If turnout is an indication of public engagement in an election, then most Canadians are distinctly disengaged from school board politics. When school board elections are held concurrently with competitive and high-turnout municipal races—as in Toronto in 2014 and Calgary in 2017, for example, both of whose municipal elections had turnout rates in the mid-50-per-cent range—turnout rates in school board elections are ten or more points below those of the concurrent mayoral race, languishing in the mid-40s (City of Toronto, 2014; City of Calgary, 2017). When school board elections are not held concurrently with municipal elections, as in Quebec, turnout rates regularly drop into the single digits (Ministère de l’Éducation et de l'Enseignement supérieur, 2014). Across Canada, voters appear to place little importance on school board elections.
Survey data tell much the same story. Results from the Calgary portion of the Canadian Municipal Election Study (CMES) reveal that Calgarians have extremely low levels of interest in school board politics. Respondents’ average reported level of interest in municipal politics was 6.8 out of 10, compared to 7.4 for provincial politics and 7.6 for federal politics—important and statistically significant differences, but all in the same general ballpark. For school board politics, however, the average level of interest was an astonishingly low 3.9 out of 10.Footnote 1 School board politics is simply not a priority for most electors.
These low levels of turnout and interest in school board elections are especially striking in light of emerging research on the composition of voting populations in low-turnout elections. Studies of school board elections in the United States, where turnout is similarly low, have noted the potential influence of well-organized groups in low-turnout elections that may have distinctly different interests from the wider population. While the effects of this phenomenon on public policy remain a matter of debate, recent research suggests that they could well be substantial: in US school board elections, school districts with off-cycle elections are associated with 3 per cent higher pay rates for teachers than those with on-cycle elections, possibly reflecting the greater influence of teacher unions in low-turnout elections (Anzia, Reference Anzia2011; Moe, Reference Moe and Howell2005). There is every reason to believe that low turnout could have similarly important policy implications in Canada.
In this research note, we undertake what we believe to be the first individual-level study of school board elections ever conducted in Canada. Using CMES survey data from the city of Calgary, we identify several factors associated with participation in school board elections. We conduct two analyses (a logistic regression model of ballot roll-off and a two-stage Heckman selection model) to account for the fact that these contests are held concurrently with mayoral and council elections, and we identify factors linked to turnout in the school board election in particular. CMES data reveal that gender, university education, left-wing ideology, Conservative partisanship, parental status and the competitiveness of the school board race are associated with participation. Some of these variables relate to distinctive features of Calgary's 2017 school board election; others, we suggest, may apply to school board elections more generally. While our findings are necessarily preliminary, they suggest new avenues of research on voter turnout in Canadian school board and other local elections and have implications for the study of down-ballot elections more generally.
Background
Across Canada, school boards are under threat. School board trustees often maintain that they form the heart of a democratically accountable education system, serving as the vital link between local needs and a distant provincial education bureaucracy (Sheppard et al., Reference Sheppard, Galway and Brown2013). However, the legitimacy of—and even the need for—school boards has been questioned in recent years, as authority over policy areas such as curriculum, teacher salaries and budgets has inexorably migrated from the local to the provincial scale. High-profile provincial interventions in big-city school boards in Canada—Vancouver, Ottawa, Toronto, Calgary—bespeak an increasing willingness on the part of the provinces to intervene directly in the work of democratically elected trustees (Junker, Reference Junker2017; Mackie, Reference Mackie2002; Sherlock, Reference Sherlock2016). More drastically, in 2018 the government of Nova Scotia announced the elimination of the province's seven elected English-language school boards, which are to be replaced by a single advisory council made up of provincial appointees (Alphonso, Reference Alphonso2018).Footnote 2 It is an open question as to whether these interventions would have occurred if local communities were more engaged in school board politics. Provincial governments may be more hesitant to involve themselves in the work of elected officials who can make a claim to a broad and deep democratic mandate.
Calgary's most recent public school board election, which took place on October 16, 2017, provides a good place to begin to explore school board politics at the level of individual voters. The election was noteworthy for its unusually high profile. Before the election, controversy about school board spending decisions and bus schedule policies prompted a number of incumbent retirements as well as the emergence of a conservative slate called “Students Count.” The slate was endorsed by Jason Kenney, a Conservative politician, and the election involved a degree of profile and expense—extensive lawn signs, substantial media coverage and even expensive tactics such as robocalls—that is not typical of school board races (Braid, Reference Braid2017; Ferguson, Reference Ferguson2017). Calgary's public school board election provides a usefully “extreme” case to explore, allowing us to understand who participates—and who declines to participate—in school board elections when the profile of those elections is relatively high.Footnote 3
An investigation of school board elections is also of interest because it contributes to the literature on second-order elections, of which there is comparatively little in Canada. The theory of second-order elections was originally developed to apply to elections to the European Parliament (Reif and Schmitt, Reference Reif and Schmitt1980) but has been expanded to also apply to elections other than those held at the national level (see Golder et al., Reference Golder, Lago, Blais, Gidengil and Gschwend2017). The label of “second order” is meant to suggest that relatively little is at stake in these elections. As a result, these tend to be associated with low levels of interest and turnout (Marsh, Reference Marsh1998; Schmitt, Reference Schmitt2005). At least by these criteria, school board elections can be considered second order; as noted above, turnout and interest in these elections was significantly lower than for federal, provincial or other municipal offices. The analysis below allows us to consider how the low-interest, low-turnout nature of these elections affects the type of individuals who choose to vote (or not).
Turnout and Ballot Roll-Off
The question of why individuals choose to vote (or not) is one of the great puzzles of political science. Studies of voter turnout number in the hundreds, and scholars have identified various types of factors to explain rates of elector participation. Among the variables studied are some institutional features of elections, such as compulsory voting, concurrent elections and voter registration requirements (Cancela and Geys, Reference Cancela and Geys2016; Stockemer, Reference Stockemer2017). Contextual variables, such as campaign expenditures and the competitiveness of election, are also robustly and positively linked to turnout (Blais, Reference Blais2006; Cancela and Geys, Reference Cancela and Geys2016; but see Stockemer, Reference Stockemer2017). At the individual level, older age, higher education, higher knowledge and interest, partisanship and believing voting to be a duty, rather than a choice, are all robustly associated with turnout (Blais, Reference Blais, Dalton and Klingemann2007; Geys, Reference Geys2006; Smets and van Ham, Reference Smets and van Ham2013).
While studies of turnout have traditionally focused upon national elections, there is an emerging interest in participation at other levels of government, as well as across levels of government and election types, on the grounds that “we cannot simply assume a general equivalence of turnout determinants irrespective of the type of election” (Cancela and Geys, Reference Cancela and Geys2016: 265). Though research suggests that many of the factors shaping turnout at higher levels are also active at the local scale (Hamilton, Reference Hamilton1971), there are variables that have a unique effect at this level of government. Contextual factors, such as population size (Breux et al., Reference Breux, Couture and Koop2017) and the nonpartisan nature of local contests (Schaffner et al., Reference Schaffner, Streb and Wright2001), have been shown to affect turnout in a manner unique to the local level. Individual-level characteristics, such as homeownership and the length of time that one has lived in a municipality, are both known to increase municipal turnout (McGregor and Spicer, Reference McGregor and Spicer2016; DiPasquale and Glaeser, Reference DiPasquale and Glaeser1999). Thus while local contests have much in common with other elections, there are nevertheless a series of factors that affect turnout at the local level exclusively.
Researchers have also considered patterns of turnout in elections where multiple races are contested simultaneously. This research is of particular relevance to the current study, given that one of the unique features of Canadian local elections is that this is the only instance where electors have the opportunity to vote in multiple elections simultaneously. In such a scenario, it is possible to cast a ballot for one position but abstain from voting in others. Most research on ballot roll-off—voting for a “top-of-ballot” position, such as mayor, while refraining from voting in concurrent “down-ballot” contests, such as for council or school board—originates in the United States, where concurrent elections are the norm for national, state, municipal and judicial elections (for examples of each, see Walker, Reference Walker1966; Nichols, Reference Nichols1998; Hall, Reference Hall2007; Bullock and Dunn, Reference Bullock and Dunn1996). This work suggests that variables such as partisanship, campaign spending, information and policy importance are important for understanding roll-off (Hall and Bonneau, Reference Hall and Bonneau2008; Lassen, Reference Lassen2005; Streb and Frederick, Reference Streb and Frederick2011). Patterns of participation in down-ballot elections have also been shown to increase when electors have high levels of information (Lassen, Reference Lassen2005; Matsusaska, Reference Matsusaka1995) and when the “stakes are high,” meaning that voters care about the consequences of the outcome (Andersen et al., Reference Andersen, Fiva and Natvik2014).
Thus far, there has been only one individual-level examination of ballot roll-off conducted in Canada (McGregor, Reference McGregor2018), and to our knowledge, no studies of turnout in Canadian school board elections have ever been undertaken. Research on this topic is important not only because participation in school board elections—and the policy consequences of such participation—is a topic of academic debate but also because it will allow us to better understand the factors that shape turnout when institutional and sociodemographic contextual factors are shared across two concurrent elections. Doing so speaks to the research agenda, articulated by Cancela and Geys (Reference Cancela and Geys2016), of identifying the determinants of turnout for specific levels of government and elected positions.
Our goal, then, is to explore the factors that are distinctively related to participation in school board elections. Since we have little research to guide us on this specific topic, we draw on the broader literature of down-ballot elections and roll-off. Research on Canadian municipal roll-off indicates that demographic variables, including age, income and ethnicity, play a role in voters’ decisions to participate in down-ballot council races (McGregor, Reference McGregor2018), and we therefore begin by exploring the sociodemographic factors that may be linked to school board turnout. These include the variables in McGregor's analysis listed above, along with two others: university education, which is connected to turnout in general (Geys, Reference Geys2006) and may be particularly important for participation in elections directly related to education; and gender, on the assumption that potentially higher levels of women's participation in school board politics (Tolley, Reference Tolley2011) may also be reflected in higher participation in elections.
A second potentially important set of factors is partisanship and ideology. Research on municipal and down-ballot races in the United States has found higher turnout rates in partisan than nonpartisan elections, as party cues make down-ballot decisions easier to make for individuals with partisan ties (Schaffner et al., Reference Schaffner, Streb and Wright2001; Schaffner and Streb, Reference Schaffner and Streb2002). While Calgary's 2017 school board election was officially nonpartisan, it involved a slate of conservative candidates called Students Count that was endorsed by a known Conservative politician (Kenney). We might therefore expect to find that partisanship is linked to participation in the Calgary school board election. We also include ideology in our analysis for two reasons: to account for the possibility that public education may be more salient for those on the left than those on the right, and because it serves as a control in our test of partisanship, rather than ideology, on school board participation.
Finally, the distinctive policy focus of school board elections prompts us to examine the role of parenthood on participation. Simply having children may provoke a connection to school board politics in what we might think of as a “sociotropic” fashion, as parents conceive of themselves as members of the community of voters for whom education policy is most relevant. Alternatively, having children presently in school may stimulate participation by virtue of a more immediate and “egocentric” connection to the school system.
Data and Methods
Our analysis is based on survey data collected from eligible voters during the 2017 Calgary municipal and school board elections.Footnote 4 The data were collected as part of the Canadian Municipal Election Study, a larger comparative study of elections in eight Canadian cities. Respondents completed a campaign-period survey as well as a post-election survey.Footnote 5 In Calgary, the CMES questionnaire includes questions on turnout in the mayoral, council and school board elections, as well as those related to our sociodemographic and attitudinal explanatory variables of interest; a full list of questions is found in online Appendix 1. A total of 1,145 respondents provided answers to the questions required for our analysis here.Footnote 6
We undertake two analyses to identify the correlates of school board turnout. The first uses a binary measure of roll-off as an outcome variable: a value of 1 indicates that the respondent voted in both the mayoral and the school board election, and 0 indicates that the respondent “rolled off,” voting in the mayoral election but not in the school board contest. We also analyze roll-off from the city council to the school board ballot. These analyses allow us to identify the factors that are distinctive to participation in school trustee elections, rather than to turnout in local elections more generally. All told, 18.6 per cent of respondents in our sample rolled off from the mayoral to the school board ballot, and 15.5 per cent rolled off from council to school board.Footnote 7
Since turnout in Calgary's 2017 mayoral race (58%) was very high, both in comparison to other municipalities (Edmonton's election on the same day saw a turnout of just 31.5%) and to Calgary's past mayoral elections (39.4% in 2013), our examination of ballot roll-off is particularly well suited to this election. In general, elections with high rates of turnout at the top of the ballot tend have high rates of roll-off to down-ballot elections. The assumption here is that a high-profile mayoral race will attract voters who might not otherwise participate. Compared to those who would vote regardless of the profile of the mayoral race, these less frequent voters are relatively unlikely to vote in down-ballot races. Calgary's 2017 roll-off rates are consistent with this argument. The aggregate roll-off rates from mayoral to school board elections in the election were higher (28.8%) than in Edmonton (14.3%) or than in Calgary in 2013 (20.9%) (City of Calgary, 2013, 2017; City of Edmonton, 2017). All else equal, higher roll-off rates create more statistical leverage with which to isolate the correlates of ballot completion, making this a good case in which to study participation in school board elections.Footnote 8
Analyzing roll-off rather than school board turnout allows us to disentangle the correlates of school board turnout from mayoral or council turnout; since every respondent in our sample who voted in school board elections also voted in the mayoral and council elections, a more straightforward analysis of school board turnout would be confounded by the factors that influence municipal election participation more generally. However, this approach comes at the cost of discarding those respondents who did not participate in the municipal election at all. We thus supplement the first analysis with a Heckman selection model, which is a two-stage modelling approach that explicitly models the process by which respondents choose to participate in the municipal election and then choose to complete their ballot by participating in the school board race (Streb and Frederick, Reference Streb and Frederick2011). Our first-stage model captures a suite of sociodemographic and attitudinal variables known to be associated with municipal turnout. Our second-stage model is meant to identify those factors associated with school board turnout, in particular, and it includes the same explanatory variables as our roll-off analysis. This approach allows us to test the correlates of school board participation while explicitly modelling the process by which school board voters are selected from among those who are voting in the top-of-ballot mayoral race.
In keeping with our discussion above, we focus on three sets of independent variables when identifying the correlates of school board turnout: sociodemographic characteristics, partisan and ideology variables, and parental status (including both having a child and having a child in public school). To account for the fact that turnout in general has been consistently linked to the closeness of a race (Cancela and Geys, Reference Cancela and Geys2016) and that roll-off rates are thought to decrease as the competitiveness of down-ballot races increases (McGregor, Reference McGregor2018), we also include a measure of competitiveness of school board races. This variable is based upon the margin of victory in each school board district, comparing the winning candidate to the second-place finisher.
Results
The results of our first analyses, summarized in Figure 1, capture the results of two logistic regression models using roll-off from mayoral to school board and council to school board elections as our respective dependent variables. Explanatory variables (all of which are coded to range from 0 to 1) include those mentioned above. Among our sociodemographic variables, age is coded in years, and all others are dummies. For partisanship, we included dummies for Conservative (those who identify with either the Albertan United Conservative party, Wildrose party or Progressive Conservative party), New Democratic party (NDP), and other partisans, while nonpartisans serve as the baseline category. Ideology ranges from 0 (left-wing) to 1 (right-wing). The variables related to children are both dummies (have children or not, and have children in school or not).Footnote 9 Finally, the competitiveness variable reflects the margin of victory of the winning candidate over the second-place finisher.
Figure 1 The Correlates of Ballot Roll-off
The figure reports point estimates for marginal effects, as well as 95 per cent confidence intervals (whiskers). Statistically significant effects (p < .05) are marked in red, while insignificant results are black. Since a value of 1 for the outcome variables indicates ballot completion (0 means school board abstention), positive entries in the figure indicate higher likelihood of participation in school board elections. To account for ward-level variation, we report standard errors clustered by ward and include ward fixed effects in our final model of ballot roll-off (in online Appendix 3). The full table of coefficients, along with additional models and robustness tests, are available in online Appendix 3.
Figure 1 points to several factors that have a statistically significant relationship with roll-off to school board elections, including two sociodemographic characteristics: gender and education. Women who vote in the mayoral election are more likely than their male counterparts to also vote in school board elections (p < .05). The direction of the effect of this variable is the same in the council model, though this result is not significant (p = .15). There are no such mixed results for education, however. University-educated respondents are significantly more likely to complete their school board ballot than are Calgarians who lack a university degree.
Two of the political variables also display a relationship with school board turnout. First, Conservative partisanship increases the probability of ballot completion by an estimated 9 percentage points, as compared to nonpartisans. Though the difference between Conservative and other partisans is insignificant, it appears conceivable that the conservative Students Count slate had the effect of mobilizing one type of provincial partisan in particular. At the same time, while NDP partisanship is not statistically significant in either model, left-wing ideology is associated with ballot completion (recall that a high value for the ideology variable indicates right-wing ideology, so the negative coefficient here indicates that those on the right side of the spectrum are relatively unlikely to complete their ballots). This variable is statistically significant in the mayoral model, and is close to being so in the council model (p = .07). This presents us with an intriguing combination of findings: while those with right-wing ideologies were less likely to participate in the school board election, Conservative partisanship increases an individual's likelihood of participating. This finding suggests to us that high-profile Conservative endorsements of the Students Count slate may have proven to motivate Conservative partisans to participate in the school board election. We return to this finding in more detail below.
Next, our parenthood variables suggest that having children, regardless of their age, proves to be more important for participation in school board elections than having children currently enrolled in school. Even in the face of controls for sociodemographics, partisanship and the competitiveness of a race, having children increases the likelihood of ballot completion by nearly 8 percentage points, a large and substantively important relationship.
Finally, the competitiveness variable has a significant effect upon school board turnout, in the direction that one would expect. That is, the closer the outcome of the school board race, the more likely it is that those individuals who cast votes in races at the top of the ballot will also vote in the school board election. This variable is significant in both the mayoral and council models.
As a robustness check on our findings, we turn to our second analysis, which uses a two-stage Heckman selection setup to first model municipal (mayoral) turnout and then turnout in the school board race; we then undertake the same analysis using council turnout and school board turnout.Footnote 10 These results are summarized in Table 1.Footnote 11
Table 1 School Board Turnout—Heckman Selection Model
Heckman selection model, dual probit. N = 990 (890 selected, 100 unselected in mayoral model;
862 selected, 128 unselected in council model).
* p < .10 ** p < .05 *** p < .01
Wald test of independent equations: p < .05 for both mayoral and council models.
The Heckman selection results are very similar to those from the roll-off models above. Again, we find evidence that women and university-educated electors are more likely to participate in school board elections, although the gender variable is once again not significant in the council / school board analysis. All other findings in Table 1 are identical in direction and significance to those in Figure 1. Conservative partisans and those on the ideological left are again especially likely to participate in school board elections. Those with children are more likely to be school board voters, but having children in school does not have an independent effect. Election closeness, as above, is significantly related to participation.
Discussion
Our analysis has identified a number of variables that are linked to participation in school board elections. We aim to use these relationships to build and test new hypotheses concerning school board turnout in future research. In the meantime, however, we suggest two broad features of our findings that may be worthy of consideration in future research.
First is the role of information in turnout. As recent roll-off studies have demonstrated (Lassen, Reference Lassen2005), the information that voters possess is clearly linked to participation; those without adequate information may refrain from voting because they consider uninformed voting to be risky (Matsusaka, Reference Matsusaka1995) or would prefer to leave the decision to those who are better informed (Feddersen and Pesendorfer, Reference Feddersen and Pesendorfer1999). Whatever the exact mechanism, one variable in our analysis—Conservative partisanship—appears to provide some support for the information thesis. Unlike many school board elections, Calgary's 2017 race was coloured by explicit and asymmetrical partisanship: Conservatives responded to the existence of the Students Count slate, while partisans of other parties did not turn out at rates any higher than did nonpartisans. One possible explanation for this difference could be that the slate served as a cue, or informational shortcut, for Conservative partisans. That is, they may have been more likely to vote because they had information that made their vote decision simpler. For opposition partisans, the partisan cue was less obvious—even if they knew that they did not want to vote for the Students Count candidate, it may have been unclear which other candidate they should support. The presence of this partisan cue may therefore have had differential effects upon supporters of different parties. Of course, we must also recognize the possibility that Conservative partisans, for some other reason, may generally have relatively high rates of turnout in school board elections. Future research on an election without such a clear party cue is necessary to determine which of these two explanations is correct.
The second important pattern in our findings and past research is that turnout is higher when more is “at stake” in a particular election (Andersen et al., Reference Andersen, Fiva and Natvik2014; Ervik, Reference Ervik2012). In the case of school boards, this may well apply to those for whom education policy is thought to be more important as a result of ideological commitments or parental status. Interestingly, however, those for whom the most would appear to be at stake—parents whose children are currently in the public system—are no more likely to participate than parents in general. Whether this is a genuine pattern in other contexts or a result of ceiling effects in our data (those with children are already very likely to participate, making it difficult to discern the additional effect of having children in the school system) is a question to be answered with a larger dataset in the future.
Finally, participation in school board elections may reflect a mix of the two factors. For instance, highly educated voters may find it easier to acquire information about school board politics, or they may feel that more is at stake in school board politics due to their own investment in education—or both. The same may be true of women, whose participation may reflect access to information (due to persistently gendered childcare roles) or, for the same reason, a higher awareness of the stakes of school board politics.
Whatever the processes that drive increased participation in school board elections may turn out to be, what we can conclude from this analysis is that the composition of the voting public in Canadian school boards is very likely to be distinct from that of municipal voters, probably tilted toward ideologically left-leaning and highly educated parents (and perhaps, depending on local cues, toward particular partisans as well). These compositional differences, having been identified, could now be investigated using aggregate election and census data, as well as survey data, in other cities. Given recent research on municipal responsiveness to local electorates (Tausanovitch and Warshaw, Reference Tausanovitch and Warshaw2014), these compositional differences may well have effects on policy (Anzia, Reference Anzia2011; Kogan, Reference Kogan, Lavertu and Peskowitz2018). Our findings in this study suggest that these possibilities are well worth exploring in the Canadian context.
Conclusion
This research note represents a first foray into the study of political behaviour in Canadian school board elections. We have identified several sociodemographic and attitudinal characteristics associated with participation in those contests, demonstrated the importance of having children for school board participation and shown that the competiveness of contests affects school board turnout rates. At a broader scale, our findings suggest that the composition of school board electorates is shaped by key demographic and life-course variables—school board electorates are likely to be more well-educated, for instance, and contain more parents than the general public—and also by ideology and, at least in Calgary's case, by partisanship.
These findings are just the beginning. Research from the United States (Moe, Reference Moe and Howell2005; Anzia, Reference Anzia2011) suggests that teachers in that country may be particularly engaged in school board elections. The influence of teachers and their unions in these elections has yet to be considered, however, in Canada. Though not available in the CMES dataset, data on occupation and school board participation would add to our understanding of school board turnout. There would also be value in studying turnout patterns in other Canadian contexts. Comparing differences in participation in settings where school system differences may be more salient for local residents—such as Catholic versus public school boards in Ontario, or English-language and French-language school boards in Quebec—will also clarify what motivates some voters to participate in school board elections. Given recent school board controversies across Canada, debates about the purpose and democratic legitimacy of school boards are likely to continue in the coming years. Academic research on overall levels of interest and engagement in school board elections, along with research on patterns of participation in school board politics, can make a valuable contribution to these debates.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0008423919000088
