It is well established that campaigns have significant potential to affect the attitudes and behaviour of electors. The 2014 Toronto mayoral campaign, a marathon compared to federal and provincial contests, lasted nearly 10 months, from January 2nd to Oct. 27th. The purpose of this blog post is to test for the presence of campaign effects during the final weeks of the campaign. That is, we test for changes in public opinion, under the assumption that such changes must be driven by some campaign period stimulus. We do so using TES data, collected during the final 38 days of the campaign period (after the candidate nomination deadline). In doing so, we find little, if any, evidence of campaign effects. All indications are that the outcome of the election would have been the same had the election been held on any of the dates covered by TES data.
The scholarly consensus on whether and how campaigns affect voters has evolved significantly in past decades. Early studies of voting behaviour (Lazarsfeld et al., 1948; Campbell et al., 1960) contended that campaigns had ‘minimal effects’ upon voters. That is, scholars widely believed that campaigns were rarely able to overcome knowledge and prejudices held by voters at the start of the campaign period. In more modern times, however, this view has been challenged to the point that it is now widely accepted that campaigns, and new information which voters might be exposed to during the course of campaigns, have the potential to significantly affect voter attitudes and behaviour (Jacobson 1983; Fournier 2004).
It has been well-established that campaigns have the potential to have a significant effect upon voter attitudes and preferences at the federal level in Canada (Johnston et al 1992, Blais et al. 2003). If campaign effects are to be observed at the municipal level in this country, one might expect them to be particularly noticeable in Toronto. Though municipal politics tend to receive significantly less attention than do other orders of government, the 2014 Toronto election was one of the most high profile municipal contests in Canadian history. Electors were bombarded with images and stories about the contest for months leading up to election day, and each one of these messages had the potential to affect voters and influence attitudes. Given the exceptional publicity afforded the race, therefore, if campaign effects are to be observed in a municipal setting, all other things equal, we might expect to see them in Toronto 2014.
Alternatively, an argument can be made that we should expect relatively little in the way of attitudinal change during the campaign, at least in the weeks leading up to election-day. The fact that the official campaign was so lengthy (nearly 10 months), and because the election and its candidates were extremely high-profile, attitudes towards the candidates may have been relatively well anchored long before election day. Zaller’s (1992) Receive-Accept-Sample (RAS) model of decision making suggests that attitudes are formed as a result of the summation of considerations, or beliefs or pieces of knowledge relevant to some decision (such as vote choice). During a campaign, new considerations are combined with existing ones. All else equal, the more considerations are present, the more strongly anchored a belief may be, and the less likely it is that new considerations will change one’s opinion. In the context of a lengthy, high-profile campaign, such as the 2014 Toronto race, it stands to reason that many voters will already have a great number of existing considerations by the nomination deadline. In such a scenario, attitude change may be particularly unlikely in the later stages of a campaign. In short, the substantial length of the Toronto campaign may very well mean that voter attitudes and preferences will solidify well in advance of election day. Campaign effects in the final weeks before election day may therefore be muted in Toronto, as compared to federal and provincial contests.
It is between these two possibilities that we adjudicate. To do so, we employ TES data to map several attitudinal measures during the course of the final weeks of the campaign (a period comparable in length to federal and provincial campaigns). Did the campaign affect voter attitudes and preferences in the final weeks of the 2014 Toronto Municipal Election, as has been found to be the case in other levels of elections in Canada, or were opinions and knowledge levels relatively stable during this time?
The pre-election wave of the TES was designed to employ a rolling cross-sectional design, meaning that surveys were conducted in such a manner than they are staggered throughout the weeks leading up to the election, with a portion of the sample being collected each week. This repeated cross-sectional approach means that new respondents were interviewed each week. In essence, this is what pollsters do when they collect new samples every few days to measure and track public opinion. For our purposes, such a design allows us to monitor potential opinion changes during the final month-and-a-half of the campaign.
We map four different types of attitudes when testing for the presence of campaign effects, including perceptions of the competitiveness of candidates, candidate ratings (‘feeling thermometer’ questions), responses to a question which asked whether there was a candidate for whom respondents would not consider voting, and finally, vote intention. Changes across time with respect to any of these attitudes could suggest the presence of campaign effects.
The figures above reveal several noteworthy findings. For example, though Olivia Chow came in third place in the election, she had the second highest rating among the three major candidates. Additionally, though Ford came in second in the race, there were more voters who would never consider voting for him than is the case for the other two candidates combined. Finally, the graphs above show that Tory dominated through the entirety of the period under consideration here. He was consistently the highest rated, had the greatest support among voters and was perceived all along as the front runner.
For the purpose of this post, however, the most important finding from above is that there is little, if any, evidence of campaign effects during this period. Competitive perceptions were fairly constant from mid-September until election day. Aside from the divergence of Chow and Ford after Sept. 18, there is little observable change to report.
The feeling thermometer scores also tell a story of consistency over time. Throughout this entire period Tory is rated most highly of the three candidates, followed by Chow, and then Ford. At no point do the candidates change positions with one another.
Responses to the question of whether there is a candidate one would not consider voting for are also remarkably consistent. Candidate ordering remains constant throughout the weeks leading up to election day, and there is little change over time in the share of the population that holds negative sentiments towards each contender. Anti-Ford sentiment is constant throughout this period (with no statistically significant change). Anti-Chow and anti-Tory attitudes see a modest decrease and increase, respectively, in the period from October 13th to 19th, though both changes are temporary and return to previous levels in the week before the election.
The flip side of anti-candidate views, and indeed the most important variable in the field of voting behaviour, is vote choice. For the most part, vote intentions largely mirror the consistency of candidate ratings and anti-candidate sentiment. The exception is the period from October 13th to 19th, where we see a temporary dip and jump in support for Tory and Chow, respectively. However, Tory is consistently ahead of the other two candidates for the entirety of the final weeks of the campaign, and the level of support pledged for each candidate is statistically indistinguishable from the first and last time period considered in the figure.
Explaining the Absence of Effects
The story of the last weeks of the 2014 Toronto mayoral campaign is that this part of the campaign had little, if any effect upon either the voter attitudes or the outcome of the election. All indications are that the outcome of the election would have been the same had the election been held on any of the dates covered by TES data.
The absence of campaign effects in the last weeks of the Toronto mayoral election is noteworthy, and bears some consideration. Indeed, the period covered by TES data is of a similar length to federal and provincial contests, during which there are often major swings in public support. We suggest that the most likely source for this relative inactivity during the late stages of the campaign is the length of the official campaign period. At nearly 10 months, the official campaign dwarfs those of other levels of government. During the eight months prior to the nomination deadline, electors had ample opportunity to learn about the candidates and their policies, and likely had fairly well formed attitudes of the candidates by the time the nomination deadline rolled around. In terms of Zaller’s (1992) RAS model (discussed above), voters had two-thirds of a year to collect and store considerations. Any new information obtained during the final weeks of the campaign that might be inconsistent with existing beliefs (and thus which may affect attitudes) thus had to compete with a tremendous number of other, older considerations.
That said, it is doubtful that that the length of the campaign itself is responsible for the absence of late-campaign effects. In municipalities with lower profile elections, voters may receive relatively little information on municipal politics. In Toronto, however, the attention paid by the media to the 2014 campaign was nothing short of unprecedented; the highly publicized antics of the outgoing mayor put the politics of Toronto in the spotlight like never before. Though the data are unavailable to test this contention, it thus stands to reason that Torontonians were relatively enlightened with respect to municipal politics, relative to their counterparts in many other cities. In sum, we suspect that it is this high level of existing knowledge, combined with the great length of the mayoral campaign, that contributed to the absence of late-campaign effects in Toronto in 2014.
 Note that there is some variation in the number of respondents interviewed during each week of the campaign. These differences, however, should have no effect upon the patterns observed here, as the figures presented here indicate confidence intervals, which are a function of sample size. The sample sizes are as follows: Sept. 19-28 = 1,000, Sept. 29-Oct. 5 = 598, Oct. 6-12 = 529, Oct. 13-19 = 392, Oct. 20-26 = 480. Note also that all data have been weighted to population average for age, gender and education for each week/group represented in all figures, to maximize representativeness of the sample and comparability between groups.