New Voters and Old Voters: Understanding Volatility in Quebecers’ Federal Election Votes between 2008 and 2019

Abstract

Federal elections between 2008 and 2019 saw a great of volatility in Quebec, with important consequences for election outcomes. The surge in New Democratic Party (NDP) support in Quebec led the party to official opposition, while Liberal gains in 2011 led the party to a majority government, and Bloc Québécois gains in 2019 helped to reduce the Liberals to a minority. To what extent was this volatility driven by voters switching parties and to what degree was it driven by voters entering and exiting the electorate? This article uses ecological inference based on riding-level data to examine the dynamics of party competition in Quebec from 2008 to 2019. We show that while voter mobilization mattered to volatility, vote switching was the important driver of changing party fortunes during this period.

Résumé

Les élections fédérales entre 2008 et 2019 ont connu une grande volatilité au Québec, avec des conséquences importantes sur les résultats électoraux. La poussée du soutien du NPD au Québec a conduit le parti à l'opposition officielle, tandis que les gains des libéraux en 2011 ont conduit le parti à un gouvernement majoritaire et que les gains du Bloc Québécois en 2019 ont contribué à réduire les libéraux à une minorité. Dans quelle mesure cette volatilité est-elle due au fait que les électeurs changent de parti et dans quelle mesure est-elle due au fait que les électeurs entrent et sortent de l'électorat ? Cet article utilise l'inférence écologique basée sur des données au niveau des circonscriptions pour examiner la dynamique de la concurrence entre les partis au Québec de 2008 à 2019. Nous montrons que, bien que la mobilisation des électeurs ait joué un rôle dans la volatilité, le changement de parti a été le principal moteur de l'évolution de la situation des partis au cours de cette période.

Introduction

The 2008–2019 period saw a great deal of volatility in federal voting patterns in Quebec. The province went from the Bloc Québécois and Liberals being the strongest parties in the province, to the orange wave of the New Democratic Party (NDP) in 2011, to a Liberal resurgence in 2015, to a return to Bloc Québécois versus Liberal competition in 2019. How much of this volatility was the result of voters switching parties and how much was a result of voters deciding whether or not to vote? There is a robust literature in the United States (for example, Hill, 2017) that finds election results are shaped by voters switching between parties and by which voters turn out to vote. The importance of vote switching and of voter turnout, however, are not always equal and depend on the contexts in which elections take place. This article investigates the relative importance of vote switching and voter mobilization during a period of high volatility in federal voting in Quebec.

We examine volatility in Quebec because of its importance to overall Canadian election results. Throughout the twentieth century Quebec was the pivot in Canadian elections; the party that won in Quebec typically won the election (Johnston, 2015; Johnston et al., 1992). While the Harper Conservatives demonstrated that they could form a government without many seats in Quebec, the province remains central to the success of the Liberals and NDP. With the Conservatives taking larger numbers of seats in Western Canada, any path to a Liberal government runs through Quebec. Similarly, that the NDP's national 2011 rise and 2015 fall mirrored the party's fortunes in Quebec demonstrates that the NDP needs large seat shares in the province if it wants to be anything more than a third party. Understanding the changing dynamics of federal elections in Quebec is therefore essential to understanding the Canadian party system's evolution.

Whether the volatility in Quebec was a result of vote switching or voter mobilization matters to how scholars understand Canadian elections. If the volatility was a result of vote switching, then the key thing to watch in elections in Quebec is parties’ ability to identify swing voters and find ways to pull those voters into their electoral coalition. For example, if the NDP made its gains by winning voters from the Bloc Québécois, then a central determinant of the NDP's future success in Quebec—and, as a result, in federal elections—will be its ability to win over such nationalists from the Bloc. However, if voter mobilization is the key driver of volatility, parties’ differing abilities to identify and mobilize supporters will be the central determinant in different parties’ success.

Using ecological inference, we find that both vote switching and voter mobilization mattered to volatility in Quebec but that they did so in different ways. Vote switching was a large factor in the rise of the NDP in 2011, the resurgence of the Liberals in 2015, and the return of Bloc Québécois to competitiveness in 2019. While the NDP in 2011 and Bloc Québécois in 2019 managed to make some gains by mobilizing voters who had not voted in previous elections, both parties made most of their gains by winning over voters from other parties. Indeed, much of the vote switching that occurred between the 2008–2019 period happened as voters moved back and forth between the Bloc Québécois and NDP, though the NDP also lost a significant number of voters to the Liberals in 2015. For the Liberals in 2015, it is less clear whether vote switching or voter mobilization mattered more. This points to vote switching as the more important explanation of volatility in Quebec during the 2008–2019 period; ¹ voter mobilization also mattered, but to a lesser degree.

Two Theories on Electoral Volatility in Quebec

Theory 1: Changing electoral coalitions

The first potential explanation for the volatility in Quebec between the 2008 and 2019 elections is vote switching, driven by changes in the preferences of the electorate. There is an extensive literature on vote choice that provides many explanations for why voters may switch parties. The Columbia school of voter behaviour, highlighted by Berelson et al. (1954), argues that voters who belong to social groups with different partisan tendencies face competing pressures and tend to switch their votes from one party to another. The Michigan school, connected to Campbell et al. (1960), presents a much more complex understanding of voter behaviour. In this model, voter decisions are influenced not just by one's social connections but also by values, views on issues, and evaluations of party leaders. Economic circumstances can also lead voters to switch their support from one party to another, as voters reward incumbents for good economies and punish them for bad ones (Lewis-Beck et al., 2008). These works have provided the foundations for models of vote choice applied to both Canada (Gidengil et al., 2012) and the United States (Miller and Shanks, 1996) that show all of these factors shape voters’ decisions. As a result, competing social pressures, as well as changes in candidate evaluation, in the issues that are important in an election, and in the views that voters have on issues, can be expected to lead to changes in vote choice.

Rational choice theories of voter behaviour offer another potential explanation for why voters switch parties. Here, Downs's (1957) foundational work argues that voters will vote for the party that offers a policy program closest to their own views. This logic suggests that as parties shift their positions, voters should shift their votes as the party closest to them changes. Blais et al. (2001) demonstrate that this proximity approach to voting is relevant in the Canadian context and that voters tend to prefer parties with positions closer to their preferences.

Complicating all of these voting models is the strategic voting that takes place in Canadian elections (Merolla and Stephenson, 2007). Because voting for a third-place party in Canada's single member plurality electoral system rarely has an impact on the outcome of an election, some voters will vote strategically for one of the two strongest parties in their region or riding (Merolla and Stephenson, 2007). This means that when polls shift to suggest that a party is starting to become stronger, voters may start to switch to that party for strategic reasons.

We take the view that all of these models offer some insight into at least some voters’ decision making. Decision making by voters is often complicated, and we expect that because different voters think about voting differently and are influenced by different factors, different models will apply better to different voters. Rather than trying to present a model of voter decision making in Quebec over this period of time (such models have been offered by Fournier at al. [2013] and Gauvin et al. [2016]), we argue that each of these models gives us some reason to expect substantial amounts of vote switching in Quebec between 2008 and 2019.

The shifting importance of sovereignty and national identity point to reasons why voters would switch parties that line up well with both the Columbia school sociological model of voting and the Michigan school model. Through the 1990s and 2000s, federal elections in Quebec tended to be contests focused on Quebec's place within Canada. These debates were well suited to both the Bloc Québécois and Liberals. In line with Berelson et al.'s (1954) discussion of social groups as influencing vote choice, the Bloc Québécois maintained an advantage among francophones while the Liberals maintained an advantage among anglophones. If one considers the more complex Campbell et al. (1960) model of vote choice, one could look to the Bloc Québécois as maintaining an advantage among separatist voters while the Liberals maintained an advantage with federalists. So long as these issue and identity divides remained central to federal politics in Quebec, it would be difficult for the Conservatives or NDP to gain a foothold in the province. They had neither the ability to appeal to a particular social group nor the ability to speak to national identity questions in a way that differed from the two strongest parties in the province. Indeed, the fundamental disagreements between many Western Canadians and Quebecers over national identity issues made it difficult for the Conservatives or NDP to speak to these debates credibly (Johnston, 2017).

The challenges that the Conservatives and NDP faced in Quebec can also be matched with spatial models of voting to explain why the Conservatives and NDP struggled to win support in Quebec in the 1990s and early 2000s. Competition structured around Quebec's place in Canada gave both the Bloc Québécois and Liberals clear bases of support (separatists for the Bloc Québécois and federalists for the Liberals). Both parties were then somewhat free to shift to the centre on national identity issues to compete over soft nationalists. The two parties’ competition over soft nationalists would have left the Conservatives and NDP little room to establish their own bases of support on national identity questions, squeezing them out of competition over one of the most important issues in Quebec during this time period.

The decline of the importance of separatism offered the Conservatives and NDP opportunities to steal votes from both the Bloc Québécois and Liberals. The failure of the 1995 referendum and the decreasing likelihood that Quebec would see another referendum on sovereignty undermined the importance of the central issue that the Bloc Québécois campaigns on. For a time, the party was able to campaign against Liberal corruption, leveraging the particular importance to Quebec voters of the sponsorship scandal (in which the Liberal government came under investigation for giving contracts to Liberal-friendly advertising firms in Quebec in exchange for little work). This allowed the Bloc to cast itself as the defenders of Quebec's interest against a Liberal government perceived as corrupt (Gagnon and Hérivault, 2004). The Conservatives and NDP, however, could also present themselves as defenders of Quebec against Liberal corruption in a way that they could not present themselves as defenders of Quebec on questions of national identity. This led to an early, though muted, Conservative breakthrough in the province in 2006. In that election, the Conservatives finished second in vote share in Quebec, with 25 per cent of the vote, and won 10 seats.

The decline of the importance of national identity questions also offered the NDP an opportunity to win support from the Bloc Québécois and Liberals. Since 2004, the NDP had been building its organizational capacity in Quebec with an eye to potential appeal of the NDP's social democratic values to large numbers of Quebec voters (Lavigne, 2013; McGrane, 2019). The adoption of the Sherbrooke Declaration in 2005, which committed the party to repealing the Clarity Act² and to allowing the province to opt out of federal programs with compensation, was part of the party's effort to strengthen their appeal in Quebec (NDP, 2005). In the early 2000s, the NDP made compromises on questions of national identity, hoping that doing so would allow them to win over left-wing Quebecers. As national identity issues declined in their importance between the early 2000s and 2011, this strategy should have become more and more effective in getting left-leaning Bloc Québécois and Liberal voters who had not backed the NDP because they wanted to support a federalist or separatist party to switch to the NDP.

Existing work on the NDP and the 2011 election suggests that the NDP was able to draw voters from the Bloc Québécois by demonstrating shared support for left-wing values and social issues. Fournier et al. (2013) show that those who were more left-leaning were more likely to support the NDP in 2011. More specifically, Gauvin et al. (2016) show that the NDP was able to exploit the fact that Bloc Québécois voters held left-leaning economic and social values to get those voters to switch their support to the NDP in 2011.

Changing evaluations of party leaders can also cause voters to switch their support from one party to another. Leadership is identified by Campbell et al. (1960) as an important driver of vote choice, and Bittner (2011) also demonstrates that leader evaluations matter to decisions voters make over which party to support.

The appeal of NDP leader Jack Layton in 2011 stands out as an explanation of why Quebec voters would switch their support to the Liberal Party. Though Layton made his political career in Toronto municipal politics, he was born and raised in the Montreal area and could present himself as having a personal connection to Quebec. Layton's appearance on the popular Quebec talk show “Tout le monde en parle” was viewed highly favourably (Fournier et al. 2013). By contrast, Bloc Québécois leader Gilles Duceppe had been leader of his party since 1997 and struggled to attract attention to his campaign. The Liberal Party also had leadership issues in Quebec going into the 2011 election. Both Michael Ignatieff and previous leader Stéphane Dion had presided over internal conflict within the Quebec wing of the federal Liberal Party, and neither had managed to establish themselves as popular in the province (Jeffrey, 2021). Ignatieff also faced questions regarding his prior support for the American invasion of Iraq, accusations that he had supported the use of torture, and a concerted campaign by the Conservatives to label him as “just visiting,” given his long career in academia outside of Canada. All of these things may have damaged his popularity.

The idea that NDP gains in 2011 were a result of vote switching fits a number of theories of vote choice. The declining importance of national identity issues made it more likely that left- wing groups would share values and issue positions with the NDP. This fits with both the Michigan school and rational choice approached to vote choice. The appeal of leader Jack Layton in Quebec would have also increased the likelihood that voters who care most about party leaders would switch over to the NDP. This leads to the following hypothesis.

H1a: Gains made by the NDP in 2011 came from voters who supported other parties in 2008 instead of voters who did not vote in 2008.

Models of vote choice also provide reasons to believe that the resurgence of the Liberals in 2015 and then the Bloc Québécois in 2019 resulted from vote switching. The return to importance of issues related to national identity may have reversed many of the gains that NDP made by connecting to voters in Quebec that were left-leaning on social and economic issues. In 2015 a mid-election court ruling reversing the Conservative government's decision to ban niqabs from citizenship ceremonies made secularism and religious accommodation an issue for voters in Quebec (Bridgman et al., 2021). The issue remained important in Quebec in 2019 as federal parties were forced to consider whether they should intervene in court challenges to provincial legislation banning government employees and those receiving provincial government services from wearing religious symbols. These issues put the NDP in an awkward position. While voters with left-wing views on economic and social issues outside of Quebec tend to favour religious accommodation and oppose Quebec's secularism legislation, the opposite is true for voters with left-wing economic and social views inside of Quebec (Turgeon et al., 2019). Debates over religious accommodation and secularism force the NDP to choose between supporting the views of their supporters in Quebec and their supporters outside of Quebec. To a large degree, the NDP has taken pro-multicultural and religious accommodation positions, aligning themselves with the views of their voters outside of Quebec and against those of many of the Québécois nationalist voters the party won over from the Bloc Québécois in 2011.

Analyses of the 2015 election suggests that the NDP lost votes as a result of their opposition to banning niqabs at citizenship ceremonies (Bridgman et al., 2021; Stephenson et al., 2019). Stephenson et al. (2019) show that the 2015 collapse in NDP support was followed by a rise in support for the Liberals only once the Liberals had become the strongest alternative to the Conservatives. Strategic voting considerations may have thus interacted with the decline of the NDP to push anti-Conservative Québécois nationalists from the NDP to the Liberals. This was the case even though the Liberals are also a pro-multicultural party. In 2019, the resurgence of the Bloc Québécois suggests competition that is much closer to what was seen prior to the rise of the NDP in 2011, with the Bloc Québécois winning the support of Québécois nationalists. In 2019, the Bloc would have been winning nationalist support because of their support for secularism and not their support for separatism. In contrast, the Liberals would have been the pro-multicultural option in Quebec, as well as the federalist option.

As with the rise of the NDP, leadership likely played a role in the resurgence of both the Liberals and the Bloc Québécois. The Liberals likely benefited from the selection of Justin Trudeau as leader. In contrast to the turmoil under Dion and Ignatieff, Trudeau was able to present himself as a charismatic leader and appeal to Canadians across the country (Jeffrey, 2021). This may have contributed to Liberal gains both inside of and outside of Quebec (Clarke at al., 2017). Trudeau also likely benefited from a native-son effect, as Johnston (2019) shows that leaders from Quebec tend to do better in the province. NDP leader Thomas Mulcair may have benefited from the same effect in 2015, given that he is also a Quebecer.

The Bloc Québécois may have benefited in 2019 from the stabilization of their leadership, with Yves-François Blanchet providing effective and charismatic leadership after the party cycled through several leaders in the period between the 2011 and 2019 elections. In contrast, leadership likely hurt the NDP in 2019. The selection of Jagmeet Singh, a Sikh who wears a turban, would have highlighted the gap between the NDP and a large number of Quebec voters on questions of religious accommodation and secularism. Indeed, Bouchard (2022) shows that Quebecers had particularly low approval ratings of Singh.

If the declining importance of national identity questions and the weakness of the Liberal leadership led voters to switch from the Bloc Québécois and Liberals to the NDP in the 2011 election, the growing importance of identity questions and the strengthening of the Bloc Québécois and Liberal leadership in 2015 and 2019 should have had the opposite effect. Quebec voters who moved to the NDP because it offered them a left-wing option on economic and social policy should have switched back to either the Bloc Québécois or the Liberals, depending on where they stood on questions of national identity. Similarly, voters who became disillusioned with the NDP because the party did not reflect their views on national identity questions should have switched to another party, likely the Bloc Québécois, instead of staying home. As with the rise of the NDP in 2011, the way the 2015 and 2019 elections played out in Quebec can be connected to different theories of vote choice to generate the following hypotheses that volatility in Quebec was driven by vote switching.

H1b: Gains made by Liberals in the 2015 came from the NDP and not from voters who did not vote in 2011.

H1c: Gains made by the Bloc Québécois in 2019 came from the NDP and not from voters who did not vote in previous elections.

Theory 2: voter mobilization and demobilization

An alternative explanation for the volatility in parties’ success in Quebec lies in changes to voter turnout. While parties can gain and lose support when voters switch from one party to another, their success is also affected by voters’ decisions to either turn out to vote or stay home. Even partisans do not always vote, and Hill (2014) shows that variations in partisan turnout can make a difference in close elections. Looking at American elections, Hill (2017) finds that changes in who votes can have as much as twice as large an effect on election results as voters switching between parties. As polarization in the United States increases, the extent to which voters switch parties may be declining (Smidt, 2017). As a result, voter mobilization may be becoming more and more important (though the multidimensionality of Quebec politics likely means polarization plays out differently in Quebec than in the United States). Parties’ ability to identify and motivate their voters to go to the polls on election day can be as important to their success as their ability to win over voters from other parties.

There is a complicated literature grappling with the different factors that affect voter turnout. The Downsian rational choice approach to voter behaviour suggests that voters’ decisions to vote depend on whether they have a strong enough preference between two candidates running to make it worth taking the time to vote. This rational choice logic also suggests that voters will be more likely to vote if they believe their vote will affect the outcome of an election, though the low likelihood that any one vote will be the difference in an election suggests that voter turnout should be very low if this model is correct (Downs, 1957).

Blais and Daoust (2020) present four explanations for why individuals vote, suggesting that the decision to vote depends on the degree to which the individual likes politics, whether they feel they have a duty to vote, whether they care about the outcome, and whether voting is easy. Three of these four explanations are unlikely to change from election to election and thus cannot be used to explain changes in election results. Whether a voter cares about the outcome, however, may change with the issues that are important in an election and which parties are competitive. A Quebec voter with strong opinions about economic redistribution may care little about an election where the main issue is national identity and the main competitors are the Bloc Québécois and Liberals. Such a voter may care a great deal about an election where economic redistribution is being debated and the NDP appears to have a chance at winning substantial numbers of seats.

Parties’ actions may also affect voter turnout. Candidates’ positions can matter, as Hall and Thompson (2018) show that extreme candidates motivate both their supporters and opponents to vote (with the opponents motivated outnumbering supporters). How aggressive parties are in their get-out-the-vote efforts also matter to turnout. Green and Gerber (2004) show that canvassing can increase voter turnout but that effective efforts to increase turnout require extensive resources. The degree to which a party is well organized and has resources to commit to mobilizing its supporters is thus likely to matter to its ability to mobilize supporters who are on the fence about whether to vote or stay home. Finally, voters may drop out of the electorate if they become dissatisfied with their preferred party (Adams et al., 2006). Voters may become disaffected by a party's poor performance in office, by problems in party leadership, or by party positions that are out of step with supporters. Such voters, however, may be too closely aligned with their party to vote for a competitor or may simply also dislike the alternatives they are presented with and choose to stay home rather than switch to another party.

There are good reasons to believe that these explanations of voter turnout would apply to turnout in Quebec between 2008 and 2019. The multidimensionality of Quebec's politics may make it more likely that different voters will see different elections as important to them. Individuals who care a great deal about national identity questions may be eager to vote when Quebec's place in Canada or identity and religious accommodation are central election themes. Such voters may be less likely to vote if those issues are not debated and instead economic or social issues are the focus of elections. The reverse may be true for voters who care less about national questions and more about economic or social issues. If changes in issue importance affect turnout, they should also affect support for parties. When individuals who care about national identity vote, support for the Bloc Québécois and Liberals should go up, as those parties are well positioned to speak to those issues (the Bloc Québécois to separatist voters, and the Liberals to federalists). When voters who care about economic or social issues and take left-leaning position on those issues turn out, support for the NDP should go up, as it is generally seen as a left-leaning party on these issues.

Voter mobilization efforts may have also affected turnout in ways that benefited some parties and hurt others. The speed and unexpectedness of the NDP's rise make it less likely that the party was able to make gains in 2011 by more effectively mobilizing voters. The comparative weakness of the NDP's local organizations in Quebec in 2011 meant that their local get-out-the- vote efforts were likely weaker than their competitors’, though the other factors discussed in this section may have helped the NDP mobilize new voters. However, the NDP may have benefited from organization problems that the Liberals had in the 2008 and 2011 elections. Both Stéphane Dion and Michael Ignatieff had problems developing strong relationships with members of the Quebec party. Both had Quebec lieutenants resign due to the neglect of local Liberal organizations in Quebec and conflicts between themselves and the party leadership (Jeffrey, 2021). These weaknesses may not have benefited the NDP directly, but lower levels of Liberal voter turnout would have reduced Liberal vote shares and made it easier for the NDP to win seats. In 2015, the Liberals made substantial improvements to their local organizations both in and outside of Quebec in the lead up to the 2015 election (Cross, 2016; Jeffrey, 2021). Revitalized Liberal local organizations could have been more effective at identifying Liberal supporters and ensuring that they got to the polls.

Changing perceptions of the parties may have finally affected whether voters were willing to turn out for the party that they normally supported. In the 2008 to 2011 period, the lingering effects of the sponsorship scandal may have depressed Liberal voter turnout. Voters frustrated by Liberal corruption who did not like the other parties running may have dropped out of the electorate during this time period. By 2015, the fading of the memory of the sponsorship scandal coupled with the revitalization of the Liberal Party may have caused these voters to turn out to vote Liberal once again.

Perceptions of the competitiveness of the Bloc Québécois and NDP at different times may have caused their voters to either stay home or enter the electorate depending on the election. Prior to 2011, Quebec voters sympathetic to the NDP may have reasonably believed that the party had such a small chance of winning seats in the province that it was not worth taking the time to vote. The sudden surge in NDP poll numbers in the middle of the 2011 election campaign may have convinced these voters that a vote for the NDP could now make a difference and that showing up on election day was worth their time. By contrast, the collapse of the Bloc Québécois and the leadership troubles the party experienced between 2011 and 2015 may have led many Bloc Québécois supporters to believe that the party was not competitive enough to make turning out to vote worth their time. Yves-François Blanchet's takeover of the party in 2019 and its perceived return to competitiveness may have led those voters to re-enter the electorate.

As with vote switching, we do not take a position on which of these explanations of changes in voter turnout are most plausible. Rather, we test changes in voter turnout as a broad explanation for volatility in Quebec against the vote switching explanations in the previous subsection of the article. We expect that the Liberals suffered from demobilization between 2008 and 2011, that the NDP made some gains as a result of mobilization in 2011, and that the increases in Liberal and Bloc Québécois support in 2015 and 2019 respectively came about as those parties were able to remobilize their supporters. These four hypotheses are expressed below.

H2a: The Liberals lost votes in 2011 because some 2008 Liberal voters did not vote in 2011.

H2b: Gains made by the NDP in 2011 came from 2008 non-voters choosing to vote NDP.

H2c: Gains made by the Liberals in 2015 came from 2011 non-voters choosing to vote Liberal.

H2d: Gains made by the Bloc Québécois in 2019 came from non-voters in previous elections choosing to vote Bloc Québécois.

Methodology

While election surveys are the most common way of examining vote choice, they present problems when evaluating the relative importance of vote switching and voter turnout to parties’ success. Election surveys almost always underestimate the number of non-voters because people who do not vote are less likely to answer survey questions and because social desirability bias often leads respondents to not admit that they did not vote (Achen, 2019). In our case, the Canadian Election Study provides the following estimates for the percentage of non-voters in Quebec: 11.7 per cent in 2019, 11.4 per cent in 2015, 15.1 per cent in 2011, and 8.9 per cent in 2008 (Fournier et al., 2015; Fournier et al., 2011; Gidengil et al., 2008; Stephenson et al., 2020). This compares with non-voter percentages reported by Elections Canada of 32.7 per cent in 2019, 32.7 per cent³ in 2015, 37.1 per cent in 2011, and 38.3 per cent in 2008 (Elections Canada, n.d.). Analysis of our hypotheses using survey data would therefore end up underestimating the percentage of the eligible voters who did not vote and therefore would overestimate the degree to which volatility in Quebec was a result of vote switching and underestimate the degree to which it was a result of voter mobilization. It is possible to reweight responses to reflect actual vote shares, but that requires an assumption that non-voters who answer surveys are representative of all non-voters. We do not feel confident making this assumption, as political disinterest is likely to affect non-voters’ willingness to answer surveys and whether they go from not voting in one election to voting in a subsequent election. Non-voters who do not vote for reasons other than disinterest (such as dissatisfaction with the parties in the election or belief that their vote does not matter) are likely to be overrepresented in elections surveys, and we expect that this would bias our analysis.

In the United States, voter files are made publicly available and are often used in analyses of voter turnout to get around the problems created by non-voters’ low response rates to elections surveys (see Hill [2017] for an example). Unfortunately, in Canada, voter files are not publicly available (Achen, 2019) and so cannot be used for this analysis. In the absence of survey data that reliably capture non-voters’ behaviour or publicly available voter files, we turn to ecological inference.

We follow Antweiler (2007) and Judge et al. (2003) in using a generalized maximum entropy (GME) estimator to estimate a Markov transition matrix between pairs of elections. This technique is used for ecological inference when we are interested in estimating the transitions between states and when we want to estimate shares that add up to exactly 1. Golan and Perloff (2002) discuss the variance of the GME estimator.

As explained in Antweiler (2007), for a pair of elections with n political parties (including non-voting as an option), we estimate an n  ×  n transition matrix. The sum of the shares of where each party's vote goes (or comes from, when relevant) are constrained to add up to 1, thus accounting for all votes exactly.

The Markov transition matrix method can be used to make estimates in both directions. In other words, we can estimate where the votes in the first election went (using the standard Markov method, as in Antweiler 2007), or we can estimate the sources of votes in the second election. For the latter calculation, the system of equations is mathematically equivalent to running the Markov process in reverse, with the first election as the independent variable and the second election as the dependent variable.

We proceed by estimating the Markov transition matrix between the first and the second election, or between the second and the first if we want to estimate vote sources as above. We use seven categories: votes for the Liberals (L), Conservatives (C), NDP (NDP), Bloc Québécois (BQ), Greens (G), all other parties (O),⁴ and non-voters (NV). We divide each total by the total number of eligible voters in the riding to obtain the share of the total electorate in each category.

Specifically, we have this system of equations as our model to estimate, where subscript 1 is the first election chronologically and subscript 2 is the second, with some error e corresponding to each party. The coefficients β _i,j correspond to the share of votes from party i going to party j.

$$\eqalign{ L_2 = & \beta _{L, L}L_1 + \beta _{C, L}C_1 + \beta _{NDP, L}NDP_1 + \beta _{BQ, L}BQ_1 + \beta _{G, L}G_1 + \beta _{O, L}O_1 \cr & + \beta _{NV, L}NV_1 + e_L}$$

$$C_2 = \beta _{L, C}L_1 + \beta _{C, C}C_1 + \beta _{NDP, C}NDP_1 + \beta _{BQ, C}BQ_1 + \beta _{G, C}G_1 + \beta _{O, C}O_1 + \beta _{NV, C}NV_1 + e_C$$

$$NDP_2 = \beta _{L, NDP}L_1 + \beta _{C, NDP}C_1 + \beta _{NDP, NDP}NDP_1 + \beta _{BQ, NDP}BQ_1 + \beta _{G, NDP}G_1 + \beta _{O, NDP}O_1 + \beta _{NV, NDP}NV_1 + e_{NDP}$$

$$BQ_2 = \beta _{L, BQ}L_1 + \beta _{C, BQ}C_1 + \beta _{NDP, BQ}NDP_1 + \beta _{BQ, BQ}BQ_1 + \beta _{G, BQ}G_1 + \beta _{O, BQ}O_1 + \beta _{NV, BQ}NV_1 + e_{BQ}$$

$$G_2 = \beta _{L, G}L_1 + \beta _{C, G}C_1 + \beta _{NDP, G}NDP_1 + \beta _{BQ, G}BQ_1 + \beta _{G, G}G_1 + \beta _{O, G}O_1 + \beta _{NV, G}NV_1 + e_G$$

$$O_2 = \beta _{L, O}L_1 + \beta _{C, O}C_1 + \beta _{NDP, O}NDP_1 + \beta _{BQ, O}BQ_1 + \beta _{G, O}G_1 + \beta _{O, O}O_1 + \beta _{NV, O}NV_1 + e_O$$

$$NV_2 = \beta _{L, NV}L_1 + \beta _{C, NV}C_1 + \beta _{NDP, NV}NDP_1 + \beta _{BQ, NV}BQ_1 + \beta _{G, NV}G_1 + \beta _{O, NV}O_1 + \beta _{NV, NV}NV_1 + e_{NV}$$

With the shares adding to one for each party:

$$\mathop \sum \limits_j^{} \beta _{i, j} = 1\;\forall \;i\;_{}$$

To find the sources of votes in the second election, simply switch subscripts 1 and 2. In that case, the sources of votes will be shares of the first election party total that add up to 1.

Data for our analysis come from election results made publicly available by Elections Canada (Elections Canada, n.d.a, n.d.b, n.d.c, n.d.d). We use the 2011–2015 riding crosswalk to adjust for boundary changes. For 2019, the People's Party of Canada is included in the “Other” category.

One simplifying assumption made, as in Antweiler (2007), is that there are no changes in the total number of eligible voters. In other words, we assume that the population of eligible voters is the same in both elections, with no entries or exits into the broader electorate. While there are individuals who become eligible to vote (either by turning 18 or becoming Canadian citizens), pass away, move to Quebec, or leave Quebec between elections, our assumption is that these changes to the number of eligible voters in each riding is not large enough to meaningfully influence results. To test this, we perform a robustness check in Appendix A. In this test we look at the change in the number of eligible voters between ridings and between elections, dropping the ridings with the largest changes in eligible number of voters between elections.

The average absolute value of the change in the number of eligible voters from 2011 to 2015 is 3,510, and from 2015 to 2019 is 1,885. Only a handful of ridings have a change larger than 5,000 in either election. By comparison, the average absolute value of the change in Conservative votes between 2011 and 2015 is 2,499; for NDP votes, it is 7,369; for Liberal votes, it is 12,529; for the Bloc Québécois, it is 2,228; and for non-voters, it is 2,291. For 2015 to 2019, the numbers are 2,023, 7,833, 2,632, 7,396 and 1,208 for the Conservative, NDP, Liberal, Bloc Québécois and non-voters, respectively. Given these numbers, changes in the number of eligible voters in a riding due to individuals turning 18, dying, gaining Canadian citizenship, or moving ridings would have to have an almost uniform effect on party vote share to explain changes in party support (for example, all voters turning 18 would have to vote for the same party). We think this is extremely unlikely. The changes in the number of eligible voters in a riding from election to election is not large enough to explain the changes in the number of votes won by each party.

There are nine ridings where the absolute value of the total change in the number of eligible voters from 2011 to 2019 was larger than 10,000 (La Prairie, Beloeil–Chambly, Pontiac, Portneuf–Jacques-Cartier, Vaudreuil–Soulanges, Ville-Marie–Le Sud-Ouest–Île-des-Soeurs, Montcalm, Rivière-du-Nord, Mirabel). Those nine ridings also show large growth over the 2011–2015 period specifically. When we drop these nine ridings from our analysis, we find that the estimated vote shifts are very similar to those found with the full data set. For the 2008–2011 analysis, we similarly drop the 10 ridings with growth in eligible voters over 5,000, and we find similar results. This is further evidence that changes in the number of eligible voters cannot explain changes in election results over the period of time we are looking at. To be sure, over the long term, demographic changes due to individuals turning 18, immigration, and deaths can affect election results. But the period we are looking at is simply too short for these demographic changes to have a meaningful effect on parties’ success.

We present our results using figures with a plot for each party showing either the proportion of its second-election voters who voted for each party in the first election in the equation or the proportion of first-election voters who voted for each party in the second election. For example, when we look at where each party's 2011 vote came from, we present a figure for each party with estimates for the proportion of its 2011 voters who voted for each different party in 2008. We present tables including estimates and standard errors, as well as estimates for the Green and other parties excluded from the figures in the main body of the article, in Appendix B.

Results and Analysis

We take the 2011 election as a turning point in federal politics in Quebec. The NDP's breakthrough in that election marks the clearest departure from previous elections. Accordingly, we organize our analysis around the NDP's rise from 2008 to 2011, looking at where the NDP came vote from, and then around the NDP's collapse in 2015 and 2019, looking at where the NDP's (as well as other) votes went. To capture the party system's reversion to Bloc Québécois versus Liberal competition, we provide a final analysis of where the Bloc Québécois’ and Liberals’ 2019 vote came from.

Where did the NDP vote in 2011 come from?

Figure 1 shows our estimates for where parties’ 2011 vote came from. In line with Fournier et al.'s (2013) findings, we find evidence that the NDP made gains from both the Conservatives and the Bloc Québécois. We, however, find little evidence that the NDP made gains from the Liberals, and we add to the Fournier et al. (2013) findings by showing that the NDP made substantial gains from 2008 non-voters. We estimate that 38 per cent of the NDP 2011 vote came from the Bloc Québécois, while 21 per cent came from 2008 NDP voters (the relatively low share of the 2011 NDP vote that came from 2008 NDP voters reflects the parties’ growth between 2008 and 2011). Our estimates for NDP gains from 2008 Conservatives and 2008 non-voters are similar, with both making up about 14 per cent of the NDP's 2011 vote. We estimate that about 8 per cent of the 2011 NDP vote came from the Liberals, though that estimate is not statistically significantly different from 0. Unsurprisingly, given the NDP was the only party that made gains in 2011, we find that most 2011 Liberal, Conservative and Bloc Québécois voters had voted for the same party in 2008. Most 2011 non-voters had also been non-voters in 2008, but a nontrivial 11 per cent had been Liberal voters in 2008.

Figure 1 Where parties’ 2011 vote came from (in 2008 votes)

Note: Bars show estimates based on Table A1 in Appendix B. Bars for the Green Party, as well as other small parties, have been omitted to improve the graph's readability.

* p < .05; ** p < .01

These results suggest there is some evidence for both H1a and H2b. In total, we estimate that 59 per cent of the NDP 2011 vote came from voters who switched to NDP after voting for the Bloc Québécois, Conservatives or Liberals in 2008. NDP gains from the Bloc Québécois were larger than NDP gains from any other party and from non-voters, and the difference between the gains the NDP made from the Bloc and the gains they made from non-voters is statistically significant at the 99 per cent confidence level.⁵ Vote switching was more important to the NDP's success in 2011 than voter mobilization. Nonetheless, voter mobilization played some role in the NDP's gains, even if it was a smaller role. The 14 per cent of the 2011 NDP vote that came from 2008 non-voters was likely enough to shift some of the closer contests in Quebec to the NDP.

We find some support for H2a, that Liberal losses in 2011 were a result of 2008 Liberal voters leaving the electorate. That 11 per cent of 2008 Liberal voters decided not to vote in 2011 is significant and can partially explain the party's struggles in that election. The number of 2008 Liberal voters who dropped out of the electorate is relatively similar to the number that switched to the NDP. We estimate that about 4.1 per cent of the total Quebec electorate in 2011 were 2008 Liberal voters who did not vote in 2011, and about 3.4 per cent of the total electorate were 2008 Liberal voters who voted NDP in 2011.⁶ Voter demobilization is not responsible for all of the decline in Liberal vote share in Quebec between 2008 and 2011, but it did matter.

Where did the 2011 vote go in 2015?

Figure 2 shows that vote shifts between 2011 and 2015 were more complicated than between 2008 and 2011. The Liberals largely held on to the voters that they won in 2011, with 73 per cent of 2011 Liberal voters sticking with the party in 2015. The NDP vote, however, went in multiple directions. We estimate that only 38 per cent of 2011 NDP voters stuck with the NDP in 2015, while 30 per cent switched to the Liberal Party and 16 per cent switched to the Bloc Québécois. This suggests some reversion back to the national identity politics structured around Bloc Québécois versus Liberal competition that was present in Quebec prior to rise of the NDP. Interestingly though, the Bloc struggled to hold on to its 2011 voters. Only 57 per cent stuck with the Bloc for the 2015 election. Despite an overall decline in its vote share, the NDP managed to pick up 21 per cent of 2011 Bloc voters in 2015, while 13 per cent did not vote. Finally, it is worth noting that the Liberals managed to mobilize a substantial number of 2011 non-voters, with 18 per cent of such voters backing the Liberals in 2015.

Figure 2 Where parties’ 2011 vote went (in 2015 votes)

Note: Bars show estimates based on Table A2 in Appendix B. Bars for the Green Party, as well as other small parties, have been omitted to improve the graph's readability.

* p < .05; ** p < .01

The vote switching that occurred in this period lines up with the expectations of hypotheses H1b. In line with H1b, the Liberals made clear gains from the NDP. There is also little evidence that NDP voters left the electorate as the NDP declined, as 2011 NDP voters make up a negligible share of 2015 non-voters. H2c is supported. The Liberals made substantial gains from 2011 non-voters, suggesting that voter mobilization played an important role in the party's 2015 recovery.

Where did the 2015 vote go in 2019?

Figure 3 shows that the main beneficiary of the NDP's 2015 to 2019 decline was the Bloc Québécois, as opposed to the Liberal Party that had been the main beneficiary of the NDP's 2011 to 2015 decline. In 2019 the Bloc won 33 per cent of voters who had voted NDP in 2015, while the NDP managed to hold on to just 26 per cent of its 2015 vote. There is some evidence that the NDP lost votes to demobilization, as about 10 per cent of 2015 NDP voters did not vote in 2019. With the exception of the NDP, the other parties in Quebec managed to hold on to the bulk of their 2015 vote, with only the Conservatives winning less than 60 per cent of their 2015 vote in 2019. There is also little evidence that there was much voter mobilization between 2015 and 2019, as we estimate that 78 per cent of 2015 non-voters remained non-voters in 2019. There is some evidence that both the Bloc Québécois and Liberals were affected by demobilization, as 16 per cent of 2015 Liberal voters and 10 per cent of 2015 Bloc Québécois voters dropped out of the electorate in 2019.

Figure 3 Where parties’ 2015 vote went (in 2019 votes)

Note: Bars show estimates based on Table A3 in Appendix B. Bars for the Green Party, as well as other small parties, have been omitted to improve the graph's readability.

* p < .05; ** p < .01

These findings provide support for hypotheses H1c. The bulk of NDP losses between 2015 and 2019 appear to be the result of vote switching, not of voters leaving the electorate. There is some evidence that NDP voters left the electorate between 2015 and 2019, but the proportion of NDP voters leaving the electorate is small relative to the proportion leaving the NDP for other parties. There is no evidence here for hypothesis 2d; gains in 2019 by the Bloc Québécois due to voter mobilization appear to have been marginal.

Where did the Bloc Québécois and Liberal 2019 vote come from?

We conclude our analysis by looking at the drivers of the resurgence of the Bloc Québécois and Liberal Party over the entirety of the 2011 to 2019 period. Figure 4 shows that vote switching and voter mobilization had different relative impacts on the Liberals and Bloc Québécois. The Liberals made substantial gains by both winning voters from the NDP and by mobilizing voters who did not vote in 2011. We estimate that only 34 per cent of 2019 Liberal voters had voted Liberal in 2011, while 29 per cent had voted NDP in 2011 and 22 per cent did not vote in 2011. Given the margins for error in our estimates (see Appendix B), we take this as evidence that the Liberal resurgence in the 2015 and 2019 elections resulted both from the party's ability to take voters from the NDP and from its ability to mobilize voters who had not voted in 2011.

Figure 4 Where parties’ 2019 votes came from (in 2011 votes)

Note: Bars show estimates based on Table A4 in Appendix B. Bars for the Green Party, as well as other small parties, have been omitted to improve the graph's readability.

* p < .05; ** p < .01

The Bloc Québécois story is different. Unsurprisingly, given that the Liberals and Bloc are on opposite sides of questions of national identity, the Bloc made no discernable gains from the Liberals. The party instead relied largely on its ability to take votes from the NDP for its 2019 resurgence. We estimate that 41 per cent of 2019 Bloc Québécois voters had been NDP voters in 2011. Our estimates of Bloc gains from non-voters are much smaller, at 10 per cent. Unsurprisingly, given they occupy a unique position on the right of federal politics, the Conservatives were most reliant on holding on to their own voters between 2011 and 2019. Of the voters who backed the Conservatives in 2019, 44 per cent had been Conservative voters in 2011. It is, however, also notable that the 2019 Conservatives won a non-negligible share of voters who had voted NDP in 2011.

The overall 2011–2019 picture provides evidence for hypotheses H1b and H1c. Both the Liberals and the Bloc Québécois were able to rebuild by taking votes from the NDP. There is little evidence that NDP voters left the electorate as their party's fortunes waned. We estimate that only 14 per cent of 2011 NDP voters were non-voters in 2019, and that estimate is not statistically significant from 0. Voter mobilization appears to have mattered more to the Liberals than the Bloc Québécois, providing support for H2c but not H2d.

Conclusion

Our analysis uses ecological inference to draw some important conclusions about volatility in federal elections in Quebec between 2008 and 2019. We present a number of theories that can explain why voters switch between parties and why mobilization or demobilization may affect party competitiveness. Unfortunately, our analysis can only distinguish between vote switching and voter mobilization; we cannot adjudicate between different theories that explain why voters switch parties or why voters turn out or stay home. Nonetheless, our estimates of the relative impacts of vote switching and voter mobilization have important consequences for scholars’ understandings of federal elections in Quebec—and because Quebec is so important to the success of parties in Canada, they have important consequences for understandings of federal election results, as well.

The first conclusion we can draw speaks to the relative importance of vote switching and voter mobilization. We show that both matter but to different degrees and to different parties. Both the gains made by the NDP in 2011 and the Bloc Québécois in 2019 were largely the product of vote switching. While both parties managed to mobilize some new voters, the bulk of their gains came from their ability to take votes from other parties. Our conclusions about the Liberal Party are less clear. Our estimates for Liberal Party gains in 2015 and 2019 from vote switching and voter mobilization are similar enough that we cannot say which is more important to the party's resurgence. Vote switching mattered in all three cases of party growth, suggesting that it is the more powerful driver of electoral volatility in Quebec. While voter mobilization appears to play a lesser role in volatility, it is not unimportant. In all three cases of party growth, at least 10 per cent of the party's vote share came from its ability to mobilize voters.

Our findings also have important implications for federal politics in Quebec and for the evolution of the broader Canadian party system as a result. We find a great deal of movement of voters from the Bloc Québécois to the NDP between 2008 and 2011 and then back from the NDP to the Bloc Québécois between 2011 and 2019. This finding lines up with Gauvin et al.'s (2016) work that suggests shared left-wing values between Bloc Québécois and NDP voters create a lot of potential for voters to shift back and forth between the two parties. However, by adding in analysis from the 2019 election, we show that the Bloc Québécois voters lost to the NDP in 2011 were not lost forever. Rather, when national identity issues re-emerged (albeit around questions of religious accommodation instead of separatism), the Bloc was able to win back many of the voters they had lost to the NDP. That there was movement between the Liberals and NDP, especially between 2011 and 2015, suggests that the Bloc and NDP's potential electoral coalitions are not complete overlaps of each other and that the Liberals and NDP are competing over at least some voters. This leaves the NDP in the awkward position of trying to appeal to Québécois nationalists choosing between the NDP and Bloc while also trying to appeal to federalists choosing between the Liberals and NDP.

The degree to which the Bloc and Liberals were able to win back their voters from the NDP once national identity re-emerged as important to Quebec politics in 2015 and 2019 suggests that future NDP success will depend on the extent to which the party can either diminish or successfully navigate question of national identity in Quebec. Our analysis finds little evidence of a left-leaning electorate in Quebec waiting to be mobilized by a federalist left-wing party like the NDP. Rather, if the NDP is to be successful, it needs to find a way to take votes from its competitors. This has consequences for the Canadian party system. An NDP that cannot win in Quebec is likely to be limited to third-party status and leave Canada as unique among rich democracies for the weakness of its social democratic party. That so many NDP voters have moved back to the Bloc leaves the Bloc Québécois in a position to prevent the Liberals from winning the seats in Quebec that they need to win a majority government.