Ballot Roll-Off in the 2021 Alberta Senate Nominee Election: Political Identities and Participation Patterns

Abstract

In 2021, Alberta held Senate nominee elections for the fifth time in the province's history. Conducted concurrently with municipal elections and multiple referenda/plebiscites, the Senate race had a much lower participation rate than any of the other votes held that day. The purpose of this research note is to identify patterns of ballot roll-off—the phenomenon whereby electors cast a ballot for one race but not another—in the Senate election. Using data from a three-wave survey of Calgarians, the note describes the attitudes of electors toward the Senate election, revealing that electors viewed it as less important than any of the other votes contested that day. It also considers the role of partisan and geographic identities in shaping participation rates. Survey data reveal that both types of identities are associated with roll-off in the Senate election but not any of the other votes with which it was held concurrently.

Résumé

En 2021, l'Alberta a tenu des élections sénatoriales pour la cinquième fois dans l'histoire de la province. Tenue en même temps que des élections municipales et de multiples référendums/plébiscites, la course au Sénat a connu un taux de participation beaucoup plus faible que les autres scrutins organisés ce jour-là. L'objectif de cette note de recherche est d'identifier les tendances d'abstention partielle—le phénomène par lequel les électeurs votent dans une course, en omettant d'exprimer leur préférence dans une autre—lors de l'élection sénatoriale. En utilisant les données d'une enquête en trois phases auprès des Calgariens, la note décrit les attitudes des électeurs à l'égard de l'élection sénatoriale, révélant qu'ils la considèrent moins importante que les autres votes disputés ce jour-là. Elle examine également l'impact des identités partisanes et géographiques sur les taux de participation. Les données d'enquête révèlent que ces deux types d'identité sont associés à l'abstention lors de l'élection sénatoriale, mais pas lors des autres scrutins qui se sont déroulés en même temps.

Despite their absence at the federal and provincial levels, concurrent elections are nevertheless a common occurrence in Canada. In most of Canada's thousands of municipalities, simultaneous elections are the norm, with elections for mayor and council (and often school boards) all held at the same time. In these races, it tends to be the case that participation rates in “top-of-the-ticket” races are several percentage points higher than those in “down-ballot” contests. Relatively little is known about this form of selective turnout.

The formal name for this behaviour—voting in one concurrent election but not in another—is “ballot roll-off.” Though the subject has received ample attention from American scholars (see Walker, 1966, for some of the earliest work on the subject, and Remmel and LaForge, 2021 for a recent overview), it has largely been overlooked in Canada (though see McGregor, 2018 and McGregor and Lucas, 2019 for exceptions). This lack of attention is noteworthy given that turnout rates in municipal elections are typically much lower than at any other order of government, and turnout in “down-ballot” races is lower still. In a context where high levels of turnout are generally viewed as desirable, and where research on turnout is so prevalent, studying turnout rates in down-ballot races is an important step toward fully understanding political participation in Canada.

An extreme instance of ballot roll-off is the 2021 Alberta Senate nominee election. The election, about which we provide further detail below, was held at the same time as Alberta's local elections. While it is not possible to precisely calculate province-wide roll-off rates for the Senate election (because some municipalities had no elections for mayor), Elections Alberta reports that, of the nearly 1.2 million Senate election ballots provided to voters at polling places across the province, a full 18.4 per cent were left blank, and another 3.7 per cent were rejected for other reasons (meaning that they were intentionally or unintentionally spoiled).¹ Thus, more than one-fifth of the Albertans who took the time to come to the polls did not cast a vote for the Senate election. In Calgary, the city which will be our focus in this research note, a total of 22.4 per cent of ballots were returned blank or spoiled. In contrast, 99.3 per cent of mayoral ballots and 96.2 per cent of council ballots were completed properly.² The decline for the Senate election is therefore distinctively large.

In the case of Calgary, electors actually had the opportunity to cast up to seven votes: for mayor, council, school board, the Senate nomination election, a municipal plebiscite on water fluoridation, and two provincial referenda on daylight saving time and the federal system of equalization payments. As we detail below, the participation rate in the Senate election was by far the lowest of any other vote held that day: turnout in the mayoral election was 46.1 per cent, whereas turnout for the Senate race among this same electorate was a mere 36.1 per cent. This 10 per cent point drop is much greater than rates of roll-off in the other contests held that day. The average turnout rate for the other elections was 42.6 per cent.³

The roll-off rate in the Senate election is not only high when compared to other elections held on the same day but also relative to the United States, where the vast majority of research on the topic originates. By our calculations, the aggregate level of roll-off in the 2020 US Senate elections was roughly 1.3 per cent, while this figure for House elections that same year was 3.1 per cent.⁴ Roll-off rates tend to be slightly higher for positions further down the ballot, but elections to national legislative bodies have significantly lower rates of roll-off than witnessed in Alberta. At the same time, the rate of roll-off we see in 2021 was not atypical of Alberta Senate races. In 2012 and 2004—the most recent Alberta Senate elections, which were both held in conjunction with provincial elections—rates of roll-off were roughly 14.9 per cent and 19.8 per cent.⁵ Alberta Senate elections, therefore, seem to consistently have uncommonly high rates of roll-off.

What might explain the extremely low level of participation in the 2021 Senate election? Using data from a three-wave survey of Calgarians fielded in 2021, this research note describes the attitudes of electors toward the Senate election, revealing that Calgarians viewed this vote as less important than any of the others that were held concurrently. We also consider the role of partisan and geographic identities in these participation rates. Our data reveal that both types of identities are associated with Senate roll-off, but that neither factor is associated with roll-off in any of the other votes held that same day.

The extraordinary rate of roll-off in the Senate election is only part of the rationale for focusing on this contest. The 2021 election is the latest of many Senate nominee elections that Alberta has held (they have occurred on four previous occasions), and it is unlikely to be the last. However, this note is also the first individual-level examination of voting behaviour in a Canadian Senate election. We see considerable value in studying atypical elections, whose unusual features can reveal important features of Canadian voting behaviour that may be obscured in more typical Canadian elections. In Alberta, Senate elections stem from long-standing debates over the very nature of the Senate and serve as a manifestation of Western (Albertan) alienation. As we outline below, our data reveal that behaviour in Alberta's Senate election is indeed atypical and is closely related to the history and politics of the Senate in Alberta.

Moreover, Senate nomination elections are not always inconsequential. There have been occasions in the past where winners have been appointed to the Senate. Given that the winners of this race were all affiliated with the Conservative Party of Canada, it is quite possible that some, or all, will eventually end up in the Senate. Finally, given the transitional nature of the Senate—which is a combination of Conservative partisan appointees and non-affiliated (Liberal) appointees and, potentially in the future, winners of the 2021 Alberta Senate election—the body is becoming an important focus of scholarly research. We also expect that the manner in which Senators are selected will be a point of public discussion, if and when there is a change in federal government. This study, therefore, has the potential to contribute to both academic and public discourse on the matter.

Alberta Senate Elections

While no Senate elections have ever occurred in Canada outside Alberta, Senate nominee elections inside the province are quite common. The 2021 election was the fifth in Alberta's history (others were held in 1989, 1998, 2004 and 2012). These elections are called sporadically by the provincial government (always by an ideologically conservative party), with the hope that, upon the opening of a vacancy in the Senate, the federal government will appoint one of the election winners (the “nominees”). The elections are consultative in nature, meaning that winners do not automatically win a place in the Senate. Prior to 2021, just half (five of ten) of previous election winners were eventually appointed to the Senate. All appointments were made by (Progressive) Conservative Prime Ministers, in some cases several years after being “elected.”⁶

Senate elections first emerged during the “Mega Constitutional Politics” era of the 1980s and 1990s (Russell, 2004). During this time, federal and provincial governments focused intensely on attempting to make wholesale constitutional changes—including to the Senate—by means of formal constitutional amendments. The notion of an elected Senate was particularly popular in Alberta, and this sentiment persisted following the failures of the Meech Lake and Charlottetown Accords.

The 2021 election employed a system of block voting; each elector was permitted to vote for up to three “senators in waiting.” Only the Conservative Party and the People's Party of Canada participated, fielding three candidates each. The seven remaining candidates were independents. It is common practice for parties of the centre and left to boycott Senate nominee elections. The Liberal Party fielded one candidate in 1989, but neither they nor the NDP have nominated candidates since. Numerous conservative parties (Progressive Conservative, Reform and Alberta Alliance at the provincial level and Conservative and People's Party of Canada federally) have fielded candidates over the years. It is also typical for numerous independent candidates to enter the race. The winners in 2021 were Pam Davidson, Erika Barootes and Myhailo Martyniouk—all three from the Conservative Party of Canada.⁷

While the Senate election is for a federal office and was held concurrently with municipal elections, it was the provincial Premier, Jason Kenney, who called the election. He did so less than four months prior to the planned municipal elections, following the passage through the Senate of a bill related to environmental assessments which he argued was harmful to the interests of Alberta (Tasker, 2021). Elections Alberta oversaw the administration of the election, though they did so by coordinating with municipal election management bodies. According to Elections Alberta, province-wide turnout in the election was estimated to be 39.6 per cent. However, this figure does not count blank or spoiled ballots as noted above. Citing the Local Authorities Election Act, Elections Alberta notes that the process to decline a ballot is to leave it blank. If such ballots are counted as non-voters, the actual rate of turnout in the senatorial election was just 32.3 per cent. By comparison, turnout in the 2019 and 2023 provincial elections was 67.5 per cent and 59.5 per cent, respectively, and turnout in Alberta in the 2021 federal election was 64.4 per cent (Elections Alberta, 2023; Elections Canada, 2021).

Identity and Political Participation

Identities are central to politics and political behaviour. Since the earliest studies of voting behaviour, scholars have understood that “who you are” affects how you behave in elections (Berelson et al., 1954; Campbell et al., 1964). Two particularly salient political identities in Canada are partisan and geographic identification.

Without a doubt, the most studied form of political identity is partisan identification. Partisanship is well-known to have wide-ranging effects in Canada and elsewhere for vote choice as well as turnout (see Rau, 2021 for an overview of the relationship with turnout). Partisans tend to be more interested in and attentive to politics, and they vote in higher numbers than non-partisans do. There is also some evidence that partisans are less likely to roll-off in both partisan and non-partisan settings (Thornburg, 2019; McGregor and Lucas, 2019).

In the context of the 2021 Senate election, we are less interested in the partisan vs. non-partisan comparison than in determining whether supporters of different parties have different roll-off rates. The fact that the Senate election was held in the first place has partisan implications; the UCP called the election, and none of the province's other major parties—the NDP, the Alberta Party and the Liberal Party—supported this decision. None of the centre/left parties at the federal level fielded candidates. The absence of the partisan cue across such a considerable range of the ideological spectrum meant that non-conservative voters had no obvious candidates to support.⁸ Presumably, their partisanship would push them away from Conservative and PPC candidates, but the absence of their preferred party from the ballot makes it difficult to identify preferred candidates from the lengthy list of independents. As a result, we expect that UCP partisans were less likely to roll-off than non-partisans as well as supporters of other parties.

Geographic identity is also deeply salient in Canada's politics. Alberta is generally thought to have a distinct political culture and high levels of regional/provincial attachment. One of the byproducts of the regional nature of Canadian politics, and the high sense of regional identification in Alberta in particular, is the emergence of the “Triple-E” Senate movement: a Senate in which each province would have equal representation, Senators would be elected, and the Senate as a whole could serve as an “effective” counterweight to the House of Commons. First suggested by the Canada West Foundation in 1981, the idea of an elected Senate quickly became popular in Alberta. It was adopted as official policy by the Reform Party, which had its political base in Alberta, in the 1980s. Later, the Conservative government of Stephen Harper—himself representing a riding in Alberta—made (ultimately unsuccessful) efforts to shift toward an elected Senate.⁹ A Senate is attractive to ideologically conservative Albertans, who believe the reform would give the province more influence over the affairs of a federal government whose House of Commons is dominated numerically by representatives from central Canada's more populous provinces.

Given this, we expect individuals with a strong provincial identification to be more likely to vote in the Senate elections, as they are more likely to be supportive of the very existence of the elections. These individuals may also see participation as a way of legitimizing the contests. Conversely, given that the notion of a Triple-E Senate is so strongly linked with the notions of Western Alienation and Alberta provincialism/regionalism, we expect that individuals with a strong Canadian identity will be relatively unlikely to vote in the Senate election in the hopes of avoiding the legitimization of the contest.¹⁰

We do not expect to see similar identity-based gaps for the other concurrent elections held that day. The only other race with seemingly clear partisan and regional implications was the equalization referendum. As with Senate reform, the subject of equalization is a perennial source of regional resentment in Alberta. The referendum simply asked if voters believed that the principle of equalization should be removed from the constitution. In this race, opponents of equalization reform had a voting option: no. With the Senate vote, however, supporters of the Liberals and NDP had no options on the ballot. We, therefore, expect the rates and correlates of roll-off in the two races to differ, and that partisan and geographic identities will matter less (if at all) for the equalization vote.

Data and Methodology

Our analysis is based on data from a three-wave panel survey of eligible voters in Calgary in 2021. Surveys were fielded in the summer (N=2,334), just prior to the October 18^th election (N=2,159), and just afterwards (N=1,400).¹¹ The questionnaires were administered online by Forum Research Inc. Respondents were recruited using random digit dialling and were then provided with a link via email for each wave of the survey. To ensure sufficient sample size, Forum recruited 1,201 new respondents in the second wave of the survey (who were then invited to complete the third wave as well). The surveys included questions on relevant political and geographic identities as well as vote choices and other attitudes. We provide a list of the questions used in the analysis below in Appendix I. Descriptive analyses are weighted for age and gender to match Calgary's population in 2021.¹²

We present our results in two stages. The first describes the perceived importance of the seven votes available to Calgarians—mayoral, council, senate, as well as the referenda and plebiscite (we do not consider school board elections because several of the races were acclaimed). We consider both absolute and relative measures of perceived importance. For the absolute measure, respondents were asked how much they “care about” the outcome of each election or referenda/plebiscite. For the relative measure, participants were asked which election they would choose if they were able to participate in just one of the available contests. These indicators provide a great deal of insight into the importance that Calgarians placed upon the various votes.

The second stage of the analysis addresses roll-off more directly. We use logistic regression models to identify the correlates of roll-off from the mayoral election to the senate election, council elections and plebiscites. The dependent variables in the models are dummy indicators of roll-off from the mayoral election (the “top-of-the-ticket” race). In each instance, the variable has a value of 1 if an individual voted in the mayoral election but not in the other election in question (that is, they “rolled off”), and 0 if the individual person voted in both contests.¹³ By creating a separate roll-off variable for each race, we are able to determine if and how the correlates of roll-off vary by contest.

Our explanatory variables of interest are partisan and geographic identity. Partisanship is a categorical variable with four categories: UCP, NDP, another party¹⁴ or non-partisan.¹⁵ We measure geographic identity using responses to a question on how close respondents feel to Canada/Alberta/Calgary—either extremely close, somewhat close, not very close, or not close at all.¹⁶ Given the multi-level nature of the Senate election—the fact that the election was called by the provincial government, contested by federal parties and held concurrently with local elections—the models also include controls for interest in politics at the three orders of government (respondents were asked their level of interest in each, ranging from 0 to 10). Finally, the models include a series of sociodemographic factors generally considered in studies of turnout and roll-off, including age, gender, education, income, immigrant status and racial identity, as well as an index that taps into knowledge of local politics, since knowledge is a well-known correlate of turnout and roll-off (Wattenberg et al., 2000).

As a robustness check of our roll-off models, we also run a series of Heckman selection models. Simple roll-off models, by definition, exclude non-voters. Heckman selection models involve first modelling the process by which respondents choose to participate in the municipal election, and then choose to vote in a down-ballot race (Streb and Frederick, 2011). The first stage of the model identifies factors associated with municipal turnout in general, while the second isolates variables of particular importance in the down-ballot races. This approach mirrors McGregor and Lucas’ (2019) study of roll-off in school board elections. If the Heckman results match those of the traditional roll-off analysis (which they do), we can be particularly confident in our findings.

Results

The perceived importance of voter decisions

To set the stage for our roll-off analysis, we begin with a descriptive overview of respondents’ interest in each of the elections for which we have data. Figure 1 reports the proportion of respondents who selected values ranging from “I don't care at all” to “I care a lot” about each election (N = 1,885 for all bars).

Figure 1. How Much Do You “Care” about Each Election?

The results in Figure 1 represent absolute measures of perceived importance—respondents are asked independently about each race, with no comparison implied in the question. Unsurprisingly, Calgarians care more about the mayoral election than any other. Nearly two-thirds of respondents (64.2%) care “a lot” about the race, while just 2.2 per cent “don't care at all.” At the other end of the spectrum is the Senate election, about which just 16.3 per cent of respondents “care a lot,” and 25.1 per cent “don't care at all.” In fact, this is the only election for which the latter category was chosen more frequently than the former. Chi-square tests reveal that individuals care less about the outcome of the Senate election than any other vote held that day; the Senate results differ from all others at p<0.001.

A second measure of perceived importance provides further evidence that Calgarians cared less about the Senate race than any other. Voters in the survey were asked which race they would vote in if they could choose only one (N = 1,220). By forcing respondents to select just one election, this measure is explicitly comparative in nature. An extremely small number of individuals would have voted in the Senate election if they could only choose one vote to cast—just 0.3 per cent of respondents. Again, the mayoral election stands at the other pole on this measure, with nearly two-thirds (63.9%) of respondents selecting this race (6.8% selected the equalization vote, 6.7% the fluoride vote, 6.4% the daylight saving time vote, and 1.2% the school board election). Admittedly, this is not a perfect measure of perceived importance because other factors, such as strategic considerations, might influence responses. However, when combined with the results from the absolute measure (Figure 1), it is clear that, in the aggregate, Calgarians viewed the Senate election as particularly unimportant.

Ballot roll-off

Prior to identifying the correlates of roll-off, we describe rates of this phenomenon in our sample. Respondents were asked if they voted, and if they did, which races they voted in.¹⁷ These data confirm that roll-off in the Senate race was particularly high—26.5 per cent of the respondents who participated in the mayoral election abstained from the Senate nominee election. The roll-off rate in the council race was just 4.5 per cent, and the corresponding values for the fluoride, daylight saving time and equalization votes were just 4.6 per cent, 3.4 per cent and 5.1 per cent, respectively. These lower rates are typical of those observed previously in other Canadian municipal elections (see McGregor, 2018; McGregor and Lucas, 2019). The roll-off rate in the 2021 Senate election was therefore exceptionally high among our respondents as it was with the population as a whole.

We now consider the correlates of roll-off using a series of logistic regression models, where the ballot roll-off measures serve as the dependent variables. We run a model for each of the down-ballot races. As noted earlier, we focus on partisan and geographic identity variables, but all models also include a suite of controls for interest and sociodemographics. Positive values on the x-axis indicate that a variable is associated with an increased likelihood of roll-off in a particular race. Figure 2 shows the marginal effects results from the models for the theoretical variables of interest. The partisanship variables are categorical (with UCP partisan as the baseline), and the geographic identity variables are interval level. All variables range from 0 to 1 and N=842 for all models. Results to the right of the zero (dotted) line in the figure suggest a positive association with roll-off. Full model results are found in Appendix II.

Figure 2. Identity and Ballot Roll-off–Marginal Effects

Figure 2 reinforces yet again the unique nature of the Senate election. For none of the other votes are any of the partisan or geographic identity variables related to roll-off. This is true even with the equalization vote, which arguably had both partisan and regional implications. The Senate model, in contrast, displays a number of significant results, suggesting that partisan and geographic identities had a unique effect on this race.

Partisan identity has a clear relationship with Senate roll-off in line with our expectations. UCP partisans have the lowest rates of roll-off of any partisan group. They were nine points less likely to roll off than NDP partisans, 16 points less likely than supporters of “other” parties, and ten points less likely than non-partisans. None of these other groups are statistically distinguishable from one another—UCP partisans roll-off at a uniquely low rate. Partisanship is a significant factor in shaping roll-off patterns. The finding that UCP partisans are comparatively unlikely to roll-off makes sense, given that their party called the Senate election and that only ideologically conservative parties fielded candidates in the race.

We also see that geographic identity has a relationship with Senate roll-off. While identification with Alberta and Calgary does not appear to have independent effects, we see evidence that identification with Canada is positively associated with Senate roll-off. We suspect that this result may reflect the unusual nature of the Senate race, in which a provincial government is attempting to exert influence over a process that traditionally falls under the jurisdiction of the federal government. Individuals high in Canadian identity may, therefore, not wish to legitimize the Senate election, instead choosing to abstain.

It is clear that both partisan and geographic identities mattered for participation patterns in the 2021 Senate race. None of these factors, however, was related to roll-off in any of the other votes held concurrently. These patterns are confirmed by the Heckman selection analysis (see Appendix III). Not only did Calgarians participate in that 2021 Senate race at a relatively low rate, but the correlates of participation in that election were markedly different than was the case in any of the other races held that day.

Conclusions

Though Alberta Senate nomination elections are held infrequently and at irregular intervals, the case of the 2021 election provides a fascinating and meaningful case study of Canadian political behaviour. Studies of concurrent elections demonstrate the extent to which, and the reasons why, Canadians behave differently toward votes of different types, providing insight into an understudied dimension of political engagement: selective participation.

As the first examination of political behaviour in a Canadian Senate election, this note highlights the value of studying unusual elections. It shows that, as a whole, Calgarians saw this non-binding and somewhat controversial election as less important than any other race held that day, and they participated at comparatively low levels. It also finds that geographic and partisan identities help to explain the low participation rate in the Senate race, though this is not the case for the other votes held concurrently that day. A key finding from this study is that the determinants of political participation in Canada can vary greatly, even when participating is essentially effortless (such is the case when electors are already in the voting booth with multiple ballots). In doing so, it confirms that Alberta Senate elections also have regional and partisan implications that are likely unique among political contests in Canada.

While this note answers some important questions about voting behaviour in this unique and controversial type of election, it also points to a series of questions worthy of future consideration. For instance, what is the role of candidate quality in shaping voter preferences vis-à-vis the power of partisanship? All three election winners represented the Conservative Party of Canada. Did electors focus entirely on partisanship, or did the quality of candidates play some role? Other questions arise about the role of political elites at both the federal and provincial levels in shaping opinion. To what extent did the actions, or inactions, of such individuals shape either turnout or vote choice in the Senate race? Next, our sample is from Calgary, but would our results hold elsewhere in the province? We expect that they would, but it is also possible that regional differences might shape behaviour in Senate elections. Similarly, our results suggest that respondents care comparatively little about the Senate election. Might these views be different if the Liberals and NDP had fielded candidates, or is it the case that Albertans just do not care about this body? Finally, what other factors not considered in this study might influence turnout and roll-off patterns? For instance, do elector opinions on the nature of the Senate (should appointments be based upon partisanship or expertise) drive behaviour? Additionally, given that just half of previous Senate election “winners” were actually appointed to the Senate, might a sense of futility depress turnout? Even among UCP respondents in our model sample, roll-off was much higher than it was for other positions. 17.4 per cent of UCPers (N=354) in our sample who voted in the mayoral race abstained in the Senate race as compared to 26.5 per cent for the sample as a whole. Why is it that, even among this group which has a comparatively high level of participation, so many voters are selectively abstaining from Senate races? While this research note makes important contributions toward understanding these fascinating elections, much remains to be learned.

As far as the eventual appointment of the election winners to the Senate is concerned, the low and uneven patterns of participation raise interesting questions about the perceived legitimacy of the outcome of the Senate election. By refusing to nominate candidates, the governing federal Liberal Party contributed to both the low level of turnout and the unequal pattern of participation in terms of both partisan and geographic identities. Given these differential patterns of turnout, questions arise as to whether the federal government should view the election results as legitimate. Yet, it is rather dubious for a party to point to a problem they helped to create as a reason not to appoint the winners. While we do not have the answer to the question of whether Alberta's Senate nomination elections are legitimate, we can say that the factors that are likely to be associated with judgments of legitimacy and illegitimacy—partisanship and geographic identity—are also associated with participation in the nomination elections themselves. For this reason, the repeated elections are unlikely to build more general momentum for an elected Senate beyond those voters who are already committed to the idea.