Understanding Voting Behaviour at the Municipal Level

It is well established that parties have a significant impact upon the way that citizens view and participate in the political process. According to Dalton (Reference Dalton2002: 125, 126), political parties “define the choices available to voters” and “shape the content of election campaigns.” Parties also provide the institutional resources for candidates to successfully appeal to voters, in the form of a “brand name” and economies of scale in campaign efforts (Aldrich, Reference Aldrich1995). Indeed, how voters relate to parties is a central component of our understanding of voting behaviour. Many voters form a long-standing, psychological attachment to parties, known as “partisanship.” Although the development of partisanship is contested—on the one hand, some argue that voters are socialized into partisanship (Campbell et al., Reference Campbell, Converse, Miller and Stokes1960); on the other, partisanship results from a “running tally” of party evaluations (Fiorina, Reference Fiorina1981)—the effects of partisanship are not disputed. Partisanship has been shown to influence not only voting behaviour (for a Canadian example, see Gidengil et al., Reference Gidengil, Nevitte, Blais, Everitt and Fournier2012), but also the manner in which voters receive and process information (Campbell et al., Reference Campbell, Converse, Miller and Stokes1960; Zaller, Reference Zaller1992). Partisanship can colour how citizens evaluate past government performance (Duch et al., Reference Duch, Palmer and Anderson2000) and is related to political engagement, interest and attentiveness (Lazarsfeld et al., Reference Lazarsfeld, Berelson and Gaudet1948; Buchanan, Reference Buchanan and Freeman1977). Indeed, the majority of research on voting behaviour makes use of voter predispositions toward parties for understanding how citizens behave politically. Partisanship can act as an “anchor,” providing a default preference over candidates, issues and policy positions.

For the most part, however, political parties do not contest municipal elections in Canada, and such is the case in Toronto. Without parties, how do voters make sense of the election? Two points are relevant. First, even without party labels, the stances of politicians can often inform voters of their ideological positions. The logic of the spatial voting model (Downs, Reference Downs1957) predicts that, when ideological positions are known, voters will prefer parties or candidates that are closer to them in ideological space. This is relevant for considerations of stated positions, in terms of left and right, but also in terms of sociodemographic characteristics which may make the interests of a voter more likely to lie on the left (more generous social spending, for example) or right (less government intervention in the economy) of the spectrum. Second, short-term factors rooted in the context of the specific election can also be important. Many voting models, from the Michigan model (Campbell et al., Reference Campbell, Converse, Miller and Stokes1960) to the bloc recursive model (Blais et al., Reference Blais, Nadeau, Gidengil and Nevitte2002), consider how more immediate considerations, such as the economy, issues and candidates, may sway a voter. In a non-partisan contest, then, it is still likely that vote choice can be understood by appealing to ideological considerations and issue preferences.

An important caveat to the above discussion is that it assumes that voters understand non-partisan municipal contests in ideological terms, simply without party labels and “teams.” This is an untested assumption. Do voters in municipal elections understand mayoral candidates in terms of their general ideologies? If not, it makes little sense to expect voters to be drawn to the candidates closest to them ideologically. We evaluate this assumption below. We also examine whether standard voting considerations, such as the economy, satisfaction with the incumbent, and issues, are relevant factors in municipal vote choice.

In so doing, this paper expands upon the work of Cutler and Matthews (Reference Cutler and Matthews2005). In the partisan Vancouver context, they found evidence that voter ideology and partisanship influenced mayoral vote choice. They also found economic considerations and social group membership to be important to some degree (among students and voters of non-European ethnicity). But for one exception—whether voters held the government responsible for problems in Vancouver's Downtown Eastside—the authors found that issues had little effect on vote choice, despite the apparent salience of issues like transit or taxes and services.

A limitation of Cutler and Matthews' study, acknowledged by the authors themselves, is that it is based on data from a partisan election and therefore lacks generalizability to most of Canada's other municipalities. We recognize that the case of Toronto is also unique. The 2014 mayoral contest was extremely high profile in nature and therefore unlikely to be as low information as many municipal contests, and two of the mayoral candidates had formal partisan ties in the past (Chow with the federal NDP and Tory with the provincial Progressive Conservatives). Although officially non-partisan, it is an open question whether partisanship structured vote preferences for Torontonians. Our study therefore enables us to consider how variation in the level of partisanizationFootnote ⁴ of an election affects the utility of standard vote models for municipal contests. We believe that this study represents a noteworthy contribution to the study of municipal elections in Canada, and a step toward understanding the generalizability of vote choice models. Before moving on to consider whether Torontonians understood the election in partisan terms, and identifying the correlates of vote choice in the election, we present some important contextual details of the mayoral contest.

Toronto 2014

The 2014 Toronto mayoral election was remarkable for a number of reasons. It was the first election in the city's history to receive significant international media attention, largely due to fascination with the incumbent mayor, Rob Ford, whose chronic substance abuse and associated problems brought widespread attention to the city's politics. Rob Ford's revelation that he had cancer, and the last minute decision to have his brother Doug replace him on the mayoral ballot, only added to the continuing circus that was Rob Ford's tenure as mayor. In essence, Doug entered the race for mayor as the pseudo-incumbent, as most voters perceived him and his brother as interchangeable; Toronto Election Study data suggest that only 3.4 per cent of respondents felt that brothers' platforms were either mostly or all different.

The election did not focus solely on the Ford brothers, however. Two high profile candidates also joined the race for mayor, eventually forcing other early and locally prominent candidates from the race. Olivia Chow was the first of the prominent candidates to officially declare her candidacy. Chow was a former city councillor and, at the time, was a member of the Canadian Parliament, sitting as an NDP MP. She was also the widow of the much-respected former leader of the federal NDP party, Jack Layton. The second high profile candidate to enter was John Tory, who had run unsuccessfully for mayor in 2003. Following that failure, he was chosen as the leader of the Ontario Progressive Conservative (PC) Party, only to step down after failing to achieve success in either the 2007 provincial election or a by-election in 2009. Aside from his past political experience, Tory also served as the CEO of one of Canada's largest telecommunication firms, Rogers Communications, and, until he announced his candidacy for mayor, hosted a radio talk show in Toronto.

On October 27, Tory carried the day, winning 40.3 per cent of the popular vote. Doug Ford was second, with the support of 33.7 per cent of voters, while Olivia Chow came in third, with 23.2 per cent of the vote (City of Toronto, 2014). If the turnout rate of 54.9 per cent is any indication, the election galvanized and motivated the public to get involved. Turnout in 2014 was significantly higher than in any election since the 1997 amalgamation;Footnote ⁵ it was also higher than the most recent Ontario provincial election. Unlike many municipal elections, this one clearly captured the interest and attention of voters.

Despite the fact that Rob Ford was no longer running for mayor, assessment of his time as mayor remained a dominant issue in the media because of Doug Ford's close association with his brother. Though the election did not, strictly speaking, include an incumbent candidate, retrospective considerations were likely among a public that viewed Doug as a proxy for, or at least very similar to, his brother. Beyond retrospective evaluations of the Fords, a number of key issues dominated the election. Transit was and continues to be a very salient issue in Toronto. In the 2010 election, Rob Ford had promised to scrap a light rail transit (LRT) system, championed by the city's former mayor David Miller, and build more subways throughout the city. Soon after being elected in 2010, Ford moved forward with his plan, scrapping his predecessor's ambitious LRT proposal in favour of a single subway line. However, after initially supporting Ford, Toronto City Council overturned its decision and came up with its own compromise solution to address transit needs. During the 2014 campaign polls conducted by newspapers consistently placed transit at or near the top of most important issues in the election (Church, Reference Church2014; Ipsos Reid, 2014; Nanos, 2014; Toronto Star, 2014a, 2014b). Doug Ford stuck to his brother's promise for more subways, while Olivia Chow supported city council's compromise and John Tory introduced plans for his “SmartTrack” commuter rail system.

Along with transit, residents identified other important issues during the campaign. Many polls identified outgoing mayor Rob Ford as the biggest “issue” next to transit. Whether to increase taxesFootnote ⁶ at, below, or above inflation, along with concerns about the city's finances, were also important issues, particularly as Doug Ford sought to build on his brother's purported success in managing the city's finances and keeping taxes low. As a corollary of the transit debate, residents and candidates also identified the issue of traffic and congestion as a major concern in the city, with both Ford and Tory promising to protect drivers' interests in the city.

Data

The analyses below draw primarily upon data from the Toronto Election Study (TES). The TES is a two-wave internet survey of Torontonians that was conducted around the time of the Oct. 27 election. Three thousand respondents were interviewed in the weeks before election day (after the candidate nomination deadline of September 12). Nearly 75 per cent of those respondents also completed the post-election wave of the survey, which was administered in the week after the election.Footnote ⁷ The TES includes a variety of questions about attitudes and behaviour, similar to those contained in many national or provincial election studies and is the first dataset of its kind that allows for a thorough consideration of voting behaviour in a non-partisan Canadian municipality.Footnote ⁸ In order to maximize the generalizability of our findings, all results discussed below are weighted for age, gender and education.Footnote ⁹

Ideological and Partisan Perceptions in a non-partisan election

As noted above, the spatial vote choice model expects that voters will prefer candidates who are closest to them ideologically. In most elections, this coincides with party preferences but, as previously mentioned, Toronto does not have partisan mayoral contests. Therefore, we need to understand how voters viewed the candidates before we can assess whether a standard vote model is relevant. To begin our analysis, then, we assess the degree to which voters differentiate the candidates in terms of their ideological positions, and compare that to recent federal and provincial elections to see if the non-partisan nature of the election was relevant for perceiving the ideology of the candidates.

For the Toronto data, we draw on a question from the TES that asks respondents to position the three major candidates on a scale from 0 (left) to 10 (right). Federal estimates are taken from Canadian Election Study data from 2011 (Fournier et al., Reference Fournier, Cutler, Soroka and Stolle2011), while for provincial estimates we draw upon data from the 2011 Ontario election collected by the Making Electoral Democracy Work project (Blais, Reference Blais2010). Table 1 includes the results of this analysis, and shows the mean and standard deviation values for the perceived ideology of the Toronto mayoral candidates and the federal and provincial parties.

The first striking observation from Table 1 is that Torontonians clearly differentiated between the mayoral candidates in terms of ideology. Chow was perceived as being on the left, while both Tory and Ford were both placed on the right. Importantly, Ford was seen to be to the right of Tory (this difference, and all other differences between candidates and parties in Table 1, is statistically significant at p < 0.01).

The ideological placements of two of the mayoral candidates were, in fact, quite similar to those of two of the parties at the federal and provincial levels. Estimates of Ford's ideology were similar to those of the provincial Progressive Conservative and federal Conservative parties: all had values above 7. On the other side of the spectrum, Chow was clearly closest to the NDP, both provincially and federally.Footnote ¹⁰ John Tory was the sole outlier. Respondents placed him at 6.53 on the scale, closer to the federal Conservative and provincial PC parties than the centrist Liberals but further from the right-wing parties than Ford. Tory's tenure as the leader of the provincial Progressive Conservative party may account in part for this placement.

There are thus two primary conclusions from Table 1. First, even in the absence of parties, voters clearly differentiated between the ideologies of the candidates; Chow, Tory and Ford were seen as ideologically distinct from one another, and the range of perceived ideologies was roughly the same as at the federal and provincial levels. Second, ideologically at least, Chow and Ford were perceived as the NDP and (Progressive) Conservative candidates, respectively, while Tory was perceived as more Conservative than Liberal, at least on a unidimensional ideological scale.

Were these ideological perceptions accompanied by corresponding partisan associations? Although the election was officially non-partisan, some of the candidates had party ties and many voters associated the candidates with parties. Table 2 shows the responses to the following TES question: Which political party, if any, would you associate with (Doug Ford/Olivia Chow/John Tory)?

The results of Table 2 reveal an intriguing association with the ideological observations in Table 1. Most voters (68.5%) associated Chow with the NDP, which is unsurprising given her past ties with the party but also because she had a similar ideological placement to both the provincial and federal parties in Table 1. The Conservative party was the modal response for both of the other candidates, though fewer people (46.2%) associated Ford with the party than Tory (55.5%) (likely due to Tory's former affiliation with the provincial PCs). Few voters associated any of the candidates with the Liberal party, though Tory was perceived as a Liberal by more respondents than Ford.

On the whole, therefore, ideological placements line up well with partisan perceptions.Footnote ¹¹ It appears that many voters understood the 2014 Toronto Election in much the same way as any other partisan contest. The left-wing candidate was associated with the leftist NDP, and the right-wing candidates were associated with the right-leaning Conservatives. Despite the absence of a formal partisan cue, most voters saw the election in partisan terms, although partisan perceptions were not uniform. Only a minority of respondents associated Ford with the Conservative Party, and almost 40 per cent did not associate him with any party (answering “none” or “don't know”). Even Olivia Chow, who resigned as an NDP MP in early 2014 in order to run in the mayoral contest, was only associated with the NDP by fewer than 70 per cent of voters; more than 1 in 5 respondents assigned no partisan affiliation. While there were clear patterns to partisan perceptions, party information was uneven, and in some cases, nonexistent.

Explaining vote choice in the 2014 Toronto Municipal Election

Of course, explaining vote choice is a much more complex process than unidimensional measures of ideology or perceived links to parties might indicate. We turn now to examine the applicability of a more complex model of vote choice to the 2014 Toronto Municipal Election. Specifically, we consider several “blocs” of factors, some of which contribute to ideological positions and some of which reflect short-term considerations. This approach is similar to the multi-stage models (drawing on the work of Miller and Shanks, Reference Miller and Shanks1996) that have been used to explain vote choice in federal elections (see Blais et al., Reference Blais, Nadeau, Gidengil and Nevitte2002; Gidengil et al., Reference Gidengil, Blais, Everitt, Fournier and Nevitte2006, Reference Gidengil, Nevitte, Blais, Everitt and Fournier2012; Nevitte et al., Reference Nevitte, Blais, Gidengil and Nadeau2000), as well as Cutler and Matthews's study of Vancouver.Footnote ¹²

Our analysis focuses on five specific types of variables: sociodemographic characteristics, ideology, partisanship, retrospective evaluations of incumbent performance and issues. The sociodemographics we include are standard, namely gender, age, income, education level, ethnicity, religion and geography (downtown Toronto versus the suburbs—Etobicoke, North York and Scarborough).Footnote ¹³ We also expand the list to include two variables which we expect may be of particular importance in the 2014 Toronto municipal election. First, we incorporate a measure of home ownership, a variable that has been found to have implications for municipal elections (Fischel, Reference Fischel2001; McGregor and Spicer, Reference McGregor and Spicer2016). Next, we introduce a measure of sexual orientation. The outgoing mayor, Rob Ford, famously avoided attending events surrounding Pride week in Toronto, and this may have impacted the vote choices of sexual minorities (Dale, Reference Dale2014; Gee, Reference Gee2012; Rider, Reference Rider2011). Although these variables are not strictly comparable to other studies, they are important for understanding the overall contribution of sociodemographic factors to vote choice in Toronto in 2014.

We account for ideology with indicators that address two dimensions of the concept: social and economic conservatism. Though these measures are related to one another,Footnote ¹⁴ they nevertheless represent two distinct dimensions of political competition. Indices were created for each dimension from a battery of TES questions (details can be found in Appendix I).

Although the election was technically non-partisan in nature, we also include a series of partisanship dummies (Conservative, Green, Liberal, NDP) in our analysis. As established above, many voters associated the mayoral candidates with parties, and such associations may have coloured voting behaviour. Partisanship variables are also standard in studies of other electoral contests in Canada and were found to be relevant in Vancouver by Cutler and Matthews (Reference Cutler and Matthews2005). The degree to which partisanship matters for vote choice is an important indicator of the partisanization of this specific election. It is possible that, because parties were not structuring the competition, partisanship itself was less of a factor than usually found in voting studies.

We also consider two types of short-term measures that have been found relevant in other studies: evaluations of the performance of the previous government and issues. Retrospective evaluations include an evaluation of the city's economy and a measure of satisfaction with the current mayor. For issues, we include attitudes towards four major campaign issues: public transit, municipal taxes, traffic and congestion, and the finances of the municipal government. Many of these issues cannot be clearly placed on a left-right continuum, so we use measures of perceived issue importance. Following from the idea of issue ownership, this will allow us to understand whether any particular candidate was seen as having the best proposal for an issue. We recognize, however, that this operationalization may underestimate the importance of issues, as individuals who prioritized an issue may have disagreed on the appropriate policy direction.

If voters approach their vote decision in municipal contests in the same way as other elections, then we expect that the model we have constructed will have considerable explanatory power, with many of the variables displaying relationships with vote choice. Note that we do not include evaluations of the candidates as is common practice in the study of partisan elections in Canada. In Canada's Westminster system of government, vote choices ultimately result in the choice of a governing party and, as evaluations of candidates and party leaders are distinct from (though undoubtedly related to) evaluations of party, it makes sense to include measures of candidate and leader ratings in explanatory models of such elections. At the municipal level, however, the choice of mayoral candidate leads directly to the election of a candidate. In such a setting, the link between candidate rating and vote choice is tautological, and the inclusion of candidate variables in an explanatory model provides no theoretical benefit.Footnote ¹⁵

In Tables 3–5 we present the results from a series of multinomial logistic regression models, where the dependent variable is vote choice. Entries report average marginal effects (calculated while leaving values of other variables unchanged), which can be interpreted as the effect of each independent variable upon the likelihood of supporting each candidate, as opposed to the other two candidates.Footnote ¹⁶ All variables are coded from 0 to 1, meaning that entries can be interpreted as the effect of each variable, moving from the minimum to maximum value (see Appendix I for a description of the coding of each variable). Such an approach allows us to easily compare the magnitude of the effect of each variable upon vote choice.

One of the advantages of the multi-stage approach to explaining vote choice is that it allows researchers to identify the manner in which factors which may tend to be relatively distal to vote choice (such as sociodemographic characteristics) are mediated by more proximate factors (such as ideology, attitudes towards candidates or issues specific to a particular election). The above analysis reveals that several of the factors shown in Model 1 in Tables 3–5 (which include sociodemographic characteristics only) are, in fact, mediated by the variables added in later iterations and models of the analysis. For example, Table 3 reveals that several sociodemographic characteristics are related to support for Olivia Chow but suggests that some of those patterns are driven by attitudinal differences related to one's sociodemographic group. For example, Model 1 suggests that home ownership and sexual orientation (two factors that receive relatively little attention in the voting behaviour literature) were associated (negatively and positively, respectively) with support for Chow. Both of these factors become insignificant, however, when ideological measures are added in Model 2 (the gender and geography variables are also significant in Model 1 but insignificant by Model 5). Such findings suggest that these groups supported (or opposed) Chow on the basis of attitudes held in greater proportion by those with specific sociodemographic characteristics, rather than because of the sociodemographic characteristics themselves.

Nevertheless, it is the full models that deserve the most attention. Tables 3–5 show clearly that the standard approach to explaining vote choice in federal and provincial elections has resonance at the municipal level. Several of the sociodemographic, ideological and partisan measures included in our models display strong relationships with vote choice. Retrospective evaluations and issues also play a significant role in vote choice at this level of government. To present the results in Tables 3–5 in a concise, intuitive manner, we present Figure 1, which shows the marginal effects of each explanatory variable upon the likelihood of support for each mayoral candidate, based upon Model 5 in Tables 3–5. Only statistically significant results are shown. Such an approach allows us to quickly identify and compare the magnitude and direction of effects for each variable and candidate.

Figure 1 The Determinants of Mayoral Candidate Support - Marginal Effects

Olivia Chow, perceived as the most left-wing candidate and associated with the NDP by most voters, receives much of the traditional left-wing vote. She performs well among young voters, those with low income and atheists. There is also modest evidence that university educated and visible minority electors support her (p < 0.10). Unsurprisingly, Chow receives little support from ideologically conservative voters, in either social or economic terms. The effect of the economic conservatism (just over 41 percentage points) is particularly great here. She also performs well among NDP partisans and poorly among Conservatives. As a non-incumbent candidate, she understandably wins support from those individuals who believe the city's economy worsened in the last year, and those who are unsatisfied with the performance of Rob Ford. Finally, issues were a significant consideration for Chow voters, who clearly prioritized public transit.

The story for the right-wing candidates, Ford and Tory, is somewhat more complex, as the two display different relationships with characteristics typically associated with voters of right-wing parties. In terms of sociodemographic characteristics, Ford receives little support from older, wealthier voters, which is rare for a right-wing candidate. He performs much better in the suburbs than he does in downtown Toronto, as would be expected of such candidates. For his part, Tory seems to receive the traditional right-wing vote, at least as far as sociodemographic characteristics are concerned. He performs well among older, wealthier, non-visible minority voters.

Another interesting area where the two right-wing candidates differ is with respect to social and economic measures of conservatism. Ford is supported by social conservatives, but the economic conservatism variable displays no relationship with a Ford vote. In contrast, a Tory vote is negatively associated with social conservatism but has a strong positive association with economic conservatism. In fact, the latter variable has the greatest impact (38 percentage points) of any included in the final model on Tory support. Such a finding suggests that Ford and Tory attracted different types of right-wing voters.

As for partisanship, although both candidates were associated with the Conservative party by a plurality of voters, the Conservative partisanship variable is associated only with one candidate (Tory) and even then this association is somewhat tenuous (p = 0.059). Interestingly, though Liberal partisanship is negatively associated with a Ford vote, it is positively associated with Tory support, perhaps due to the fact that Tory was viewed as tied to the Liberals by some voters. NDP partisanship is negatively associated with support for Tory but has no relationship with a Ford vote. Considering that Ford's association with the Conservatives was the weakest of all candidate associations (both perceived and actual), these findings are quite interesting. It suggests that partisanship may more be useful in municipal vote models when there are clear ties between candidates and parties.

Retrospective evaluations and issue positions are also significantly related to support for the two right-wing candidates. Economic evaluations and opinions of Rob Ford's performance work much as one would expect, with both measures positively associated with support for Doug Ford and negatively associated with a Tory vote. The issue variables suggest that Ford voters believed transit to be relatively unimportant but prioritized the state of the city's finances. This resonates with the Ford brothers' focus on leaner government. In contrast, the finances issue was relatively unimportant to Tory voters, who instead believed taxes were an important issue. The various issue results suggest that, in the aggregate, Chow was seen as the best to deal with public transit, Tory for taxes and Ford for municipal finances.

As a whole, the results above show that our version of a standard vote model has resonance at the municipal level in a non-partisan, yet-partisanized mayoral election. A great number of the variables included in the model are statistically significant, and many operate in a logical fashion, based upon the perceived ideological and partisan placements of the candidates. Even without formal parties structuring political competition, voters clearly differentiated between left- and right-wing candidates, and many variables work much the same way as they do at other levels of political competition in Canada.