Steven B Wolinetz. Handbook of Party Politics. Editor: Richard S Katz & William Crotty. 2006. Sage Publishing.
Political parties competing with each other for elective office and control of government form a party system. Party systems have been a key factor in the study of political parties and more broadly in comparative analysis. Reasons for this are not difficult to fathom: the number of parties contesting elections shapes the menu of choices which voters face when they cast ballots. The number of parties winning seats in legislative elections affects the ease with which governments can be formed in parliamentary systems and the ease with which political executives can find support in presidential systems. Because party systems are so closely linked to democratic control and government formation, political scientists have sought not only to characterize them, but also to understand their causes and consequences, particularly their sources in electoral laws and cleavage structures and their effects on cabinet and system stability, and more broadly, the quality of democracy.
Research on party systems falls into different streams or literatures. We can distinguish an American and a comparative literature. The former is concerned primarily with the American two-party system and the ways in which it has changed over time, as well as ways in which state party systems have differed from each other and from the larger national party system. A portion of the American literature focuses on partisan realignment and the extent to which the two national parties reflect or blur different lines of cleavage. Changes in cleavage structures and partisan balance over time have been central concerns, and the term ‘party system’ is used to denote periods of time, often a generation or more in length, exhibiting different cleavage structures and patterns of party strength (Key 1964; Burnham, 1970; Sundquist, 1983). A separate literature considers the extent to which state party systems reflect national patterns. Particularly in the long period in which the South was solidly Democratic, students of politics such as Key (1949) documented variation in state party systems, particularly patterns of factional competition in the dominant party. A third stream has focused on the quality of democracy within the American party system, particularly the perceived need for ‘a more responsible two-party system’ (American Political Science Association, 1950; Schattschneider, 1960).
The comparative literature has moved in different directions. Here the primary concern has been variation among national party systems, particularly differences in numbers of political parties, patterns of competition, and what difference they make. Initially, the central distinctions were either between two-party and multiparty systems, or among one-party two-party, and multiparty systems. However, scholars such as Holcombe (1933) and Almond (1956) argued that one-party systems were qualitatively different, while Neumann (1956) argued that one-party systems were a ‘a contradiction in terms.’ Insisting that the term ‘party’ implied parts of a larger whole and that systems had to be made up of regularly interacting parts, Sartori (1976) made the point even more strongly: although he had no difficulty conceiving of a party-state system in which one party monopolized political life, parties were, by definition, parts of a larger whole. A party system (see below) had to be made up of more than one party.
Thinking about party systems has paralleled the development of comparative analysis. Initially, political scientists focused on a limited range of countries: the United States and Britain with two-party systems, and countries such as France or Germany with multiparty competition. Distinctions between two-party and multiparty systems were attributed to electoral systems, and multipartyism was associated with—and in the view of authors such as Hermens (1941) and Duverger (1954) caused by—proportional representation. Two-party systems were typically associated with strong, effective, and decisive government, multiparty competition with cabinet and system instability (cf. Hermens, 1941). Almond (1956) found it necessary to distinguish between continental multiparty systems, more typically clogged and unstable, and ‘working multiparty systems’ (see also Almond and Coleman, 1960; Almond and Powell, 1978). Earlier assumptions were revised in the 1960s and 1970s. Political scientists began to take account of a broader range of liberal democracies. When the scope of comparative politics broadened in the 1960s, political scientists developed more complex typologies distinguishing party systems according to patterns of opposition (Dahl, 1966), the relative size and strength of parties (Blondel, 1968; Rokkan, 1970), or, in the case of Sartori (1966, 1976), the number of parties and the degree of ideological polarization among them (see Mair, 1996, 2002; and Ware, 1996). Sartori’s work provided a way to separate cases of polarized pluralism, wracked by centrifugal tendencies and cabinet instability, from moderate pluralism, in which the direction of competition was centripetal and stable multiparty competition was the norm.
The focus of the literature has changed over time. The initial preoccupation with cabinet and system stability reflected the tumult of the interwar experience, and the fact that countries with extreme multiparty systems, such as Weimar Germany (1919-33) or Second Republic Spain (1931-36) had seen the collapse of liberal democracy. As the interwar period faded, emphases shifted. Taking as gospel Lipset and Rokkan’s (1967) observation that the party systems of the 1960s reflected those of the 1920s, students of party systems focused on continuity and change. Initially, the emphasis was on continuity; more recently it has been on change. In addition, transitions to democracy have sparked interest in how party systems become entrenched or institutionalized.
This chapter explores thinking about party systems and the ways in which they have developed over time. We begin by examining the definition of a party system, then consider efforts to order complexity and discover patterns of interaction, as well as their causes and consequences.
The Systemic Dimension: Party System Properties
A party system consists of regular and recurring interactions among its component parties. Although the term ‘party system’ came into use well before he wrote (see, for example, Holcombe, 1933), one of the first ‘systemic’ uses of the term can be found in Duverger’s Political Parties. Duverger (1954: 203) argues that:
With the exception of the single-party states, several parties co-exist in each country: the forms and modes of their coexistence define the ‘party system’ of the particular country being considered.
In addition to characteristics of the parties, these include
new elements that do not exist for each party community considered in isolation: numbers, respective sizes, alliances, geographical localization, political distribution and so on. A party system is defined by a particular relationship amongst all these characteristics. (Duverger, 1954: 203)
Although the definition of party system is not separate from the characteristics of the parties themselves, Duverger’s reference to ‘forms and modes of their coexistence’ and ‘characteristics that do not exist for each party community considered in isolation’ indicates the importance of interaction. Sartori (1976: 44) argues:
Parties make for a ‘system’ only when they are parts (in the plural); and a party system is precisely the system of interactions resulting from inter-party competition. That is, the system in question bears on the relatedness of parties to each other, on how each party is a function (in a mathematical sense) of the other parties and reacts, competitively or otherwise, to the other parties.
As such, a party system is distinct and different from the parties forming the system.
Sartori’s insistence on the systemic properties not only enables him to separate party-state systems monopolized by a single party from party systems in which there is competition for government, but also provides a basis for examining their most important features. These are relational and arise both from their competition for elective office and interaction in between elections in both the formation and support of governments and the legislative process. Parties compete for a share of the vote and, in doing so, try both to shore up their own support and pry votes from their competitors. The strategy and tactics which they employ are influenced by what other parties have done in the past and expectations about what they will do in the future. The ability of parties to cooperate with each other after elections will depend not only on their size and relative strength, but also on their distance from each other on key issues and the ways in which they present themselves during elections. Equally parties may discover that choices made in political office—e.g. decisions to participate in or remain aloof from coalitions, as well as policies pursued in government or opposition -can affect their ability to win electoral support.
Party systems have a number of distinct features which arise from electoral competition and parties’ relation to each other. These include the number of parties contesting elections and winning legislative seats, their relative size and strength, the number of dimensions on which they compete, the distance which separates them on key issues, and their willingness to work with each other in government formation and the process of governing. Party systems can vary on any or all of these. Voters, politicians, and political analysts often think of parties divided along a left-right spectrum, but it is not unusual for party systems, at least in their origins, to reflect multiple dimensions of conflict. European party systems, for example, often reflect not only economic or distributional issues, but also religion and religiosity and, in certain instances, urban-rural cleavages (Lijphart, 1982). Party systems can be more or less polarized on any or all of these dimensions. Other features on which party systems may differ include the degree to which their competition for government is open to all parties or closed—restricted only to certain parties or combinations of parties (Mair, 1996, 2002)—and the degree to which the party system itself is institutionalized or entrenched (Mainwaring and Scully, 1995). However, this latter facet reflects not so the much the ways in which parties relate to each other, as the degree to which parties, taken together, are able to enlist durable support and structure the electorate.
Because party systems can vary on any or all of these features, students of political parties often try to simplify the world around them by grouping them into distinct types. The most common classifications usually differentiate party systems according to the number of parties winning seats and one or more relational features, such as size and relative strength or the ability of parties to work with each other. The most obvious distinctions are between two-party systems and multiparty systems, but two-party systems may be more or less polarized, and not all multiparty systems are necessarily the same: there is considerable difference between a party system with three or four parties and one with six or seven or eight. Even so, this depends on how parties are counted and what weights are assigned to different sizes of parties. Typically, classifications count major parties, but, as Sartori (1976) has pointed out, clear rules are needed to determine which parties should be counted and which should be excluded. Once this is done, other questions remain: whether the number of parties is a sufficient criterion, or whether relative sizes and strengths of parties and mechanics (or direction of competition) should be taken into account as well. Efforts to do so have given rise to distinct typologies, as well as continuous measures, such as Laakso and Taagepera’s (1979) effective number of political parties, which weights parties according to their size.
Counting Parties
The oldest distinctions are among one-party two-party and multiparty systems; almost all classifications of party systems make distinctions on the basis of number. However, decisions have to be made about whether to consider all parties contesting elections, only those winning seats in the legislature, or only those involved in government formation. Although continuous measures such as Rae’s fractionalization index (Rae, 1967; Rae and Taylor, 1970) or Laakso and Taagepera’s (1979) effective number of political parties can be used to measure the number of parties contesting elections, counts of political parties are usually based on the number of parties winning seats in parliament. In the 2005 British general election, a total of 14 parties, one local list and one non-partisan group ran candidates for parliament. Of these, 12 won seats in parliament and the overwhelming share of the vote was won by three national parties. Except for the anti-war coalition, Respect, which won one seat, all other parties winning seats were regionally based: the Scottish National Party, Plaid Cymru (Welsh Nationalists), and three Northern Irish parties. The Social Democratic and Labour Party and one local list also won seats. However, no one would term the British party system a 12-party system, and in view of the regional concentration of the vote for smaller parties, few would characterize it as a ten-party system. More problematic is whether the Liberal Democrats’ 22% of the vote makes Britain a three-party system rather than a two-party system.
Once a decision has been made to focus on parties winning seats in the national parliament, further decisions must be made about which parties to count. This can be done in several ways: All parties can be counted (although in the British case this would lead to results which are counter-intuitive) or some can be excluded on the basis of either size or standards of relevance. Many characterizations of the number of political parties focus implicitly only on major political parties. However, this presumes some kind of criterion. Alan Ware (1996) excludes all parties with less than 3% of the vote. In contrast, Sartori (1976), argues that relevance should be assessed according to coalition potential and blackmail potential. Smaller parties are counted only if their seats in parliament are needed to form coalitions, or alternatively if they have sufficient seats to block the formation of coalitions. If we follow Ware, we would call Britain a three party system. If we follow Sartori, Britain remains a two party system because, despite winning almost 20% in most elections since 1974, the Liberal Democrats have rarely been able to affect government formation. Only in the late 1970s were their seats in parliament needed to keep a Labour government in office. Nevertheless, we could argue that three way competition in individual districts makes them electorally relevant; both Labour and Conservatives need to worry about third party candidates depriving them of seats they might otherwise win.
In other political systems, parties with considerably less than the British Liberals’ 22% are counted, typically because proportional representation gives them a similar percentage of seats in parliament. In Germany, Free Democratic Party (FDP) support has ranged from a high of 12-13% of the vote and seats in parliament to a low of 6-7%. In contrast to the British Liberals, the FDP has been particularly relevant: through 1998, it could often determine whether the Federal Republic of Germany would have a center-left or center-right government. The presence of the FDP ensured that the Federal Republic after 1957 would be considered a three-party or in some instances a two-and-a-half-party system (see below). After 1983, a fourth party, the Greens, leapt the 5% threshold, winning 5.6-8.6% of the vote and seats in the Bundestag. If we were to follow Ware’s criteria, we would consider Germany a four-party system after 1983, and with the entry of the Party of Democratic Socialism (PDS) in 1990 a five-party system, because the PDS had more than 3% of the vote. In contrast, Sartori would argue that the Greens only became relevant when they began joining provincial and later federal coalitions in the 1990s. The PDS would not be considered relevant because their votes have not been needed to form coalitions, they have not been able to block the formation of coalitions, and their presence has not altered the direction of competition.
Weighted or disaggregated measures provide an alternative to simple counting, with or without explicit cutoffs for smaller or irrelevant parties. Two have been used in the parties literature: Rae’s fractionalization index (for electoral fractionalization or for legislative fractionalization) and Laakso and Taagepera’s effective number of political parties. Fractionalization does not measure the number of political parties directly, but estimates the probability that any two randomly chosen voters or legislators will be of the same party (Rae, 1967). The effective number of parties is measured by dividing 1 by the sum of the squares of proportions of votes (effective number of electoral parties, ENEP) or seats won by each party (effective number of parliamentary parties, ENPP). This results in a number which is typically smaller than the actual number of parties contesting elections or represented in parliament. Squaring the decimal shares of votes or seats won gives additional weight to larger political parties. Smaller parties which would be excluded under Sartori’s decision rules are counted, but they do not count for very much: a party like Plaid Cymru, with 0.6% of the vote only adds to 0.000036 to the denominator of the measure.
Measures like the effective number of political parties finesse the problem of exclusion or inclusion and provide a continuous measure which can be used in correlation and regression. This has been particularly useful in assessing the effects of electoral systems and can be used to examine changes in the number of parties over time. Compressing the actual numbers of political parties, measurement of the effective number of electoral or legislative parties produces a series of decimals ranging from 1.8 or 1.9 for systems with two parties to 5 or more for systems with eight or more parties in parliament. On the other hand, Dunleavy and Boucek (2003) argue not only that Laakso and Taagepera’s index and related measures obscure variations in relative size and strengths of parties, but also that the index behaves quirkily rather than continously for certain values.
Party System Types
Students of party systems have moved beyond number and attempted to construct typologies which capture relationships and interactions. This can be done in different ways: combining numbers of parties with information about their relative size and strength, as Jean Blondel (1968) and Alan Siaroff (2000) have done, or looking at patterns of government formation and party interaction, as Rokkan (1970) and Dahl (1966) and more recently Peter Mair (1996, 2002) do, or, in the case of Sartori (1966, 1976), considering polarization and internal dynamics as well as the number of parties.
Classification on the Basis of Relative Strength and Size of Parties
Jean Blondel (1968) was one of the first to move beyond simple counting and consider the relative size or strengths of political parties. Blondel used the share of the vote won by political parties in elections from 1945 through 1966 to construct a fourfold typology. He distinguishes two-party systems, two-and-a-half-party systems, multiparty systems with a predominant party and multiparty systems without a predominant party. His typology is derived by looking at the average share of the vote won by the largest two parties and then considering the ratio of the first party’s share to the second and third parties. In the five two-party systems (the United States, New Zealand, Australia, the United Kingdom, and Austria), the two-party share was greater than 89% and closely balanced between the two parties. In the next cluster, the two party share ranged from 75% to 80% of the vote cast but there was a wider average difference (10.5%) between the first and second parties. Although these could be considered three-party systems, Blondel categorizes them as two-and-a-half-party systems to take account of the imbalance in parties’ share of the vote. These include Canada, the Federal Republic of Germany, and Ireland. Blondel then distinguishes among party systems with four or more major parties: those with one larger party winning 40% or more of the vote and typically twice as much as the second party in the system are multiparty systems with a predominant party (e.g. Sweden, Norway, Denmark, Italy, and Iceland) or, if this is not the case, multiparty systems without a predominant party (Netherlands, Switzerland, France, Finland).
Blondel’s typology is useful both because it permits us to distinguish among different types of multiparty systems and brings out differences and similarities among pure two-party systems and systems like the Federal Republic of Germany, with two larger parties and a relatively smaller party sometimes able to play a balancing role between them. Although a refinement over simple counting, the scheme is problematic. As Mair (1996, 2002) points out, Blondel’s scheme disaggregates the multiparty category, but his categories bring together party systems whose dynamics are not necessarily the same. Multiparty systems with a dominant party include both Norway and Sweden, with predominant social democratic parties, and the much more polarized pre-1993 Italian party system. Multiparty systems without a dominant party include consociational democracies such as the Netherlands and Switzerland, in which elite cooperation is said to outweigh centrifugal tendencies, and more polarized party systems such as France and Finland.
The designation of two-and-a-half-party systems captures differences between pure two-party systems, on one hand, and moderate multiparty systems, on the other. However, the two-and-a-half-party category brings together party systems in which the role of the smaller party differs considerably. As Siaroff (2003) notes, the role of the ‘half party’ varies from hinge parties, located between two larger parties, such as the German Free Democrats, influential because their votes were needed to make parliamentary majorities, and ‘wing parties’ such as the Canadian New Democratic Party, less influential because their votes are rarely needed either to form coalitions or ensure that legislation is passed. In the first instance, the hinge party determines who governs; in the second, the wing party’s influence is at best confined to agenda setting and proposing policies which may be taken over by larger parties. There is also a question of why smaller ‘half parties’ should be highlighted in what otherwise would be three-party systems but not in multiparty systems with a larger number of parties.
Patterns of Government Formation
Looking at the relative size and strength of parties is only one way to refine classifications based on number. Patterns of government formation and party interaction can also be considered. Examining patterns of opposition in Western democracies, Dahl (1966) uses parties’ behavior in electoral and legislative arenas to develop a fourfold scheme. Patterns of opposition can be strictly competitive (Britain), cooperative and competitive (the USA, France and Italy), coalescent and competitive (Austria and wartime Britain), or strictly coalescent (Colombia). Where necessary, each of these types can be further broken down into two-party and multiparty categories. Dahl’s scheme is not a classification of party systems, per se, but of patterns of opposition. The scheme demonstrates that two-party and multiparty systems need not be as different as either simple counting or standard typologies assume. Both the competitive and coalescent and competitive categories bring together party systems which might otherwise be categorized as two-party and multiparty systems.
Rokkan (1970) uses patterns of government formation to classify the party systems of smaller democracies. Rokkan distinguishes among party systems, such as Austria and Ireland, which display a 1 vs. 1 + 1 format, akin to a British and German pattern, Scandinavian 1 vs. 3-4 pattern (Norway, Sweden, Denmark), and ‘even multiparty systems’ which display a one vs. one vs. two-three (1 vs. 1 vs. 1+2-3) pattern of competition. Like Blondel’s scheme, this is an attempt to disaggregate the multiparty category (Mair, 1996, 2002), but the organizing principle is patterns of government and opposition rather than relative size.
Sartori’s Typology: Moderate versus Polarized Pluralism
Sartori argues that the standard distinction among one-party two-party, and multiparty competition is too crude to explain very real differences among party systems. After separating out party-state systems, he proceeds to establish rules which tell practitioners which parties to count and which to exclude. The next steps are to select cutoff points, establish classes, and take account of special cases like segmented societies. Classes are then distilled into distinct types. Sartori ends up with a typology based on numbers (properly counted), whose principal distinction is not number as such, but rather the degree of polarization and whether party competition, and thus the mechanics of the system, are centripetal or centrifugal.
Sartori begins by establishing explicit counting rules. He argues that the criterion by which parties, large or small, should be counted is their effect on party competition. Smaller parties are relevant when they have either coalition potential or blackmail potential. Coalition potential depends on parties having sufficient seats to make coalitions feasible and is measured by their having participated in or made cabinet coalitions possible; parties whose seats are never needed are deemed irrelevant. The second criterion is their impact on the direction of party competition: parties, large or small, are relevant when their existence alters the direction of party competition leftward or rightward, changing the direction of competition from centripetal to centrifugal.
Sartori’s next step is to establish classes of party systems. He begins by breaking down what he describes as the one party and multiparty ‘lumps.’ The first consists of a mixed bag of one-party and hegemonic party political systems, not properly competitive and predominant party systems in which one party that regularly wins 50% of the seats in parliament predominates over a number of smaller parties; no other party can govern because of the predominant position of the first. Multiparty systems are grouped into two classes: limited pluralism, with three, four, or five relevant parties, and extreme pluralism, with six, seven, or eight. Finally, Sartori adds a residual category, atomized party systems, which are so fragmented that the addition of one more party makes no difference to the pattern of competition. These party systems are insufficiently structured or consolidated to be considered.
Sartori then refines the multiparty categories. Here no party has or is likely to obtain an absolute majority. Sartori argues power structures (relations among the parties) are important and then proceeds to differentiate party systems according to their mechanical predisposition, or, more specifically, relations among the parties. Doing so enables him to establish criteria for moderate and polarized pluralism. The crucial factors are the direction and character of competition: competition under moderate pluralism resembles competition in two-party systems. The system is bipolar and competition is centripetal: parties on either side of the spectrum compete for votes in the center. Polarized pluralism is different. Although the center is occupied, the dynamics of the system are centrifugal rather than centripetal. Anti-system parties at the extremes compete with parties in the center, pulling parties and voters toward them. Because bilateral oppositions located ‘two poles apart’ cannot coalesce, parties in the center govern without the benefit of an alternative government which can replace them. As such, the system is characterized by ideological divisions, centrifugal drives, ‘irresponsible oppositions’ and a politics of ‘outbidding or over-promising.’
The initial criterion for distinguishing between moderate and polarized pluralism is the number of political parties, but the cutoff point, five or more, is in Sartori’s view an artifact. Segmented systems characterized by elite accommodation are cases of moderate pluralism because the mechanics of the system are centripetal rather than centrifugal. The mechanics of competition and particularly the extent of polarization are more important than the number of relevant parties. Sartori ends up with a fourfold typology: predominant party systems, two-party systems, moderate pluralism, and polarized pluralism.
More Recent Schemata
Since the typologies which we have been considering were developed and refined in the 1960s and 1970s, transitions to democracy in different parts of the world have given us a larger range of political systems to take into account, and party spectra in older liberal democracies have become increasingly crowded by the addition of green and new politics parties, and by the entry and growth of new right and neo-populist parties. Nevertheless, few of the latter can be characterized as fundamentally opposed to liberal democracy. Instead, as Mair (1996, 2002) has observed, Sartori’s polarized pluralism has emptied out, while moderate pluralism has become increasingly crowded. Included are not only the German, Austrian and Scandinavian party systems, as well as the Dutch and Belgian, but also the French and the post-1993 Italian party system. Like its predecessor, the post-1993 Italian party system contains a large number of relevant political parties, but it lacks anti-system parties at its extremes.2 There are also fewer two-party systems: following a change in its electoral law in 1994, New Zealand changed from a pure two-party system to a multiparty system. Dominated by a single party from 1979 to 1997, Britain in the Thatcher-Major era should be classified as a predominant party system (Mair, 1996, 2002).
Students of party systems have yet to come to grips with the changed situation, let alone refine moderate pluralism. Building on Dahl and Rokkan, Peter Mair (1996, 2002) has suggested using competition for government as a device for distinguishing among party systems. In party systems in which competition for government is closed, there is either wholesale alternation between parties or groups of parties, governing formulae are familiar rather than novel or innovative, and access to government is typically restricted to only a few parties. In contrast, in systems in which the structure of competition is open, there is partial alternation: some parties rotate in and out of government, while others remain, and as new parties appear, there is frequent recourse to innovate governing formulas. Closed structures of competition were typical of the United Kingdom, Japan, pre-1994 New Zealand, and Ireland over the period 1948-89. Open patterns of competition characterize both the Netherlands and Denmark, as well as newly emerging party systems: in the Netherlands, new parties have been incorporated into governing coalitions; in Denmark, novel coalitions and new forms of minority governments were used to accommodate changes in the number of parties (Mair, 1996, 2002).
Mair argues that focusing on structures of competition not only directs attention to key relationships among political parties, but also allows the party system to function as an independent variable to which parties and voters may respond. He illustrates his point by demonstrating the ways in which changes in Irish coalition patterns—the willingness of Fianna Fáil to enlist coalition partners after refusing to do so since the 1940s—paved the way for shifts in voting alignments and further shifts in coalition patterns (Mair, 1996, 2002). Using open or closed competition for government is novel, but its full potential has not yet been explored.
An alternative approach is to sort moderate pluralism according to the size and relative strength of parties. Alan Siaroff (2000) does this by refining and building on Blondel’s earlier typology. Siaroff uses multiple measures to tap the relative size and strength of political parties winning more than 3% of the seats. He ends up with an eightfold classification, distinguishing: (1) pure two-party systems, with a mean two party share of 95%; (2) moderate multiparty systems with three to five parties above 3% (which he argues are in fact two-and-a-half-party systems); (3) moderate multiparty systems with one dominant party; (4) moderate multiparty systems with two main parties, (5) moderate multiparty systems with a balance among parties; (6) extreme multiparty systems with one dominant party; (7) extreme multiparty systems with two main parties; and (8) extreme multiparty systems with a balance among parties. The resulting scheme categorizes party systems according to the number of parties (two-party systems, moderate multiparty systems with three to five parties, and extreme multiparty systems with six to eight) and the relative balance among parties (one dominant party among others, two main parties, or an even or nearly even balance among them), which can then be related to electoral systems, length of cabinet formation, type of cabinet (e.g., minimum winning or not), as well as duration of governments. As Table 6.2 demonstrates, Siaroff’s categories tap variations in the effective number of parliamentary parties.
If the aim is to disaggregate moderate pluralism, then Siaroff has succeeded. In place of a single overloaded category, we now have a more refined scheme with several categories. The large number of categories also permits Siaroff to analyze changes over time. However, some ‘party systems’ last no longer than a single election period. This is difficult to accept if, following Sartori, we believe that party systems consist of recurring rather than one-off relationships. Siaroff is in fact referring not to party systems but to patterns of party strengths which have resulted from particular election outcomes. This difficulty can be overcome either by changing the terminology, so that we are referring to patterns of party competition, some more permanent than others, rather than party systems, or by averaging results over two or more elections to tap more durable features. More problematic is the complexity of the scheme. With eight categories more or less arrayed on two dimensions, Siaroff’s scheme lacks simplicity or parsimony. Whether it will gain acceptance remains to be seen.
New Directions
One of the more surprising features of this exercise is the absence of new typologies.3 Little has occurred since Sartori (1976). In some respects, this is a testament to his success. More than its predecessors, Sartori’s typology sorted the available cases, and it did so in a meaningful way. Nevertheless, its utility is increasingly problematic. We now have almost no cases of polarized pluralism, save for the now historical instances for which it was developed, and moderate pluralism is increasingly overcrowded (Mair, 1996, 2002). The party systems of most stable liberal democracies fall within its reach. If we believe that there are no significant differences among these party systems, there is no cause for concern. If not, then we need to emulate Sartori and consider how relevant cases can be sorted.
The number of parties does matter. There is considerable difference between countries with two, three, or perhaps four parties, and those with six or eight or more. Voters in the former face simpler choices than voters in the latter. Similarly, politicians—assuming that we are talking about a parliamentary system -find the task of forming governments easier when there are fewer parties. However, this depends not only on the number of parties, but also on the degree of polarization and the extent to which parties cluster together, forming durable coalitions and alliances. How can we distinguish such systems?
One strategy is to use Laakso and Taagepera’s index. Both ENEP, the effective number of legislative parties, and ENPP, the effective number of parliamentary parties, have been used to great advantage in analyses of the effects of different types of electoral laws (Lijphart, 1994; Taagepera and Shugart, 1989; Cox, 1997). However, despite their advantage for correlation and regression, ENEP and ENPP blur distinctions and tell us little about relationships among parties or the dynamics of different types of party systems. One advantage of simple counting, with or without explicit cutoffs for smaller parties, is that it produces outcomes which are readily (if not always correctly) understood.
Space does not permit development of a new typology, but it is possible to suggest features which one should display. Typologies work when they sort the available cases into types which are mutually exclusive and can be understood easily (Lange and Meadwell, 1991). Siaroff’s scheme, a by-product of a larger effort at data collection, falls short because it has too many categories, and does not explain why the relationships which it captures are relevant. The most important features of party systems are those on which Sartori and successive scholars have focused: numbers of parties and relationships among them. The greater the number of parties, the more complex their interrelations are likely to be. Equally important is the degree of polarization: parties in most party systems may no longer be two poles apart, but some party systems are more polarized than others. There is a considerable difference between the more polarized pre-and post-1993 Italian party systems, and the Dutch party system, which continues to have a large number of parties in parliament, but is rarely so polarized that parties are unable to work with each other.
Relative size may also be important (Blondel, 1968; Rokkan, 1970; Siaroff, 2000), but the relative size or strength of parties is a tertiary characteristic less likely, in and of itself, to shape relationships among parties. In addition, in a period of pronounced electoral volatility, in which fewer and fewer parties can count on automatic support from loyal electorates, the size and strength of parties may be too variable to reflect the durable systemic relationships at the core of the study of party systems. More important in systems with six or more parties competing are relationships among parties: for example, do parties compete around the center, the mode of competition at the core of Sartori’s category of moderate pluralism, or is competition more centrifugal, centering around two poles, even if not as thoroughly polarized as the Weimar Republic or First Republic Italy, as Sartori understood it? Equally important, do parties compete as independent entities or cluster into semi-permanent alliances, as parties in Fifth Republic France, Israel, or Italy after 1993 have done? Clustering is important because it mitigates some, but not all, of the effects of multipartyism. In systems like Fifth Republic France or Italy after 1993, parties still face competition on their flanks, and voters are still presented with a wide array of choices. However, when parties cluster into distinct blocs—left and right in France, Olive Tree and the House of Liberty in Italy—voters receive additional information about how parties are likely to behave after elections. The number of alternatives is reduced, simplifying some choices, while making others more complex. Clustering into distinct blocs also structures and sometimes simplifies post-election processes of government formation.
A new typology, refining Sartori’s moderate pluralism, should consider the number of parties, their interrelationships, and the presence or absence of clustering, as well as centripetal versus centrifugal drives. Clear, neutral labels are needed. Like Siaroff (2000), students of party systems frequently distinguish moderate and extreme multipartyism. Moderate multipartyism—typically three to five parties, centripetal drives and competition around the center—is clear enough, but extreme multipartyism is more problematic. We typically mean multiparty competition with more than three to five parties. However, ‘extreme’ conjures up other implications: multiparty competition with extremist or anti-system parties and, of course, polarized pluralism. In an era in which there are fewer and fewer viable or presentable alternatives to liberal democracy, few (if any) of the left libertarian or neo-populist parties which have crowded political spectra since the 1970s are opposed to liberal democracy. These parties, to be sure, oppose some of the policies and practices of older and more established parties, but not the system itself (Abedi, 2002; Zaslove, 2003a, 2003b). Their appearance and relative success have made a difference—new right populist parties have strained the boundaries of political correctness and forced other parties to take up some of their claims—but this has been done working within the boundaries of liberal democracy. If we are going to use labels like extreme multipartyism, then we must neuter the term, stripping it of its earlier connotations. If not, then we should substitute more neutral terms, such as extended rather than extreme multipartyism.
Finally, new typologies should be based on parties and their interrelationships, rather than on properties of the parties themselves. This is difficult because properties of party systems can never be entirely separate from the parties which populate them. Relationships depend on numbers. Examining the ways in which Scott Mainwaring (1999; see also Mainwaring and Scully, 1995) has approached party systems in transitional democracies illustrates the problem. He has developed measures to compare the degree to which Latin American party systems are institutionalized or entrenched in their societies. These include the age of parties—how long individual parties as distinct organizations have been around—as well as aggregate electoral volatility as a measure of the collective ability of parties in a party system to maintain stable bases of support. Using these measures, he is able to show considerable difference between more institutionalized party systems, such as Argentina, Mexico, and Venezuela, and less institutionalized systems such as Peru, Bolivia, Ecuador, or Brazil. Similar comparisons can be made between the increasingly institutionalized systems of newer central European democracies such as Hungary, the Czech Republic, and more recently Poland, and inchoate party systems, like that of Russia or other parts of the former Soviet Union. Measuring the average age of party alternatives says something about the degree to which party systems are entrenched.
Using electoral volatility as an index of party system institutionalization is questionable. Electoral volatility measures the ability of parties to build loyal followings and collectively structure the electorate. These are properties of parties, individually and collectively, rather than aspects of the party system—that is, parties and the ways in which they relate to each other. Rates of electoral volatility have increased in well-established party systems, such as the Netherlands and Austria, in part because older lines of cleavage have weakened and established parties have had less loyal electorates than in the past. This is a new development, which may reflect changes in the media and the ways in which parties approach voters. However, the diminishing ability of parties to hold voters does not necessarily mean that the party system is becoming less entrenched or institutionalized: even if they have lost support for a time, older parties retain resources, which enable them to continue and often recover in subsequent elections. We still have few examples where established parties have disappeared or have in large measure been replaced. Most are found in Italy, where most parties in the pre-1993 party system have been supplanted. However, the Italian case remains an exception rather than the rule.
Thus far, Mainwaring’s set of measures are the only ones brought forth. They are useful in that they link to measures already in use in the parties literature, but problematic because of the presumptions made about the degree to which voters should—or in the future are likely to—have stable party preferences in a world dominated by rapid electronic media. Nevertheless, Mainwaring’s measures provide a starting point from which comparisons can be drawn. Clearly, we need ways to take account of variation in party systems. At issue are not only the number of parties and the ways in which they compete, but also, in a period in which multilevel governance is increasingly prominent, ways of describing and categorizing links among party systems at local, regional, national, and transnational levels of governance. Also important is the impact of institutions—whether the system is presidential, semi-presidential or parliamentary—on parties and party systems. Clearly, new research is needed, if not new categories.