Recent work by Robin Best (Reference Best2010) notes a divergence at the national level in the number of parties winning legislative representation and in the number of parties competing for and winning votes. In both majoritarian and proportional systems, the number of parties winning votes has increased since the Second World War. However, the number of parties winning seats in majoritarian systems has remained closer to two than in proportional systems, where the number of parties winning seats has increased in relative proportion to the number of parties winning votes.
Best (Reference Best2010) notes the possibility that the discrepancy in majoritarian systems between the number of parties winning seats and the far greater number of parties winning votes reflects ‘increasing irrationality’ on the part of voters in these systems, with voters knowingly supporting parties that do not have any chance of winning representation. It remains to be seen whether this is the case. It could be that these national-level results appear ‘irrational’ only in the aggregate, while, at the district and individual levels, voters may still conform to the two-party expectations of strategic voting theory. As Gary Cox notes (1997, 1999), parties may encounter ‘coordination failures’ (which refer to the inability of parties to compete effectively across all districts) that lead to their total replacement by a locally strong third party at the district level, thereby leaving only two parties competing in each district while competition at the national level features more than two parties.
This article probes these issues further. Specifically, it looks at the number of parties competing for and winning votes at the district level in order to determine whether the number of parties conforms to the two-party expectations consistent with strategic voting theory, or whether voters are displaying patterns of ‘increasing irrationality’ and supporting multiparty competition at the district level. The results are important because they speak to the larger debate between the two major explanations for the number of parties observed cross-nationally. One explanation – the institutional approach – maintains that the number of parties is determined by the proportionality of the electoral system, predicting that the number of parties in each district will converge on two in countries operating under first-past-the-post. The other explanation – the social cleavage approach – maintains that the number of parties is determined by the number of cleavages: as the number of cleavages grows, the number of parties increases. Should the number of parties converge on two, this would demonstrate that the institutional approach remains valid despite the national-level evidence of increasing party system size presented by Best. However, should the number of parties exceed two at the district level, this would provide evidence in favour of the social cleavage approach.
To preview the findings, this article demonstrates that the number of parties choosing to contest elections and winning votes increasingly exceeds two-party predictions in three countries operating under first-past-the-post: Canada, the United Kingdom and New Zealand (prior to 1993). In order to explain these findings, I develop an explanation that is rooted in the social cleavage approach. The number of parties has increased in these countries as the number of cleavages articulated by party elites has increased. Third parties have pursued strategies that are irrational according to the strategic voting literature, competing in most districts in order to represent groups of voters who are present to some degree in every district and who are not well represented by the existing parties.
Electoral Systems, Strategic Voting and Coordination Failures
The conventional wisdom maintains that the number of parties observed cross-nationally owes much to the proportionality of the electoral system. Following the seminal work by Duverger (Reference Duverger1954), this literature holds that the number of parties tends towards two in countries using first-past-the-post, while multiparty systems tend to be found in countries using proportional representation electoral systems. Duverger posited two reasons for this. One is the mechanical effect of electoral systems, or how votes are translated into seats: first-past-the-post systems tend to produce more disproportional seats-to-votes ratios, while proportional representation systems tend to produce seat shares that more proportionally reflect parties’ vote shares (Gallagher Reference Gallagher1991; Lijphart Reference Lijphart1999; Rae Reference Rae1971). The other and perhaps more powerful effect is psychological. Anticipating the mechanical effect, voters will strategically desert their most preferred candidates in favour of less preferred candidates if they anticipate that votes for their most preferred candidates will be wasted (and possibly resulting in victory for their least preferred candidates). For their part, parties will forgo expenditures on districts they do not expect to win (McKelvey and Ordeshook Reference McKelvey and Ordeshook1972). The prediction that strategic voting will produce two-party convergence has been (re)affirmed by a number of works (Chhibber and Kollman 2004; Cox Reference Cox1997). Because only one candidate is elected per district, the psychological effect should be strongest in first-past-the-post systems, diminishing as district magnitude increases.
A considerable body of evidence has been amassed demonstrating this strategic voting phenomenon among voters (Cox Reference Cox1997; Lanoue and Bowler Reference Lanoue and Bowler1992; Merolla and Stephenson Reference Merolla and Stephenson2007; Niemi et al. Reference Niemi, Whitten and Franklin1992). Examining the system-level consequences of strategic voting, numerous studies have found that the number of parties is greater in countries operating under proportional representation than in majoritarian systems such as first-past-the-post (Clark and Golder Reference Clark and Golder2006; Taagepera Reference Taagepera2007; Taagepera and Shugart Reference Taagepera and Shugart1989). While most studies focus on the number of parties at the national level, the psychological effect applies most clearly at the district level. As Cox (Reference Cox1997, Reference Cox1999) has noted, the number of parties competing at the national level may exceed two, yet the number of parties competing in any one district may still conform to the predictions of strategic voting theory and remain at (or below) two. Referring to coordination failures, Cox notes that one of the two major parties may strategically exit a district in which a third party with strong local support – support that makes the party more competitive than the two major parties – also contests that district.
Although evidence examining the number of parties at the district level has found that the institutional effects apply at the district level much as they do at the national level (Singer and Stephenson Reference Singer and Stephenson2009), other studies have found that district-level party systems exceed the two-party predictions made by the institutional approach. Examining the case of Canada, Brian Gaines (Reference Gaines1999; see also Johnston and Cutler Reference Gaines2009) noted that the number of parties competing at the district level exceeded two throughout most of the twentieth century. Gaines (Reference Gaines2009) found similar evidence when looking at the United Kingdom in the period following the 1960s. Others have found that the number of parties at the district level often exceeds two in India, where the number of parties appears to be affected significantly by the choices made by party elites to contest or not contest certain districts (Chhibber and Murali Reference Chhibber and Murali2006; see also Diwakar Reference Diwakar2007). Thus, while the institutional approach predicts that the number of parties competing in any one district in countries operating under first-past-the-post should not exceed two, there is reason to believe that the number of parties at the district level exceeds two much as Best (Reference Best2010) has noted with regard to national-level party systems.
Elite Agency and the Social Cleavage Approach
The major alternative explanation maintains that the number of parties is determined more by the number of social cleavages than by the proportionality of the electoral system (Bartolini and Mair Reference Bartolini and Mair1990; Lipset and Rokkan Reference Lipset and Rokkan1967; Rokkan Reference Rokkan1970). Because parties emerge to represent the interests of particular social groups to the exclusion of others, increases in the number of cleavages place strains on the existing parties’ ability to represent these new groups of voters. This can be seen early in the twentieth century with the extension of universal (male) suffrage, which incorporated working-class voters who were essential to the fortunes of socialist and labour parties targeting their support (and difficult for the existing liberal and conservative parties to represent credibly) (Lipset and Rokkan Reference Lipset and Rokkan1967). More recently, the emergence of postmaterial voters has put pressure on the existing parties either to adapt to represent this new base of voters and risk alienating their existing bases of support, or to risk losing this potential base of voters to another (third) party (Inglehart Reference Inglehart1977, Reference Inglehart1990). Given these challenges of representation, the number of parties may increase as the number of cleavages present in a given district increases, despite the incentives of the electoral system.
While much of the early work in this tradition assumed that the articulation of cleavages by party elites would be automatic, a growing body of literature suggests that elites have considerable agency in determining which cleavages to articulate (Enyedi Reference Enyedi2005; Przeworski and Sprague Reference Przeworski and Sprague1986; Sartori Reference Sartori1969; Torcal and Mainwaring Reference Torcal and Mainwaring2003). The interplay of social cleavages and elite articulation can be seen with regard to the evolution of party competition in the nineteenth and early twentieth centuries. Caramani (Reference Caramani2003, Reference Caramani2004, Reference Caramani2005) notes that the emergence of competitive elections was determined by the decisions taken by party elites to represent certain cleavages over others. As the incentives for controlling the machinery of government became more alluring, parties chose to de-emphasize the issues associated with the territorial (urban–rural, regional, and so on) cleavages they had traditionally represented in favour of issues related to functional (class, religious and so on) cleavages that were present across most districts. This choice allowed them to compete effectively for votes in most districts, thus improving their chances of winning a majority of seats. While this understanding of cleavage articulation differs from definitions found in some of the most prominent works (for example, Bartolini and Mair Reference Bartolini and Mair1990), it corresponds with more recent definitions of the term ‘cleavage’ in ways that retain the focus on parties’ representation of issues associated with certain social groups to the exclusion of others, while recognizing the considerable agency possessed by party actors in deciding which issues to articulate and which groups to represent (Enyedi Reference Enyedi2008).
This agency may have an impact on the number of parties. For instance, because parties represent certain groups to the exclusion of others, the major parties may choose not to articulate cleavages that do not fit well with their existing bases of support, particularly if these parties are already able to control the machinery of government with their existing bases of support (as has been the case with the two largest parties in each country operating under first-past-the-post examined below). Additionally, if third parties do not articulate issues associated with these cleavages, the would-be support for these third parties goes to the existing parties (or these voters fail to turn out); if third parties do choose to articulate issues associated with these new cleavages, the number of parties will increase beyond two.
In sum, the literature rooted in the social cleavage approach provides an explanation for why the increasing size of Western party systems observed by Best at the national level may be mirrored at the district level. If the number of parties competing for and winning votes at the district level converges on two, this would provide evidence in favour of the institutional approach. If the number of parties exceeds two, particularly if sustained over an extended period, this would provide evidence in favour of the social cleavage approach.
The Number of Parties Winning Votes
Turning to the empirical evidence, this section examines whether the number of parties winning votes at the district level exceeds the two-party predictions made by the institutional approach in first-past-the-post systems. There is need for additional research that focuses directly on first-past-the-post systems. Previous research has concluded that institutional and social cleavage effects appear to interact with one another (see, for example, Clark and Golder Reference Clark and Golder2006; Ordeshook and Shvetsova Reference Ordeshook and Shvetsova1994; Singer and Stephenson Reference Singer and Stephenson2009). However, while these studies help to explain differences in the number of parties at the same level of electoral system proportionality in proportional representation systems, the interactive effects seen in these studies lead us to the conclusion that the effects of social cleavages do not produce multiparty systems in first-past-the-post systems. Consequently, these studies fail to explain why the number of parties might exceed two in first-past-the-post systems. Thus, further study of first-past-the-post systems is needed. To this end, the data in this study focus on three countries operating under first-past-the-post: Canada (covering the years 1867–2011), the United Kingdom (1832–2010) and New Zealand (1905–93, prior to the adoption of a mixed-member proportional system).Footnote 1 Focusing on Canada and the United Kingdom (where previous research has found that the number of parties exceeds two) in addition to New Zealand (which has received little attention in previous research) allows one to examine whether the forces thought to lead the number of parties to exceed two in Canada and the United Kingdom are also present in New Zealand.
Despite the small number of countries examined here, the findings from these cases are quite generalizable. Because these three countries have relied exclusively on the same first-past-the-post electoral system throughout most of the period under study (particularly the period following the Second World War, when the number of parties at the national level increased beyond two),Footnote 2 the primary institutional effect argued to be associated with party system size – the proportionality of the electoral system – is held constant. As a result, any deviations from two-party predictions in these three countries would suggest that these processes are capable of occurring in more proportional systems, where the electoral system poses less of a barrier to new party entry. Additionally, while focusing solely on first-past-the-post countries helps to minimize electoral system variance, these three countries do vary in terms of other variables thought to be necessary for the emergence of multiparty systems. Namely, these three cases allow us to examine what effects federalism, party financing and partisan dealignment have had on the number of parties. If district-level party systems in all three countries exceed two-party predictions, then these additional variables cannot explain the emergence of multiparty systems at the district level.
To measure the number of parties, I use the effective number of electoral parties (ENEP) (Laakso and Taagepera Reference Laakso and Taagepera1979). While the institutional literature predicts that the number of parties will converge on two, few would accept the presence of third parties winning scant votes as evidence that is sufficiently contradictory of Duverger's law. Because ENEP discounts the number of parties by weighting each party according to its vote share (thus leaving only the ‘effective’ number of parties), it provides a more conservative measure of the number of parties, biased in favour of the predictions made by the institutional literature. Previous studies testing the predictions made by the institutional literature have made use of this measure for precisely the same reasons (see, for example, Cox Reference Cox1997; Singer and Stephenson Reference Singer and Stephenson2009; Taagepera Reference Taagepera2007).
Figure 1 displays the mean ENEP at the district level in each election.Footnote 3 During the early period in each country's time series, mean district-level ENEP converges on two in each country.Footnote 4 At several points, however, mean ENEP exceeds two; even allowing for some deviation from the two-party prediction by using a conservative measure of the number of parties such as ENEP, the fact remains that the number of parties increasingly exceeds the two-party predictions made by the institutional literature. This occurs early in the time series in New Zealand (1919), but later converges on two from 1938 to 1954, before increasing consistently beyond two (beginning in 1954) until the end of the time series. In Canada, the mean ENEP has not fallen below two since 1935, and often the number of constituencies with effectively two or fewer parties is less than 10 per cent. While mean ENEP converges on two between 1931 and 1959 in the United Kingdom, mean ENEP begins to exceed two between 1918 and 1929 and then even more significantly from 1974 onwards. The most consistent deviations from two-party predictions in each country occur following the Second World War, the same period in which Best (Reference Best2010) finds that the number of parties exceeded two at the national level. These deviations from two-party predictions are sufficient to conclude that the number of parties at the district level exceeds the expectations of the institutional literature.
Figure 1 Mean District-level ENEP in Canada, New Zealand and the United Kingdom
Note: single-member districts only.
Before proceeding further, one should note that these patterns rule out three other prominent explanations for district-level party systems in excess of two-party predictions. Several recent works argue that because issues differ at the federal and provincial levels, federalism may give rise to third parties with support bases in certain regions that allow them to compete successfully at the district level with the larger, more nationalized parties (Chhibber and Kolman Reference Chhibber and Kolman1998, Reference Chhibber and Kolman2004; Gaines Reference Gaines1999). However, the fact that Canada is the only federal system of the three countries examined here – paired with the fact that mean district-level ENEP exceeds two-party expectations in each country – demonstrates that federalism is not a necessary cause of multiparty systems. While Gaines (Reference Gaines2009) has extended the multilevel elections concept to the United Kingdom, where European Parliament elections have added a federal-type component in the past few decades, the fact that New Zealand (which has no such second level of elections) exceeds two-party predictions provides further evidence that multilevel elections are not the driving force behind the party fractionalization observed in these three countries.
Additionally, arguments rooted in party financing are insufficient explanations of multiparty systems in these three cases. With an increasing number of parties dependent upon the state for finance cross-nationally (van Biezen and Kopecký Reference van Biezen and Kopecký2007; Katz and Mair Reference Katz and Mair1995), government-subsidized campaign finance may provide third parties with enough money to enable them to persist in the face of strategic desertion by voters and donors. However, the empirical record casts doubt on such arguments as explanations for why multiparty systems have emerged in first-past-the-post systems. Johnston and Cutler (Reference Johnston and Cutler2009) note that while party subsidization in Canada dates back to 1968 (though the most consequential campaign finance laws date back only to 2004), it may have been more a consequence rather than a cause of increased fractionalization, as district-level ENEP exceeded two-party predictions throughout (and prior to) the 1960s. Additionally, the United Kingdom and New Zealand demonstrate that party subsidies alone cannot explain the increasing patterns of fractionalization seen here because multiparty systems emerged at the district level in both countries despite the lack of subsidization in the United Kingdom and before the advent of party subsidization in New Zealand.Footnote 5
Finally, these patterns are inconsistent with an explanation that stresses dealignment as the root cause of party fractionalization at the district level. Particularly regarding electoral competition since the 1970s, one interpretation of the dealignment literature would predict that, as voters have become less attached to the parties and more issue oriented (Dalton Reference Dalton1984; Dalton et al. Reference Dalton, Flanagan and Beck1984; Dalton and Wattenberg Reference Dalton and Wattenberg2000), third parties have benefited and the number of parties increased beyond two. As one can see, however, mean ENEP exceeded two-party predictions in all three countries before the period of dealignment. While major-party dealignment probably improved the chances for third parties to appeal to voters dissatisfied with the two larger parties, the data presented here clearly show that dealignment was not necessary for the development of multiparty competition.
To summarize: the data presented here clearly show that the number of parties exceeds the two-party predictions made by the institutional approach for district-level races. While the data presented in this section do not decisively support the social cleavage approach, the fact that the patterns of party competition at the district level do not conform to the institutional approach and do not appear to conform to other explanations for deviations from two-party predictions provides tentative support. In the next section, I examine the emergence of some of the key third parties in each country to show more conclusively how the social cleavage approach is at work, leading the number of parties to exceed two at the district level.
The Evolution of Multiparty Competition
Evidence in favour of the social cleavage approach can be seen when looking at the patterns of district contestation, particularly for some of the major third parties, over time. Specifically, I examine the proportion of districts contested by the major parties in each country, similar to Caramani (Reference Caramani2004), who notes that the nationalization of parties’ electoral strategies and support reflects the types of cleavage groups targeted. Figures 2, 3 and 4 present the proportion of districts contested by the major parties in Canada, the United Kingdom and New Zealand, respectively. Before proceeding to the country-specific analyses, one should note some general patterns. At the beginning of the time series in each country, the proportion of districts contested by each party is quite variable, ranging from highs of around 0.9 for the Liberal Party in New Zealand to a low of 0.35 for the Canadian Liberals in 1867. Over time, the two largest parties in each country begin to contest nearly every district, consistent with the discussion in Caramani (2003) that notes the evolution of competitive elections resulting from the decisions of elites to target more functional (instead of territorial) cleavages. During the period following the Second World War examined by Best (Reference Best2010), the two largest parties in each country contest nearly every district. By as early as 1954 and as late as 1974, the major third party examined in each country also contested the same proportion of districts as the two largest parties.
Figure 2 The Evolution of Party Competition in Canada
Note: single-member districts only. CCF/NDP = Cooperative Commonwealth Party/New Democratic Party.
Figure 3 The Evolution of Party Competition in the United Kingdom
Note: single-member districts only.
Figure 4 The Evolution of Party Competition in New Zealand
Closer examination of each country explains why third parties began to contest the same proportion of districts as the two larger parties. Beginning with the Canadian case in Figure 2, arguably the most important third party in Canadian politics began competing with both the Liberal and Conservative parties in most districts shortly after its inception. The Cooperative Commonwealth Federation (later to be renamed the New Democratic Party in 1962) emerged in 1935 contesting over 40 per cent of the total number of districts, and by 1949 was competing in over 80 per cent. It is interesting to note that prior to the emergence of the Cooperative Commonwealth Federation in 1935, the mean district-level ENEP was just over two; since then, mean district ENEP has rarely fallen below 2.5.Footnote 6 The incentive to contest most, if not all, districts originated in the fact that the party appealed to working-class voters (and others sympathetic to the party's social-democratic appeals) in all professions – voters who are present to some degree in every district.Footnote 7 To borrow Caramani's language, the Cooperative Commonwealth Federation/New Democratic Party appeals to a functional cleavage that the two larger parties have chosen not to articulate,Footnote 8 leaving this sizeable base of voters open to the Cooperative Commonwealth Federation/New Democratic Party's social-democratic appeals. Unlike the major third party preceding the Cooperative Commonwealth Federation/New Democratic Party – the Progressive Party, which appealed primarily to Western agricultural interests and only secondarily to those in manufacturing (Morton Reference Morton1950; Sharp Reference Sharp1948) – the Cooperative Commonwealth Federation/New Democratic Party has consciously chosen to appeal to working-class voters in all occupations so as not to face the same fate as the short-lived Progressive Party.
The evolution of multiparty competition in the United Kingdom occurs in two distinct periods, as seen in Figure 3. The first occurs between 1918 and 1929, when all three parties – Conservative, Liberal and Labour – competed in nearly every district. The gradual emergence of the Labour Party between 1900 and 1918 immediately preceded the slow decay of the Liberal Party, whose bases of support had begun to shrink as the cleavage between Conformist and Nonconformist voters had likewise begun to decline (Matthew et al. Reference Matthew, McKibbin and Kay1976; Wilson Reference Wilson1966). By 1931, the Liberal Party was ultimately displaced by Labour as one of the two largest parties, and the Liberals entered a long period of decline, contesting fewer seats (for instance, losing all but nine of the party's candidate deposits despite the large proportion of contested seats in 1950). However, the party slowly began to recover its stake in elections so that the party was able to contest 90 per cent or more of the districts from 1974 on. Such a strategy is in keeping with the party's decision to target educated middle-class voters, particularly those with postmaterial values (Evans et al. Reference Evans, Heath and Lalljee1996), who find both the Labour and Conservative parties unacceptable (Heath et al. Reference Heath, Jowell and Curtice1985; Rose and McAllister Reference Rose and McAllister1990). Because these voters are present to some degree in every district, the Liberals have contested most districts as they have chosen to articulate issues related to this functional cleavage. It is no coincidence that the sharp and lasting upturn in mean ENEP noted in Figure 1 (rising from 2.19 to 2.62 between 1970 and 1974) occurs at the same time as the Liberals (and later, the Liberal Democrats) began catering to this functional cleavage and contesting nearly every district.
The patterns of party competition observed in New Zealand begin in quite dramatic fashion. While the Liberal (later United) Party saw little organized opposition in 1905, the dramatic rise of Reform in 1908 ushered a brief period of two-party competition that evolved into three-party competition as the Labour Party gradually emerged as a force strong enough to give incentive to the parties of the centre and right (United and Reform) to merge, forming the National Party.Footnote 9 As seen in Figure 1, these two parties effectively dominated elections as the number of parties converged on two from 1935 to 1951. This changed in the election of 1954 as the Social Credit Political League broke through, contesting nearly every seat and winning 11 per cent of the vote. Drawing support from small-business owners, farmers and family-oriented religious voters – whose concerns on issues of moral traditionalism and economic self-sufficiency were not well-represented by Labour or National (James Reference James1980; Miller Reference Miller2005) – the party contested every district from 1957 until 1987, before merging with several smaller third parties to form the Alliance in 1991. Although the party experienced some setbacks during these years, the party averaged nearly 10 per cent of the national vote during the period from 1954 to 1990, never polling below 5 per cent prior to 1990 and polling as high as 20 per cent in 1978. From its initial success in 1954 until 1987, the presence of Social Credit in nearly every district is associated with district-level ENEP values significantly in excess of two.Footnote 10
In sum, these figures show that the increases in district-level ENEP observed in Figure 1 coincide with increases in the proportions of districts contested by the major third party in each country. In keeping with the social cleavage approach, the fact that each third party has chosen to articulate and represent functional cleavages not represented by the existing parties reflects an increase in the number of cleavages in each country; in other words, the increases in district-level ENEP appear to be due to increases in the number of cleavages articulated by parties. Because each of the third parties has chosen to articulate functional cleavages – with voters belonging to these cleavages present to some degree in most districts – these parties have defied the expectations of strategic voting theory and contested nearly every district in order to represent their targeted base of voters.
Conclusion
Inspired by the findings presented in Best (Reference Best2010), noting that the number of parties competing for – and winning – votes at the national level in countries operating under first-past-the-post has increasingly exceeded two over the past several decades, this article has sought to determine whether the same processes have occurred at the district level. While the institutional literature predicts that the number of parties should not exceed two at the district level even if the number of parties exceeds two nationally (Cox Reference Cox1997, Reference Cox1999), the evidence presented here is more consistent with the social cleavage approach, which predicts that the number of parties may increase beyond two as the number of cleavages increases. Even if one allows for some deviation from two-party predictions, and even if one uses a measure of the number of parties that is biased in favour of the institutional approach, the fact remains that the number of parties increasingly deviates from two-party expectations.
Examining the patterns of party competition of the most important third parties behind the increases in the number of parties demonstrates that these parties are pursuing strategies that allow them to compete for voters belonging to the cleavages they represent, competing in nearly every district despite the fact that such strategies are ‘irrational’ according to strategic voting theory. Although the data presented here do not bring evidence to bear that directly measures variables associated with the social cleavage approach, previous studies have shown that patterns of party competition reflect the social-group bases of parties’ support (Caramani Reference Caramani2003); because each third party examined here represents issues and values that resonate with certain groups of voters more than others, and because the patterns of third-party competition presented above match the findings of previous work, the conclusions reached here accord with previous literature.
While the evidence clearly shows that the number of parties exceeds the expectations of the institutional approach (and the strategic voting literature on which this literature is based), the mechanisms preventing strategic voting incentives from limiting the number of parties to two in any one district are not fully understood. The strategic voting literature acknowledges several assumptions; without meeting these assumptions, strategic voting will not occur and the number of parties will exceed two.Footnote 11 However, there is precious little research specifically examining whether the absence of these factors does indeed prevent strategic voting, let alone the frequency and degree to which these assumptions are met.Footnote 12 Further research is needed to determine which of these assumptions is most crucial for two-party convergence, or whether other factors are preventing the emergence of multiparty competition, in order to understand why third parties such as those examined here have not been strategically deserted by their supporters.
