This file is also available in Adobe Acrobat PDF format
THE SAME POLITICAL PARTIES dominate contemporary American politics at the national level and in nearly every state. Despite a few well-publicized independent candidates and politicians and, in recent years, a smattering of celebrities from the Reform Party or Independent Party, such as Ross Perot, Patrick Buchanan, and Jesse Ventura, the Democratic and Republican parties control congressional delegations from all the states, majorities in the state legislatures, and governorships in all but a few states.1 Since the early twentieth century, the United States has displayed a pattern of virtually complete two-partism--that is, two national parties compete and win seats in every major region in the nation. Two-partism, however, has not always been characteristic of the United States, at least not for congressional and state elections. Throughout most of the nineteenth century, electoral support was spread across more than two parties, and some parties were competitive only in a few states. In certain regions, such as the South from the 1890s to the 1970s, one party predominated for long periods.
Compared with the contemporary era in the United States when two major parties compete and win seats in every major region in the nation, the 1850s marked a more highly regionalized, fractious, and turbulent decade in American party politics. After the collapse of the Whig Party early in the decade and subsequent attempts by various competitors to win cross-regional support, it was not clear whether another national party would emerge to compete with the Democrats for control of the national government. For a time the Know-Nothings, a shadowy political group dedicated to anti-immigrant and anti-Catholic principles, sought to be that party. But the Know-Nothings foundered by mid-decade, not only because their northern and southern factions split over the slavery issue but also because they could not agree on a national policy program to deal with issues that were essentially local in American society to that time: regulation of liquor, authorizations for Catholic schools, and legal tolerance of immigrant laborers (Gienapp 1987). Because these issues were addressed either by state or local governments, party politics for several elections was characterized by electoral competition among state-level or regional parties.
By creating an intraparty consensus around a national policy program opposing the expansion of slavery into new territories, the Republicans ultimately forged a national party. Republican leaders came to recognize that electoral success required presenting to the electorate national policy solutions to the great questions of the age. As a result, the young party won the presidency and dominated the congressional elections in 1860. It was a humiliating loss for the Democrats and for the political leaders in the South, leading to southern secession and civil war.2 In the 1990s politicians and voters in India and Canada faced difficulties in forming national coalitions not unlike those of nineteenth-century Americans. Numerous political parties fielded candidates in India, but in 2002 none had been able to craft a national majority to replace the Congress Party, which had lost its dominant status in the late 1980s. As of this writing, only two parties since 1991 have contested for national power, the Congress Party and the Bharatiya Janata Party (BJP), a right-wing Hindu nationalist party. Neither party, however, has been successful in articulating a national policy program that unites disparate regional factions within its party or absorbs enough regional or state parties to form a winning national party. And neither party can overcome the most difficult obstacle to national power--namely, the perception that the delivery of public goods and services such as electricity and clean water are the perview of state governments.
Canadian voters faced a loss of a national opposition to the Liberals in the 1990s. The Conservatives went from the position of majority party in one election to a party with only three seats in the House of Commons after the 1993 elections. By the turn of the millennium, few could predict which party, if any, would coordinate enough voters across the provinces to provide a serious challenge to the Liberals for national control. Ideological and regional divisions among former Conservative voters prevented them from rallying behind a single national leader or party label, and attempts to build national coalitions by other parties of the right such as Reform and Alliance parties foundered. As a result, for the first time since World War II, a Canadian prime minister, Jean Chrétien from the Liberal Party, won a third consecutive parliamentary majority in 2000.
In Britain, two national parties competed to form majorities in the House of Commons for much of the twentieth century. Nevertheless, unlike the situation in the United States, there has nearly always been the persistent presence of a third party, which draws a substantial share of the vote away from Labour and the Conservatives. At times, it was the Liberals or the Social Democrats. Today, it is the Liberal Democrats. In Britain, minor party strength has fluctuated. During World War I, for example, British politics was in dramatic flux, and as many as five parties won at least ten seats to the House of Commons in 1918.
The current U.S. phenomenon of two national parties competing everywhere in the country does not exist in Canada, Great Britain, and India, even though they share many of the same electoral rules. Two-partism did not even exist in previous eras in the United States, although its electoral rules have stayed relatively constant. In the three other countries, the national legislatures seat politicians from parties that are strong only in particular provinces, regions, and states. In Canada and India, provincial and state politics are often dominated by parties that have little or no national standing. In contemporary Britain, after the 1997 devolution and creation of independent assemblies in regions, regional political parties such as Plaid Cymru and the Scottish Nationals have gained prominence.
Modern American two-partism not only looks unusual in comparison with party politics in other countries with similar electoral laws but also when compared with party politics in other eras in American history. Why is it that two national parties dominate the American political landscape in modern times? And why does this pattern not exist in other countries such as Canada, Great Britain, and India? In this book we seek to explain not only such differences across these countries but also to explain within each country why and when national parties emerged and why regional parties have drawn significant vote shares.
We show that, although these four countries have similar electoral systems--single-member, simple-plurality voting systems for the lower houses of parliament--party systems vary not only across these countries but also over time within these countries. Using historical data, we attribute changes in the party systems in these nations to the changing role of the state. In particular, we examine the relationship between the national (federal) and provincial (state) governments. Our claim that the nature of federalism influences the dynamics and stability of a party system differs from previous party system theories that stress the significance of social cleavages, electoral laws, and other constitutional features.
PARTY SYSTEM CHANGE IN CANADA, GREAT BRITAIN, INDIA, AND THE UNITED STATES
Political parties and party systems are vital to the functioning of modern democratic politics. We define a political party in this book as a group of candidates running for election under the same label. (We acknowledge, though, that parties accomplish far more than that. A more extensive discussion appears in chapter 3.) Parties provide a means to organize and coordinate voters, candidates, political donors, legislators, executives, and interest groups around common goals.
A party system is an enduring pattern of electoral competition between parties for public office. There are marked differences in party systems across countries, including the number of parties that compete regularly at the national and lower levels, the stability of the governing coalitions and opposition, the durability of party loyalties within electorates, and the frequency of new-party formation.
Our primary interest in this book is the formation of national party systems. We define a national party system as one in which the same parties compete at different levels of vote aggregation. In practice, this means that party systems at the constituency level, or at the state or provincial levels, look similar to national party systems. Our understanding of the nationalization of party systems is similar to that of Carmani (2000; 2004), Cox (1999), and Jones and Mainwaring (2003). According to Jones and Mainwaring (2003, 140), a party system is "highly nationalized . . . [when] the major parties' respective vote shares do not differ much from one province to the next. In a weakly nationalized party system, the major parties' vote shares vary widely across provinces."
One widely studied component of party systems is the number of parties. The number of parties contesting seats in lower-house elections in Canada, Great Britain, India, and the United States has differed over time. In Britain, for example, twelve parties won seats to the House of Commons, and five parties won at least ten seats each in the 1918 elections. In 1992, five parties won seats, and three parties won at least ten seats. For elections to the Canadian House of Commons, the number of parties winning more than 10 percent of the national vote has fluctuated from as low as two (in 1925, for example) to as high as four in 1979 and 1993, while the number of parties seated in Parliament has ranged from two to as high as twelve.
In most political systems with free and fair elections, there can be dozens, if not hundreds, of parties competing in elections, most of which have little or no bearing on governments and policy outcomes. There are often hundreds of independent candidates who avoid party labels altogether. This is especially true for single-member district systems such as in the four countries examined here. In the United States, for example, the Prohibition Party ran candidates in hundreds of districts across the country from the 1870s to the 1950s. For a thirty-year stretch, from the 1880s to the 1910s, it won between 0.5 and 3.0 percent of the national vote in congressional elections, and fielded candidates in approximately half of the nation's congressional districts, although it never elected a single member to Congress. While still fielding hundreds of candidates into the 1950s, it won no more than 0.5 percent of the national vote in any decade after the 1910s. And this is a party that actually may have had some bearing on national policy during its peak in the early part of the twentieth century, especially in pressuring the major parties to adopt the policy goal of prohibition.3 Far more numerous, and less consequential, are the candidates who have run under obscure or humorous labels, such as the Umoja Party, the Miller High Life Party, or the Politicians Are Crooks Party, to give some examples from congressional elections in recent years.
Similar examples exist in most countries, and as these parties have little bearing on either the vote share of major parties or who governs, political scientists do not simply count the number of parties that contest elections as indicative of how many parties are competitive in a party system. Instead, political scientists use measures that calculate how many parties actually influence political outcomes. The most commonly used measure is Laakso and Taagepera's (1979) "effective number of parties" index, or N, which gives increasing weight to parties that get higher proportions of the vote. The formula is the inverse of the sum of the squared proportions of the vote or of the seats. For n parties receiving votes, and for pi representing the proportion of popular votes received by party i,
If one is using votes to calculate proportions, as we do throughout this book, when 2 parties share 98 percent of the vote equally between them, and 100 tiny parties win the remaining 2 percent, the measure will be very close to 2. When 2 parties each win 44 percent of the vote and a third party wins 12 percent of the vote, however, the measure will be close to 2.5.4
Figures 1.1-1.4 show the fluctuations in the effective number of national parties competing in national lower-house elections in Canada, Great Britain, India, and the United States. First notice the United States in Figure 1.4. Prior to the New Deal, and most certainly throughout the nineteenth century, more than 2 parties regularly won substantial portions of the vote in elections to the House of Representatives. In Canada (Figure 1.1) the effective number of parties getting votes in national elections was 3.5 in the first national elections held in 1867. In 1917 only 2 parties competed in elections for the lower house, whereas in 1995 the effective number of parties getting votes was approximately 4. In India (Figure 1.3) the number of parties competing in national elections also has fluctuated, although there has been a steady increase since the 1970s with the effective number of parties rising from 3 in 1977 to almost 7 for the 1996 and 1999 elections. In Britain (Figure 1.2) the effective number of parties receiving votes in 1885 (after the Second Reform Act and the adoption of single-member districts for much of Britain) was 2.17, increasing to 4.43 in 1918, and then settling to about 2.5 for much of postwar period before rising above 3 in the 1990s.
There are also significant differences across the countries in the number of parties. The United States has had the fewest number of parties in the contemporary period, although in the first half of the nineteenth century the party system in the United States resembled party systems in the other countries. India and Canada have more parties on average than either the United States or Great Britain. India in recent decades has had an unusually large number of parties receiving votes for the Lok Sabha (the lower house).
The fact that these countries have not experienced consistent two-partism at the national level represents a departure from well-known assertions made about the number of parties in countries with single-member district systems. Duverger ( 1963) wrote that countries with single-member districts tend to have two dominant national parties, and a large volume of literature has followed his original book. All four of these countries had single-member, simple-plurality electoral systems for lower-house elections during the periods represented in Figures 1.1-1.4. Yet the effective number of parties has varied over time in each of these nations. Although many qualifications and addendums have been made to Duverger's "Law"--such as William Riker's (1982) revised version, which seeks to explain why Canada and India did not conform to the prediction of two-partism, or Sartori's (1976) version that links the number of serious competitors to the ideological differences among the major parties--none has explained successfully both the changes over time in the number of parties in such systems and the variation in the number of parties across these countries.
We show in chapter 2, however, that two-partism continues to be a robust phenomenon at the district or constituency level. If the effective number of national parties is considerably above 2 while the average effective number of district or constituency level parties is near 2--a common occurrence in these countries--it follows that the same parties do not compete across local levels. In recent elections in India, for example, the effective number of national parties has been near 7, while the average effective number of parties at the constituency level is much lower, near 2.5. This suggests that different parties are getting significant shares of votes across constituencies and that many parties get votes only in particular locales. By our definition of a national party system, the degree to which the gap between national and local party systems exists is the extent to which the party system deviates from a pure national party system.5
In the remainder of this book, we offer an explanation for when a national party system will be formed and detail conditions under which the party system may not be national. We also suggest reasons for why we do not find the same number of national parties across these countries. In the remainder of this introductory chapter, we discuss previous explanations for changes in party systems, provide an introduction to our explanation, and lay out the plan for the book. We also describe the electoral data we use.
APPROACHES TO THE STUDY OF PARTIES AND PARTY SYSTEMS
Party systems have real consequences, affecting the quality and nature of democratic representation, economic policies, and the stability of governments and political systems.6 There is evidence, for example, that the number of parties in governing coalitions--which is related to the number of parties in the party system--affects the ability of governments to respond to economic shocks (Franzese 2002). Likewise, voter turnout across countries is positively correlated with various aspects of party systems, including the number of parties (Blais and Dobrzynska 1998). And some scholars have linked the success of regional parties to secessionist pressures (Filippov, Ordeshook, and Shvetsova 2004).
One could cite evidence that party systems are shaped to some extent by electoral systems and other features of governmental institutions and that the consequences of having national parties or a certain number of parties are not nearly as important as the consequences of having in place a parliamentary, majoritarian, or presidential system of government (see Powell 2000 for a summary of these arguments). In other words, party system differences could be merely epiphenomenal, reflecting other differences in institutions that are more important in determining political outcomes and policies.
We believe that it is valuable to study the causes and consequences of party system change because these changes are themselves consequential. Cross-national research on a variety of topics often includes the number of parties in the parliament--an imperfect measure of the nationalization of the party system in cross-national studies--as a proxy for the degree of difficulty in introducing significant policy change, and these variables often are statistically significant in predicting policy outcomes, even when controlling for whether the country has one chamber of the legislature or two, has an elected president, or is federal (Persson and Tabellini 2003).
Although national party systems may not be inherently more desirable than party systems with regional parties, we agree with Sartori (1976; 1986) that having national parties, as opposed to fragmented, localized parties, tends to channel the choices of voters and politicians into a smaller number of coalitions and to force governments to confront national-level problems. Few would deny, for example, that if recent Indian governments had consisted of single-party majorities, such as in Britain, economic reforms arguably necessary for development would be easier to pass into law. Instead, Indian policy making was hampered in the early years of the twenty-first century by the constant struggle by the prime minister to keep in place his governing coalition, the National Democratic Alliance (NDA), which had well over a dozen parties (Bardhan 2002; Nayar 1999; Saez 2002).7 In contrast, it would have been very difficult for Tony Blair, as prime minister of Britain, to pursue reforms if he had not had a single-party majority but rather had had to contend with leaders of diverse leftist and center leftist parties to maintain a majority. Further, it is hard to disagree with Schattschneider's ( 1975) view that national, organized political parties are the most important countervailing powers to wealthy special interests in modern democracies.
Because parties and party systems are so central to democratic politics, their features and behavior have been the subject of research across many subfields in political science, including the study of voting, elections, legislatures, presidents and executives, bureaucracies, courts, electoral systems, and international relations. Scholars have used various methodologies and theoretical paradigms to analyze parties and party systems. Among these, three general approaches have dominated the literature.
Party Systems as Reflections of Social Cleavages
The first approach in analyzing party systems, and by far the most prominent in comparative politics, focuses on the nature of social cleavages that manifest themselves in party politics. Scholars seek to understand which groups in society political parties represent. Lipset and Rokkan (1967) offer a well-known thesis in this research tradition. They argue that deeply rooted, stable social cleavages lead to stable party systems. Mid-twentieth-century voting patterns in Europe reflected the economic, social, and religious divisions that arose as a result of the national and industrial revolutions many decades earlier (Caramani 2004; Katz and Mair 1994). It takes major social changes, such as postindustrialization (Inglehart 1997), civil war, depression, or massive population shifts, to alter those patterns significantly (Burnham 1970). Caramani (2004) demonstrates how pre-World War I electoral cleavages remained relatively stable throughout the twentieth century in Europe.
In this approach, social cleavages shape party systems in an almost axiomatic way. While political leaders can try to shift groups of like-minded voters into and out of parties to serve partisan or political ends, these efforts can be difficult. For some scholars, whom Torcal and Mainwaring (2001) term as the "objective" social relations interpreters, parties represent societal interests, and these interests are ontologically prior to partisan debates. Numerous country studies use this perspective to account for the nature of the party system. Often political scientists will either use or control for a set of social categories when attempting to explain developments in party politics.8
The literature on party systems in Canada, Great Britain, India, and the United States is predominantly rooted in this tradition. British electoral politics typically are described in terms of the class and regional cleavages that exist in Britain (Crewe, Fox, and Day 1995). Class forms the basis of the party system with the working class voting overwhelmingly for the Labour Party, whereas regionalism stokes the success of the Scottish and Welsh parties (Butler and Stokes 1970; Mughan 1986; Rose 1974a). In Canada, region and language also have been the dominant factors used to explain the number of parties and the fluctuations in partisan fortunes. Ethnic divisions began to play a large role in Canadian party politics in the 1890s and were, according to some, critical in explaining the first electoral success of the Liberals (Schwartz 1974; Martin 1974). Regional parties, such as the Social Credit of Alberta and the Progressives of Saskatchewan, emerged because of the ostensible neglect of farming issues by the industry-centered politics of eastern Canada (Martin 1974; Blais 1997; LeDuc 1985). The rise of linguistic separatism in Quebec also has reshaped the Canadian party system with the Parti Quebecois and later the Bloc Quebecois emerging as powerful electoral forces in Quebec. For Canada, as a result, a host of scholars continues to stress the continuing role of ethno-religious divisions such as Catholics and Protestants, French versus English, and urban-rural divisions in structuring the Canadian party system (see, for instance, Johnston et al. 1992, chap. 2).
Indian electoral politics, because of well-known social divisions in language, religion, and caste, have been analyzed largely in terms of the role played by existing social cleavages. Despite some notable exceptions (Kothari 1964), analysts have focused on the impact of caste in structuring much of the party system (Brass 1965; 1984; Rudolph and Rudolph 1987; Yadav 1996). Religion and "religiosity" have been regarded as a major social cleavage in contemporary Indian politics insofar as they lie at the base of support for the BJP (Jaffrelot 1995). Regionalism has been long present in Indian party politics, and the contemporary Indian party system is seen as following regional patterns rather than divisions based on ideologies or different preferences over national policies (Wallace 2000).
Multiple lines of research on American party politics also place social cleavages at the center of many analyses. Kleppner (1970) and Holt (1999) use the "objective" social cleavage of ethnicity--especially national origin--to account for developments in U.S. party politics in the nineteenth century. The study of party realignments, influential not only in research on the United States but also in research on politics in many countries, traces the changing nature of partisan coalitions back to preceding social and political crises that disrupt formerly stable alliances among social groups (Key 1949; Schattschneider  1975; Burnham 1970; Petrocik 1981; Sundquist 1983). For example, the Great Depression severed the long-standing allegiance of many northern whites and southern blacks to the Republican Party, allowing Franklin Roosevelt to forge a national Democratic coalition that endured for at least four decades. Research into partisan attachments among voters emphasizes the stability of voting patterns among large social groups in the United States (Campbell et al. 1960; Converse 1964). The voting patterns of a vast majority of southerners, Catholics, Jews, African Americans, urban residents, and suburban whites are predictable over many elections and change slowly, if at all. (The similarity of conclusions between the microlevel study in American Voter, by Campbell et al. , and the macrolevel study in Lipset and Rokkan's  research on European partisan coalitions is striking.) A relatively antiquated tradition in the study of the American two-party system explains the weaknessof third parties on the cultural dualism of American society, and the lack of a strong labor movement in the late nineteenth and early twentieth centuries (Lipset, Trow, and Coleman 1956; Charlesworth 1948; Hartz 1955). This line of research in American politics emphasizes how social cleavages, while malleable, are mostly durable and tend to shape the American party system.
The inverse effect, namely the impact of the machinations of party competition on recognizable and salient social cleavages, has been less emphasized (see Riker 1986; 1993). Some argue that political considerations have an influence on the party system somewhat independent of social cleavages, or that the cleavages themselves are formed as a result of party competition or interelite politics (Torcal and Mainwaring 2001; Bartolini 2000; Bartolini and Mair 1990). While social cleavages shape party systems, political leaders can try to shift groups of like-minded voters into and out of parties to serve partisan or political ends, although these efforts can be difficult. Voters develop strong loyalties for parties, politicians, and ideological labels, and they develop habits in partisan voting that are difficult to change. Accordingly, social cleavages can become politicized in interrelated ways that affect party systems (Bartolini 2000; Katz and Mair 1994; Kirchheimer 1966; Pederson 1979; Webb 2000). Politicians who seek public office campaign and form party coalitions that shape the partisanship of the electorate, which in turn solidifies party coalitions. Recent studies of Chile and Spain show that the role of political elites is indeed critical to cementing a relationship between political parties and social cleavages (Mainwaring and Torcal 2001). More generally, studies point to interelite politics as the cause of party systems based on social cleavages (Chhibber and Torcal 1997; Torcal and Mainwaring 2001). These recent inquiries that offer more autonomy to the "political" in structuring the relationship between social cleavages and the party system mostly stress the role of political elites in politicizing some cleavages over others; hence, interelite conflict lays the basis for the party system.
Government policies and the development of state bureaucracies also can create winners and losers that form or solidify partisan alliances (Kitschelt 1999; Maravall 1997). These policies interact with other social factors to form cleavages that remain durable over decades. Bartolini (2000), in a detailed analysis of the political mobilization of the left in European electoral politics for the twentieth century, suggests that the relationship of class to the party system is not axiomatic. Rather, the politicization of the class cleavages was determined by many features of European society and politics. The consolidation of external boundaries, the bureaucratization and centralization of the state, the degree of cultural heterogeneity, the educational level of the population and the extent of intrastate communication, the role of the church, economic transformation, and democratization all played roles in determining whether class was a partisan cleavage in particular countries. For Bartolini, the "economic-functional conflicts tended to prevail over other divisions with the formation and consolidation of the economic and administrative center. It was only with the development of the modern state and with the integration of different groups . . . [that] conflicts between these groups . . . [were] centralized" (18-19).
Parties as Solutions to Collective Dilemmas
The second approach, which has much in common with recent literature in many of the social sciences on the origins of institutions (North 1990), is relatively new and is by far the smallest of the three in terms of the number of published articles and books that rely on it. The approach begins with the fundamental question, Why have parties at all? Scholars seek to explain the origins and existence of political parties in the self-interested behaviors of voters, candidates, or legislators. For most scholars writing in this literature, parties have their origins in legislatures (Aldrich 1995; Cox and McCubbins 1993). Collective dilemmas such as cycling majorities and collective action problems are inherent in democratic politics, especially in legislatures. As a result, entrepreneurial politicians have strong incentives to set up long-term commitment devices. Such devices could take several forms, but as Aldrich (1995, 186) writes, "there are more or less continual incentives for ambitious politicians to consider party organizations as means to achieve their goals. In the most general terms, these incentives flow from the very nature of liberal democracies in an extended republic, and in an immediate sense that means the ability to fashion and hold majorities."
Demonstrating that parties will be stable over time is a theoretical challenge for these scholars. Within the framework adopted by these authors it is possible to conclude that parties will eventually unravel. Individual, self-interested legislators may want to defect from any short-term agreement made to support the goals of their party. What binds politicians over the long term to parties? To explain the persistence of political parties in this approach, parties must be presented as "equilibrium solutions"--that is, organizational solutions from which legislators would rather not defect continuously.9
We discuss some of these theories in more detail in chapter 3. For now we note that theories stressing the role of elites, whether they focus on the politicization of social cleavages or party formation, leave unanswered a series of questions about the nature of the party system. How many parties will be competitive? Will these parties be national or regional? The next approach offers some answers to these questions.
Party Systems as Reflections of Institutional Rules
The third approach, of which Duverger ( 1963) and Cox (1997) are two major bookends, focuses attention on the influence of electoral laws on party systems (Rae 1971; Riker 1982; Lijphart 1994; Taagepera and Shugart 1989). This literature is vast, and we address it in greater detail in later chapters. The approach tends to be prominent in cross-national comparative studies of party systems. (The more sociological approach that we summarized first tends to be used in single-country studies, although seminal works such as by Lipset and Rokkan and by Inglehart are cross-national.) For this third, more institutional approach, the main aspect of party systems to be explained is number of political parties that contest either for seats in the national parliament or for executive power in presidential systems, or for both. The many ways in which votes are counted and seats are allocated affect the number of parties. Likewise, different methods for choosing presidents affect the number of serious presidential contestants (Shugart and Carey 1992; Lijphart 1994).
Recently, this approach has become enriched by recognition of the importance of other factors, such as the nature of social cleavages (Amorim and Cox 1997; Ordeshook and Shvetsova 1994), the role of presidential elections (Cox 1997), the relative timing of presidential and parliamentary elections (Mainwaring and Shugart 1997), and the degree of fiscal centralization (Chhibber and Kollman 1998). Nevertheless, researchers typically begin from the premise that electoral rules go a long way in explaining the number of parties and the nature of party competition, and that other factors, especially those highlighted in the sociological tradition, can help to explain exceptions, or interact with electoral rules to play a causal role.
This approach complements the other two approaches, although it is more methodologically compatible with the second, microeconomic perspective on party formation. The second and third approaches are strongly related in that they both emphasize the importance of formal institutions that constrain the self-interested behavior of politicians, voters, and legislators. Both see parties as solving collective dilemmas; however, the second approach emphasizes how parties solve n-person prisoner's dilemma problems or voting cycle problems within legislatures, whereas the third stresses how parties solve coordination problems among politicians and voters trying to decide which candidates or parties to support. Cox's Making Votes Count (1997) is an attempt to combine these two approaches, although more attention is paid to the latter than the former.
The first and third approaches are also related, but more in terms of their conclusions than in their underlying premises. A major theme of these two approaches is the durability of party systems. Following Lipset and Rokkan (1967), many researchers have studied the variability in party strength over time in democratic systems all over the world. The relative lack of significant fluctuations in party strength has been a consistent finding (e.g., Bartolini and Mair 1990; Caramani 2004). In the political-institutional tradition, fixed formal institutions lead to stable party systems because the particular electoral system, first-past-the-post or various kinds of proportional representation or multimember districts, leads to an equilibrium number of parties (Lijphart 1994; Sartori 1986).
Some Omissions in the Literature
All three approaches highlight certain features of social and political life and leave out other features. Significantly, none of these approaches seeks to explain how party systems change over time within a country that has not changed its electoral system or has not undergone significant social changes commensurate with changes in the party system.
Several other minor shortcomings of the existing literature deserve some mention. Although the social cleavage approach to the study of party systems has a long and distinguished lineage and has proved valuable in analyzing some aspects of party systems, noticeably absent is a theory of social action underlying much of the research. Which cleavages will become salient enough for political leaders to exploit? How do voters and candidates decide among alternative strategies for winning representation and political power? How many parties will form in response to social cleavages? The conceptual tools used in this literature offer little leverage for answering these questions.
More specifically, in Britain and Canada, although race and the national origin of naturalized immigrants conceivably could lead to voting blocs that split along ethnic lines--for example, specific Asian immigrant communities developing new parties in western Canada or in London, England--these cleavages actually have small, direct influences on the party systems. In India, with literally thousands of castes, only a few large agglomerations have any bearing on the party system (more on this in chapter 2). Similarly, in the United States there are many more social cleavages than there are political parties. In none of our four countries can the number of social cleavages explain the number of parties or predict when a new party based on a social division will come into existence.
Almost all explanations of the rise of parties based on social cleavages are "after the fact." No theory emphasizing the primacy of social cleavages can link all potential political cleavages based on social cleavages to existing political parties. In fact, research on the origin of parties based on new social divisions by Kitschelt (1989), Kalyvas (1996), Rudig (1990), and Rosenstone (1983) shows that one needs far more than a social cleavage to form a party, much less a party system. As Bartolini (2000) argues, the influence of class cleavage--the cleavage that structured European party politics for almost a century--on the party system was a result of the confluence of a multitude of factors, and the cleavage itself did not automatically form the basis of the party system. Similarly, Hug (2001) demonstrates the interactive effects of new social cleavages and institutional factors, including ballot access laws and electoral rules, in explaining the rise of new parties.
The research linking electoral systems to party systems does, in contrast, have an underlying theory of individual choice, although researchers have had a notoriously difficult time matching the theories to real data. Witness all the amendments to Duverger's Law to improve the track record of the argument (Riker 1982; Sartori 1986; Cox 1994). The correlations between district size and the number of parties across long-standing democracies are reasonably tight, but they improve considerably when scholars include social cleavages in the analyses (Amorim and Cox 1997; Ordeshook and Shvetsova 1994; Golder 2003). Ironically, when incorporated into cross-national studies that control for electoral system factors, social cleavages end up having important effects and enhance the correlations between electoral system factors and the number of parties. This begs the question of whether these researchers in the electoral institutions tradition have assessed the causal relationship accurately. An alternative hypothesis--that electoral systems are chosen by political leaders to reflect or even deflect social cleavages, and that the formal rules are determined by, rather than determine, political or partisan alliances--is always present (Stokes 1963; Boix 1998; Jones-Luong 2002; Robinson 2002). In Canada, Great Britain, India, and the United States electoral rules have been more or less the same in the period under consideration, while the party systems have changed (Figures 1.1-1.4). Therefore, to explain the dynamics of party systems, one must focus some attention on factors that change over time as well.
Our concerns are similar for the second approach on parties as solutions to collective action problems. The approach does not do well (and, in fairness, it does not seek to) in explaining variations in party systems. The argument can generally explain the existence of one governing party or perhaps the existence of two parties in a legislature, a majority and an opposition, but it has so far provided little leverage in explaining the existence of more than two parties. For this reason it has focused almost exclusively on American-style political systems (Aldrich 1995). When it has been applied, it has been used to explain the existence of parties in the United States.
These are sins of omission by the literature on parties, not deep flaws that require us to start anew. The literature on parties and party systems is vast, diverse, and valuable. Admittedly, our research here does not escape all of these criticisms.
To explain variation in party systems, we draw on concepts from all three approaches. Party systems, we suggest, are shaped by social cleavages, electoral rules, political entrepreneurs, and a fourth element that interacts with all three of these others and creates incentives for candidates and elected officials to link voters in disparate geographic locations under common party labels. That element is the distribution of authority across different levels of government.10
FEDERALISM, CENTRALIZATION, AND PARTY SYSTEMS
Developing the logic of our argument and providing empirical support for it constitute much of the content of this book. For now, we offer a brief overview. The argument begins with a focus on electoral competition within a single electoral district. As mentioned, in all four of our countries members of the lower house of the national parliament are elected in single-member, simple-plurality districts.11 Only the candidate with the most votes from a districtwide election attains a seat in the legislature, and there are no runoff elections.
In predicting the number of candidates or parties at the district level under single-member, simple-plurality systems, Duverger's Law has considerable bite (Cox 1994; 1997; Feddersen 1992; Palfrey 1989). Duverger's Law holds that this type of electoral rule leads to two-partism. Theoretically, the most compelling arguments in favor of the law rely on the premise that voters in a single election tend to vote for the candidates who have a chance of winning. This not only reduces votes for candidates expected to finish third or worse but also diminishes the incentives for candidates to join the contest for election if they do not think they can finish in the two top positions. Likewise, it reduces incentives for funders to provide money or other resources to candidates who are unlikely to finish near the top of the heap. Empirically, the evidence for the existence of two parties (or two candidates winning nearly all the vote) in district elections under this electoral system is quite strong (Chhibber and Kollman 1998; and see chapter 2), although there are noticeable exceptions, as shown in the next chapter. Even when districts are remarkably heterogeneous socially, as in India, in a vast majority of district elections to the national parliament two candidates receive nearly all the votes.
Whether national party systems will resemble the predictions in Duverger's Law is entirely another matter. It depends on the degree to which candidates and parties make linkages across districts to establish larger political groupings or organizations based on common party labels. While there may be two parties or two candidates receiving most of the votes in district elections, if there are D districts in the country, there could be as many as 2D parties at the national level if each candidate uses a unique party label on the ballot. Candidate Francine Jones could herself become the sole representative of the Francine Jones Party, and Gerry Smith could be the sole representative of the Gerry Smith Party, and so on.
Of course, since the beginning of modern representative democracy, there have been local, regional, and national parties. Regardless of electoral rules, politicians have always seen it in their collective and individual interests to establish linkages across district lines, to aggregate their votes across districts to create regional or national parties that can influence policy or run the government. We call this process party aggregation. Cox (1997) has called it linkage. Even in those countries with an unusually high number of parties, such as the Netherlands or Israel (with the purest forms of national proportional representation and low thresholds), there are vastly more candidates winning seats to the parliament than national parties, indicating that most candidates prefer to aggregate their votes into party totals. However, the number of national parties is a function, to a large degree, of the incentives for candidates and politicians to coalesce around the same labels as politicians from different regions who have different ideologies and have different loyalties to previous and current government policies and leaders.
Consider two situations. It could be the case that politicians feel it necessary to join parties that represent only their particular region or particular narrow slice of the ideological spectrum. Or it could be the case the politicians feel it necessary to join parties that link districts across the entire country or across a very wide range of the ideological spectrum. Which kind of situation prevails--whether minor parties can survive because politicians are comfortable in alliances with less-than-national groupings, or whether minor parties cannot survive because politicians are only comfortable in national groupings--in turn depends on which level of government controls resources that voters care about.
In the first situation, if local governments make most decisions that affect voters, then it may be relatively unimportant to politicians that party labels communicate to voters the national party group that the politician will work with once in office. In fact, it may be crucial that the party label communicate the local or regional group that the candidate will work with once in office. The national label and the local or regional label may coincide, but they don't have to. Under decentralized political or economic systems, candidates will have fewer pressures to join broad, national parties because voters will know that local or regional governments make the important decisions anyway.
In the second situation, if national governments make most decisions that affect voters, then it becomes important for candidates to communicate to voters the policy position of the candidate relative to national government policies, and also the possibility that the candidate, once elected, could become part of the government. For both functions, national-party labels, especially labels of parties that may be expected to become part of the government, will be valuable.
This discussion so far has presented the possibilities starkly, as though only two scenarios are possible. In one, politicians are comfortable with minor-party labels; in the other, politicians want to have the label of a major, national party in nearly every case. These are extremes, of course, and the degree of political and economic centralization and, by extension, the incentives of candidates to adopt only major-party labels can vary by matters of degree. For example, as we argue in later chapters, the United States in the current era represents the one extreme, where serious candidates for the House of Representatives feel compelled to take either the Republican or Democratic label almost without exception. In the nineteenth and early twentieth centuries, however, candidates often felt comfortable adopting minor-party labels. This comfort level, judging by the proportion of competitive candidates who ran as neither Democrats, Republicans (or Whigs in the earlier era), fluctuated in tune with the degree of political and economic centralization. Over time, the number of parties competing for House seats changed, sometimes gradually, sometimes suddenly, but always in keeping with the notion that the incentives of politicians respond to which level of government was responsible for the policies that voters cared about.
To summarize our argument in brief, electoral system effects are most prominent in district elections, but party aggregation depends on the policies and role of the national government in relation to subnational governments. Federal policies of the national governments hinder or help minor, regional-based parties to survive on the national scene and, therefore, affect the nature of party coalitions and the party systems. Party systems become more national as governments centralize authority; in contrast, there are more opportunities for regional, state, or provincial parties to thrive as provincial or state governments gain more authority relative to the national government.
It could be argued that political parties, which control governments, are actually the instruments of centralization and decentralization, and that the party system has a bearing on which level of government has more influence rather than the other way around. We address this issue of reciprocal causation explicitly in the final chapter. For now, we merely note that the evident trends toward centralization and decentralization (or the level of government at which most decisions are made, central versus provincial or state) are actually the consequences of larger forces that work mostly independently of the party system.
We shall leave our theoretical arguments there for now and develop them in more detail in the third chapter. Let us now turn to a description of our data and summarize the chapters that follow.
We have collected data on electoral returns for lower-house contests going back to the beginning of democratic elections in Canada, India, and the United States, and going back to 1885 in Great Britain. For Canada the data begin in 1867, for India in 1957, and for the United States in 1789. (Details on data collection and organization are in the appendix.) The data report district-level election results, with the generic term "district" referring to ridings in Canada, constituencies in Great Britain and India, and congressional districts in the United States. When we measure national party totals, we aggregate vote totals for parties across districts.
For the most part, the party label information of candidates in our data is quite good. Party labels from early elections in Canada and especially in the United States are spotty, and so our inferences from these data are less confident than from later eras. Party labels are missing for nearly 90 percent of U.S. House candidates in decades before the 1830s, for example. Beginning in the 1840s, however, we have labels for nearly every candidate.
The treatment of independent candidates (i.e., independent of party label), especially in India, deserves some comment. The orientation we adopt is that candidates join a party when they adopt the party label on the ballot. If a candidate chooses to run as an independent, then he or she is running under a party label, but the label is his or hers alone. In other words, an independent candidate is a party, albeit a degenerate one. It only has one member. To see where this becomes relevant, consider the process of counting the number of parties in a national party system. Let us say that across all district elections in a given year, party A receives 45 percent of the national vote, party B 20 percent, and party C 15 percent. The remaining 20 percent of the national vote is split among dozens of independent candidates. This is similar (in spirit) to what occurred in India in the early elections following the advent of democratic elections after independence. In the elections for the Lok Sabha in 1957 and 1962 there were hundreds of independent candidates who did well in district elections. When we count the number of parties, using the measure N, for example, we consider each independent candidate to be a party. Thus, if in our hypothetical example the 20 percent of the remaining votes were split among nine candidates, then the formula for the number of parties will have vote counts of the three top parties, plus the nine other parties. In other words, the vote proportions of nine parties will be entered into the formula. (We shall just make the obvious point for clarity: the formula does not imply that there are twelve parties in the party system; the formula weights the parties by their vote share.)
Besides election results (measuring our main dependent variables), to measure our major explanatory factor--the relative authority of central governments and subnational governments--we collected and analyzed original data and analyzed voluminous secondary literature on a variety of government activities related to federal relations in our four countries. A summary of our collection and coding criteria is contained in chapter 5, and we discuss many of our sources in the appendix. Briefly, our most complete information includes data on changes in constitutional or legal authority between levels of government, threats to the integrity of the nation-state from secession or other nations, alterations in the relative economic role of the national and/or provincial governments, and fluctuations in the size of national and provincial governments.
We have selected Canada, Great Britian, India, and the United States for this study mostly because of their similar electoral systems. Each country has a single-member, simple-plurality electoral system in the period under study, and so for cross-national comparisons we are able to control for the effects of formal, electoral rules, especially the one variable that has loomed large in studies of party systems, the district magnitude or size. The size of the district in this literature refers not to population size but to the number of legislative seats awarded from the district. These four countries have district magnitudes (or sizes) of one. Three of the countries--Canada, India, and the United States--are federal, meaning that, by constitutional mandate, they grant significant powers to states or provinces. Britain, in contrast, is not federal under this definition, at least not for the period under study.
Since our concern is how voters and candidates link across geographic regions, single-member district systems arguably offer the hardest cases of coordination or vote aggregation across districts. In our four countries of interest, no national districts or party lists automatically pool party votes to award extra seats to parties or majorities. The closest institutional device that accomplishes this kind of pooling is the U.S. presidential election. But for lower-house elections in all four countries, voters, candidates, or regional parties have to combine their votes themselves by using party labels, without any built-in features of the electoral rules. This is in contrast to systems in Germany, Russia, or South Korea, which have special portions of their legislatures reserved for pooled party votes. Coordination mechanisms in those countries are written into law, whereas in our four countries aggregation has to occur without the aid of national districts or bonus seats.
Our set of cases includes the two largest democracies in the world (India and the United States), the clearest example of a long-standing democracy facing constant threats of secession (Canada), and probably the countries with the most developed and richest party systems at the state or province levels (Canada, India, and the United States; the only country that comes close to these three from among long-standing democracies is Germany). The countries span a range of party systems found among long-standing democracies, from nearly the largest number of parties represented in a national legislature anywhere (India) to the smallest number among democracies (the United States), to somewhere in between (Britain and Canada).
There are other single-member, simple-plurality countries, including the Bahamas, Barbados, Belize, Botswana, Dominica, Ghana, Grenada, Jamaica, Kenya, New Zealand, St. Kitts-Nevis, St. Lucia, St. Vincent, Trinidad, and Zimbabwe. None of these has had a lengthy history of democratic elections under single-member, simple-plurality rules (New Zealand is an exception) and, as far as we know, offers available data on electoral returns at the lowest level spanning more than a few years. It is fortuitous that our four countries keep excellent election statistics.
Although the electoral systems for national parliamentary elections are virtually identical for our four countries, some caveats on case selection are in order. Differences across these countries may make it easy to challenge our claims that cross-national variation is caused by different degrees of centralization. Most glaring, perhaps, is the fact that the United States has a presidential system with separation of powers, whereas Canada, Great Britain, and India have parliamentary government. This is no minor difference according to research in comparative politics. Various scholars, including Cox (1997), Samuels (2002), Shugart and Carey (1992) , and Jones (1994), have pointed to the existence of a presidential system as having a large effect on a party system, and most often the claims are that presidentialism reduces the number of parties. So, for example, one explanation for the persistence of a highly nationalized two-party system in the United States, when other single-member, simple-plurality countries have had more fractionalized party systems on a regular basis, is that presidential elections encourage politicians at other levels of government or in other branches of government to link themselves to one of the candidates who has a chance to win the grand prize. Because the presidential election system is, in effect, a contest within a single, national district, the number of serious contestants--following the logic of Duverger's Law--will generally be two. Therefore, if politicians are going to link themselves to one of the serious presidential candidates, they will typically have two options (e.g., Republican or Democrat). Without this linkage, the argument goes, it leaves more room for other parties. It is worth noting, however, that while the number of parties competing for seats to the U.S. House of Representatives has fluctuated over time--in the first half of the nineteenth century, for example, many parties received votes in elections to the House of Representatives--there have always been presidential elections in the United States. So, although the presidential system may explain some cross-national differences, it can hardly explain the over-time variance in the number of parties getting votes in elections to the House of Representatives.
Changes in the significance of the presidential office may explain changes in the party system, and this corresponds to the main argument in this book. As the presidential office becomes more significant and powerful, something that leads to increased centralization of authority in the country, it, in turn, may lead to greater incentives for politicians across the country to link themselves to presidents or presidential candidates. McCormick (1975) makes the argument that this is what occurred in the 1830s in the United States. As the presidency became more prominent under Andrew Jackson, congressional candidates increasingly adopted national-party labels. Our data bear this out. While more congressional candidates linked themselves to the party labels of presidential candidates after 1832 than they did in the previous two decades, they did not confine themselves to two party labels right away. Early in the decade, they linked to various party labels, making the all-important distinctions on party ballots between Democrats, Jackson Democrats, and Democrat-Republicans. Yet over the course of the 1830s, as shown in Figure 1.4, the Democrats aggregated into a truly national party.
Presidentialism in the United States is not the only noticeable difference among our four political systems. One could focus on a number of other differences. For example, the United States has a stronger, more independent judiciary and a stronger upper house of parliament than the other three, and India is larger (in population) and has more diverse electoral districts than the other three countries. Canada and the United States grew gradually, with westward expansion appending new local party systems onto the national systems piecemeal. As many of these states and provinces applied to become part of the nation, perhaps this gave incentives for local politicians to adopt party labels that mimic national-party labels, in order to win over national parliamentarians and encourage them to vote in favor of statehood or admission to the Canadian nation.
As noted, by our definition, Britain was not federal prior to 1998. We include Britain in the study because it provides some small leverage to examine the effects of federalism among countries with similar electoral systems. The British cases allow us to explore whether patterns discovered in the electoral data occur in spite of federalism, because of federalism, or because of some interaction between federalism and other factors. For example, fluctuations over time in the strength of regional, ethnic-based (or nationalist) parties are less abrupt in Britain in comparison with those in Canada or India. We suggest later that these differences can be attributed to the fact that, for example, in Canada, provincial assembly elections provide a platform for nationalist parties to operate prior to competing for national seats, whereas such opportunities for lower-level parliamentary election victories did not exist in Britain when it did not have regional assemblies prior to 1998. The city council elections that did exist in Britain gave small boosts to minor parties in some instances. As Margaret Thatcher's actions in the 1980s proved, however, when she abolished several of the councils, those councils were not sovereign governments protected by constitutional laws in the same way that American and Indian state parliaments and Canadian provincial parliaments are protected.
Despite these caveats, we believe that our four countries effectively illustrate our arguments. Moreover, we actually have more than four cases. The "cases" in our data are not actually countries but moments of time in the party systems within countries; to the extent that we can compare party systems over time within the same country, we have in effect many cases.
SUMMARY OF REMAINING CHAPTERS
In the next chapter we focus on district-level elections. We summarize the research around Duverger's Law and evaluate the formal theoretical models in this tradition. We then present electoral data from our four countries that are broadly consistent with the law, at least as it applies to district elections. There are enough exceptions within districts, however, to warrant a closer analysis, and so we examine these exceptions with an eye toward confirming or challenging the fundamental premises behind the law. The results are consistent with the notion that voters most of the time adopt some weak version of strategic voting at the district level.
Chapter 3 presents the main theoretical arguments of the book. We discuss the incentives for voters and candidates to link across districts and to aggregate their votes into regional or national party vote totals. Federalism and the degree of fiscal and political centralization play a large role in party aggregation. The fourth chapter focuses on the establishment of party formation and partisan voting in district elections in all four countries and discusses the gradual disappearance of independent candidates. Parties in all four countries were formed by politicians who sought to influence public policy. As control of policy by a group of legislators grew in importance, independent candidates lost their raison d'etre for the voters. Several trends are identified and discussed, including the changes in the number of independent candidates running for office and winning substantial numbers of votes.
In the fifth, sixth, and seventh chapters we examine economic data and electoral data to link centralization and party system change. We categorize periods of centralization and provincialization in our four countries in chapter 5, and then in chapters 6 and 7 we show that periods of change in party systems correspond as we predict for these periods.12 Evidence strongly supports the idea that centralization leads to national party systems, an effect that is especially pronounced in federal systems. We also devote considerable attention to the degree of regionalization of the party system and of parties themselves. How well regional parties survive depends on the nature and the degree of the centralization.
In chapter 8 we discuss alternative explanations for the patterns in our data, addressing whether party systems can account for patterns of centralization and provincialization. We conclude with an analysis of cross-national data from more than forty countries, discussing the implications of these results for extending this research program to countries with a variety of electoral systems.
Return to Book Description
File created: 8/7/2007
Questions and comments to: email@example.com
Princeton University Press