<<

Title:

Invalid in an Apolitical and Low Turnout Context: A Case Study of One Canadian Municipal

Authors/ Auteur(e)s

Sandra Breux* Professeur-chercheur Institut National de Recherche Scientifique Centre Urbanisation Culture et Société 385, rue Sherbrooke Est Montréal (Québec) H2X 1E3 Canada Email : [email protected] Tel : 514-499-4059 * Auteur à contacter / Corresponding author

Jérôme Couture PhD Candidat Assistant de recherche Université Laval and INRS-UCS 385, rue Sherbrooke Est Montréal (Québec) H2X 1E3 Canada Email : [email protected] Tel : 514-499-4059

Abstract

Canadian municipal has several distinctive features: 1) a lower than other levels of , 2) of political parties, 3) a high rate of invalid votes. This study seeks to understand the reasons for the high invalid vote rate in Canadian municipal politics. For this, we test three commonly-accepted hypotheses on invalid voting for the city of Montreal using electoral and socio-demographic data published by the city in 2009. The results show that the institutional complexity and social structure of a given borough are likely to influence invalid voting. The analysis moreover highlights the need to measure the cost, in terms of effort and difficulty, for voters to obtain information at the local level.

Keywords: invalid votes, low turnout, municipal election, Canada, non-politicised context

1 Introduction

The presence of invalid votes, that is to say, blank or spoilt votes, is a constant in Western (Uggla, 2008). However, invalid voting has been the subject of a relatively small number of studies, in particular concerning the municipal level (Trounstine, 2009; Faure, 2009). Yet, it is at this level that invalid voting is generally highest, even if the overall voter turnout there is often comparatively lower. For Reif and Schmitt (1980), these phenomena, alongside a drop in support of the power in place and strategic voting, are two of the main attributes of so- called “second-order” . According to these authors, voters consider these types of elections to be less important than others as they usually involve less high-stake issues. Thus, compared with national elections, municipal elections are seen to concern issues of lesser importance. A high percentage of invalid votes would therefore reflect this lower significance of municipal elections.

However, this classification of election types assumes that national elections are connected to the other types of elections and that, therefore, voters’ range of choices is essentially the same at the different levels of government. In that sense, some speak of a “nationalization” of elections at the municipal level (Parodi, 1983; Parodi, 2004). However, in North America, and more specifically in Canada, there is no direct link between the municipal and higher levels of government. For this reason, the theory of second-order elections does not adequately explain the presence of invalid voting. In fact, in the provinces with partisan municipal systems, as is the case with Quebec, municipal political parties have no affiliation with the other levels of government and can be characterized as apolitical (Bherer & Breux, 2012). In such a context, voters cannot express their approval or disapproval of the upper levels of government through their votes in municipal elections.

In Quebec, invalid voting in municipal elections is prevalent. For example, in the 2009 Montreal municipal elections, the number of invalid votes was 58,362,1 compared to 53,749 rejected votes in the 2012 Quebec provincial election and 51,430 in the 2011 federal election. This means that the number of invalid votes in the 2009 municipal elections in Montreal was greater than those for the provincial and federal elections. Moreover, for 18 of the 105 elected positions (17.1%), the number of invalid votes was equal to or higher than the victory margin of the winning candidate. This phenomenon is not unique to Montreal and can in fact be observed for many of the province’s big cities. Yet, it does lead us to question why the rate of invalid voting is higher at the municipal level, which is characterized by an apolitical context with limited electoral mobilization, than at the other levels of government.

Studies on the model of second-order elections and the dynamics underlying invalid voting have shed new light on this topic. Rallings and Thrasher (2005), for example, emphasize that second- order elections are far from homogeneous, while Henderson and McEwen (2010) highlight the importance of context for understanding this type of election. Studies on invalid voting likewise underlined the importance of context, in particular of the social and political institutional context (Power & Garand, 2007). However, many studies on invalid voting focus specifically on

1 According to the last official census of 2011 realized by Statistics Canada, the city of Montreal had 1,649,519 inhabitants and the province of Quebec 8,054,800.

2 systems, which does not apply to North America. Thus, these studies are only of limited value for explaining why some people voluntarily take the time and make the effort to cast that will then be invalidated.

With this backdrop, the purpose of this article is to discuss invalid voting in the context of voluntary voter mobilization, limited turnout and apoliticism. First, we provide explanations for invalid voting, both as a general phenomenon and pertaining specifically to North American and Quebec municipal elections. Secondly, we test three hypotheses that may explain invalid voting in the 2009 Montreal municipal elections. In conclusion, in the presentation of results, we will describe certain specificities of municipal electoral behavior in more detail.

1. Invalid voting: a little-studied phenomenon

A discussion on invalid voting gives rise to three observations. First, very few studies have addressed this issue. This is partly due to the difficulty of capturing invalid votes in their different forms. Essentially, they comprise various types of votes, including those in which the invalidation results from an error on the part of the voter rather than a deliberate action or omission. It is often difficult for researchers to distinguish the latter kind of ballots from those that were voluntarily invalidated. Moreover, reports on election results do not always make this distinction: “In the United States, for instance, invalid votes are not even reported in official statistics. Similarly, the European Social Survey does not include the alternative of the blank when pondering respondents’ choices during the most recent elections in their countries” (Uggla, 2008).

Further, because invalid votes often represent such an insignificant percentage of all ballots, researchers generally prefer to focus on valid votes or on . The lack of research attention given to this phenomenon may also be explained by the still relatively prevalent view that invalid voting is an expression of voter incompetence: “[Invalid ballots] may have something to do with an assumption that they belong to a residual category that indicates the incompetence of an art of the electorate and, hence, lacks interests from a political perspective” (Uggla, 2008). Another paradox is that the profile of the type of person who casts an invalid ballot is generally grouped into the same category with those who abstain from voting, yet that invalid ballots are nevertheless included in the voter turnout calculation. This alone is sufficient grounds for advocating a more in-depth study of the topic.

The second observation is that studies on invalid voting seek to grasp the underlying rationale behind the phenomenon. Here, three main approaches stand out, being the institutional, socio- economic and dissenter approach. For some, the institutional context, in influencing the cost, in terms of effort, of voting for the voter, largely explains this type of voting behavior. This approach is based on the idea that voter turnout is highly correlated to the institutional context (Perez-Linian, 2001). For others, invalid voting is the result of a specific social structure (McAllister & Makkai, 1993), while still others see in it a sign of protest (Zulfikarpasic, 2001). More recent works also seek to interconnect all three of these distinct models within their studies. Power and Garand (2007), in their analysis of 18 Latin American countries between 1980 and 2000, state that: Socioeconomic variables: such as urbanization and income inequality are associated with levels of invalid voting, while institutional variables such as compulsory voting, electoral

3 disproportionality, and the combination of high district magnitude and a personalized voting system tend to increase blank and spoiled ballots. Moreover, regime-level factors such as political violence and the level and the direction of democratic change also shape the rates of invalid voting. This view is shared by Uggla (2008). Thirdly, of the few studies that have addressed this issue, most focus only on a context of compulsory voting, in which the invalid vote takes on a whole new meaning (Pion, 2010). In addition, most neglect the local level, despite a number of indicators that call for a more in-depth examination of the meaning behind invalid voting at this level of government.

2. The specificities of the municipal level in the United States and Canada

In keeping with the overall lack of research on invalid voting, little research exists on the North American municipal level. This is due to neglect of this governmental level by and to the difficulty of obtaining usable data (Trounstine, 2009). Thus, research on North American municipal , including on how the local political arena is structured, is lacking. Of the few studies on this topic, most focus on voter turnout and its specificities.

There are three main elements to consider when looking at this level of government. One, voter turnout is often lower at the municipal level than at the other levels of government. Overall, voter participation at the local level appears to be of a different order, a phenomenon that is not specific to North America. In the United States, municipal voter participation is often around 27%, while in some Canadian provinces it is around 40% or 45% (Cutler & Matthews, 2005). Some attribute this limited voter participation, at least in part, to the institutional structure. Factors likely to influence voter turnout are the degree of decentralization (Sellers & Lidström, 2008), the , especially with regard to territorial division (Wood, 2002), and modalities for registering to vote (Trounstine, 2013). In addition, the re-election of incumbents is a key element of the political landscape in both the United States and Canada, and especially in Quebec (Mévellec, 2011).

Secondly, at this level of government there are few or no political parties. Overall, the usefulness of political parties and their potential influence on voter turnout at the municipal level is a contentious subject, with many maintaining that there is no place for ideological debates at this electoral level (Peterson, 1981). Some researchers also posit that the North American local political context suffers from an information deficit, or a poor flow of information. As such, Cutler and Matthews (2005) argue that: In comparison with national elections, municipal elections provide voters with less political information—i.e., information relating to issues, candidates and the like! And the information that is provided tends not to be well organized into coherent partisan packages. The municipal voter, then, faces challenges of both informational quantity and quality. (pp. 360) According to these authors, this information deficit is partly due to the virtual absence of political parties at this level. This observation also brings Elmendorf and Schleicher (2012) to propose, with regard to the United States, the creation of non-partisan political parties such as is found in the province of Quebec in order to improve the access to and dissemination of information.

4 In the case of Quebec, these features take on a different character since this province has non- partisan political parties. It is often argued that it is due to this political structure that voter turnout at the local level is higher in Quebec than in the United States, although this assumption has not been proven. That said, non-partisan political parties at this level of government are indeed advantageous for the creation of an opposition and the dissemination of information, in addition to allowing the recruitment of candidates that are more representative of the population (Lightbody, 2006). However, in comparison to political parties at the provincial level, they are poorly organized, their life span tends to be short, sometimes not even lasting until the election date, and their candidates tend to have shifting allegiances (Bherer et al., 2008). In addition, the names of these parties, generally composed of the name of the municipality followed by the name of the head of the party, often make it difficult for voters to associate a specific vision or with them. Moreover, the political agendas of these parties can be difficult to distinguish from one another (Breux & Bherer, 2013).

In short, at the North American municipal level, it appears that the political, institutional and social contexts play a significant role in the formation of individual voting behavior (Trounstine, 2009). To the extent that this may also apply to invalid voting, it would be important to investigate the factors that contribute to invalid voting at the municipal level. This seems all the more necessary in a context where elections are voluntary and where voter mobilization is weaker compared to other levels of government. In other words, the question to be examined is: Why would some one go to a polling station only to have their vote nullified?

3. Data and method

3.1 The dependent variable

In general, the invalid votes of an election are represented as a percentage of the number of voters registered on the electoral list. Like the voter turnout rate, it is possible to identify the rate of invalid ballots, expressed by the proportion of voters whose ballot is not valid to the total number of potential voters. However, this equation has the drawback of underestimating the impact of the invalid vote on electoral results, especially in a context of limited voter mobilization, as is the case at the municipal level.

It would therefore be more appropriate to calculate the proportion of the number of invalid ballots to the number of valid ballots. In other words, this ratio expresses the proportion of voters whose vote is invalid relative to the voters whose votes contributed to determine the winner of an election. This measure allows for comparison with the margin of victory of the winning candidate and for assessing the impact of voter turnout in light of the invalid vote. Equation 1 shows the operationalization of this variable.

Invalid voting = !"#$%!& !"#$ � 100 !"#$% !"#$

This measure of invalid voting moreover allows comparing the rate of invalid votes of Montreal with those of other elections that took place in Quebec. Table 1 presents invalid vote rates for provincial and federal elections in Quebec as well as for municipal elections of the province’s

5 five largest cities, whereby Montreal ranked highest.2 Invalid voting at municipal elections in Montreal was also nearly three times higher than at Quebec provincial and federal elections and, more generally, invalid voting was higher in municipal elections than in provincial and federal elections. It should be noted, moreover, that our decision to focus on a single municipality resulted from the diversity of electoral and institutional systems in place in Quebec, as elsewhere, which complicates a comparative analysis of invalid voting. Nevertheless, the city of Montreal offers great potential for conducting a quantitative analysis of this phenomenon. (Table 1 about here)

The invalid vote rate in the 2009 Montreal municipal election varied considerably across the territory. Taking the elected offices of the electoral districts as our unit of analysis, we calculated the invalid votes cast for each office of the city’s 58 electoral districts. The regression analysis presented in Table 3 (further in the text) is based on panel data of 201 observations that examined invalid voting with regard to different elected offices; 58 focused on the office of mayor, 54 on the office of borough mayor, 50 on the office of city councilor, and 38 on the office of borough councilor. According to these observations, the average invalid vote rate for the city of Montreal is 3.89%. Normally distributed, this methodological variable ranges from a minimum of 1.42% to a maximum of 9.92%. The descriptive analysis of the different variables is presented in Table 2.

Furthermore, the detailed analysis in Table 1 also reveals that elections in Montreal are more complex than in other cities in Quebec, which contributes to the value of this case study. The existence of four different types of elected offices appears to have an impact on invalid voting. More specifically, the invalid vote rate tends to decrease with the level of office. For example, less invalid votes are cast for the offices of city mayor (3.32%) and borough mayor (3.73%) than for the offices of city councilor (4.00%) and borough councilor (4.15%). These differences between the elected offices are moreover statistically significant.3 Finally, these distinctions also apply to other major Quebec cities.

3.2 Model estimation

As suggested in the above, we base our approach on the premise that invalid voting is determined by the institutional arrangement, social structure and presence of protest voting. Institutional arrangement is assessed on the basis of the complexity of the electoral system and a comparison of the invalid vote rates of the different types of elected offices. Social structure is evaluated for each in terms of its homeownership rate, median income, proportion of immigrants and proportion of university graduates. And protest voting is seen as a function of electoral performance of the party of the outgoing mayor, the number of contending candidates, voter turnout and district magnitude. A distribution of all variables of these three factors is shown in Table 2, while details concerning their operationalization are provided in the Annex. Our model can be described with the following equation:

2 In Quebec municipal elections, an invalid vote is considered to be a ballot that “has not been furnished by the deputy returning officer; has not been marked or has not been marked properly; has been marked in favor of more than one candidate; has been marked in favor of a person who is not a candidate; bears fanciful or injurious entries; bears a mark by which the elector can be identified; has been marked otherwise than with the pencil given to the elector by the deputy returning officer” (Act Respecting Elections and in Municipalities, RSQ) 3 Differences in invalid voting between the different elected offices in each of the electoral districts are statistically significant, be it when measured with a matched-sample t test or a Wilcoxon test.

6

������� ������ = � + �1 ����� ����� ����������� + �2 ������ �� ���������� + �3 ��������� ������������� + �4 �������� ��������� + �5 ��������� ���������� + �6 ������ℎ ����� + �7 ���� ��������� + �8 ������ℎ ��������� + �9 �������� ������ + �10 �������� ������ℎ�� + �11 �������� ����������� + �12 (�������� ����� �� ���������)

Given that our research was based on different elected offices within multiple electoral districts, we decided against applying an OLS (ordinary least square) regression type analysis. In fact, with this type of analysis there would have been the risk that our estimates become affected given the presence of autocorrelations in observations of one and the same electoral district as well as heteroscedasticity between electoral districts. This would have, in turn, contaminated the statistical significance tests of our estimates. Therefore, we use the regression with corrected standard errors for clustered samples method, a type of regression analysis that interprets data in the same way as OLS but that corrects estimates by taking into account the structure of the panel data. Developed by William(1993), this approach is designed to correct such problems. Thus, our model corrects estimates in assuming a heteroscedasticity of the error term across the panel data and a correlation within electoral districts. Table 2 presents the descriptive statistics of the variables in our analysis model, while the Annex provides more detailed definitions of each variable. In the following paragraphs, we offer an overview of the theoretical expectations for each of these variables.

(Table 2 about here)

3.3 The protest approach

The notion that invalid voting is an expression of protest was measured in lines 1 to 4 of the regression table (Table 3). The protest approach sees invalid voting as a means for voters to express their discontent with the government. From that perspective, applying the theoretical framework developed by Hirschman (1970), we see invalid votes as having a voice. Here, invalid ballots are distinguished from abstentions, whereby abstentions are seen as an “exit” version of voting behavior. The act of distinguishing between these two types of ballots suggests that invalid voting should be seen as a type of with voter participation.

Moreover, using a dummy variable, we analyzed the effects which the victory of an incumbent district mayor’s party has on invalid voting. With this variable, we wanted to find out whether invalid voting could be considered as a strategic vote on the part of voters. Our assumption was that if the objective of the voter is to send a message of dissatisfaction to those in power, invalid voting should be weaker in districts where the party of the incumbent mayor is strong, as voters in these districts would vote for the opposition candidate rather than waste their vote. By contrast, in districts where the party of the mayor is weaker, invalid voting should be higher, as it is here more useful as an expression of disapproval.

7

The phenomenon of voters deliberately casting invalid ballots for the purpose of making a statement is also the subject of two other hypotheses. Using a dummy variable, we measured the impact of having only three competing candidates in an election. The results were that the three candidates were always those of the three major parties at Montreal City Hall. Invalid voting should thus be higher when there are only three contending candidates, as the invalid vote here is an actual means to express disapproval of the party in power. Furthermore, invalid voting becomes a less useful option when there are more competing candidates. The effect of district magnitude on invalid voting, for its part, depends on whether or not a voter’s ballot can make a difference in an election. The larger the district, the less impact a vote will have. We thereby expect that the more voters in an electoral district, the greater the invalid vote rate.

3.4 The institutional approach

Invalid voting as a result of the institutional environment was measured by four variables, as shown in lines 5 to 8 of the regression table (Table 3). The complexity is measured with a dummy variable that compares districts where voters had to vote for four or five elected offices with those whose voters only had to vote for two or three offices. The assumption here is that the more complex an electoral system, the higher the invalid vote rate should be, because the level of information required of a voter increases. In addition, we used three dummies to compare invalid voting between types of elected offices. The reference value for interpreting these three variables is the invalid vote rate for the office of mayor. In this comparison, the assumption is that voters may attribute a different relative importance to each of these offices and that the invalid vote rate may be indicative of this phenomenon, as in the case of second-order elections.

3.5 The socio-demographic approach

Finally, in lines 9 to 12 of the regression table (Table 3), the impact of the socio-demographic factors was measured for each of the districts. The socio-demographic approach assesses voters’ level of competency. Invalid voting should be greater when an electorate is less competent, particularly with regard to the level of information, which is also related to the interest of a voter to become informed. Thus, we expect invalid voting to vary according to the social structure of each district. Variables taken into account are median income, home ownership, level of education and the proportion of the immigrant population.

4. Empirical results

The regression analysis in Table 3 clarifies some possible explanations of the mechanisms behind the formation of invalid voting. The model predicts, quite well, invalid voting in the 2009 Montreal municipal elections (R2= 0.64; F = 21.99; prob < 0.001) and explains the 64% variance in invalid voting by the independent variables. In addition, 10 of the 12 variables in the model have a statistically significant impact on invalid voting, namely in line with our theoretical expectations. Therefore, each of our three main categories of explanation had an impact on invalid voting. An examination of the standardized regression coefficient (ᵦ) sheds light on the relative weight of each of these categories. For example, the social structure of the electoral district was shown to have the most impact on invalid voting, followed by institutional

8 arrangement and protest voting. The following table provides more detailed information on the impact of each independent variable.

Table 3 (about here)

4.1 The protest approach

All other things remaining equal, cases where the incumbent mayor’s party received the most votes for an office within the district level presented a lower level of invalid voting than the other cases (b = -0.56; t = 2.19). This result is statistically significant (p <0.05) and is consistent with the idea that voters are more comfortable with canceling their vote if the incumbent party is defeated in an election. Thus, while an invalid vote may be a , it is not exempt from strategic considerations.

Additionally, cases where there are only three contending candidates—namely in Montreal, where they represent the three main parties—show a lower level of invalid voting than others (b = -0.26; t = -1.89n.s.). This result is not consistent with the hypothesis, because the greater the choice for voters, the less useful protest voting should be. However, this result is not statistically significant. Hence, as an indicator of voter protest, invalid voting does not seem to be associated with a rejection of the system in place.

Furthermore, an increase in voter turnout increases invalid voting (b = 0.05; t = 2.84). This result is statistically significant (p <0.01). When voter mobilization is strong, invalid voting is more pronounced. This result supports the idea that the profile of people who do not vote is not necessarily the same as that of voters who nullify their vote. In other words, a stronger voter mobilization generates higher protest voting.

Finally, an increase of 1,000 voters within an electoral district was shown to decrease invalid voting (b = -0,02; t = -1,78 n.s.). This result does not support our hypothesis, which states that the incentive to vote at all decreases with rising district magnitude, as the relative weight of each vote thereby decreases, and that this very situation then provides incentive to cast a protest vote. However, this result that runs counter to our hypothesis is not at all statistically significant.

4.2 The institutional approach

The complexity of Montreal’s electoral system has an influence on invalid voting. For example, cases where voters must vote for four or five different elected offices show a higher level of invalid voting than others (b = 0.64; t = 2.95). This result is statistically significant (p <0.01). Therefore, the complexity of the electoral system may partially explain the considerable presence of invalid voting in Montreal compared to other cities in Quebec.

All other things remaining equal, invalid voting varies depending on the type of elected office. In all cases, invalid voting is lower for city mayor than for any of the other city or borough offices. Invalid voting is highest for the office of borough councilor (b = 0.97; t = 5.61; p < 0.001), followed by the offices of city councilor (b = 0.70; t = 3.56; p < 0.01) and borough mayor (b = 0.52; t = 4.79; p < 0.001). These three results are statistically significant. This means that the

9 institutional complexity of the city of Montreal impacts invalid voting with regard to two aspects. First there is the effect resulting from the number of votes a voter must cast, and then there is the phenomenon that the invalid vote rate tends to differ for each office. This suggests that voters probably do not attribute the same importance to each of these offices. In sum, the statistical effect of the institutional variables is greater than that of the variables related to protest voting.

4.3 The socio-demographic approach

Of all variables, the socio-demographic variables have the most significant statistical effect on invalid voting. As presented in the regression table (Table 3), the logarithm of a district’s median income is positively associated with invalid voting (b = 2.30; t = 3.22; p < 0.01). The more affluent the social structure of an electoral district, the higher the invalid vote rate. Moreover, the higher the number of homeowners, the lower the invalid vote rate (b = -1.59; t = -4.03; p < 0.001). The proportion of immigrants (b = 0.06; t = 5.98; p < 0.001) and the proportion of university graduates in a district (b = -0.10; t = -8.06; p < 0.001) also contribute to a statistically significant explanation of invalid voting. In fact, based on the value of the standardized regression coefficient (ᵦ), these latter two variables have the greatest statistical effects of the model.

In sum, each of these three approaches explain how invalid votes are generated, yet from different angles. Overall, although socio-demographic factors are predominant, the weight of institutional factors cannot be ignored. And, the protest approach, less pronounced, may be indicative of a strategic dimension of voting behavior more so than a rejection on the part of voters of the range of choices.

5. Discussion

Our results allow us to conclude, following Uggla (2008), that invalid voting is the result of a combination of variables. The Montreal case study clearly shows that the institutional design and the electoral system play a role in the formation of invalid voting. The more complex the system, the higher the percentage of invalid voting is likely to be. This result echoes the findings of McAllister and Makkai (1993), who, on the topic of Australian national elections emphasize that “[i]f informal voting is to be reduced, more consideration needs to be given to simplifying the act of voting, or at the very least applying consistent rules for the plethora of elections that are conducted at the various levels of government.”

In addition, our study showed the presence of a hierarchy among the elected offices. For example, elected officials likely to being closer to voters, such as the office of borough councilor, was shown to receive the largest share of invalid votes. The latter implies that for some voters the office of city mayor remains the most important. There is thus a ranking based on the importance accorded to the type of elected office. These findings suggest, following Elmendorf et al. (2013), that municipal elections have their own logic within which offices are categorized according to first-order and second-order offices. The topic of hierarchization in turn raises the following question that merit further research: Is the hierarchization determined primarily by the design or structure of the ballot, by a lack of understanding of the powers held by these offices, by a more general lack of information, or by the lesser importance accorded to the offices of councilor?

10

Both institutional complexity and socio-economic variables were shown to influence invalid voting, whereby the specific role of socio-economic variables remains difficult to identify. If anything, income appears to play a role. For example, in districts with a more affluent social structure, the number of invalid votes is higher. Although this relation may seem surprising, there is a link between voting and the redistribution of wealth (Mahler, 2008). The local level has a regressive taxation system because the main tax base relies primarily on property tax (Binet & Pentecost, 2004). It is therefore consistent to find less invalid voting in poorer electoral districts as voters there have less to gain from canceling their vote. In addition, invalid voting is correlated negatively with the rate of homeownership. With property tax billed directly to homeowners, this variable is a key factor at the local level (Manturuk et al., 2009). Due to their tax account, homeowners have more electoral information at their disposal than renters. Finally, the value of their property is often affected by decisions made by local elected officials. It is therefore logical to assume that they are less willing to spoil their vote.

The variable of number of university graduates by district also affects invalid voting, namely, the greater the number of university graduates, the lower the invalid vote rate. Given the absence of data on individual voting behavior, this finding is difficult to interpret. Nevertheless, several questions may be formulated. Is it because those without university degrees do not vote? Or because university graduates have more access to information or take the time to become informed? Similar questions emerge with regard to social structures with a strong presence of immigrants. Does language play a role? Is the range of choices in elections inadequate for these communities? Are immigrants more prone to cancel their votes than non-immigrants? What role does information play in this equation?

Our study raised many questions, in particular because electoral behavior at the local level, as a whole, is little researched. We were able to identify institutional context, social structure and political context as factors that influence voters with regard to invalid voting. Together, our findings confirm the need to know more about the cost, in terms of effort, of obtaining information and of going to the polling station. Indeed, as pointed out by McAllister and Makkai (1993), “[i]n voluntary voting systems, eligible voters who are less interested in casting a ballot would normally be expected to abstain completely from the act of voting.” Or is it that voters’ sense of duty is strong enough to drive them to the polling stations, yet that they are then likely to invalidate their vote if they lack sufficient information or understanding of the issues? In any case, all of these elements bring us back to the central question of information and the cost, or effort, required for obtaining it. This topic should thereby constitute the focus of future research.

This research has also shown that in certain contexts, such as the municipal level, the invalid vote rate merits further investigation, especially when it is higher than the victory margin of the winning candidate. Here, the prevalence of invalid voting casts doubt on the legitimacy of certain elected officials. This position is supported by recent revelations of corruption in Quebec municipal politics and the subsequent succession of resignations of elected officials in the province throughout 2013, including the mayor of Montreal. Finally, the phenomenon of invalid voting may also be explained by citizens’ loss of confidence in the itself.

11 References

Bherer, L., Breux, S. 2012. L’apolitisme municipal. Bulletin d’Histoire politique 21 (1), pp. 170- 184. Bherer, L., et al., 2008. La participation et la représentation politique des femmes au sein des instances démocratiques municipales. Institut national de la recherche scientifique - Urbanisation Culture Société, Montreal. Binet, M.-E., Pentecôte, J.-S. 2004. Tax degression and the political budget cycle in French municipalities. Applied Economics Letters 11 (14), pp. 905-908. Breux, S., Bherer, L., 2011. Les élections municipales au Québec : enjeux et perspectives. Presses de l’Université Laval, Quebec. Breux, S., Bherer, L., 2013. The Democratic Contributions on Local Political Parties : A View from Canada. Political Studies Association, Cardiff, G.B. Cutler, F., Matthews, J. 2005. The challenge of municipal voting : Vancouver 2002. Canadian Journal of Political Science 38 (2), pp. 359-382. Elmendorf, C., et al., S. 2013. Lost in space ? Shorcuts and spatial voting in low-information elections. UC Davis Legal Studies Research Paper no 285, (328), pp. 1-45. Faure, A., 2009. Les élus locaux changent-ils la politique ? Méthode, controverses, énigmes…. In: Bidégaray, C., Cadiou, S., Pina, C. (Eds.) L’élu local aujourd’hui. Presses Universitaires de Grenoble, Grenoble, pp. 29-39. Henderson, A., McEwen, N., 2010. A comparative analysis of voter turnout in regional elections. Electoral Studies 29, pp. 405-416. Lightbody, J., 2006. City Politics in Canada. University of Toronto Press, Toronto. Loi sur les élections et les référendums dans les municipalités (LERM) 2012 (1978). Mahler, V.A., 2008. Electoral turnout and income redistribution by the state: A cross-national analysis of the developed democracies. European Journal of Political Research 47 (2), pp. 161-183. Manturuk, K., et al., 2009. Homeownership and local voting in disadvantaged urban neighborhoods. Cityscape 11 (3), pp. 213-230. McAllister, I., Makkai, T., 1993. Institutions, society, or protest ? Explaining invalid votes in Australian elections. Electoral Studies 12, pp. 23-40. Mévellec, A., 2011. Les élections municipales de 2009 dans les villes moyennes du Québec : entre changement et reconduction. In: Breux, S., Bherer, L. (Eds.) Les élections municipales au Québec : enjeux et perspectives. Presses de l'Université Laval, Quebec, pp. 289-310. Parodi, J.-L., 1983. Dans la logique des élections intermédiaires. Revue Politique et Parlementaire (903), pp. 42-72. Parodi, J.-L., 2004. Les élections “intermédiaires” du printemps 2004. Revue française de science politique 54 (4), pp. 533-543. Perez-Linian, A., 2001. Neoinstitutionnal accounts of voter turnout : moving behond industrial democracies. Electoral Studies 20, pp. 281-297. Peterson, P.E., 1981. City Limits. The University of Chicago Press, Chicago, London. Pion, G., 2010. Le vote blanc et nul en Wallonie : analyse écologique et individuelle. Belgeo 3, pp. 249-264. Power, T.J., Garand, J.C., 2007. Determinants of invalid voting in . Electoral Studies 26, pp. 432-444.

12 Rallings, C., Thrasher, M., 2005. Not all “second order” contests are the same : turnout and party choice at the concurrent 2004 local and European Parliament elections in England. The British Journal of Politics and 2005 (7), pp. 584-597. Reif, K., Schmitt, H., 1980. Nine second-order national elections – a conceptual framework for the analysis of European election results. European Journal of Political Research 8, pp. 3- 44. William,R.H. 1993. « Regression standard errors in clustered samples.» Stata Technical Bulletin 13, pp. 19-23.

Sellers, J., Lidström, A., 2008. Decentralization, local government and the welfare state. Governance 20, pp. 609-632. Trounstine, J., 2009. All politics is local : the reemergence of the study of city politics. Perspectives on Politics 7 (3), pp. 611-618. Trounstine, J., 2013. Turnout and incumbency in local elections. Urban Affairs Review 49 (2), pp. 167-189. Uggla, F., 2008. Incompetence, alienation, or calculation ? : Explaining levelds of invalid ballots and extra-parliamentary votes. Comparative Political Studies 41 (8), pp. 1141-1164. Wood, C., 2002. Voter turnout in city elections. Urban Affairs Review 38 (2), pp. 209-231. Zulfikarpasic, A., 2001. Le vote blanc : civique ou expression politique ? Revue française de science politique 51 (1-2), pp. 247-268.

13 Annexes

Table 1 Invalid voting in Quebec in the five largest cities and in provincial (2012) and federal (2011) elections

Election Elected office Invalid voting (%) Montreal Mayor 3.32 Borough Mayor 3.73 City Councilor 4.00 Borough Councilor 4.15 Provincial Island of Montreal constituency (33) 1.12 election 2012 Entire constituency (125) 1.23 Federal election Island of Montreal constituency (17) 1.24 2011 Entire constituency (75) 1.35 Gatineau Mayor 1.61 City Councilor 2.18 Laval Mayor 2.47 City Councilor 3.20 Longueuil Mayor 2.93 City Councilor 2.65 Quebec City Mayor 1.61 City Councilor 1.65

14

Table 2 Descriptive statistics Variable Mean Minim Maximu Standa Skewn Kurtos um m rd ess is deviati on Invalid voting (%) 3.89 1.42 9.92 1.36 0.96 1.33 Voter participation (%) 40.54 28.32 58.28 6.32 0.51 0.17 District magnitude (by 19.29 2.58 29.03 7.63 -1.01 -0.26 thousand) Number of candidates (only 0.31 0 1 0.46 - - 3) Complexity (4 or 5 votes by 0.54 0 1 0.50 - - voters) Mayor party performance 0.56 0 1 0.50 - - (win district) District median income ($) 56,778 35,327 148,753 20,206 2.35 6.45 Log district median income 10.90 10.47 11.91 0.29 1.38 1.95 District ownership (%) 38.99 15.00 97.90 17.37 1.36 1.74 Log district ownership 3.58 2.71 4.58 0.41 0.37 -0.16 District education (academic 30.40 11.3 69.40 13.77 0.85 0.03 degree in %) District immigration (%) 28.13 8.60 59.80 12.32 0.74 -0.11 Source: The 2009 Rapport officiel du recensement des votes of the city of Montreal and website of Ville de Montréal.

15

Table 3 Regression analysis Dependant variable: Invalid voting (%)

Expected sign in parenthesis ( ) b t ᵦ Intercept -16.19 -2.45* 0.0 Independant variable Protest variables 1) ����� ����� ����������� (-) -0.56 -2.19* -0.21 2) ������ �� ���������� (+) -0.26 -1.89n.s. -0.09 3) ��������� ������������� (+) 0.05 2.84** 0.23 4) �������� ��������� (+) -0.02 -1.78n.s. -0.11 Institutional variables 5) ��������� ���������� (+) 0.64 2.95** 0.24 6) ������ℎ ����� (+) 0.52 4.79*** 0.17 7) ���� ��������� (+) 0.70 3.56*** 0.22 8) ������ℎ ��������� (+) 0.97 5.61*** 0.40 Socio-demographic variables 9) Log �������� ������ (+) 2.30 3.22** 0.49 10) Log �������� ������ℎ�� (-) -1.59 -4.03*** -0.48 11) �������� ����������� (+) 0.06 5.98*** 0.61 12) �������� ��������� (-) -0.10 -8.06*** -1.01 (N) (201) R2 0.64 F 21.99*** OLS regression with corrected standard errors for electoral district affiliation (William,1993) estimated with Stata. The standardized regression coefficient (ᵦ) is calculated by multiplying the unstandardized coefficient (b) with the ratio of the standard deviation of each independent variable to the standard deviation of the dependent variable (i.e., sX/sY). * p < 0.05; ** p < 0.01; *** p < 0.001; n.s.: not significant

16 Table 4 Description of variables Dependent variable Operationalization Invalid voting ������� ����� � 100 ����� ����� Independent variable Voter participation (%) ����� ���� � 100 ���������� ������ District magnitude (by thousand) ���������� ������

1000 Number of candidates (only 3) = 1 if only three candidates compete for this office Complexity (4 or 5 votes by voters) = 1 if the voter of the district must vote for 4 or 5 different offices Mayor party performance (win district) = 1 if the winner of the district for this office belongs to the party of the incumbent mayor (Union Montréal) ������ℎ ����� = 1 if it concerns an election for the mayor of the borough ���� ��������� = 1 if it concerns an election for city councilor ������ℎ ��������� = 1 if it concerns an election for a borough councilor Log district median income Log district median income ������ �� �������� Log district ownership (%) ��� � 100 !"!#$ ��������� �� �������� District immigration (%) !"#!$" !"#$ !"#$%&' !"#"$" � 100 ����� �������� !"!#$%&'"( District education (academic degree in !"#$%"&$ !"#!$" !"#! !"!#$%&" !"#$"" � 100 %) ����� �������� !"!#$%&'"( Source: The 2009 Rapport officiel du recensement des votes of the city of Montreal and website of Ville de Montréal.

17