Party Switching in Comparative Context

The empirical study of party switching is relatively sparse, and most of the existing research focuses on legislators' incentives and motivations to switch parties. McElroy (Reference McElroy, Giannetti and Benoit2008) argues that career advancement is key to understanding party switching in the European Parliament. She finds that party leaders, senior members and members on key committees are less likely to defect than backbench MEPs or those on low-ranking committees, while legislators are much more likely to defect to a party represented in the Parliamentary Bureau. Yoshinaka (Reference Yoshinaka2005) has found a similar trend among United States congressmen and congresswomen, where committee assignments provide a strong incentive for defection. While most committee placements are allocated by both parties according to seniority, party switchers in the House of Representatives are more likely to be rewarded with a committee assignment that violates the seniority norm. Office seeking motivations have been a strong determinant of party switching in Mexico where ambitious but dissatisfied representatives abandoned their parties to take up attractive leadership roles or senior committee assignments across the aisle (Kerevel, Reference Kerevel2014). Likewise, Kam (Reference Kam2009) found that Westminster MPs whose progressive ambition has been curtailed by a lack of opportunity within their existing parties may be more likely to switch. Conversely, many MPs anchored in leadership or prestige positions are less likely to abandon their parties (Castle and Fett, Reference Castle, Fett and Bianco2000, McElroy, Reference McElroy, Giannetti and Benoit2008). Sometimes the benefits of party switching are even more tangible; for example, Desposato (Reference Desposato2006) finds that access to “pork,” such as lucrative contracts and jobs for supporters and rewards for campaign workers, is a key variable in the calculus of party switching in Brazil.

Party switching has also been used to examine legislator voting behaviour. Desposato (Reference Desposato2003) finds evidence that party membership matters in Brazil by demonstrating that switchers in the 49^th Legislature voted with their original parties less after a switch while voting more with their new party. Nokken (Reference Nokken2000) finds a similar effect in the United States Congress, where legislators who switch parties vote like members of their new party, even in systems where the enforcement of party discipline is low. Following from this, Heller and Mershon (Reference Heller and Mershon2008) find that more disciplined parties in Italy experience higher levels of party switching, and MPs whose ideal points in dimensional space were furthest away from their parties were most likely to defect to another party. Relatedly, McMenamin and Gwiazda (Reference McMenamin and Gwiazda2011) found that party switching and an examination of party switching type can be used to study and measure party institutionalization in Poland.

The motivations of party switchers may stem from a genuine disconnect between the representative and their party. Castle and Fett (Reference Castle, Fett and Bianco2000) argue that in two-party systems representatives situated at the edge of their party's ideological policy space may be tempted to switch affiliation if a neighbouring party's ideal point is closer to their own.

With respect to the consequences associated with party switching, Grose and Yoshinaka (Reference Grose and Yoshinaka2003: 57) and Yoshinaka (Reference Yoshinaka2015: ch. 5) demonstrate that incumbent legislators in the US who switch parties suffer in both general and primary elections, while both their “new” and “old” parties subsequently hold more competitive primary contests. Even though party switchers experience a decline in vote margin during primary elections, they find that this effect dissipates over time. This impact of switching is more pronounced in general elections, where switchers' vote shares are typically lower in all elections after their switch (Grose and Yoshinaka, Reference Grose and Yoshinaka2003: 69). This is related to the costs of party defection: “Party switchers may alienate past supporters, including general election voters, primary supporters, and partisan elites, and be unable to fully compensate with the support of new partisans” (Grose and Yoshinaka, Reference Grose and Yoshinaka2003: 69, 57).

Relatedly, Evans and colleagues (Reference Evans, Peterson and Hadley2012), in a survey following the 2010 midterm switch of Republican Arlen Specter to the Democrats, established that most Pennsylvanians did not buy Specter's stated reason—that an increasingly conservative Republican party no longer aligned with his own ideological position—and figured that he was motivated by electoral considerations. Where variation took place, it fell along established party lines: Democrats believe the switch was policy motivated, Republicans and independents believed Specter to be an opportunist. It is worth noting that Specter lost in his bid for re-election.

Given there is no shortage of research on legislator behaviour in Canada, it is surprising that little has been written on party switching in this context. In an appendix to a 2005 report published by the House of Commons Ethics Commissioner and subsequently republished in the Canadian Parliamentary Review, Morton paints a broad (and brief) picture of the Canadian party switching landscape. He finds that “lists of ‘switchers’ include a disproportionate number of Quebec and Western MPs expressing their discontent with party labels they seldom controlled or which, in the case of the Progressives and, later, Reform, virtually legitimized an independent spirit” (Reference Morton2006: 6). But Morton's findings are more descriptive than prescriptive.

Docherty followed Morton with a more detailed descriptive analysis of Canadian party switching. He finds that, “not surprising…party switching is most likely to occur when political parties themselves are undergoing periods of transition” (Reference Docherty2011: 198). Docherty sought to dispel the popular notion that party switchers commit electoral suicide; “Despite conventional wisdom that floor crossing is not rewarded we find that 57% of floor crossers who run for re-election under their new party banner win their bid” (200). While a valuable challenge to the conventional wisdom, the remaining 43 per cent of switchers who run and lose remain markedly different from the 20 per cent of all MPs who exit parliament involuntarily (Kerby and Blidook, Reference Kerby and Blidook2011: 628). Party switching has also attracted some descriptive and inferential attention more recently from Snagovsky (Reference Snagovsky2015), Cochrane and Sevi (Reference Cochrane and Sevi2016) and Snagovsky and Kerby (Reference Snagovsky, Kerby, Farhat and Poirier2017).

Theory and Design

In Canada, parties control almost all levels of political career advancement, from committee assignments, to cabinet appointments, to electoral labels and funding. MPs who vote against the party line do so with risk to their career ambitions and even their future electoral prospects (Kam, Reference Kam2009: 150). Legislators act rationally in response to both proximate and distant incentives; for example, they join and remain loyal to political parties with the expectation that the benefits of party membership outweigh the costs. When they pursue policy preferences, upward career mobility or votes, partisan affiliation is not a foregone conclusion or an end in and of itself, but rather an instrumental tool applied in pursuit of other goals. Thus, “[not] only do politicians choose to identify with a major party in the first place, they choose to maintain or alter that identification. Their choice, therefore, must be the consequence of the theory, rather than be assumed by it” (Aldrich and Bianco, Reference Aldrich and Bianco1992: 103).

The existing literature on political careers and party switching tells a story of the decisions made by ambitious legislators. These actors use party switching to further their interests, such as advancing a certain policy agenda, attaining higher political office such as a seat at the cabinet table or simply getting re-elected. Representative democracies are based on ambition: citizens elect representatives on their behalf, who in turn serve them both for mutual gain and out of fear of dismissal. As Schlesinger writes, “A political system unable to kindle ambitions for office is as much in danger of breaking down as one unable to restrain ambitions…No more irresponsible government is imaginable than one of high-minded men unconcerned for their political futures” (Reference Schlesinger1966: 2). Regardless of whether ambition is a desirable feature of our political system, it is a reality of it; ambition cannot simply be ignored or treated “as a human failing to be suppressed” (Reference Schlesinger1966: 2).

Following Grose and Yoshinaka's approach (Reference Grose and Yoshinaka2003) to the case of the US Congress, we present a model of the electoral consequences of party switching in the Canadian House of Commons. However, we also argue that legislators are motivated by three core “desires”: policy, office and votes (Müller and Strøm, Reference Müller and Strøm1999). MPs seek votes in the political arena as an instrumental means to realize a combination of policy outcomes and office-related prestige. In turn, they consider the costs and benefits of each “desire” to maximize their total expected payoffs.

As we highlight in our review, Muller and Strøm's policy office and votes lens has been used explicitly and implicitly to consider the determinants and motivations of party switching. In this respect, our choice of independent variable is not path breaking. The novelty and contribution of this study lies in our direct examination of the electoral consequence of party switching for each of the policy office and votes motivations. And, in this way, we bridge both the motivations and consequences literature using an up-until-now unexamined case: Canada.

Data

We constructed an original longitudinal dataset of individual-level MP electoral results for the period 1945–2011 to conduct our analysis of the electoral consequences of party switching. The records are drawn from the Canadian Library of Parliament's ParlInfo database which contains biographical, election and party membership data for the population of incumbent MPs elected and defeated during the period of interest. Supplementary data on the nature and timing of MPs' party switches were also drawn from ParlInfo, as well as Canadian newspaper archives and assorted political biographies. A complete list of the party switchers, the reasons for their switches and relevant sources can be found in supplementary Online Appendix 1.

The unit of analysis is incumbent election result recorded at the time of a general election. Therefore, each observation contains an MP identifier, the election date, the MP's party label at the time of the election and whether they retained or lost their seat. We also record whether the MP switched parties in the parliament preceding the election. By definition, we only consider those party switchers who ran in the election following their change of party affiliation. Those MPs who switched parties and resigned or chose not to run in the next election were dropped from the analysis. We chose to omit those MPs who switched out of and then switched back into the same party (for example, the Democratic Representative Caucus MPs who briefly left and then returned to the Canadian Alliance in 2001). We also do not consider MPs whose party labels change due to party mergers and re-brandings, such as the Co-operative Commonwealth Federation's transformation into the New Democratic Party (NDP) in 1961 or the merger of the Canadian Alliance and Progressive Conservative parties in 2003. Table 1 breaks down the possible outcomes for switcher and non-switcher incumbents on election day. It demonstrates that the incumbency rate of MPs who do not switch parties is different from those who change their party affiliations. Indeed, the proportion of MPs who switch parties and fail in their bids for re-election is considerably higher than it is for those do not switch at all. This difference is statistically significant at the .05 level.

Table 1 Frequency Table of Party Switch and Electoral Result (incumbents 1945–2011)

Pearson chi²(1) =  41.7792 Pr = 0.000

We look only at MPs' party affiliation at the time of an election. Thus, Robert Toupin, who switched from PC to independent to NDP and then back to independent between 1984 and 1987, is coded as having switched from PC to independent, the party labels he held in the 1984 and 1988 elections. Likewise, if an MP follows an “independent intermediate” trajectory where she leaves one party to sit briefly as an independent before joining another, then the intermediate step is excluded. So, when Lucien Bouchard moved from the PCs to the newly created Bloc Québécois after spending seven months as an independent, he is coded as having moved from PC to BQ.

Independent variables

Like the dependent variable, we take several approaches to coding party switches. In the first instance, we use a dichotomous variable—“switched”—that is coded as 1 for “switched” and “0” for “not switched” in the preceding parliament. In the second instance, “switched” takes one of four values. The first three are based on the Müller and Strøm (Reference Müller and Strøm1999) “policy, office and votes” framework which was originally used to describe party motivations, but which in recent years has been used to describe the motivations of other political actors at both the individual and aggregate levels of analysis (Martin, Reference Martin2014; Mershon and Shvetsova, Reference Mershon and Shvetsova2008; O'Brien and Shomer, Reference O'Brien and Shomer2013). The fourth category, “booted,” is assigned to those party switchers who were expelled from their political parties. Each category was matched to the appropriate party switchers according to the official stated reason in ParlInfo and/or the reasons reported most often in the press. Some MPs were more difficult to code than others as they switched for multiple reasons. In these instances, the switches were coded according to the most cited motivation. The frequency of each switch type is reported in Table 2.

Table 2 Frequency Table of Party Switching Types (Incumbents: 1945–2011)

We operationalize “policy” switches as any event where an MP leaves caucus voluntarily on a point of principle, as a response to her party's position on a certain policy issue or because she identified a new party as being stronger on a certain issue than her old one. For example, Maurice Bourget left the Liberal party in 1944 to protest the government's conscription policy (“Bourget bolts government,” 1944)Footnote ¹, while Lucien Bouchard resigned from the PC cabinet and caucus over the failure to pass the Meech Lake Accord (Freeman, Reference Freeman1990b).

“Office” switches refer to those cases where MPs leave their party to join the government party and are appointed to cabinet or become parliamentary secretaries within one year of the switch. Belinda Stronach's high-profile defection from the Conservative party to the governing Liberals in 2005 saw her immediately appointed Minister of Human Resources and Skills Development, as well as Minister Responsible for Democratic Renewal (Howlett, Reference Howlett2006). Likewise, Jack Horner took his seat at the cabinet table on April 21, 1977, the day after he became a Liberal (Hawthorn, Reference Hawthorn2004). Both Stronach and Horner are coded as office switchers.

“Vote” switches take place when MPs cite electoral motivations as the determining factor for their switch or when there was strong indication that they left for fear of a negative result in an approaching election. For example, Gerard Girouard left the Social Credit party to join the Progressive Conservatives because “the Social Credit party [had] committed suicide by its internal disputes and hopeless division” (Westall, Reference Westall1964). Disaffected with his party leadership, Rémi Paul left the Progressive Conservative caucus in 1965. Upon his exit, he told the press that “Not only would I be defeated as a Conservative candidate under Mr. Diefenbaker, I would lose my deposit” (“Quebec Tory Defector Decides To Quit Politics at Federal Level”, 1965). Other MPs, like Raymond Rock, may face stiff nomination challenges and likely fear that they will not get their parties' subsequent endorsement (Wills, Reference Wills1972). Still others switch between specific parties and unique labels such as “Independent Liberal” or “Liberal Labour,” to retain a party association but avoid the electoral wrath of their constituents. This was electoral strategy used by a number of MPs in the 1940s and 1950s (Stevens, Reference Stevens1966).

Some MPs jump ship; others are pushed. Any instance where an MP lost a nomination battle, was ejected from caucus or was forced to step down by the party leadership was coded as “booted.” For example, Jean Guy Carignan resigned from the Liberal caucus after being sentenced to jail time for a hit-and-run accident (LeBlanc, Reference LeBlanc2001), while David Kilgour and Alex Kindy were expelled from the PC caucus for voting against the establishment of the Goods and Services Tax (Freeman, Reference Freeman1990a).

While the proportion of those incumbents who switch parties and are re-elected is not nearly as high as the proportion of non-switchers who are re-elected, Table 3 presents some interesting variation across the switching categories. Nearly three-quarters (72.73%) of MPs who switch for office reasons are re-elected. This is closely followed by those who switch for policy reasons (65.71%) and then (ironically) the more evenly balanced MPs who switch to enhance their electoral chances (54.55%). Unsurprisingly, only 18.52 per cent of MPs ejected from their parties are successfully re-elected in the next election.

Table 3 Frequency Table of Switch Type and Electoral Result (incumbents 1945–2011)

While the “policy, office, or votes” framework helps to demarcate party switcher motivations in a more nuanced way, it is possible that the voters who cast judgment on party switchers are less concerned with the specific categories. Therefore, we include a third, in-between approach to coding party switching types that employs an “opportunist” variable that combines the office and vote switching motivations into a single category. MPs who do not switch parties are coded as “0” and treated as the reference category in all the model specifications.

Control variables

We include several control variables to account for the alternative explanations of incumbents' changes in vote share. Our model accounts for incumbency effects, that is, the advantage or disadvantage that incumbent legislators have when seeking re-election. Incumbents may be more familiar to voters than challengers; MPs appear at community events, advertise in local newspapers and provide services to their constituents. As a result, incumbent legislators seeking re-election may have an advantage over other candidates. This notion is consistent with the literature, which suggests incumbents have an advantage over challengers in Canadian elections (Blais et al., Reference Blais, Gidengil, Dobrzynska, Nevitte and Nadeau2003; Kendall and Rekkas, Reference Kendall and Rekkas2012; Krashinsky and Milne, Reference Krashinsky, Milne and Seidle1991, Reference Krashinsky and Milne1985). In the context of party switching, MPs may switch parties because they think they are well-known by their constituents from years of service. Since all party switches in our model are incumbents, we cannot simply include an “incumbent” dummy variable. Instead, we code “experience” as the number of years an MP has served in parliament since their first election. This makes our measure more robust, since the longer an MP holds office, the greater the familiarity their constituents should have with them, and therefore the greater their potential advantage. In a similar vein, the “minister” variable is coded as “1” for incumbents who run in an election as a sitting cabinet ministers and “0” for those who are not. This is intended to capture potential voter preferences for representation by a member of cabinet, perhaps thinking this gives their riding more of a say in policy matters. Since several of our MPs switch for office-related reasons, this consideration is important.

While local candidates matter, so do parties. In the context of party switching, it is important to understand whether switchers are elected on their new party's coattails, or by contrast do better than another candidate would have in their place. Further, we are interested in whether switching to an unpopular party makes re-election unlikely or whether the MP's personal popularity still carries them to victory. To control for party effects and partisan swings, we consider the popularity of MPs' political parties at the time of the general election. This is operationalized as the percentage of the popular vote the MP's party received in their province at the time of the election, since party support can vary dramatically from province to province. We also control for whether the MP ran as an independent, and thus had no party support to rely on.

In addition to these control variables we also control for MPs' vote share in the previous election and we include a series of interaction variables which consider the multiplicative effect of previous vote share for independents and for party switchers who run as independents in the election following their switch. We do this in order to disentangle independents' lack of party vote (captured by the party provincial vote share) from their fellow (yet separate) independents who share the same label but not a common affiliation.

Variable descriptions and summary statistics for the dependent, independent and control variables are found in Tables 4 and 5.

Table 4 Variable Descriptions

Table 5 Summary Statistics

Data and Method

Our dataset consists of the population of incumbent MPs elected to the House of Commons for the period 1945–2011 (less those MPs who were dropped for the reasons outlined above). Each MP is observed at the time of a general election. Thus, we possess data for 2014 unique MPs who are observed over the course of multiple elections. The data are organized as a panel (N = 5153 MP-elections) to examine both cross-sectional (across individuals) and time series (across time) dimensions. While it is tempting to employ a fixed-effects, or multilevel mixed effects linear regression model, the inclusion of the lagged dependent variable on the right-hand side of the equation requires us to deal with the problem of serial correlation by using an Arellano-Bond dynamic panel model (Arellano and Bond, Reference Arellano and Bond1991). The class of econometric model is precisely useful when there is a risk of an endogeneity problem, the likelihood of serial correlation caused by the inclusion of the lagged dependent variable as predictor of the dependent variable and a panel that is organized such that the time dimension (elections) is significantly smaller than the observation (MP) dimension. In order to control for serial correlation, we included two lags of the dependent variable on the right-hand side of the equation.

We estimate six specifications of the model that use MPs' vote shares as the dependent variable. We take this approach to examine the different effects party switching has on electoral popularity. We use three variations of the party switch independent variable: the first employs the binary “switched” variable; the second model subsets “switched” by motivation (policy, office, votes or booted–hereafter abbreviated as POVB); and finally, the third uses the “opportunist” variable in place of “office” and “votes.” The additional three models build on these specifications with the use of interactions, as described below.