Hostname: page-component-7b9c58cd5d-6tpvb Total loading time: 0 Render date: 2025-03-16T09:00:18.955Z Has data issue: false hasContentIssue false

Polarized Pluralism in the Canadian Party System: Presidential Address to the Canadian Political Science Association, June 5, 2008

Published online by Cambridge University Press:  17 December 2008

Richard Johnston*
Affiliation:
University of Pennsylvania
*
Richard Johnston, University of Pennsylvania, 208 S. 37th Street, Room 206, Philadelphia PA,19104-6215, rgcj@sas.upenn.edu
Rights & Permissions [Opens in a new window]

Abstract

Abstract. The Canadian party system is an example “polarized pluralism,” whose key feature is domination by a party of the centre. Centrist domination induces centrifugal tendencies elsewhere in the system. Polarized pluralism accounts for several of the system's peculiar features: three-party competition in individual ridings, contrary to the predictions of Duverger's Law; the existence and episodic dynamics of sectional parties; boom and bust cycles in Conservative party electoral history; and the large gaps between federal and provincial outcomes within many provinces. But domination by a centrist party itself demands explanation.

Résumé. Le système canadien des partis politiques exemplifie le «pluralisme polarisé», dont l'élément clé est la force dominante d'un parti du centre. Une telle domination induit des tendances centrifuges ailleurs dans le système. Pour sa part, le pluralisme polarisé explique plusieurs éléments du système canadien : la présence de trois partis compétitifs au niveau de la circonscription, malgré les prédictions du Duverger; l'existence même et la dynamique des partis sectionalistes; l'alternance entre force et faiblesse dans l'histoire du Parti conservateur; et les grands écarts qui se présentent dans plusieurs provinces entre les élections fédérales et les élections provinciales. Dans ce contexte, la force d'un parti centriste se doit d'être expliquée.

Type
Research Article
Copyright
Copyright © Canadian Political Science Association 2008

Canada's party system is a bundle of seeming contradictions: multi-party competition but single-party government; durable cabinets but massive electoral volatility; roots that are deep but archaic and possibly irrelevant; tight agenda control but inflammatory rhetoric; and sharp discontinuities between federal and provincial elections within many provinces. Not surprisingly, claims about the system's essential character are as contradictory as the empirical patterns. I want to argue that the patterns are not so much contradictory as complementary and that a single model accounts for much of what we see. As with any good model, it raises as many questions as it answers.

In brief, the Canadian party system must be seen as an example of polarized pluralism, an ideal type first anatomized by Giovanni Sartori (Reference Sartori, LaPalombara and Weiner1966Reference Sartori1976).Footnote 1 He saw it as a rare and unhealthy form of party competition, which juxtaposes multipartism to domination by a party of the centre. Ironically, the very power of the centre party gives the rest of the system a centrifugal logic. All of Sartori's examples were of proportional representation (PR) systems. He regarded Canada as a garden-variety two-party system, as would be expected under our first-past-the-post (FPP) electoral formula. Even as Sartori was writing, however, the Canadian system was starting to exhibit pathologies that now must be recognized as symptomatic of the syndrome. Recognition of this is aided by events of the last two decades, in particular the electoral earthquake of 1993.

My argument operates at the level of the party system. In doing so, it attempts to revive a mode of analysis that was never very fashionable and for Canadians in the last 30 years, almost nonexistent. The pedigree is distinguished, however, as the very first volume of the Canadian Journal of Political Science featured two remarkably influential articles pitched at the systemic level, Alan Cairns's “The Electoral System and the Party System in Canada, 1921–1965” and Jean Blondel's “Party Systems and Patterns of Government in Western Democracies.” For all that these articles are widely cited and influential, they stand out as exceptional even for their time. In the decades since, focus on party systems as systems has been rare and not just in the Canadian literature (Bardi and Mair, Reference Bardi and Mair2008). Cairns certainly spawned controversy (Lovink, Reference Lovink1970; Johnston and Ballantyne, Reference Johnston and Ballantyne1977; Bakvis and Macpherson, Reference Bakvis and Macpherson1995) and inspired reform proposals (Irvine, Reference Irvine1979), but further contributions were more about geographic incentives in the electoral system than about the fundamental structure of party competition. There is also a Canadian literature about the succession of social bases and organizational forms on the partisan landscape (Carty, Reference Carty and Perlin1988; Smith, Reference Smith and Aucoin1985; Johnston et al., Reference Johnston, Blais, Brady and Crête1992). Individuals' behaviour and individual parties continue to be objects of study. Debate over the psychology of Canadian voters, as affected by stylized facts about the party system, has been a cottage industry. But little of this work considers systemic causes of systemic effects.

The Electoral System and the Number of Parties

The main exception is the study of the impact of the electoral system on the number of parties. This is a question in comparative politics, for claims about systemic cause and effect require comparison across systems. The driving force of the literature is elaboration on Duverger's Law and Hypothesis (Duverger, 1954/Reference Duverger, North and North1963).Footnote 2 The law states that FPP always produces two-party politics, or at least produces dynamic tendencies pointing in that direction. The hypothesis, a weaker statement, says that PR may produce multipartism.Footnote 3 The standing claim is that a complex social structure can produce a complex party system only if the electoral system allows it to (Ordeshook and Shvetsova, Reference Ordeshook and Shvetsova1994; Amorim Neto and Cox, Reference Amorim Neto and Cox1997). And only “weak” electoral systems—systems that do not punish co-ordination failure among kindred political groups or parties—PR basically—allow this to happen.Footnote 4 This also makes PR a necessary but not sufficient condition for multipartism. A “strong” electoral system—a system that does punish co-ordination failure—constrains the number of parties to slightly more than two, regardless of the complexity of the underlying social structure. As Amorim Neto and Cox (Reference Amorim Neto and Cox1997) put it, “A polity can tend toward bipartism either because it has a strong electoral system or because it has few cleavages. Multipartism arises as the joint product of many exploitable cleavages and a permissive electoral system” (167). The problem with Canada, of course, is that it fits none of these boxes. It combines the quintessentially strong electoral system, FPP, with a high level of electoral fractionalization.Footnote 5

The extent of fractionalization appears in Figure 1, which shows the “effective number of parties” in every election since 1878. It is no longer conventional to represent the number by a simple integer or by fractional approximations such as Blondel's “two-and-a-half” category (Reference Blondel1968). Now the indicator of choice is a continuous one, the “effective number of parties” (Laakso and Taagepera, Reference Laakso and Taagepera1979). The indicator captures the intuition that the fractionalization of a party system reflects not just the number of discrete party labels but their relative sizes as well.Footnote 6 The magic number is, of course, two, the theoretical ideal of parties in an FPP system.

Figure 1 Components of Electoral Fractionalization

Before 1921, the effective number of parties in the Canadian electorate oscillated right around that magic number.Footnote 7 The number surged in 1921 with the Progressive breakthrough, fell back as the Conservative party recovered, and then grew permanently in 1935, when the system acquired the equivalent of roughly one extra party. In fact, two discrete parties—the Co-operative Commonwealth Federation (CCF) and Social Credit—appeared and endured. In the 1990s, the equivalent of one more party was added to the system. Again, several new players—most prominently, the Bloc Québécois and Reform/Alliance—gained prominence. The system consolidated in 2000, but only slightly. Although the shifts in 2004 and 2006 significantly altered the partisan balance, they left the system as fractionalized as before.

The scale of fractionalization made Canada an outlier. Although all comparable systems fragmented in the 1910s and 1920s, all subsequently purged themselves, so to speak, as parties on the centre and right consolidated to block the threat from the labour/social democratic left. This reconsolidation is dramatic evidence of the power of Duverger's Law. In Canada, however, multipartism became a way of life. Down to the 1970s, the gap between Canada and its comparators was about one party. Some gaps then closed, but the Canadian system pulled away from its remaining comparators again in the 1990s.Footnote 8

All this happened notwithstanding a powerful and continuing contrary effect from the translation of votes into seats. As in other FPP systems, parliament is less fractionalized than the electorate, reflecting what Duverger (1954/Reference Duverger, North and North1963) called the system's mechanical effect.Footnote 9 The Canadian mechanics do not seem unusual.

What is unusual is the apparent weakness of the follow-on psychological effect, also posited by Duverger. Electoral-system theory says that a strong system induces strategic action to move the number of electoral parties down, such that sooner or later voters and parties will consolidate into two roughly co-equal blocs.Footnote 10 There is a hint in Figure 1 that the system exhibits Duvergerian equilibration, in that sudden gains in effective number of parties (ENP) are mitigated eventually. At least, one can impose such a reading on the data.Footnote 11 But electoral consolidation arrives late and typically lasts for only one parliament. As the ensuing breakup is usually greater than the preceding consolidation and each successive mitigation is weaker than the last one, electoral and parliamentary fractionalization have each ratcheted upward. In the electorate the system went from two parties to four. In parliament, the effective number increased from two parties to three.

Unpacking Fractionalization

The standing explanation for Canada's embarrassment of parties starts with the observation that although governments are formed nationally, votes are counted locally. Lipset made this point as early as Reference Lipset1954 and it lies at the heart of Cairns's (Reference Cairns1968) critique. Rae put the Canadian case front and centre in the comparative literature as the exception that proves the Duverger rule (Reference Rae1971). His stylization of the Canadian pattern—local bipartism, national multipartism—has stuck. It hints that Duverger's Law works only at the constituency level, a proposition now taken as canonical.Footnote 12 By implication, co-ordination across locales requires some other force. The critical extra-local factor may be the centralization of the policy agenda, as argued by Cox (Reference Cox1987). In most countries power has flowed to the centre, such that partisan co-ordination across geographic subunits becomes imperative. In Canada this logic turns on its head: the decreasing importance of the federal government weakens the imperative for cross-district co-ordination. Thus it is possible for Duverger's Law to apply even as the national party system breaks down.

This argument has been made most forcefully by Chhibber and Kollman (Reference Chhibber and Kollman2004). Although Chhibber and Kollman admit difficulties with the Canadian case, they stare the difficulties straight in the face and pass on by. They notice some local fractionalization—contra Duverger—and also admit that the flux in cross-district breakdown does not admit an unambiguous interpretation.Footnote 13Figure 1 puts the problem front and centre, by splitting total fractionalization into local and extra-local components. The line labeled “local” is the average constituency-level effective number of parties. Cross-district failure is indicated by the “local-national” gap.Footnote 14 Local co-ordination failure is indicated by the difference between the observed local ENP values and the theoretically indicated ENP of two. On the Duvergerian account, the local line should exhibit no net upward movement. The big shifts should be on the extra-local side, and these shifts should explain most of the gains on the national line.

In a word, they do not. In fact, each component explains about half the total fractionalization gain. The typical constituency harboured one more “party” at the end of the century than at the beginning. Before 1921—really, before 1935—the effective number of parties within a typical riding was under two, as riding contests were often one-sided to the point of acclamation. From 1935 on, however, the local ENP was always larger than two. By the turn of the century, it was about 2.8. Likewise, the extra-local contribution grew by about one “party.” The extra-local value is positive right from the start, as it is arithmetically required to be. The real excitement starts with a large gain in 1921. The line dips over the next few years but surges again in 1935. The 1945 election brought another pulse, and the surge in 1993 is stunning.

The two components of fractionalization exhibit qualitatively different dynamics and each dynamic reveals a puzzle.Footnote 15 Local fractionalization, notwithstanding modest discontinuities, follows a trend. The dynamic force behind this trend is the CCF/NDP. The appearance of the CCF was the biggest contributor to the 1935 lift in the local line, as from the beginning the party contested about 60 per cent of all ridings.Footnote 16 Starting in 1962, the NDP completed the move toward universal tripartism, and by 1968 it, like the traditional parties, was contesting every seat. The spread of candidacies induced a spread of votes. In this respect, the CCF and NDP merely extended a late nineteenth-century pattern of diffusion initiated by the Conservatives and Liberals. Relative to the early years of the twentieth century, the vote for the system's three core parties has become more nationalized, consistent with the pattern identified by Caramani for Europe (Reference Caramani2004). Extra-local flux, in contrast, is episodic. And critically, there are four elections—1930, 1958, 1984, and 1988—where the gain in the extra-local component, relative to the nineteenth-century starting point, is effectively zero.

The narrative poses difficult questions. The spread of local tripartism arguably has only expanded the total volume of electoral futility. Why would Liberals and Conservatives divide the centre-right and risk capture of seats by the left? Why would the Liberals and NDP divide the centre-left and risk capture of seats by the right? Equally hard to square with the usual Duvergerian story is the system's episodic extra-local dynamic. The just-mentioned four elections that take sectional differences back to the nineteenth-century starting point are also the instances, referred to earlier in the paper, of modest equilibration. If so, the equilibration is not what we would expect from the current formulation of Duvergerian logic. On that logic, consolidation should occur locally, where the co-ordination problem seems more tractable and where the penalty for co-ordination failure is immediate. Instead, when Canadian parties and voters get their act together, they do so through convergence among regions, leaving local tripartism largely untouched. And each moment of extra-local convergence has one thing in common: these are the only elections since 1917 to produce Conservative majority governments.

The Dominant Centre

Both peculiarities—local fractionalization and episodic extra-local flux—stem from the dominance of the system by the Liberal party. And the critical thing about the Liberal party is that it is a party of the centre, the only such party to dominate an FPP-based consolidated party system.Footnote 17 Liberals command the centre on at least two key dimensions of choice, the left-right ideological axis and the “national question.” Left and right are conventional categories, organizing a big fraction of party politics almost everywhere. Most countries also have a form of national question, often a variant that poses a fundamental challenge to their territorial integrity. But no other country forces secessionist and anti-secessionist politics through the FPP electoral formula.

A two-dimensional stylization of party preference is outlined with 2004 and 2006 Canadian Election Study survey data in Figure 2. The figure locates party supporters rather than the parties as such, for my argument requires that I split party support groups up. The horizontal axis deploys an indicator of left-right self-placement. The vertical axis displays support for Quebec's aspirations.Footnote 18 For visual simplicity, each measure is scaled to a −1,+1 range. Because one dimension is framed in terms of Quebec, party supporters are separated into Quebeckers and non-Quebeckers.

Figure 2 The Policy Space, 2004–6

Each point indicates the mean position of each party support group simultaneously on the left/right and pro-/anti-Quebec axes. Scanning horizontally, we see that each party's supporters locate themselves from left to right “correctly.”Footnote 19 Distances among party-support groups seem small, especially in Quebec. This is a fairly standard finding, however; in most electorates, supporters are not as polarized as party elites.

In this context, the width of Quebec/non-Quebec gaps within parties is all the more striking. In itself, this should not be surprising, given the very question. The Bloc, of course, is the most pro-Quebec; if things were otherwise we would distrust the indicator. Within Quebec, supporters of the pan-Canadian parties differ hardly at all over left and right but diverge sharply over the Quebec question. Outside Quebec, the gap between Liberal and Conservative supporters is greater over Quebec than over left and right. Most critically, each pan-Canadian party group is sharply divided internally. Among serious parties in Quebec, Conservative supporters are second only to Blocistes in pro-Quebec orientation. Outside Quebec Conservative supporters are the most anti-Quebec group. The Quebec/non-Quebec gap is equally great among New Democrats although for this party Quebec support is all but fictional. Liberals are also quite divided, but the gap within the Liberal camp is only slightly more than half as wide as among Conservatives. On this dimension, the Liberals are not the only party to seek the centre but they have the least difficulty holding it.Footnote 20

The Conservative marriage is of outright opposites: francophones and francophobes. The pattern is not a peculiarity of 2004 and 2006. Rather, it reflects several decades of Conservative manoeuvring on the dimension. Outside Quebec, much of the party's support is predicated on opposition to that province's claims. This was even true in 1988, notwithstanding the party's lead role on the Meech Lake Accord (Johnston et al., Reference Johnston, Blais, Brady and Crête1992). When Conservatives seek Quebec votes, however, the only ones available to them are located toward the nationalist end of the spectrum, as the Liberals already control the ground closer to the pro-Canada pole.

The Idea of Polarized Pluralism

All this is to say that party competition in Canada takes the form of polarized pluralism. The most extended account of this “ideal type” of party competition is Sartori (Reference Sartori1976).Footnote 21 The model is not deductive but instead emerges from Sartori's meditation on Weimar Germany, postwar Italy, Fourth-Republic France, and pre-1973 Chile. Its pivotal feature is that a large party controls the centre.Footnote 22 This point cannot be stressed too much. In the standing theory of committees and elections the centre exerts a powerful attraction on political competitors. This is the fundamental Downsian point in spatial analyses (Downs, Reference Downs1957; Enelow and Hinich, Reference Enelow and Hinich1984). Empirically, however, the attractive power of the centre is exerted on parties or ideological families that bracket the centre; they do not actually occupy it. Here is Duverger (1954/Reference Duverger, North and North1963) on the point:

Political choice takes the form of a choice between two alternatives. A duality of parties does not always exist, but there is almost always a duality of tendencies…. This is equivalent to saying that the centre does not exist in politics…. The term “center” is applied to the geometrical spot at which the moderates of opposed tendencies meet…. Every Center is divided against itself and remains separated into two halves, Left-Center and Right-Center. For the Center is nothing more than the artificial grouping of the right wing of the Left and the left wing of the Right. The fate of the Center is to be torn asunder. (215)

So, if parties are typically pulled to the centre by competitive considerations, they rarely start there.Footnote 23 But the centre is exactly where the Liberal party of Canada starts.

Where a major party commands the centre, opposition is forced to be bilateral, coming from both ends of an ideological or policy spectrum. For Sartori, because the centre is occupied it is “out of competition” in that “the very existence of a center party … discourages … the centripetal drives of the political system” (Reference Sartori1976: 135). Oppositions are likely to be irresponsible and engage in a politics of outbidding. In itself this may not be bad if in the long run emptying the centre creates the conditions for the ideal form of competition, that is, off-centre parties responding to the pull of the centre. But also typifying polarized pluralism, indeed probably a critical factor in its very existence, is the presence of one or more “anti-system” parties. Such parties do not see themselves as engaged in the struggle for power under the existing rules but rather as committed to changing those rules. Classic examples are Communist and fascist parties. Votes and seats for these parties are subtracted from the zone of true competition, yet their presence limits the scope for coalition building. At the same time, their threat to the system encourages concentration on the centre by voters concerned to maintain the overarching polity. Even so, the centre cannot hold: “centrifugal drives [prevail] over centripetal ones” leading to “enfeeblement of the centre, a persistent loss of votes to one of the extreme ends (or even to both)” (Sartori, Reference Sartori1976: 136).Footnote 24

Sartori did not see any of this applying to Canada. For him, the critical thing about Canada is that governments are formed by one party only. The fact that some of these are minority governments testifies all the more to the powerful logic of the Westminster system: “This pattern [minority governments] attests … to the force of the inner, systemic logic of twopartism. One could also say—with respect to the ‘conventions’ of the constitution—that the Canadians are more British than the British themselves” (188–89). Much of what Sartori saw as characterizing the Canadian case still holds. Governments are still formed by only one party and only four times since he wrote has the governing party been in the minority. But many features of the polarized pluralism model echo recent—in some cases, abiding—tendencies in Canadian politics. And it is the key to understanding both peculiarities—the rise of riding-level three-party competition and the cycles in sectionalism—embedded in Canadian multipartism.

Left-Right Ideology and Local Tripartism

Earlier I alluded to the risks of co-ordination failure on the centre-right and centre-left. In fact, such risks are low. Failure on the centre right has never produced a federal government of the left. Conservatives may complain that the Liberal party really sits on the left, not at the centre, but that is not a complaint about a failure to co-ordinate. The NDP finishes first or second—usually second—in about one riding in four, one in three outside Quebec. So the centre right rarely faces a threat from the left. The threat from the right is usually greater. The 1990s aside, Conservatives, Progressive or otherwise, finish first or second in 60 to 80 per cent of ridings, 80 to 90 per cent outside Quebec. The Conservatives did even better in 1984 and 1988. And Conservative majorities, when they occur, are usually so overwhelming that co-ordination elsewhere on the spectrum hardly seems relevant. Only in 1988 was it plausible that centre-left co-ordination failure yielded a perverse result. For all that, the Liberals routinely do best of all: they win or place in 70 to 90 per cent of all ridings, even when their backs are to the wall. Voters who prefer the Liberal party rarely need to move to block an unacceptable outcome. Ironically, Liberal party strength permits voters on the right to support a conservative alternative without risking victory by the party furthest left. The Liberals do the same for voters on the left, although less often. When there is a real threat from the right, it tends to be the Liberals, not the NDP, who benefit from strategic consolidation.

In most of Canada, the only other strategic option available to Conservatives and New Democrats is to combine with each other. Merely to say this is to restate the problem: the presence of the strong, centrist Liberals. To the extent that politics is organized on a left-right basis, an NDP-Conservative combination is implausible, Red Tory nostalgia notwithstanding. In general, supporters of parties that are ideologically disconnected—separated from each other by one or more intermediate parties—do not coalesce. The Liberals are the “Condorcet winner,” the party that beats all others in a straight fight. As such, they will be the co-ordination point when either ideological extreme threatens to take power.

The centre has shrunk, however, consistently with the polarized pluralism's logic. Liberal vote shares shrank after 1960, even though the party continued to win seat majorities. Compared to earlier winning years, the Liberal share shrank further in 1993. From then until 2004, Liberal parliamentary majorities rested on a narrower electoral base than formerly had typified minority governments, and the tiniest of perturbations would have deprived them of their majorities.

The National Question and Governmental Succession

By their very existence as a strong party of the centre, the Liberals are also responsible for the system's peculiar sectional dynamics. This is a story about the national question, if we admit that the question takes us beyond the boundaries of Quebec. The dynamics partake of two things: the existence of anti-system parties or tendencies; and the continuing imperative, given FPP, of single-party governments.

Figure 3 shows how cyclical sectionalism maps onto the history of Conservative success and failure. It does so by portraying the geographic basis of party coalitions in three parliamentary situations. The horizontal axis arrays the provinces from West to East. The vertical axis gives average values of the federal vote within each province. The portrayal is for all years since the completion of the nine-province system in 1908, with Newfoundland dating from 1949. The nuance washed over by such draconian pooling is not central to my argument. Besides, surprisingly little nuance is lost: the picture remains remarkably stable over many temporal groupings.

Figure 3 Geographical Inclusiveness of Electoral Coalitions

To underscore just how remarkable the Conservative pattern is, consider first the Liberal one. The popular basis of Liberal governments has always been highly differentiated geographically. The East-West gradient steepened in the second half of the twentieth century, but changes since 1921 have been very modest. When the Liberals retreat from power differences sharpen only at the margin. When they lose their majority but hold on as a minority government, losses are typically outside Quebec, and Quebec sustains them in power. Only when they suffer a Quebec reverse do they actually lose power. Although this pattern may not indicate healthy, broad support for a purportedly national government, its dynamics are at least moderate and contained.

The Conservative Party, in contrast, constantly flirts with the edge. When in opposition, Conservative support follows an East-West gradient altogether like that for the Liberals. This was least true from 1963 to 1984, although even in those years Conservative support was lower in Manitoba, Saskatchewan and BC than in Atlantic Canada. Conservative support seems more uniform when the party forms minority governments. This reflects the fact that all Conservative minorities postdate 1957.Footnote 25 Indeed, Conservative strength in the West seems to be a necessary condition for the party even to approach power.

The West does not suffice to give Conservatives parliamentary majorities, however. For that, capturing Quebec is the necessary final step. True, in achieving majority status, the party makes gains generally proportional to former weakness. But the gains in Quebec are always stunning. This was the pattern for the 1984 Conservative landslide: a Quebec swing of 38 points, compared to 10 points elsewhere. The 1984 pattern magnified but otherwise repeated earlier episodes: in 1958, the Quebec gain was 19 points, compared 15 elsewhere; in 1930, 10 points, compared to 2; and in 1911, 5 points, compared to 2.5. Figure 3 portrays not just gains, but losses as well, and the flipside of stunning growth can be sudden collapse. With the exception of 1957–1963, each Conservative experience of power left the party worse off.

The asymmetry between the parties implies that Conservative gains and losses come in exchange not so much with the Liberals as with third parties, in particular with parties representing sectional interests. Many of these parties qualify for Sartori's designation as “anti-system.” The parties that qualify, some only barely, include the Progressives (especially the United Farmers of Alberta), Reconstruction, Social Credit, the Bloc Populaire Canadien, the Bloc Québécois, and Reform (but probably not the Alliance).Footnote 26 The key, in my view, is not that these parties always reject Westminster parliamentarism or the Canadian union. It is that at critical moments, these parties dangle a metacritique of the system and that voters rise to the bait, whether or not these voters know what they are doing or intend much by the act. The rhetoric is, in Sartori's terms, one of outbidding. Even if little comes of the rhetoric, the bar is raised for the next round. Under FPP, anti-system insurgency can have compounded effect (Cairns, Reference Cairns1968). And as long as an anti-system party persists, it compresses the scope for government formation.Footnote 27

Reflections

To recapitulate, the key to the puzzle of Canadian elections is their domination by a party of the centre. At the riding level this accounts for three-party competition, where outside Quebec the centrist Liberals dominate a left-right ordering. Given strategic circumstances, the only option for consolidation involves an ends-against-the-middle coalition, and this is unrealistic. Such coalition building is possible across regions, however, as the Conservative record shows. Joining the opposites sometimes involves absorbing anti-system parties or anti-system tendencies. The resulting incoherence accounts for the short life and commonly dire fate of Conservative majority governments. The fall of Conservative governments is often the midwife of anti-system parties, with the Bloc and Reform as only the most recent examples. Conservative boom and bust is the source of the episodic volatility in Canadian elections and is in fact the complement of Liberal longevity.

The logic outlined in this paper also helps account for federal–provincial divergence. The three-party dynamic in federal constituencies does not extend to provincial ones, not in most provinces at least. Where circumstances favour growth on the left, as they first did in the West, jockeying will ensue for survival on the centre right and only one party can remain as an alternative to the NDP. Where circumstances do not favour the NDP, as has traditionally been the case in Atlantic Canada, no centre-right consolidation is required in provincial elections, but neither is three-party competition the norm.

Why does a party of the centre survive? Many voters are undoubtedly happy that it does, and its survival makes Canada the sole majoritarian system in which the governing party routinely covers the median (Powell, Reference Powell2000). It also makes Canada the only system among the obvious comparators not dominated by a party of the right. Most likely, the key to Liberal dominance is the national question. It is a commonplace to observe that the Liberals have been historically better than others at managing that question. But their survival as a dominant player requires that they control a pole on at least one major dimension of choice and, as I have shown that for the electorate as a whole, they do not control such a pole, not even on the national question. The electorate is not a unitary entity, however; it is segmented between, at a minimum, Quebec and the rest of Canada. If Liberal supporters are a guide, their party controls a pole within each segment, but the opposite one between segments: the pro-Quebec pole outside Quebec; the pro-Canada one inside Quebec. This does not require the Liberals to be peculiarly inconsistent, only that the electorate be segmented.

The origins lie, I believe, in the period in which the national question was outward looking—Canada in Empire—and the Liberal party controlled the anti-imperial pole. Imperial relations were the only policy domain in which party and religious interests aligned cleanly. All other issues with linguistic or religious content divided each party against itself (Crunican, Reference Crunican1974; Brown, Reference Brown1975; Miller, Reference Miller1979). Although Quebec was the chief locus of anti-imperial sentiment, it was so as the most Catholic place. Outside Quebec, Catholic persons and places also gravitated to anti-imperial appeals. The 1900 election, with the South African War on its agenda, was the test run for the twentieth century. Even when external relations dominated party choice, the potential for polarized pluralism was detectable. The 1911 election is commonly thought of as mainly about commercial policy, the failed Canada–US Reciprocity Agreement. But it was also a referendum on imperial relations more generally, with the 1910 Naval Bill as the case in point. The Laurier government's creation of the Canadian Naval Service was a compromise between British demands for dreadnoughts and Quebec's resistance to any naval policy whatsoever. Outside Quebec, the policy was castigated as insufficiently imperialist. In Quebec, it was condemned for opposed reasons. The beneficiary of both critiques was the Conservative party: ends against the middle.Footnote 28

This pattern became clearer as the national question turned inward, to the place of Quebec in or out of Canada. The Liberal party was able to cash out its older record on continuing credibility in Quebec, at least for a time. But with the inward turn came a “stretching,” so to speak, of the Quebec end of this dimension, and options hitherto unimaginable—or at least, unspeakable—became concrete possibilities. This still leaves the Liberals in the centre. But can the Liberals sustain their position? As mentioned, the middle has been shrinking. Perhaps a centre party can remain a major player only so long as it is the dominant one. If so, the Conservative objective must be to keep the Liberals out long enough that their claim to superiority in managing the Quebec–Canada relationship founders on the mere fact of distance from power.

If I am right for the Canadian case, then it follows that comparative analysis must move beyond studying electoral systems in a context- and history-free vacuum. It does not suffice to invoke the mere fact of diversity, even in an interactive setup in the manner of Ordeshook and Shvetsova (Reference Ordeshook and Shvetsova1994) or Amorim Neto and Cox (Reference Amorim Neto and Cox1997). After all, the Canadian case does not feature a weak electoral system that accommodates diversity. Rather, the Canadian system punishes co-ordination failure as severely as any in the world, and yet such failure persists. Comparative analysis should also, on my argument, address systemic relations among parties. Does a strong party of the centre in fact necessitate centrifugal appeals by the other players, or does that logic apply only to anti-system parties? Or is it the presence of credible anti-system parties or of potentially exploitable anti-system tendencies that in turn privileges the centre? To the extent that two-party competition remains normative among our comparators, understanding its breakdown is necessary to identify the contingencies that underpin bipartism where it survives. These contingencies may be disappearing, such that multipartism will become the norm even under strong electoral systems. If so, it is all the more critical to understand Canada, the country ahead of the curve.

Footnotes

1 I am not the first to make this claim for Canada. Pride of place goes to William Dobell (Reference Dobell1986).

2 Although Duverger was not the first to formulate the proposition that FPP constrains party systems to a low number, it was he who claimed lawlike status for it (Riker, Reference Riker1982).

3 Rae (Reference Rae1971) provided the first systematic empirical demonstration. Taagepera and Shugart (Reference Taagepera and Shugart1989) shifted emphasis among electoral institutions and raised the level of empirical sophistication several notches. Lijphart (Reference Lijphart1994) produced the most complete inventory of patterns. Cox (Reference Cox1994) supplied the pivotal theoretical elaboration and Cox (Reference Cox1997), the complementary empirics.

4 The notion of an electoral system's strength or weakness originates with Sartori (Reference Sartori, Montgomery and Hirschmann1968).

5 Riker (Reference Riker1982) and Benoit (Reference Benoit2006) are key intellectual histories of the Duvergerian idea, and this paragraph depends heavily on these accounts. Also important is the careful summary in Chhibber and Kollman (Reference Chhibber and Kollman2004), chapter 2.

6 The calculation is as follows:


$$
ENP_t  = {1 \over {\sum\limits_i {p_u^2 } }}
$$

where:

ENPt is the effective number of parties in the election at t; and

p it is the i-th party's proportion of seats or votes in the election at t; 0 < p it < 1 ; ∑ip it = 1

The intuition behind the indicator is exactly the same as in Rae's index of fractionalization (Reference Rae1971). If p it gives the probability of randomly selecting a supporter of the i-th party at t, p it2 gives the probability of selecting a supporter of that party on successive independent draws. If the system has two parties of exactly equal size, the probability of choosing one party's supporter on a given draw is 0.5 and of choosing such persons on successive draws is 0.25. In this situation, the same probabilities hold for the other party. The sum of the two probabilities is 0.5: one has a 0.50 probability of choosing supporters of the same party, 0.25 for each party. If the system has two parties but one is larger than the other, the probability of choosing two supporters of the same party is greater than 0.5; on the arithmetic, the diminished probability of choosing supporters of the smaller party is more than offset by the enhanced probability of selecting from the larger group. Contrariwise, if the number of alternatives is more than two, the probability of choosing from the same party is diminished, all the more so as the number of alternatives increases and as each alternative approaches equiprobability with the others. If we now take all this logic and, literally, turn it on its head, by taking the reciprocal of ∑ip it2, we get an indicator with a minimum value of one. Where ∑ip it2 = 0.5, its reciprocal equals 2, and so on. For an acerbic critique of ENP, see Dunleavy and Boucek (Reference Dunleavy and Boucek2003).

7 For the record, the effective number for the first few elections is smaller in my series than in the series reported in Chhibber and Kollman (Reference Chhibber and Kollman2004). This reflects coding decisions about partisan affinities among candidates: I tend to see affinities where Chhibber and Kollman do not. The difference matters little for the argument in this paper, which is about trends over the twentieth century, by which point my series and the Chhibber-Kollman one coincide.

8 For all this, Canada was not the extreme case among FPP systems. Over the postwar period, the Indian electorate was even more fractionalized than Canada's. Generally speaking, India's ENP averaged about 4 when Canada's ENP averaged about 3. The further fractionalization of the Canadian system in the 1990s closed the gap considerably (but not entirely, especially as the Indian system fractionalized a bit more in the late 1990s).

9 Over the full period since 1878, the electoral system dismissed the equivalent of 0.5 parties. Since 1935, 0.75 parties have been stripped, and in two periods—1935 to 1958 and 1984 to 2000—the electoral system dismissed the equivalent of more than one party.

10 The empirical number seems to be 2.5 (Taagepera and Shugart, Reference Taagepera and Shugart1989).

11 The 1930 election undid nearly all of the 1921 damage, 1958 did so for much of the 1935 breakdown, 1984 corrected some of 1962, and 2000 reversed a small part of 1993.

12 It was made so by Riker (Reference Riker1982) and Cox (Reference Cox1994, Reference Cox1997).

13 See in particular their chapter 2. Where this chapter's first half summarizes decades of intense theoretical work on the link between electoral formula and number of district parties, the second half is a frank discussion of the failure of the Duvergerian two-party prediction as well as of Cox's non-Duvergerian equilibrium (Reference Cox1994). After all that, however, they conclude the chapter by noting that the plurality formula nonetheless exerts a defractionalizing pressure. They interpret this as mandating their emphasis in the rest of the book on cross-district co-ordination failure (60).

14 Strictly speaking, cross-district failure could be as much within provinces as between them, but as Chhibber and Kollman (Reference Chhibber and Kollman2004), Figure 6-6 shows, within-province variation in ENP has decreased.

15 That said, local and extra-local ENP tend to move together in the short run, as spikes and drops in one tend to be accompanied by like shifts in the other. The correlation between the two series is 0.72. Some of this is the result of serial correlation in the disturbances of each series, but AR(1) estimations indicate that a unit shift in local ENP induces a 0.71 shift in extra-local ENP. Impact from a shift in the other direction is weaker, about 0.55, as the extra-local series is markedly more autoregressive than the local one. The point stands, however, that the two series share a large fraction of short-term variation.

16 Social Credit was also a factor, but only in a handful of places.

17 This is not quite—or not always—true. Until the consolidation of the Bharatiya Janata Party (BJP), Congress controlled the middle of Indian party spectrum and posed its rivals with a co-ordination challenge altogether like that in Canada. This was a central point in Riker (Reference Riker1976), which contrasted Canadian and Indian multipartism. Canada has now converged on the Indian case, or the countries have switched places.

18 The left-right indicator is based on the following item in the 2004 Canadian Election Study (CES) mailback: “In politics people sometimes talk of left and right. Where would you place yourself on the scale below?”

The scale ranges from 0 to 10. The indicator of support for Quebec is: “How much do you think should be done for Quebec: much more, somewhat more, about the same as now, somewhat less, or much less?”

This question was asked of 2006 CES respondents. Party preference is indicated by reported 2006 vote in the CES. The juxtaposition of 2004 left-right and 2006 Quebec-orientation and party-support data was made possible by the fact that the 2004 and 2006 CES surveys had a panel component that comprised roughly half the sample each year.

19 The ordering conforms to the patterns for party platforms in Budge et al. (Reference Budge, Klingemann, Volkens, Bara and Tanenbaum2001) and Benoit and Laver (Reference Benoit and Laver2006).

20 Benoit and Laver (Reference Benoit and Laver2006) have a “Quebec” dimension that locates parties as pro- or anti-sovereignty and place all parties other than the Bloc at the same spot. The Bloc and the Liberal party are coded as placing more emphasis on the issue, however.

21 The first account, however, seems to be Sartori (Reference Sartori, LaPalombara and Weiner1966).

22 The centre is not defined by some absolute standard but by relative positioning:

“when one speaks of a centre-based system, one is concerned only with a centre positioning, not with centre doctrines, ideologies, and opinions—whatever these may be”

(Sartori, Reference Sartori1976: 134).

23 This is the dynamic tension captured by Aldrich (Reference Aldrich1983).

24 The pattern can appear only when there are more than two parties. Indeed, Sartori's conjecture, based on PR examples, is that at least six (nominal, not “effective”) parties are required. The Canadian case does not quite reach this threshold, especially in the early years but I want to argue that this particular threshold is not critical.

25 The short-lived 1926 Meighen government does not count for my purposes, as the party was in opposition for the bulk of the relevant parliament.

26 Flanagan (Reference Flanagan1995) shows that the Reform program had other elements in contradiction with this aim, and these other elements ultimately prevailed, as did the imperative to co-ordinate the right-wing vote across provinces.

27 The Canadian picture of massive, recurring swings has particular affinities with one of Sartori's polarized pluralism cases, Chile: “The election of the president by universal suffrage imposes alliances that—in a polarized system—generate strong tensions within the centre group of parties and induce acrobatic ideological leaps along the left-right dimension” (Sartori, Reference Sartori1976: 159). The critical fact about Chile is that, unlike the other three cases, it is a true presidential system, that is, a system with a directly elected national office. Strictly speaking, where no candidate received an outright majority—the usual case—Congress was required to choose between the top two vote getters. But no matter how close the vote or what might have been the second choices of supporters of the eliminated candidates, Congress always chose the first-place candidate. This produced striking dynamics and the eventual election, with 37 per cent of the vote, of Salvador Allende. So the Chilean formula is, in effect, FPP for its most important office.

28 See in particular, Brown (Reference Brown1975: 163ff, 235 passim).

References

Aldrich, John H. 1983. “A Downsian Spatial Model with Party Activism.” American Political Science Review 77: 974–90.CrossRefGoogle Scholar
Amorim Neto, Octavio and Cox, Gary W.. 1997. “Electoral Institutions, Cleavage Structures, and the Number of Parties.” American Journal of Political Science 41: 149–74.CrossRefGoogle Scholar
Bakvis, Herman and Macpherson, Laura G.. 1995. “Quebec Block Voting and the Canadian Electoral System.” Canadian Journal of Political Science 28: 659–92.CrossRefGoogle Scholar
Bardi, Luciano and Mair, Peter. 2008. “The Parameters of Party Systems.” Party Politics 14: 147–66.CrossRefGoogle Scholar
Benoit, Kenneth. 2006. “Duverger's Law and the Study of Electoral Systems.” French Politics 4: 6983.CrossRefGoogle Scholar
Benoit, Kenneth and Laver, Michael. 2006. Party Policy in Modern Democracies. London: Routledge.CrossRefGoogle Scholar
Blondel, Jean. 1968. “Party Systems and Patterns of Government in Western Democracies.” Canadian Journal of Political Science 1: 180203.CrossRefGoogle Scholar
Brown, Robert Craig. 1975. Robert Laird Borden, A Biography. Volume I: 1854–1914. Toronto: Macmillan.Google Scholar
Budge, Ian, Klingemann, Hans-Dieter, Volkens, Andrea, Bara, Judith and Tanenbaum, Eric. 2001. Mapping Policy Preferences: Estimates for Parties, Electors, and Governments, 1945–1998. Oxford: Oxford University Press.CrossRefGoogle Scholar
Cairns, Alan C. 1968. “The Electoral System and the Party System in Canada, 1921–1965.” Canadian Journal of Political Science 1: 5580.CrossRefGoogle Scholar
Caramani, Daniele. 2004. The Nationalization of Politics: The Formation of National Electorates and Party Systems in Western Europe. Cambridge: Cambridge University Press.CrossRefGoogle Scholar
Carty, R.K. 1988. “Three Canadian Party Systems: An Interpretation of the Development of National Politics.” In Party Democracy in Canada, ed. Perlin, G.C.. Toronto: Prentice-Hall.Google Scholar
Chhibber, Pradeep K. and Kollman, Ken. 2004. The Formation of National Party Systems: Federalism and Party Competition in Canada, Great Britain, India, and the United States. Princeton, NJ: Princeton University Press.Google Scholar
Cox, Gary W. 1987. The Efficient Secret: The Cabinet and the Development of Political Parties in Victorian England. Cambridge: Cambridge University Press.CrossRefGoogle Scholar
Cox, Gary W. 1994. “Strategic Voting Equilibria under the Single Nontransferable Vote.” American Political Science Review 88: 608–21.CrossRefGoogle Scholar
Cox, Gary W. 1997. Making Votes Count: Strategic Coordination in the World's Electoral Systems. Cambridge: Cambridge University Press.CrossRefGoogle Scholar
Crunican, Paul. 1974. Priests and Politicians: Manitoba Schools and the Election of 1896. Toronto: University of Toronto Press.CrossRefGoogle Scholar
Dobell, W.M. 1986. “Updating Duverger's Law.” Canadian Journal of Political Science 19: 585–95.CrossRefGoogle Scholar
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row.Google Scholar
Dunleavy, Patrick and Boucek, Françoise. 2003. “Constructing the Number of Parties.” Party Politics 9: 291315.CrossRefGoogle Scholar
Duverger, Maurice. [1954] 1963. Political Parties: Their Organization and Activity in the Modern State, trans. North, Barbara and North, Robert. New York: Wiley.Google Scholar
Enelow, James N. and Hinich, Melvin J.. 1984. The Spatial Theory of Voting: An Introduction. Cambridge: Cambridge University Press.Google Scholar
Flanagan, Thomas. 1995. Waiting for the Wave: The Reform Party and Preston Manning. Toronto: Stoddart.Google Scholar
Irvine, William P. 1979. Does Canada Need a New Electoral System? Kingston, ON: Queen's University Institute for Intergovernmental Relations.Google Scholar
Johnston, Richard and Ballantyne, Janet. 1977. “Geography and the Electoral System.” Canadian Journal of Political Science 10: 857–66.CrossRefGoogle Scholar
Johnston, Richard, Blais, André, Brady, Henry E. and Crête, Jean. 1992. Letting the People Decide: Dynamics of a Canadian Election. Montreal and Kingston: McGill-Queen's University Press.CrossRefGoogle Scholar
Laakso, Markku and Taagepera, Rein. 1979. “Effective Number of Parties: A Measure with Application to West Europe.” Comparative Political Studies 12: 327.CrossRefGoogle Scholar
Lijphart, Arend. 1994. Electoral Systems and Party Systems: A Study of Twenty-seven Democracies. Oxford: Oxford University Press.CrossRefGoogle Scholar
Lipset, Seymour Martin. 1954. “Democracy in Alberta.” Canadian Forum November: 175–77; December: 196–98.Google Scholar
Lovink, J.A.A. 1970. “On Analyzing the Impact of the Electoral System on the Party System in Canada.” Canadian Journal of Political Science 3: 497516.CrossRefGoogle Scholar
Miller, J.R. 1979. Equal Rights: The Jesuits' Estates Act Controversy. Montreal: McGill-Queen's University Press.Google Scholar
Ordeshook, Peter C., and Shvetsova, Olga. 1994. “Ethnic Heterogeneity, District Magnitude, and the Number of Parties.” American Journal of Political Science 38: 100–23.CrossRefGoogle Scholar
Powell, G. Bingham. 2000. Elections as Instruments of Democracy: Majoritarian and Proportional Visions. New Haven and London: Yale University Press.Google Scholar
Rae, Douglas W. 1971. The Political Consequences of Electoral Laws. 2nd ed.New Haven and London: Yale University Press.Google Scholar
Riker, William H. 1976. “The Number of Political Parties: A Re-examination of Duverger's Law.” Comparative Politics 9: 93106.CrossRefGoogle Scholar
Riker, William H. 1982. “The Two-Party System and Duverger's Law: An Essay on the History of Political Science.” American Political Science Review 76: 753–66.CrossRefGoogle Scholar
Sartori, Giovanni. 1966. “European Political Parties: The Case of Polarized Pluralism.” In Political Parties and Political Development, ed. LaPalombara, Joseph and Weiner, Myron. Princeton, NJ: Princeton University Press.Google Scholar
Sartori, Giovanni. 1968. “Political Development and Political Engineering.” In Public Policy, ed. Montgomery, John D. and Hirschmann, Albert O.. Cambridge: Cambridge University Press.Google Scholar
Sartori, Giovanni. 1976. Parties and Party Systems: A Framework for Analysis. Cambridge: Cambridge University Press.Google Scholar
Smith, David E. 1985. “Party Government, Representation, and National Integration in Canada.” In Party Government and Regional Representation in Canada, ed. Aucoin, Peter. Toronto: University of Toronto Press.Google Scholar
Taagepera, Rein and Shugart, Matthew Soberg. 1989. Seats and Votes: The Effects and Determinants of Electoral Systems. New Haven, CT: Yale University Press.Google Scholar
Figure 0

Figure 1 Components of Electoral Fractionalization

Figure 1

Figure 2 The Policy Space, 2004–6

Figure 2

Figure 3 Geographical Inclusiveness of Electoral Coalitions