fits with the conventional wisdom that Reform and Conservative partisans are on the right of the ideological spectrum, Liberals in the centre, and NDP and Green supporters on the left.

17. One reason for this outlier could be that by 2006 the party had ruled uninterrupted for 13 years, and crises such as the sponsorship scandal were dragging the party's popularity down enough to lead to an increase in NPID. While our focus here is upon ideological and competitive explanations of NPID, we acknowledge that there are other factors that could also influence negative partisan attitudes.

18. Results of the 1993 and 2011 elections contradict campaign period expectations. Despite the fact that the Conservatives won only two seats in 1993, CES respondents thought them to be the second-most competitive party (they were given a 50% chance of victory on average, as compared to 61% for the Liberals and 24% for the Reform Party). Additionally, while the NDP did form the official opposition in 2011, few respondents expected this result. A CES question on the expected outcome of the election reveals that 85.8% and 13.4% of respondents expected Conservative and Liberal victories, respectively. Only 0.8% thought that the NDP would win.

19. Note that this analysis considers the 1988, 1993 and 2006 elections only as the necessary information on competitive expectations is unavailable for 2008 and 2011 (these elections do not have information on which party is the second-most competitive nationally). The Green Party is excluded from the analysis due to a limited sample size.

20. National-level competitive considerations are considered.

21. Appropriate questions include the “would never vote for” question, feeling thermometers, or as mentioned in footnote 7, “propensity to vote” questions.