Divisive primaries and sore losers

Once an American phenomenon, political parties around the world are increasingly adopting primaries (Kenig et al. Reference Kenig, Cross, Pruysers and Rahat2015a) as a means of selecting party leaders and general election candidates (Cross and Blais Reference Cross and Blais2012a; Cross et al. Reference Cross, Kenig, Pruysers and Rahat2016; Sandri et al. Reference Sandri, Seddone and Venturino2015). According to Jean-Benoit Pilet and William Cross (Reference Pilet and Cross2014), primaries have been among the fastest-growing methods of selection over the last 50 years. While these intra-party elections are rarely as inclusive as traditional American primaries (Ware Reference Ware2002),Footnote ¹ they do come close in so far as they typically invite eligible party members, and in some cases registered supporters, to cast a ballot for their preferred candidate. What's more, a variety of parties have already begun experimenting with expanding the pool of eligible voters beyond party members. In 2013, for example, the Canadian Liberals adopted a method in which both traditional dues-paying party members and registered supporters could participate in the selection of the party's leader (Cross Reference Cross, Pilet and Cross2014).

Many parties, it is argued, adopted primaries as part of a broader bundle of reforms that were designed to enhance intra-party democracy. Such reforms were meant to arrest, or even reverse, party decline (Cross and Blais Reference Cross and Blais2012b; Katz Reference Katz, Cross and Katz2013).Footnote ² Commenting on the decline of parties, Fulvio Venturino and Antonella Seddone (Reference Venturino and Seddone2022: 233) explain that political parties have been faced with ‘an unrelenting crisis, heralded by declining membership and waning levels of trust among citizens’. The adoption of primaries offered a potential solution to many of the challenges that parties were facing. First, primaries afford parties and their candidates with opportunities to mobilize and recruit thousands of new supporters into the party. In this sense primaries can help grow the party and combat declining memberships. Second, primaries enhance the legitimacy of parties by providing ordinary grassroots members with a meaningful opportunity to participate in important internal party decisions. Finally, primaries can generate an excitement and buzz around the party as they garner the attention of both media and the general public. In this regard, primaries offered political parties a great amount of ‘promise’ (Cross et al. Reference Cross, Kenig, Pruysers and Rahat2016). This promise, however, has not always fully materialized. Indeed, numerous scholars contend that many of the theorized benefits (i.e. enhanced legitimacy and public support) of greater intra-party democracy have often been short-lived or overstated (Pedersen and Schumacher Reference Pedersen, Schumacher, Cross and Pilet2015; Wauters and Kern Reference Wauters and Kern2021). Although primaries were meant, at least in part, to address decline, a growing body of literature has now explored some of their negative consequences. Such consequences include the strengthening of party leadership at the expense of the grassroots (Katz and Mair Reference Katz and Mair1995), the gendered dynamics of selection and removal (Astudillo and Paneque Reference Astudillo and Paneque2022; O'Neill et al. Reference O'Neill, Pruysers and Stewart2021) and the manipulation of intra-party rules (Cross et al. Reference Cross, Kenig, Pruysers and Rahat2016).

One potential consequence of primaries that has received little attention outside of the United States is their propensity to create internal divisions and therefore create a divide between winners and losers within the same party. In the American case this literature has often fallen under the umbrella of the ‘divisive’-primaries hypothesis. Here the argument is that primary elections, particularly competitive/divisive ones, can have negative implications for party cohesion as well as the success of the party/nominee in the subsequent general election. Much of this stems from the fact that primaries pit individuals of the same party against one another. James Lengle et al. (Reference Lengle, Owen and Sonner1995: 372) capture this sentiment when they write,

In the wake of an intra-party battle, supporters of the losing candidates, according to the divisive-primaries hypothesis, are expected to exhibit less loyalty to their party in the general election than their winning counterparts. This is expected to take the form of individuals withdrawing from the electoral process (i.e. abstaining from voting) or defecting (i.e. voting for an alternative party).

Patrick Kenney (Reference Kenney1988) describes two mechanisms through which this might work. First, there is a psychological perspective. In this view primaries encourage individuals to establish a positive affect towards a specific candidate. Over the course of an intra-party election, the psychological bond creates a sense of loyalty among members of the ingroup (i.e. supporters of a particular candidate) and fosters a sense of hostility towards members of the outgroup (i.e. supporters of the other candidates). Here it is argued that the bond between candidate and supporter is such that if an individual's preferred candidate loses the nomination, they may withdraw from the party as their attachment to other candidates is weak or non-existent. This may be especially likely if the individual was mobilized into the party apparatus as a result of the leadership contest – as many donors and members are – and had no pre-existing relationship (financial or otherwise) with the party. The second perspective is the ‘sore losers’ or ‘sour grapes’ explanation. Here the emphasis is on more general discontent or disgruntlement as opposed to deep psychological attachments and loyalties. As Kenney (Reference Kenney1988: 765) explains, ‘Individuals may simply become angry at the winning candidate and find it impossible to support the nominees in the general election, even though they may not be psychologically attached to another candidate. The source of that displeasure may be as sophisticated as policy differences or as simple as concerns over style and personality.’ Dissatisfied with the outcome, these individuals turn away from the successful candidate, either abstaining or finding a home in an alternative party.

Such expectations are corroborated by the broader political behaviour literature. Here a large body of research, focused on general elections, has demonstrated a consistent ‘winner–loser’ gap in terms of satisfaction with democracy, general support for the political system and subsequent voter turnout (Anderson et al. Reference Anderson, Blais, Bowler, Donovan and Listhaug2005; Blais and Gélineau Reference Blais and Gélineau2007; Dahlberg and Linde Reference Dahlberg and Linde2017; Howell and Justwan Reference Howell and Justwan2013). As André Blais and François Gélineau (Reference Blais and Gélineau2007: 426) explain,

In short, losers of general elections tend to express less satisfaction and less support for the political system than winners.

Whether it is the psychological and emotional mechanisms outlined by Kenney (Reference Kenney1988) or the utility-maximizing and cognitive-consistency mechanisms outlined by Christopher Anderson et al. (Reference Anderson, Blais, Bowler, Donovan and Listhaug2005), the divisive-primaries hypothesis suggests that there may be a cost for parties when holding a primary. Discontent, disgruntlement and even disloyalty may all befall the party in the general election as the intra-party contest creates a divide between winners and losers. While the logic and theory are simple enough, empirical validation of the hypothesis has been somewhat mixed. On the one hand, a number of studies have argued that the dynamics of primary elections do indeed impact the fortunes of political parties during the subsequent general election (Fouirnaies and Hall Reference Fouirnaies and Hall2014; Kenney and Rice Reference Kenney and Rice1987; Lengle et al. Reference Lengle, Owen and Sonner1995; Makse and Sokhey Reference Makse and Sokhey2010; Southwell Reference Southwell1986). However, many who have found support for the thesis nonetheless suggest that the magnitude of the effect has been overstated (Henderson et al. Reference Henderson, Hillygus and Tompson2010; Kenney Reference Kenney1988). While much of the sore-losers work considers voters, Donald Johnson and James Gibson (Reference Johnson and Gibson1974) extend this to include party activists. The authors provide compelling evidence that supporters of a winning candidate remained more committed than did supporters of the losing candidate. What's more, a sizeable minority of intra-party losers (21%) reported that they intended to volunteer for an opposing party. On the other hand, there are many who are less convinced by the sore-losers hypothesis (Lazarus Reference Lazarus2005; Wichowsky and Niebler Reference Wichowsky and Niebler2010). Piereson and Smith (Reference Piereson and Smith1975: 562), for instance, write that ‘a candidate's primary election experience bears little relationship to his success in the general election’. While much has been written on the subject, no clear consensus has emerged.

Despite the growing adoption of primaries, the sore-losers literature has been largely limited to the United States. There are, however, at least two notable exceptions. First, William Cross and Scott Pruysers (Reference Cross and Pruysers2019) provide an analysis of the general election campaign activities and satisfaction of party members following participation in a candidate nomination. Using survey data of Canadian party members, the authors find that individuals who supported a losing nomination candidate were significantly less likely to engage in high-intensity activism during the subsequent general election (i.e. go door-to-door canvassing); were more likely to report that they were unsatisfied with their party membership; and were considerably less likely to report that they would renew their membership moving forward. While levels of high-intensity election activism and membership satisfaction differed significantly between winners and losers, the analysis found no support for the idea that intra-party losers were disloyal during the general election (both cohorts supported their party at equally high rates). Their analysis, therefore, finds support for the sore-losers hypothesis outside of the United States, with the caveat that the significant results revolved around activism and satisfaction and not voting behaviour.Footnote ³

The second exception is recent research by Venturino and Seddone (Reference Venturino and Seddone2022), who explore similar questions in the Italian context. Here the authors rely on a series of surveys (exit polls) administered after four leadership selection primaries within the Italian Partito Democratico (PD): 2009, 2013, 2017 and 2019. In doing so the authors consider the (anticipated) general election loyalty of individuals who participated in the primary election. Consistent with the American literature, their analysis provides rather mixed results: in two instances intra-party winners were more likely to vote for the party in the general election whereas the reverse was also true in two instances. As Venturino and Seddone (Reference Venturino and Seddone2022: 243) explain,

While important to the advancement of our understanding of the consequences of primaries beyond the American context, both studies referenced in the previous section (Cross and Pruysers Reference Cross and Pruysers2019; Venturino and Seddone Reference Venturino and Seddone2022) rely on self-report data, which carries with it concerns of reliability (i.e. memory recall) and generalizability (i.e. how representative was the sample?).Footnote ⁴ The study presented here, by contrast, provides a new test of the sore-losers hypothesis using observational data. To do so, we turn our attention away from party members and voters to a different cohort of party actors: financial donors. Doing so allows us to draw upon actual donation patterns that are publicly disclosed by political parties and their candidates.

More than just the availability of observational data, donors are remarkably important to democratic politics. Indeed, money is often referred to as the ‘lifeblood’ of elections, and such a statement is only becoming more true in an age of professionalized and increasingly costly campaigns (Norris Reference Norris2000). There is no shortage of literature demonstrating the link between financial resources and electoral success (Carty and Eagles Reference Carty and Eagles2005; Currinder Reference Currinder2009; Johnston and Pattie Reference Johnston and Pattie2014). Given the link, it is not surprising that political parties work tirelessly during the election and inter-election periods to solicit new donors and keep them engaged with party politics. This is especially true in the Canadian case, where other sources of funding have been removed and individual contributions have been capped (Tolley et al. Reference Tolley, Besco and Sevi2022; Young and Jansen Reference Young, Jansen, Young and Jansen2011).

Donors, however, should not be assumed to be identical to party members. Although the categories of party member and financial donor clearly overlap, they ‘are not synonymous’ (Garnett et al. Reference Garnett, Pruysers, Young and Cross2022: 423). Cross et al. (Reference Cross, Pruysers and Currie-Wood2022), for example, report that just two-thirds of party members make a financial donation to their party (outside of membership fees). In other words, there are many donors who are not members, and many members who are not donors. Likewise, research suggests that party members may be more motivated by solidary incentives compared to financial donors who are more transactional/material in their motivations (Garnett et al. Reference Garnett, Pruysers, Young and Cross2022). Despite this, donors remain the most understudied cohort of party actors in Canada. Given the importance of money to electoral outcomes, as well as the need to study donors separately from members, this study explores what happens to leadership election donors whose preferred candidate lost. In particular, it advances a single hypothesis. Namely that intra-party losers will be exhibit less commitment/loyalty than will winners.

Data, methods and case

Are ‘losers’ less likely to display subsequent financial commitment to their party than are ‘winners’? To answer this question, we draw upon a unique dataset of tens of thousands of donors to recent Canadian federal leadership elections. This includes contributors to the following party's internal leadership elections: Bloc Québécois (2011, 2014), Conservative (2017), Green (2006), Liberal (2013) and New Democrat (2012, 2017).Footnote ⁵ Data are derived from the financial reports that political parties are required to file with Elections Canada.Footnote ⁶ In all cases the data are from the ‘reviewed’ as opposed to ‘as submitted’ reports.Footnote ⁷ These financial records include information such as the name and postal code of the individual donor, the amount donated, and the recipient of the donation (i.e. leadership contender, registered party, electoral district association, etc.). Usage of this administrative financial data is not uncommon in scholarly work, and it has been used to examine questions of gendered donation patterns, racial affinity and personalization (see Besco and Tolley Reference Besco and Tolley2022; Cross et al. Reference Cross, Currie-Wood and Pruysers2020; Tolley et al. Reference Tolley, Besco and Sevi2022). In the case of this article, we use these data to explore whether a ‘sore losers’ pattern applies to party leadership donors in recent Canadian intra-party elections.

To construct our dependent variable of interest – whether a leadership donor also donated to their party in the year following the leadership election – we rely on two sources of Elections Canada data. First, we captured donations made to leadership contestants from the official start of the internal election up until the final day of voting.Footnote ⁸ Next, we captured donations made to the registered political party in the year following the leadership election. Take, for example, the 2013 Liberal leadership election. The internal election officially began on 14 November 2012 and the final day of voting in the leadership election was 14 April 2013. In this case we begin by capturing all the donations to leadership contestants during the 14 November to 14 April period. Next, we capture donations to the Liberal Party for the period of 15 April 2013 to 15 April 2014 (the year following the internal leadership election). We are then able to use these two separate lists of financial donors (leadership and party donors) in order to establish a variable that identifies whether a leadership donor made a donation to the party in the year following the intra-party election. This variable is coded dichotomously (1 = leadership donor made a party donation in the year following; 0 = leadership donor did not make a party donation in the year following).Footnote ⁹ We repeat this process for each of the seven elections for which we have data and merge the data into a single dataset for analysis.Footnote ¹⁰

Our key independent variable of interest is whether the donor's preferred leadership candidate won the internal party election. To create this variable, we look at the leadership contestant to which the donor contributed financially. We take this donation as a strong indication of their expressed preference for who should win the party's leadership.Footnote ¹¹ We then code whether the leadership contestant to whom the donation was made either won or lost the internal election. This too is a dichotomous variable (1 = donor supported winning candidate; 0 = donor supported losing candidate). Recall that our central hypothesis is that winners will remain committed and will donate to the party at higher rates than will losers.

In addition to our primary independent variable, we include a small number of controls in our analysis. First, we include the amount that the donor contributed to the leadership candidate. This data is provided in the Elections Canada reports as parties are required to disclose all donations above $100. When donors made multiple donations, these values were summed to create an overall donation amount.Footnote ¹² Here we might have competing expectations. On the one hand, larger amounts donated may reflect a strong commitment to the party, a fact that would likely encourage future giving, irrespective of whether the donor's preferred candidate won or lost the leadership. On the other hand, large donations may reflect a particularly strong commitment to a specific candidate, which may encourage a sore-losers effect in the event that another candidate won the election.Footnote ¹³

Next, we include two variables that capture how competitive the internal leadership election was. Highly competitive internal elections may be considered more divisive, and as a result, may have greater implications for party cohesion and future donation patterns. If an individual's preferred leadership candidate loses in a hotly contested and close race, it may signal that the party is internally divided. A landslide victory, by contrast, suggests more internal cohesion with a single candidate for party members (and supporters) to rally around. Patrick Howell and Florian Justwan (Reference Howell and Justwan2013), however, reveal that general election winners express the highest satisfaction in ‘nail-biter’ elections (i.e. highly competitive ones). We operationalize competitiveness by including variables for the number of leadership contestants on the ballot as well as the margin between the top two candidates on the first ballot. For the former, the more candidates the more competitive the election. For the latter, the smaller the margin the more competitive the election.

We also include a variable for the gender of the leadership contestant. Many of the financial aspects of party politics have been found to be gendered, with donations to women candidates following an affinity pattern (Tolley et al. Reference Tolley, Besco and Sevi2022) and with women donating less in general (McMahon et al. Reference McMahon, Sayers and Alcantara2021).Footnote ¹⁴ If donors are motivated by affinity more than a connection with the party, those who supported a woman leadership candidate may be less likely to continue donating to the party afterwards. This is especially the case given that the vast majority of women candidates in the data lost their election (Elizabeth May in the 2006 Green election being the sole exception). Finally, we also control for political party using a series of dummy variables with the Liberals as the reference category. The multivariate model, therefore, includes the following: supported winner, donation amount, number of candidates, margin between top two candidates on first ballot, candidate gender and political party.

We end this section with a brief discussion of the merits of studying the Canadian case. The Canadian case is a useful one to study intra-party politics for a number of reasons. First, Canadian parties have a long history of internally democratic elections (Courtney Reference Courtney1995; Cross and Blais Reference Cross and Blais2012b; Cross et al. Reference Cross, Kenig, Pruysers and Rahat2016; Pilet and Cross Reference Pilet and Cross2014; Pruysers and Stewart Reference Pruysers and Stewart2018). While a number of parties around the world have recently moved towards more participatory selection methods (Pilet and Cross Reference Pilet and Cross2014; Sandri et al. Reference Sandri, Seddone and Venturino2015), Canadian parties generally have more experience and a longer history with inclusive selection methods (Cross et al. Reference Cross, Kenig, Pruysers and Rahat2016; Pruysers and Cross Reference Pruysers and Cross2016a).

Second, Canadian leadership races are relatively competitive affairs. The average number of leadership candidates in our dataset is 5.5 per election. While five candidates may not seem particularly high, Ofer Kenig and colleagues (Reference Kenig, Rahat, Tuttnauer, Cross and Pilet2015b) find that more than half of all leadership elections in their study of 14 parliamentary democracies resulted in acclamation (i.e. only one candidate). In fact, Canada has the lowest rate of ‘coronations’ of any of the 14 countries that Kenig and his colleagues studied. The competitive nature of Canadian leadership elections, therefore, provides a particularly useful case to study the intra-party dynamics of winners and losers.

Finally, as Christine Cheng and Margit Tavits (Reference Cheng and Tavits2011) suggest, the Canadian case is more generalizable to other Western democracies than is a country like the United States. Cross and Pruysers (Reference Cross and Pruysers2019: 486) make a similar argument, noting that ‘Canadian parties share many institutional features with political parties in Europe, Oceania, and Israel (i.e. parliamentary democracy, multi-party competition, membership-based organizations, etc.).’ Thus, while phenomena like the divisive-primaries thesis may be most closely associated with the United States, Canada offers a more generalizable case, which is particularly important given the spread of democratic leadership selection methods worldwide.