Introduction
In Canadian municipal elections, incumbent city councillors have an almost insurmountable advantage—the rate of municipal incumbent re-election in this country is regularly above 90 per cent (CBC, 2017). In the United States, research shows an estimated national re-election rate of 84 per cent for city council (DeSantis and Renner, Reference DeSantis and Renner1994). Though high rates of incumbent success are not limited to municipal politics (Kendall and Rekkas, Reference Kendall and Rekkas2012; Friedman and Holden, Reference Friedman and Holden2009), the incumbency advantage is nevertheless particularly pronounced locally (Lucas, Reference Lucas2019). As Moore et al. (2015: 88) succinctly put it, “Incumbents are king” in municipal politics.
This article takes a new approach to improving our understanding of one of the potential sources of municipal incumbency advantage—by considering the ability of electors to correctly identify the two most competitive candidates in one's ward and the factors associated with being able to do so. It has been suggested that the nonpartisan nature and low-information environments of most local elections combine to create an environment in which voters are not easily able to identify the strongest challenger to an incumbent, thus providing a significant advantage to sitting city councillors (Moore et al., Reference Moore, McGregor and Stephenson2015; Taylor and McEleney, Reference Taylor and McEleney2017). The absence of party labels (and the competitive information relayed by these labels) means that this important shortcut cannot be used to inform one's competitive understanding of an election (Schaffner et al., Reference Schaffner, Streb and Wright2001). Coupled with the fact that readily available information about ward races is often abysmally low (Krebs, Reference Krebs1998; Trounstine, Reference Trounstine2008), voters face a serious challenge in trying to understand the competitive landscape in council races. Under these circumstances, name recognition of the incumbent councillor can serve as an exceptionally strong cue for voters.
We suggest that this same combination of factors can make it difficult for voters to identify the challenger with the highest likelihood of defeating, or at the very least credibly challenging, a sitting councillor. Such a challenge is not a concern for those individuals who are satisfied with the incumbent; for those who are not, however, it poses a significant strategic problem. If opposition voters are unable to coordinate behind a single candidate, incumbents stand to do extremely well, especially in light of all the other advantages incumbents already enjoy.
Using survey data from the Canadian Municipal Election Study (CMES), we consider the case of the 2018 municipal elections in Mississauga, Ontario.Footnote 1 CMES survey respondents from this city were asked to identify the two most competitive candidates in their local ward races, and the accuracy of responses can be determined by comparing answers to actual election outcomes.Footnote 2 We focus on the top two candidates in an effort to determine the extent to which voters who may conceivably wish to see change at the ward level are able to coalesce behind a single challenger. We seek to explain the ability of voters to correctly identify competitive candidates in their wards through a consideration of individual-level factors (local rootedness, political awareness and knowledge) and ward-level features (including the number of individuals contesting an election and candidate spending). The results of our multilevel logistic regression analyses reveal a number of significant correlates. In particular, both individual-level (voter) and contextual-level (ward) factors appear to play an important role in shaping the ability of voters to identify either the incumbent or second-place finisher.
Mississauga (the country's sixth most populous municipality) is an excellent case with which to consider these questions. The city, like most, has high rates of incumbent success, at both the mayoral and council level. Its longest-serving mayor, Hazel McCallion, was elected to 12 consecutive terms, often receiving over 90 per cent of votes cast (Urbaniak, Reference Urbaniak2009). At the council level, incumbent re-election rates are also very high (though not out of line with other North American cities); since 2010, 92.6 per cent of incumbents seeking re-election have been successful.Footnote 3 In 2018, all 9 incumbents who sought re-election were successful. The city is a good example of a low-information context and is much more representative of other municipalities than larger cities such as Toronto on Montreal in this regard. Despite its size (the city has over 720,000 inhabitants), and largely due to its proximity to neighbouring Toronto, Mississauga has no major local media to provide detailed reporting and information about local politics and the competitive dynamics of municipal races during elections—a situation typical of many municipalities in territory covered by the 905 area code, which surrounds Toronto and includes millions of residents.Footnote 4
As a result, Mississauga is a particularly useful case with which to first study the conditions under which electors are able to identify competitive city council challengers. Since the city is a low-information setting where incumbents have a strong track record, it should be relatively easy for Mississaugans to identify incumbents/winners but much harder for them to identify likely second-place candidates.Footnote 5 For the many Canadian cities that share these characteristics, Mississauga serves as a typical case, from which results can be generalized. There is a significant amount of variation, however, among the thousands of municipalities in Canada, so findings from this setting can also be tested later in other settings, including larger or smaller cities where levels of information might be different. Regardless of generalizability, we believe there is value in learning about this large and understudied city in Canada. This first examination of our research questions, however, speaks to the large literature on the paramountcy of name recognition and incumbent advantage in municipal politics and sheds light on how voters might successfully navigate the nonpartisan, low-information environments of these elections.
The Incumbency Advantage
Research suggests that incumbency is one of the most important predictors (if not the most important predictor) of candidate success at the municipal level in both Canada and the United States (de Benedictis-Kessner, Reference de Benedictis-Kessner2018; Krebs, Reference Krebs1998; Kushner et al., Reference Kushner, Siegel and Stanwick1997, 2001; Moore et al., Reference Moore, McGregor and Stephenson2015; Stanwick, Reference Stanwick2000; but see Boyne et al., Reference Boyne, James, John and Petrovsky2009 for a discussion of negativity bias).Footnote 6 Trounstine (Reference Trounstine2011), for instance, finds that incumbency increases the expected total vote share of a candidate by more than 20 percentage points. A series of recent studies employing regression discontinuity designs reveal that incumbency increases the probability that a municipal candidate will run and win the next election by more than 30 percentage points (de Benedictis-Kessner, Reference de Benedictis-Kessner2018; Lucas, Reference Lucas2019; Trounstine, Reference Trounstine2011; Warshaw, Reference Warshaw2019). In the Canadian case, the incumbency advantage appears to be most pronounced at the municipal level. Whereas Lucas (Reference Lucas2019) estimates that incumbency increases the probability of a candidate winning by 30 percentage points municipally, Kendall and Rekkas (Reference Kendall and Rekkas2012) estimate the advantage to be about 10 percentage points for incumbent members of Parliament.Footnote 7 This is in contrast to Warshaw (Reference Warshaw2019) who finds that the incumbent advantage is very similar across “top of the ticket” offices such as mayor, governor and senator in the United States.
The advantages that incumbents have over their challengers during elections (municipal and otherwise) are numerous. First and foremost is name recognition (Kam and Zechmeister, Reference Kam and Zechmeister2013). As Kushner et al. (Reference Kushner, Siegel and Stanwick1997: 543) suggest, when there is no party affiliation on the ballot, as is the case in most Canadian municipal elections, “having a well-known name can be particularly helpful.” Spicer et al. (Reference Spicer, McGregor and Alcantara2017) make a similar argument, noting that without political parties to structure the vote, citizens often use incumbency as a heuristic when casting their ballot, and Taylor and McEleney (Reference Taylor and McEleney2017: 213) write that “the absence of party names on the ballot increases the cost to the voter of acquiring political information, increasing reliance on the name recognition heuristic.”Footnote 8 In addition to name recognition, incumbents also tend to have more extensive local networks than non-incumbents, allowing them to attract more financial and human (volunteer) resources for their re-election bids (Kushner et al., Reference Kushner, Siegel and Stanwick1997; Stanwick, Reference Stanwick2000; Trounstine, Reference Trounstine2008, Reference Trounstine2011). At the municipal level, this often includes important relationships with influential local ratepayer associations, business improvement areas and grassroots organizations. Finally, incumbents can also rely on the resources of municipal office (that is, media and publicity) and tout a multi-year record of successfully holding office and assisting voters with a variety of local problems and initiatives (Kushner et al., Reference Kushner, Siegel and Stanwick2001; see also Mayhew, Reference Mayhew1974). Given that city councillors often want to continue to serve beyond the next election, they are sensitive to the demands of their constituents and engage in the kinds of casework that is rewarded during elections (Trounstine, Reference Trounstine2008). In this regard, incumbents are able to cultivate a personal vote based on their “homestyle” (Ansolabehere et al., Reference Ansolabehere, Snyder and Stewart2000; Fenno, Reference Fenno1978). As Moore et al. (Reference Moore, McGregor and Stephenson2015) note, when combined, these individual incumbent advantages often deter high-quality challengers from entering the race, therefore further contributing to incumbent success (for more on incumbent “scare-off” effects, see Cox and Katz, Reference Cox and Katz1996; Hall and Snyder, Reference Hall and Snyder2015; Sidman, Reference Sidman and Warren2008).
The incumbency advantage has consequences for the health and functioning of representative democracy. Many argue that the high rates of re-election for incumbents pose serious obstacles for women and ethnic minority candidates, both municipally and at other levels (Palmer and Simon, Reference Palmer and Simon2001; but see Spicer et al., Reference Spicer, McGregor and Alcantara2017, for competing evidence). At the very least, the incumbency advantage slows turnover and reduces the ability for marginalized groups to make considerable electoral breakthroughs from one election to the next (Carbert, Reference Carbert2012). Recent evidence reveals that voters tend to tune out of campaigns dominated by incumbents, undermining the quality of participation and engagement with local elections (Moore et al., Reference Moore, McGregor and Stephenson2015). Finally, there is also some evidence to suggest that incumbency is, at least partly, related to declines in legislative responsiveness (King and Gelman, Reference King and Gelman1991).
The extent of the incumbency advantage at the municipal level, coupled with its potentially harmful consequences for democratic practice, raises important questions about voters’ ability to identify and coordinate around a single meaningful candidate who is capable of challenging an incumbent. Are voters in a low-information election able to identify and rally behind the correct candidate to challenge the incumbent? According to scholars such as Taylor and McEleney (Reference Taylor and McEleney2017: 210), the answer is likely no. As they write, “crowded nonpartisan races are low-information environments in which candidates, donors, and voters cannot assess challenger quality.” In this article, we empirically test this assumption and consider the factors that are associated with the ability to identify leading challengers.
Voter Predictions and Perceptions of Competitiveness
Research shows that electors do not always vote for their preferred candidate or political party. As Blais and Bodet (Reference Blais and Bodet2006) note, the fact that some voters support the expected winner of the election (bandwagon effect; Bartels, Reference Bartels1988), while others choose the weakest of the options (underdog effect; Simon, Reference Simon1954), and still others vote strategicallyFootnote 9 to defeat their most disliked option (strategic voting; Downs, Reference Downs1957; Cox, Reference Cox1997; Blais et al., Reference Blais, Gidengil, Fournier, Nevitte and Hicks2008; Alvarez and Nagler, Reference Alvarez and Nagler2000; Blais and Nadeau, Reference Blais and Nadeau1996; McGregor, Reference McGregor2012) suggests that voters are indeed able to identify the competitive position of various political actors when making their ballot decision. At the very least, voters believe they can make these predictions, as strategic, underdog and bandwagon voting would be impossible otherwise.
There is good reason to believe that voters are generally equipped to predict election winners, at least at the aggregate level. The idea of “wisdom of the crowds” or “citizen forecasting” (Murr, Reference Murr2011) has found support in Canada (Temporao et al., Reference Temporão, Dufresne, Savoie and van der Linden2019), as well as a variety of other settings, including the UK (Lewis-Beck and Stegmaier, Reference Lewis-Beck and Stegmaier2011), Germany (Leiter et al., Reference Leiter, Murr, Ramirez and Stegmaier2018) and the United States (Lewis-Beck and Tien, Reference Lewis-Beck and Tien1999). This research shows that citizens, at least in the aggregate, can predict election outcomes with a fair degree of accuracy.
So how do individual voters form perceptions of candidate or party competitiveness, especially when the literature has consistently shown that voters have generally low levels of political knowledge and interest (Lazarsfeld et al., Reference Lazarsfeld, Berelson and Gaudet1944; Campbell et al., Reference Campbell, Converse, Miller and Stokes1960; Delli Carpini and Keeter, Reference Carpini and Keeter1997; Pruysers and Blais, Reference Pruysers and Blais2014)? Voters adapt to their context and make use of what information or informational shortcuts are available (Alvarez et al., Reference Alvarez, Hall and Levin2018). Using these heuristic cues leads to what Popkin (Reference Popkin1991) termed “low-information rationality” and is what allows voters to make reasonable decisions without being completely informed. In partisan elections, party labels are the most obvious cue that voters look for when determining who is the most competitive candidate in their district. In nonpartisan elections, by contrast, voters tend to rely overwhelmingly on incumbency (Hajnal et al., Reference Hajnal, Lewis and Louch2002; Moore et al., Reference Moore, McGregor and Stephenson2015).
Blais and Bodet (Reference Blais and Bodet2006) find that a number of factors are related to voters’ ability to predict party and candidate competitiveness. This includes objective information like incumbency and the availability of polling data, as well as more subjective information like partisan predispositions. What about elections where this information may not be readily available? In the absence of party labels or polling information, voters can look to other sources of information (for example, talking with other electors or estimating rally attendance) to assess the relative competitiveness of candidates (Clough, Reference Clough2007).
Voters might, for example, use candidate race or sex to infer (unofficial) party affiliation and extrapolate which candidates are the most competitive in the district. In the United States, women and Black candidates are stereotyped as more liberal and more Democratic than their white counterparts (Huddy and Terkildsen, Reference Huddy and Terkildsen1993; McDermott, Reference McDermott1998). Voters also rely on name recognition to infer candidate quality, equating recognition with viability (Kam and Zechmeister, Reference Kam and Zechmeister2013). Recent research also suggests that voters also rely on candidate appearance (that is, attractiveness) to inform their voting decisions, as candidates who are more attractive or capable-looking are significantly more likely to win elections (Antonakis and Dalgas, Reference Antonakis and Dalgas2009; Banducci et al., Reference Banducci, Karp, Thrasher and Rallings2008). Todorov et al. (Reference Todorov, Mandisodza, Goren and Hall2005), for example, find that electors can correctly predict winning candidates for the US Senate more than 70 per cent of the time based on a simple black-and-white photograph of the top two candidates. Even in the absence of party and polling data (like we might find at the municipal level), voters are able to form perceptions of candidate competitiveness by relying on more readily available sources of information.
In a rare study of the ability of voters to identify riding candidate competitiveness, Blais and Turgeon (Reference Blais and Turgeon2004) find that less than half (47%) of respondents are able to correctly identify the party/candidate that would finish third in their local constituency. A number of factors are identified as being associated with this ability. The first is timing. Voters are somewhat better able to predict competitiveness as the election progresses, suggesting that there is some level of learning happening during the campaign. The second is the closeness of the race. The more competitive the race, the more difficult it is for voters to disentangle who will finish first, second, third—and so on—in the riding. The third is political knowledge. Unsurprisingly, more knowledgeable voters are able to better predict competitiveness and correctly identify the weakest candidate. Finally, there is party identification. Voters who identify with a losing party/candidate tend to overestimate their chances of winning (see also Blais, Reference Blais2002; Uhlaner and Grofman, Reference Uhlaner and Grofman1986).
We build on the work of Blais and Turgeon (Reference Blais and Turgeon2004) in this article and consider the extent to which voters are able to correctly predict the top two municipal candidates in their ward.
Expectations
In this section, we detail our expectations with regard to the factors that are associated with correctly predicting the top two competitive candidates in a ward. In doing so, we consider two categories of hypotheses: factors related to the individual voter and to the ward itself.
At the individual level, the first factor relates to an individual's level of connectedness to the local community, as measured by the length of time lived in the city. Second, those who have devoted more attention to following the local election campaign are expected to be better able to identify the first-place and second-place candidates. These individuals should have more information upon which to base their assessment of competitive candidates, whether derived from living in the ward for many years or through learning about the candidates due to higher levels of campaign attentiveness.
H1 Electors who have lived in the city for a long period of time will be more likely to correctly identify the first-place and second-place candidates than will those electors who have lived in the city for less time.
H2 Electors who paid more attention to the local election campaign will be more likely to correctly identify the first-place and second-place candidates than will those electors who paid less attention to their local election.
The degree to which electors are mobilized into or engaged with (electoral) politics in general should also matter. Those who are more informed about politics, for instance, should be better equipped to predict which municipal candidates will perform the best on election day (Blais and Turgeon, Reference Blais and Turgeon2004; Miller et al., Reference Miller, Wang, Kulkarni, Poor and Osherson2012).
H3 Electors who are more knowledgeable about politics in general will be more likely to correctly identify the first-place and second-place candidates than those electors who are less knowledgeable.
There is also good reason to expect that the campaign itself should be related to the ability of electors to correctly identify competitive candidates; individuals should become better equipped to identify leading candidates as the campaign progresses. Election campaigns provide voters with the opportunity to become better informed and learn about the various candidates through exposure to media reports, polling data and other campaign communications. The CMES was in the field for nearly a month prior to election day. Respondents who completed the survey later in the campaign should have more information on which to base their judgments regarding candidate competitiveness than those who responded to the survey earlier in the campaign period. Such an effect has been observed with respect to federal elections in Canada (Blais and Turgeon Reference Blais and Turgeon2004; Temporao et al., Reference Temporão, Dufresne, Savoie and van der Linden2019).
H4 Respondents surveyed later in the campaign will be more likely to correctly identify the first-place and second-place candidates than respondents who completed the survey earlier in the campaign.
The nature of the election in the ward should also influence the ability of electors to identify leading candidates. The greater the number of candidates contesting an election, the more difficult it should be for electors to identify leading candidates. Every additional candidate running in the ward raises the informational costs for electors and should therefore weaken their ability to correctly identify the first-place and second-place finishers.
H5 Electors in wards with increasing numbers of candidates contesting the election will be less likely to correctly identify the first-place and second-place candidates than electors in wards with fewer candidates.
The final ward-level explanatory factor we consider is candidate spending. Spending in local elections is generally focussed on advertising (lawn signs, brochures, etc.). It is limited by legislation, and candidates must file detailed spending reports after the election. Campaign spending is positively associated with campaign intensity and the ability for candidates to produce information leaflets, campaign signs and other forms of campaign advertising (Cross and Young, Reference Cross and Young2011). We expect that spending will be interpreted by voters as a sign of competitiveness.
We consider the effect of candidate spending with two measures: a comparison of campaign spending between the first-place and second-place candidates and between the second-place and third-place candidates. As the spending gap between the first- and second-place candidate grows, we expect that it will be easier for electors to identify the first- place candidate and harder for electors to correctly identify the second-place candidate. Specifically, the lived experience of the campaign and the differential campaign intensity of the candidates should contribute to the impression of the relative electoral competitiveness. When the gap in campaign activity/spending grows between the first-place and second-place candidates, electors should be better able to determine who the first-place candidate is likely to be. Similarly, as this gap grows, it may become more difficult for voters to differentiate the next most-competitive candidate among a field of candidates who are all being outspent by the first-place candidate.
H6a As the spending gap between the first-place and second-place candidates grows, electors will be more likely to correctly identify the first-place candidate and less likely to identify the second-place candidate.
Following the same logic, the second iteration is the spending gap between the second-place and third-place candidates. In this case, when the gap is small (that is, the race is more competitive), it will be harder to identify the second-place candidate running in the ward election, but as that gap in spending increases between the second-place and third-place candidates, electors should be better able to identify the second-place candidate.
H6b As the spending gap between the second-place and third-place candidates grows, electors will be more likely to correctly identify the second-place candidate.
Data and Methodology
The analysis below is based upon survey data from the CMES, as well as publicly available contextual data in Mississauga's ward elections. The CMES dataset includes pre- and post-election survey results from respondents in eight cities.Footnote 10 The surveys included a series of questions standard in election studies, though many questions are meant to address the unique nature of local politics. Respondents in Mississauga were asked the following question, meant to tap into their perceptions of the competitive dynamics of their local ward race: “Which TWO of the following candidates do you think have the best chance of winning the COUNCIL election in your ward?”Footnote 11 Note that respondents chose from a list of all municipal candidates contesting the election in their ward.
Since we know which ward respondents live in, as well as the eventual outcome of the council elections, based on official election results (see City of Mississauga, 2018), we are able to determine if survey respondents were able to correctly identify the two top finishing candidates in their ward. Using this information, we have created two dummy variables that categorize respondents according to whether they identified the incumbent candidate (which, in this instance, is the same as identifying the winner) and the second-place candidate.
Expectations are tested through a series of three multivariate models. We consider the correlates of being able to identify the winning and second-place candidates in the first two models, respectively. The third model is again focussed upon identifying the second-place candidate but is limited to those respondents who also identified the winner. This final model provides insight into the ability of those respondents who are aware of the fact that the incumbent has a strong chance of winning to consistently identify the strongest challenger (meaning that they could coalesce behind a single challenger if they wished to oust the incumbent).
Due to the stratified nature of our data, we employ a multilevel approach. Voters are nested into wards, each of which has a unique dynamic that can affect competitive expectations. At the same time, we expect that individual-level characteristics might affect one's ability to identify the most competitive candidates in a race. A multilevel approach allows us to consider both individual-level and ward-level effects appropriately.
The explanatory variables are based upon a combination of survey data and actual election results. The individual-level characteristics meant to test H1–H4 are based upon survey data. These include an indicator of the length of time lived in a city (respondents are categorized as long-term residents or not), a measure of attentiveness to the city council campaign and an index of general political knowledge (based upon questions from the three levels of Canadian government). Finally, the CMES notes the date of completion of the pre-election questionnaire, and this information is employed to test H4 (surveys were fielded from September 25 to October 22).Footnote 12 We include a series of individual-level controls for age, gender and university education.
The remaining expectations (H5–H6) are related to ward-level characteristics, and these indicators are based upon publicly available data about the election. First, wards are coded according to the number of candidates contesting the race. Next, candidate spending data is considered through two variables, representing the difference in spending of the first- and second-place candidates, and second- and third-place candidates, respectively. Spending data are available via financial statements submitted by all candidates following the election (City of Mississauga, 2019).Footnote 13 These variables (in 1,000s of dollars) reflect the absolute difference in spending between the top three candidates.Footnote 14 Recall that we expect that electors use information on candidate spending to make inferences about the relative competitiveness of each contestant. We also include two additional control variables to account for the actual competitive circumstances in the election. Similar to the spending difference variables, we include two “distance” indicators that represent differences in vote share between the first- and second-place finishers, and second- and third-place. These variables are meant to account for ward-level variation in the actual competitiveness of the election outcome, allowing us to properly isolate the effects of spending and the number of candidates upon competitive evaluations. Note that summary statistics for all explanatory variables considered here are found in Appendix II.
As a primary purpose of this study is to determine if electors are able to identify the strongest challenger to an incumbent and what factors are associated with being able to correctly do so, our analysis is limited to wards in which an incumbent candidate was seeking re-election in 2018. This applies to 9 of Mississauga's 11 wards. Though we provide descriptive information on residents in the two wards without an incumbent, the multivariate analysis excludes these respondents.
Results
Prior to testing our hypotheses, we wish to provide some relevant descriptive information. Two-thirds (65.9%) of respondents in wards where incumbents were present (N = 915) were able to correctly identify the incumbent (and eventual winner) as one of the two most competitive candidates in the ward, but only 29.8 per cent correctly listed the second-place finisher. A total of 27.5 per cent of respondents correctly identified the top two challengers, meaning that the vast majority (92.3%) of individuals who identified the second-place candidate also named the incumbent. These results point to a population largely unable to identify (and potentially coalesce) around a single challenger.Footnote 15
In those wards where an incumbent was present, the average number of competitors the incumbent faced was 6.9 (this figure ranged from 2 to 10), and all 9 incumbents were re-elected. The average percentage of the vote won by the nine incumbents was 70 per cent, and some won with more than 90 per cent of the vote. While rare, there were wards where the incumbent was not particularly safe. In Ward 6, for example, incumbent Ron Starr won with only 36 per cent of the vote (less than 3% ahead of the runner-up).Footnote 16 Overall, however, Mississauga's council elections played out as most elections in Canadian municipalities do—with sitting councillors outperforming all challengers.Footnote 17
Incumbents also tended to greatly outspend their opponents: the average spending gap between the winner/incumbent and second-place candidate was $22,943 (the mean spending level of incumbents was $39,331). Only in Ward 4 was an incumbent outspent: second-place finisher Grant Gorchynski spent $9,409 more than incumbent John Kovac. Spending by the second- and third-place finishers is less straightforward. On average, the second-place candidate spent $2,761 more than the third-place finisher. In only two wards (4 and 11), however, did the second-place candidate actually outspend the third. In three wards (3, 5, and 6), the opposite was true, while in another four (2, 8, 9 and 10), there was no difference in the spending of the second- and third-place finishers (this lack of a difference is due to the fact that numerous candidates reported spending no money on their campaigns in their financial statements). These results point to a weak relationship between differences in spending and election results for the second- and third-place finishers. If candidate spending somehow influences competitive perceptions, there seems significant potential here for this variable to lead electors astray.
We turn, now, to test our expectations using multivariate analysis of CMES data. As noted above, we run a series of multilevel logistic regression models, where the outcome variables represent correctly identifying the (1) winning candidate/incumbent, (2) the second-place candidate and (3) the second-place candidate, among those respondents who also identified the winner. All variables are scaled to range from 0 (minimum) to 1 (maximum). The results of this analysis are found in Table 1. Entries in the table show marginal odds ratios and standard errors (in parentheses).
Table 1 Competitive Expectations: Multilevel Logistic Results
Entries report odds ratios and standard errors (in parentheses).
* p < .10; ** p < .05; *** p < .01.
Perhaps the most striking feature of the results in Table 1 is that different types of factors appear to be associated with being able to identify the winner and the challenger. In Model A, we see that two individual-level characteristics—being a resident of the city for 10 years or more and attentiveness to the campaign—tend to increase the likelihood of naming the incumbent as competitive. However, neither factor is associated with being able to identify the strongest challenger. In fact, none of the individual-level variables are associated with correctly identifying the challenger (Models B and C).
In contrast, only ward-level factors are statistically significant in the latter two models. Results suggest that the number of candidates decreases the likelihood of correctly identifying the challenger (or the top two finishers) by about a third. As the difference in spending between the first- and second-place finisher increases, it becomes more difficult to identify the challenger. However, contrary to expectations, as the gap in spending between the second- and third-place candidates increases, the probability of correctly naming the second-place candidate decreases. Such a finding is likely due to the fact that, at least among the second- and third-place finishers, the relationship between spending and election results is not only weak but also negative.
With respect to our formal expectations, we see mixed results. We find support for H1 (length of tenure in city) and H2 (attention to campaign) but only for the winner/incumbent model. There is no support for either H3 (general knowledge) or H4 (date of interview). The latter null finding is particularly noteworthy, given that previous work has found this factor to be a significant predictor of correctly understanding the competitive circumstances of federal election races (Blais and Turgeon, Reference Blais and Turgeon2004; Temparao et al., Reference Temporão, Dufresne, Savoie and van der Linden2019). This could suggest that campaigns have different effects upon learning in municipal elections than they do at other orders of government.
The ward-level expectations see similarly disparate levels of support. Though results suggest that a high number of candidates (H5) decreases the ability of electors to identify the second-place finisher, it has no effect upon the incumbent model; this latter null finding is no doubt a result of the name recognition that comes with being a sitting city councillor. Finally, the spending difference variables exhibit a complex and unanticipated relationship with the competitive perceptions of electors. Results indicate that neither variable has an apparent effect upon identifying the winning candidate (again, likely due to the strong name recognition of this candidate). Differences in spending are, however, significant in both Models B and C (though not always in the manner predicted by H6a and H6b).
In terms of the control variables, two results stand out. At the individual level, women are less likely than men to be able to identify the challenger (but not the winner). Such a finding fits with a vast literature (Delli Carpini and Keeter, Reference Carpini and Keeter1997; Pruysers and Blais, Reference Pruysers and Blais2014) that suggests that women have comparatively low levels of political knowledge. At the ward level, we find that the eventual competitiveness of the race, with respect to the gap between the second- and third-place finishers, does indeed have an effect upon the ability of electors to identify the challenger. Voters in wards where the second-place candidate finishes well ahead of the third are indeed comparatively likely to correctly predict the election outcome.
To further flesh out the results in Table 1, we present Figures 1 through 3, which show the estimated effect of the three ward-level factors upon being able to correctly identify the second-place finisher, across the range of each explanatory variable. Post-estimation results are based upon Model C because it is this subset of the population that we are theoretically focussed on: those respondents who are already able to identify the incumbent.
Figures 1, 2 and 3 show clearly that all three ward-level variables are negatively associated with being able to correctly identify the second-place candidate. As the number of candidates increases from 3 to 11, the likelihood of correctly identifying the challenger drops from 55.0 per cent to 10.7 per cent. The difference in spending between the top two candidates has a similar effect, with a range from 68.6 per cent to 27.4 per cent. The variable with the greatest effect, however, is the difference in spending between the second- and third-place candidates: the predicted rate of identifying the second-place candidate drops from 72.9 per cent to 6.1 per cent across the range of this variable.Footnote 18
Figure 1 Number of Candidates and Ability to Correctly Identify Challenger
Figure 2 Differences in Spending between First-Place and Second-Place Candidates and Ability to Correctly Identify Challenger
Figure 3 Differences in Spending between Second-Place and Third-Place Candidates and Ability to Correctly Identify Challenger
As expected, results suggest that an increase in the number of candidates and an increase in spending gap between the top two candidates decrease the likelihood that electors are able to identify the second-place finisher. Assuming that incumbents have significant advantages in terms of fundraising, the spending result in Figure 2, in particular, helps to account for incumbent dominance. The results in Figure 3, however, only add to the puzzle of why it is that the second spending variable works in the direction opposite to that which was anticipated. As the second-place candidate outspends the third-place finisher, electors actually become less likely to correctly identify the challenger, a finding that reflects the apparently weak relationship between spending and election standing for these candidates.
These findings could conceivably reflect unique circumstances in the 2018 Mississauga elections, and we can think of no theoretical reason why this variable has such a strong relationship in the direction that it does. Regardless of the reason for this finding, however, the data suggest that spending by a challenger is far from a guarantee of being competitive. If voters are immune to the effects of spending by challengers, this may once again help to explain the dominant position of incumbents.
Discussion and Conclusion
As a growing body of research suggests, the power of incumbency in city council elections is undeniable, and re-election rates of sitting councillors are staggeringly high (de Benedictis-Kessner, Reference de Benedictis-Kessner2018; Lucas, Reference Lucas2019; Trounstine, Reference Trounstine2011; Warshaw, Reference Warshaw2019). The City of Mississauga represents an excellent case of incumbent advantage, with over 92 per cent success rates for sitting incumbents in the last three elections. One of the many possible explanations for this phenomenon is the difficulty encountered by voters in identifying, and subsequently unifying behind, a single candidate with the best chances of unseating, or at least effectively challenging, an incumbent. In the absence of party systems, the strongest challenger to an incumbent can be difficult to identify, thus dividing the vote of those who might want to oust a sitting councillor. CMES data reveal that less than a third of Mississaugans who live in a ward where an incumbent sought re-election in 2018 were able to correctly identify the top two finishers. Absent an incumbent, electors were better able to name the top two challengers. Rallying around a credible challenger is therefore a particularly difficult task for voters in wards dominated by an incumbent.
This study has identified a series of individual-level and contextual characteristics that are related to the ability of electors to identify the most competitive candidates in a ward-level city council race. We find evidence consistent with the interpretation that both types of factors matter. Individuals who are long-term residents of a city and who are attentive to council campaigns are comparatively more likely to be able to name the incumbent as one of the two most likely candidates for victory. It is ward-level characteristics, however, that may most affect the ability of electors to correctly identify the second-place finisher (or the strongest challenger). Both candidate spending and the number of candidates in a race have a relationship with the ability of electors to accurately predict election outcomes in their wards.
High rates of incumbent success have implications for accountability, the representation of women, visible minorities and other marginalized groups, and for voter turnout and engagement. The current study adds to our understanding of incumbency advantage by exploring one of its many potential sources but does little to suggest ways to reduce it. Candidate spending appears to have mixed effects upon the ability of voters to identify (and conceivably coalesce behind) a single challenger, so making changes to spending limits might have little effect. CMES data do suggest that having a lower number of candidates may help in this respect, but it is certainly unreasonable for municipalities to dictate the number of candidates that must contest each council race.
Further, while Mississauga exhibits many useful features as a case for consideration, the findings of individual- and ward-level effects should be extended to other empirical contexts—including both those that might be larger in size (for example, Toronto or Vancouver) and smaller in size (for example, Moncton or Brandon). Variations in media and information environment, economic size and the role of personal discussion networks may serve as important contrasts to the electoral context observed in Mississauga. Future research may also consider more factors related to the candidates, in addition to the voter and ward characteristics examined here. While this study included campaign spending behaviour, other factors, including candidate campaign tactics and voter contact strategies, may also influence candidates’ general visibility—and therefore voters’ ability to identify their competitive position. A greater emphasis on candidates is therefore warranted.
Regardless of the absence of any obvious policy recommendations that might be aimed at addressing the incumbency advantage, this article adds significantly to our understanding of the individual-level and contextual factors that affect the ability of electors to identify both the winner and second-place finisher in low-information, nonpartisan local elections. The article demonstrates the relative inability of voters to correctly identify likely challengers to the incumbent, shows that there are a few individual and contextual factors that influence this ability and generally paints a picture of an enduring incumbent advantage in local elections.
There is one other ray of sunshine, however, worth highlighting. As noted following the regression models, the actual closeness of the election (as measured by the eventual difference in vote share of the second- and third-place candidates) was negatively related to the ability of electors to name the second-place candidate. Electors in wards where the runner-up performed much better than the third-place finisher did therefore appear to be able to pick up on some type of environmental factors to inform their competitive assessments. Voters are therefore able to pick up on some kind of indicators to improve their competitive understanding of a contest. Future research can investigate what these other types of cues are that voters use to inform their competitive perceptions, with the goal of better understanding one of the underlying sources of the of incumbency advantage.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0008423920000384.
