Challenges to established parties: The effects of party system features on the electoral fortunes of anti-political-establishment parties

European Journal of Political Research 41: 551 583, 2002 551 Challenges to established parties: The effects of party system features on the electoral fortunes of anti-political-establishment parties AMIR ABEDI Department of Political Science, University of British Columbia, Canada Abstract. The rise of parties that challenge the political establishment has recently sparked the interest of political scientists. Scholars have identified several factors that lie behind the success of such anti-political-establishment parties. Most empirical studies, however, have concentrated their attention either on the importance of electoral system features or on the effects of socioeconomic conditions. This article focuses instead on the role that party system factors play in the electoral success of these parties. Using three data sets from studies conducted in three different time periods it tests two seemingly contradictory hypotheses. On the one hand, the claim that where the established parties have converged toward centrist positions and thus fail to present voters with an identity that is noticeably different from their established competitors, the electorate will be more susceptible to the markedly different policies put forward by anti-political-establishment parties. On the other hand, there is the argument that these parties profit more from increasing polarization and the subsequent enlargement of the political space than from a convergence toward the median. The results of the analyses show that anti-political-establishment parties generally profit from a close positioning of the establishment parties on the left-right scale. However, there is no consistent support for the notion that party system polarization by itself is associated with an increase in the support for parties that challenge the political establishment. Theoretical background and methodology The literature that deals with parties and party systems from a comparative perspective has grown considerably in recent years. Many studies have included party system features as their independent or dependent variables in analyses of party system change. Examples include Bartolini & Mair (1990), Broughton & Donovan (1999), Crepaz (1990), Daalder & Mair (1983), Hazan (1997), Ieraci (1992), Keman (1994, 1997), Mair (1997), Pennings & Lane (1998), Powell (1982, 1986) and Smith & Mair (1989). The notion that the party system should not only be used as a dependent variable but also as an independent variable has in no small part been a result of Giovanni Sartori s (1976) seminal work on parties and party systems. He Published by Blackwell Publishers, 108 Cowley Road, Oxford, OX4 1JF and 350 Main Street, Malden, MA 02148, USA

552 amir abedi was the first scholar who combined the numerical criterion with an ideological criterion in order to develop his typology of party systems. Using the numerical criterion, Sartori (1976: 273 293) distinguished between two-party systems and multiparty systems, which are either of a limited pluralist or an extreme pluralist nature. He then went on to examine the either small or large ideological distance between the extreme parties of a party system. Combining these two criteria, Sartori arrived at his famous typology which includes three types of party systems: (1) two-party systems that are characterized by two parties and a small ideological gap between the extremes; (2) moderate pluralist party systems that consist of three to five parties and evidence a relatively small ideological distance between the extreme parties; and (3) polarized pluralist party systems characterized by more than five parties and by a large ideological distance between the extremes. Sartori (1976: 316 317) maintains that the numerical criterion is not independent of the ideological criterion that is, the number of parties in a party system affects the degree of polarization. Thus, the more parties a party system contains, the more polarized it will be. This, Sartori (1976: 316 317) argues, ultimately also influences the stability of a political system. While the centripetal competition that characterizes two-party systems and moderate pluralist party systems is conducive to governmental stability, the centrifugal competition that is characteristic of polarized pluralist systems leads to instability. As Peter Mair (1997: 204) has pointed out, one of the main contributions of Sartori s typology is the fact that it underlines the influence exerted by systemic properties, and by the party system, on electoral behaviour and electoral outcomes. Sartori s work has made clear that the party system can serve as an independent variable, limiting and leading electoral preferences. Consequently, scholars began to examine the relationship between, for example, polarization and democratic stability (e.g., Dodd 1976; Powell 1986; Sani & Sartori 1983) or the association between polarization and voter turnout (Crepaz 1990). Other studies have explored the relationship between centre parties and various party system features (e.g., Hazan 1997; Ieraci 1992; Keman 1994, 1997). However, another area of research that has also experienced some growth in recent years namely the literature that focuses on the rise of various parties that challenge the political establishment has so far neglected to take into account the possible impact of party system features on the electoral fortunes of these anti-political-establishment parties. 1 The emergence of various Green parties, as well as the success of several right-wing extremist parties in Western Europe, sparked the interest of scholars in this phenomenon. This led to a number of studies that either examined particular countries in which these parties were successful or looked at certain

challenges to established parties 553 types of parties, such as left-libertarian parties, Greens, populist parties and/or right-wing extremist parties (e.g., Betz 1994; Betz & Immerfall 1998; Bomberg 1998; Dalton & Kuechler 1990; Ignazi 1997; Kitschelt & McGann 1995; Mackie 1995; Müller-Rommel 1989; O Neill 1997; Richardson & Rootes 1995; Taggart 1996). However, empirical studies that have explored possible reasons behind the emergence and electoral success of these parties have neglected to examine the role of party system features. Instead, they have focused their attention on other factors. Numerous studies have, for example, tested the effects of certain electoral system features such as proportionality, district magnitude and effective thresholds on the electoral fortunes of parties that challenge the political establishment (e.g., Harmel & Robertson 1985; Jackman & Volpert 1996; Willey 1998). Socioeconomic characteristics, such as the size of the welfare state, the degree of post-industrialism as well as economic growth and the impact that they might have on the level of electoral success of anti-political-establishment parties, have also been explored in many studies (e.g., Taggart 1996; Kitschelt & McGann 1995; Knigge 1998). However, there is a surprising lack of empirical research when it comes to assessing the role of party system features. While several scholars have discussed the effects that establishment party divergence and party system polarization might have on the electoral fortunes of anti-political-establishment parties, those studies have for the most part been theoretical in outlook. Let us take a closer look at these theoretical arguments. Peter Mair (1995: 48 51) claims that the lack of a recognisable difference between the traditional parties in other words, the fact that many establishment parties increasingly fail to present voters with an identity that is noticeably different from that of their competitors contributes to the growing alienation that some of the electorate feels toward these parties. The establishment parties are seen as components of a basically undifferentiated political class. Voters are consequently more susceptible to the markedly different policies put forward by antipolitical-establishment parties. Kitschelt & McGann (1995: 17, 20 23, 48) also argue that where moderate left and right parties have converged toward centrist positions and may even have cooperated in government coalitions the chances for populist antistatist parties as well as parties of the New Radical Right to be electorally successful rise considerably. Elsewhere, Kitschelt (1988: 215 216, 232; see also, Kitschelt 1989; Kitschelt & Hellemans 1990) contends that left-libertarian parties were the beneficiaries of the socialist and social democratic parties move toward the centre in the 1960s and 1970s. Finally, Paul Hainsworth (1992a: 11) maintains that the decreased ideological distance which results from the convergence of the main parties of the right and left creates a vacuum that is exploited by parties situated on the extremes of the political spectrum.

554 amir abedi This argument is, however, challenged by Piero Ignazi (1992: 19 25, 1996: 557 559) who maintains that new right-wing parties profit more from increasing polarization and the subsequent enlargement of the political space than from a convergence toward the median. He explains the electoral success of extreme right-wing parties in the 1980s by the emergence of left-libertarian parties and the resulting radicalization of establishment parties of the Left on the one hand, and by the move of establishment parties of the Right toward radical neo-conservatism (which he characterizes as a mixture of laissez-faire economic ideas and revitalized traditional values), on the other. Ignazi thereby contends that the latter occurred because neo-conservatism provided not only a challenge to the post-war consensus on the Keynesian political economy of the collectivist age and the costs of the welfare system (Ignazi 1996: 558), but also a non-materialist answer to the post-materialist agenda of the New Left. The ensuing polarization of the political system not only legitimized many of the extremist parties political ideas but, by permanently enlarging the political space, also enabled them to keep a sizeable share of the electorate even after the traditional parties began to move closer to the centre again. Thus, according to Ignazi (1997: 316 319), an increasing political distance between the extremes favour[s] institutionalized extreme-right parties and Green parties. Of the scholars mentioned above, only Kitschelt & McGann (1995: 53 56) have attempted to test these theoretical arguments empirically. However, his methodology is open to criticism. First, they use data collected at a single point in time (1990) in their study, thus neglecting dynamic effects of party system change. This article uses a different approach. While it does test the argument that the closer the establishment parties are placed on the left-right scale at a specific point in time the more certain voters are inclined to vote for parties that challenge the political establishment. It also explores another related argument namely, that it is not the degree of closeness at a certain point in time that influences voters but rather the development over time in the positioning of establishment parties on the left-right scale that motivates a voter s decision. In other words, voters are more likely to vote for anti-politicalestablishment parties if the establishment parties actually converge over time with the result that the electorate feels that it has become increasingly difficult to differentiate between them. A second problem with Kitschelt s methodology is related to the way in which he attains the divergence score for each country. In order to determine how far apart from each other the establishment parties are positioned on the left-right scale, he examines the difference between the scores for the two major parties of the Left and Right in each country, thus overlooking the fact that other smaller establishment parties might have scores that place them

challenges to established parties 555 further apart from those two main establishment parties. In doing so Kitschelt risks producing a score that underestimates true levels of divergence. It can be argued that voters would be less likely to support anti-politicalestablishment parties if they were presented with the choice of voting for an establishment party, regardless of its size, that is noticeably different from its main competitors. More precisely, while the two major parties of the Left and Right might lose votes as a result of their convergence, these losses would not necessarily lead to increased support for anti-political-establishment parties as long as there are other establishment parties that offer the voter a clear alternative. It is argued here that to more effectively measure how far apart the establishment parties are positioned on the left-right scale one should instead start out by clearly defining the term establishment parties. The definition of this party type used here is inspired by one of Giovanni Sartori s (1976: 122 123, 300 301) criteria of party relevance namely the governing potential criterion. According to Sartori this criterion consists of two measures, the governing potential of a party and its actual governmental relevance. In this article, the establishment party is viewed as consisting of: (a) all of those parties that have participated in government, or alternatively those parties that the governing parties regard as suitable partners for government formation; and (b) parties that are willing to cooperate with the main governing parties by joining them in a coalition government. These are very broad definitional features that are nevertheless clear enough to allow us to identify parties as belonging to the political establishment depending on whether they fulfil both criteria or not. The divergence score is then calculated by subtracting the score for the leftmost establishment party from that of the rightmost establishment party. In addition to assessing more effectively the convergence hypothesis, this article also tests Ignazi s polarization hypothesis. This requires us not only to ascertain whether the degree of polarization of a party system is related to the level of electoral success of anti-political-establishment parties, but also to consider whether a change over time in the degree of polarization of a party system has an effect on the level of support for this type of party. Party system polarization has been measured in different ways. Lee Sigelman and Syng Nam Yough (1978), for example, used the US State Department s rankings of parties on a 4-point left-right scale. First, they calculated each national party system s weighted mean by multiplying each party s percentage of the vote by its left-right score and summing those products for all parties in the system (Sigelman & Nam Yough 1978: 367). The polarization scores were based on the variance of party scores about the mean. Moreover, each party s left-right score was plotted against its vote percentage in a national election. Consequently, a party system would receive the maximum

556 amir abedi polarization score whenever the parties at the left and right ends of the political spectrum each received 50 per cent of the votes. The minimum polarization score would be obtained whenever parties located at a single point on the left-right continuum received 100 per cent of the votes. By contrast, Hazan (1997: 18 19) based his measure of polarization on the parliamentary strength of a party instead of its electoral performance. He multiplied each party s share of parliamentary seats (in %) by its score on the left-right political scale (as determined by Castles & Mair 1984) and summed the results to create a weighted party system mean of left-right seat distributions for a specific election. The weighted system mean of seat distributions (for that particular election) was then subtracted from the left-right position score of each party. Finally, the results for all the parties were added up to arrive at the measure of party system polarization for a particular country and election. In developing their measure of polarization, Giacomo Sani and Giovanni Sartori (1983) do not take party strength into account. Instead, they concentrate on what Sartori (1976: 342 343) has termed the space elasticity or the space of competition and use as their measure of polarization the distance between any two groups of partisans as measured by the (absolute) difference between their mean self location [in the 1976 Eurobarometer Survey s 10-point scale] divided by the theoretical maximum (i.e., 9) (Sani & Sartori 1983: 321). The measure used in this article also focuses on the distance between the leftmost and rightmost party in a party system and does not take party strength in terms of vote share into account. Thus, this study concerns itself purely with the ideological distance between political parties in a party system. Consequently, the polarization score used here is determined by subtracting the score of the leftmost party in a party system (regardless of whether it is an establishment party or an anti-political-establishment party) from that of the rightmost one. This measure of polarization is very similar to measures developed by Hans Keman (1997: 88, 111), Markus M.L. Crepaz (1990: 192), and Peter Mair and Francis G. Castles (1997: 154 155). Hans Keman (1997) arrives at what he terms the range of the party system by calculating the difference between the most left-wing and right-wing parties in a party system. The other two studies also use the distance between the leftmost and rightmost party in a party system to obtain their respective polarization scores. For our dependent variable, a party is defined as an anti-politicalestablishment party if it fulfils all of the following three criteria: 1. it perceives itself as a challenger to the parties that make-up the political establishment; 2. it asserts that a fundamental divide exists between the political

challenges to established parties 557 establishment and the people (implying that all establishment parties, whether in government or in opposition, are essentially the same (cf. Mudde 1996; Schedler 1996; Smith 1987)); and 3. it challenges the status quo in terms of major policy issues and political system issues (cf. Ignazi 1992; Mackie 1995; Schedler 1996; Smith 1987). It should be pointed out that the definitions of the establishment party and the anti-political-establishment party are not meant to be exhaustive. There are certainly parties that are neither anti-political-establishment parties nor a part of the political establishment as defined above. Most of these other (residual) parties are not politically relevant in the Sartori sense. This refers to parties that have no realistic chance of participating in government and, at the same time, do not see themselves as challengers to the political establishment. Moreover, it has to be stressed that the definitional criteria for the establishment party and the anti-political-establishment party can and should be assessed over time for a given party. The decisive factor as to whether a party can be classified as an anti-political-establishment party is thus whether it fulfils all of the above-mentioned criteria at a particular point in time or not. Appendix A provides a list of establishment parties and antipolitical-establishment parties for the 16 democracies included in this study. 2 The score forming the dependent variable is then obtained by simply adding up the percentages of the votes that all anti-political-establishment parties received in an election for each country. 3 It is maintained here that rather than simply including only extreme right-wing parties and/or New Politics parties, it is preferable to include all parties for which challenging the political establishment lies at the heart of their agenda because it better reflects the true range of alternatives that are available to a voter who wants to cast his/her vote against the political establishment. The fact that the antipolitical-establishment party category includes any party regardless of its organizational structure, size, ideology and placement on the left-right political scale prevents the introduction of a selection bias in the dependent variable that is, the selection of parties that only fall on a particular spot on the left-right scale. Data The hypotheses set out above were tested using data sets from three different time periods that rely on expert judgements to place parties on the left-right political scale in a series of countries. Lawrence C. Dodd (1976) collated information from a range of published sources to position parties on the left-right

558 amir abedi scale for the 1945 to 1974 period. All parties are placed in a fixed ordering for the entire period. The only exception is France, for which Dodd produced separate scales for the first five postwar elections. Data for 1982 are contained in the study of Francis G. Castles and Peter Mair (1984), while John Huber and Ronald Inglehart s (1995) study provides data for 1993. Both of these last two data sets used comprehensive expert surveys to arrive at party positions on the left-right scale for a number of countries. The data on the left-right locations of the parties are thus comparable across countries and over time. 4 Using a unidimensional spatial model (i.e., the left-right ideological continuum) to assess the degree of party system polarization and establishment party divergence is certainly controversial. Several scholars have criticized Downs s (1957) unidimensional left-right scale. Ronald Inglehart (1977, 1997), for example, has pointed to the emergence of a new dimension of political conflict distinct from the traditional left-right continuum centred around qualityof-life issues. A new axis based on the polarization between Postmodern and Fundamentalist worldviews (Inglehart 1997: 265) has emerged. It pits culturally conservative, often xenophobic, parties disproportionally supported by Materialists, against change-oriented parties, often emphasizing environmental protection, and disproportionately supported by Postmaterialists (Inglehart 1997: 237 238). However, Inglehart s findings do not necessarily imply that the left-right continuum has become obsolescent. A number of studies have shown that, while other dimensions of conflict might exist, the left-right dimension is still dominant. For example, Ian Budge and David Robertson examined political parties election programmes in 19 democracies and concluded that the generalizability and predominance of the left-right cleavage has been resoundingly confirmed (Budge & Robertson 1987: 393). Other scholars have pointed out that the left-right continuum not only captures traditional economic policy conflicts (Beyme 1985; Crepaz 1990; Hazan 1997; Laponce 1981; Mair 1997; Sani & Sartori 1983; Smith 1989), but also its flexibility and catch-all character (Mair 1997: 26). Thus, the left-right continuum is an overall dimension capturing a multitude of different political conflicts (Beyme 1985: 257). Moreover, Mair (1997: 26) points out that the absorption capacity of left and right now also appears to have accommodated many of the new politics concerns. Crepaz (1990: 191) adds that the left-right dimension does not only tap differences in ideology; it taps deeply seated differences over a number of dimensions and therefore it mirrors fairly well the people s stand on major conflict domains. The left-right dimension has proven its usefulness in a number of studies (e.g., Crepaz 1990; Hazan 1997; Keman 1994, 1997). For the analytical purpose of this study, too, the unidimensional left-right scale appears to be the most appropriate methodological tool.

challenges to established parties 559 This article focuses on 16 advanced-industrial democracies for which data for 1945 1974, 1982 and 1993 are available. The dependent variables in this study are the average of the combined anti-political-establishment party vote scored in parliamentary elections between 1945 and 1974; the combined antipolitical-establishment party vote scored in the election closest to 1982 and 1993, respectively (see Appendix B); and the changes in that vote share from 1945 1974 to 1982, 1982 to 1993 and 1945 1974 to 1993. 5 The independent variables are the polarization of the party system at the same points in time and the changes in polarization over time, and the divergence of the establishment parties (i.e., the distance between the leftmost and rightmost of these parties) and the changes in divergence over time. Tables 1 and 2 show the scores for party system polarization, establishment party divergence as well as the combined anti-political-establishment party scores for the 16 countries examined in this study. A first glance at the data, particularly at the changes over time, yield three important observations. First, there seems to be a general trend toward polarization in the party systems between 1945 1974 and 1993, with Iceland being the main exception. Second, the establishment parties in the 16 countries under study have generally moved further apart on the left-right political scale between 1945 1974 and 1982. However, between 1982 and 1993 they have for the most part displayed a trend toward convergence. The main outliers in this case are, on the one hand, Iceland, Italy and New Zealand where the establishment parties have moved closer together between 1945 1974 and 1982 as well as between 1982 and 1993 and, on the other hand, Canada, Denmark and Germany where the opposite was true. Finally, the electoral support for anti-politicalestablishment parties has on average decreased between 1945 1974 and 1982 but then increased between 1982 and 1993. Denmark shows the reverse pattern. Australia, Germany, Iceland, Ireland, Italy, the Netherlands, New Zealand and the United Kingdom all experienced a steady increase in support for anti-political-establishment parties between 1945 1974 and 1993, while the opposite was the case in Sweden. The observations for the 1982 to 1993 period basically support similar conclusions reached by Mair & Castles (1997). Examining the relationship between party system features and electoral support for anti-political-establishment parties The aim of this article is to test empirically the hypotheses outlined in the first part of the article. From these hypotheses one can derive six theoretical expectations. The first three relate to establishment party divergence.

Table 1. Combined anti-political-establishment party scores, party system polarization and establishment party divergence in 16 democracies Combined APE* party score Party system polarization Establishment party divergence Country 1945 1974 1982 1993 1945 1974 1982 1993 1945 1974 1982 1993 Australia 1.2 5.0 5.4 3.2 4.2 5.5 3.2 4.2 3.7 Austria 9.1 2.0 30.6 4.5 5.6 5.7 3.2 3.4 1.5 Belgium 9.9 9.4 12.2 5.1 7.5 6.4 3.2 5.7 3.9 Canada 12.4 2.7 32.2 4.5 4.1 6.1 2.5 3.0 4.4 Denmark 6.1 12.6 9.5 5.8 7.1 7.1 4.5 4.9 5.2 Finland 17.4 14.1 14.6 4.5 4.9 5.5 3.8 4.9 3.9 France 42.8 1.1 21.2 5.2 7.5 9.0 3.9 6.1 5.6 Germany 4.1 5.6 6.8 5.1 5.8 7.8 3.2 4.1 4.4 Iceland 0.2 5.5 9.5 4.5 4.4 4.3 4.5 4.4 4.3 Ireland 3.0 3.2 3.7 3.9 4.5 5.6 3.9 2.9 4.2 Italy 35.2 40.4 42.5 5.2 7.7 8.3 2.6 2.5 2.3 Netherlands 11.2 11.5 16.6 5.2 7.8 7.7 3.9 5.3 3.0 New Zealand 6.7 20.7 29.2 3.2 2.0 3.8 3.2 2.0 1.5 Norway 4.6 4.5 7.4 4.5 7.4 7.9 4.5 5.8 5.4 Sweden 3.2 1.7 1.2 4.5 5.8 6.5 3.2 5.8 5.7 United Kingdom 1.0 1.1 2.9 4.5 5.4 6.3 3.2 5.4 4.4 Averages 10.5 8.8 15.3 4.6 5.7 6.5 3.5 4.4 4.0 560 amir abedi Note: Since the extreme right Vlaams Blok was not listed in the 1993 survey for Belgium, its score has been estimated as that recorded in 1982. Since Iceland was not included in the Castles & Mair study, the polarization and divergence scores for 1982 have been estimated as being the average of the 1945 1974 and 1993 scores. In the French case, all the data for the 1945 1974 period refer specifically to 1956. In the British case, the scores refer to the United Kingdom as a whole, thus including Ulster Unionists and Irish Nationalists. * anti-political-establishment. Sources: Castles & Mair (1984), Dodd (1976), Huber & Inglehart (1995) and Mackie & Rose (1991, 1997).

Table 2. Changes over time in combined anti-political-establishment party scores, party system polarization and establishment party divergence Combined APE* party score Party system polarization Establishment party divergence 1982 to 1993 to 1982 to 1993 to 1982 to 1993 to Country 1945 1974 1993 1982 1945 1974 1945 1974 1993 1982 1945 1974 1945 1974 1993 1982 1945 1974 Australia 3.8 0.4 4.2 1.0 1.3 2.3 1.0-0.5 0.5 Austria -7.1 28.6 21.5 1.1 0.1 1.2 0.2-1.9-1.7 Belgium -0.5 2.8 2.3 2.4-1.1 1.3 2.5-1.8 0.7 Canada -9.7 29.5 19.8-0.4 2.0 1.6 0.5 1.4 1.9 Denmark 6.5-3.1 3.4 1.3 0.0 1.3 0.4 0.3 0.7 Finland -3.3 0.5-2.8 0.4 0.6 1.0 1.1-1.0 0.1 France -41.7 20.1-16.6 2.3 1.5 3.8 2.2-0.5 1.7 Germany 1.5 1.2 2.7 0.7 2.0 2.7 0.9 0.3 1.2 Iceland 5.3 4.0 9.3-0.1-0.1-0.2-0.1-0.1-0.2 Ireland 0.2 0.5 0.7 0.6 1.1 1.7-1.0 1.3 0.3 Italy 5.2 2.1 7.3 2.5 0.6 3.1-0.1-0.2-0.3 Netherlands 0.3 5.1 5.4 2.6-0.1 2.5 1.4-2.3-0.9 New Zealand 14.0 8.5 22.5-1.2 1.8 0.6-1.2-0.5-1.7 Norway -0.1 2.9 2.8 2.9 0.5 3.4 1.3-0.4 0.9 Sweden -1.5-0.5-2.0 1.3 0.7 2.0 2.6-0.1 2.5 United Kingdom 0.1 1.8 1.9 0.9 0.9 1.8 2.2-1.0 1.2 * anti-political-establishment. Sources: Castles & Mair (1984), Dodd (1976), Huber & Inglehart (1995) and Mackie & Rose (1991, 1997). challenges to established parties 561

562 amir abedi I. The closer the establishment parties are located to each other on the left-right dimension at a certain point in time, the greater the success of anti-political-establishment parties in the election closest to that point in time. II. The more the establishment parties converge on the left-right scale over time, the greater the success of the anti-political-establishment parties in the election closest to the end of that same time span. III. The more the establishment parties converge on the left-right scale over time, the more the electoral support for anti-politicalestablishment parties will increase over that same time span. The coefficients showing the bivariate relationships between party system features and anti-political-establishment party support (see Table 3) appear to support theoretical expectation I. The closer the establishment parties are located to each other on the left-right dimension at a certain point in time, the stronger the performance of the anti-political-establishment parties in the election closest to that point in time. The correlation coefficients indicate a strong negative relationship between the two variables that is, the less divergent the establishment parties were on the left-right political scale in both 1982 and 1993, the higher the combined anti-political-establishment party vote share. The result for 1945 1974 is less encouraging in that the association between the two variables is clearly weaker. However, the fact that the sign of the correlation coefficient is negative as expected lends some support to theoretical expectation I. The same cannot be said of theoretical expectation II which posits that the more the establishment parties converge over time, the better the anti-political-establishment parties will do in the election closest to the end of that same time span. While all the correlation coefficients show the anticipated negative association between the two variables, they are somewhat weaker than the coefficients for the crosssectional relationship. Moreover, the only statistically significant relationship is that between the change in establishment party divergence between 1945 1974 and 1993 and the combined anti-political-establishment party score in 1993. Theoretical expectation III, which asserts that the more the establishment parties converge over time, the more the electoral support for anti-politicalestablishment parties will increase over that same time span, seems to receive slightly stronger backing from the data than theoretical expectation II. All the coefficients indicate the expected negative association between the two variables. Furthermore, two of the three relationships are fairly strong and statistically significant. Only the relationship between change in establishment party divergence for the 1982 to 1993 time period and the change in combined

Table 3. Bivariate correlations among party system features and combined anti-political-establishment party scores Combined APE* party score Party system features 1945 1974 1982 1993 1982 to 1945 1974 1993 1982 1993 to 1945 1974 Establishment party divergence 1945 1974-0.18 1982-0.47** 1993-0.64*** 1982 to 1945 1974 difference -0.39-0.45** 1993 1982 difference -0.07-0.02 1993 to 1945 1974 difference -0.44** -0.57** Party system polarization 1945 1974 0.40 1982 0.14 1993 0.08 1982 to 1945 1974 difference 0.11-0.30 1993 1982 difference 0.13 0.26 1993 to 1945 1974 difference 0.04-0.52** * anti-political-establishment; ** p < 0.05; *** p < 0.01. challenges to established parties 563

564 amir abedi anti-political-establishment party score over that same time span is very weak and not significant. Overall the results of the above correlation analyses lend some support to those hypotheses which claim that the distance between the establishment parties on the left-right political scale has an effect on the electoral fortunes of anti-political-establishment parties. The results also underline the negative direction of that relationship. Thus, the less ideological space the establishment parties cover, the more successful anti-political-establishment parties are in the electoral arena. Moreover, the results suggest that the relationship between these two variables is not only cross-sectional, but also longitudinal. Theoretical expectations IV to VI concern party system polarization. IV. The more polarized a party system is at a certain point in time, the higher the percentage of the votes will be for the anti-politicalestablishment parties in the election closest to that point in time. V. The more a party system polarizes over time, the higher the percentage of the votes for the anti-political-establishment parties will be in the election closest to the end of that same time span. VI. The more a party system polarizes over time, the more the electoral support for anti-political-establishment-parties will increase over that same time span. The results indicate that theoretical expectation IV, which posits that the more polarized a party system is at a certain point in time, the higher will the support for anti-political-establishment parties be in the election closest to that point in time, does not receive strong support. While the correlation is positive as expected, two of the coefficients (for 1982 and 1993) are very weak. Moreover, none of the three coefficients is statistically significant. The same can be said for theoretical expectation V, which maintains that the more a party system polarizes over time, the higher will be the support for anti-politicalestablishment parties in the election closest to the end of that same time span. Again, while the correlation coefficients have the expected positive sign, they all indicate a rather weak relationship between the two variables. Theoretical expectation VI asserts that the more a party system polarizes over time, the larger the increase in electoral support for anti-political-establishment parties over that same time span. Here, too, the results of the analysis are not supportive. The coefficients for the 1945 1974 to 1982 and 1945 1974 to 1993 time periods show the association between the change in party system polarization and the change in combined anti-political-establishment party score to be negative and not positive as we would have expected. Only the coefficient for the 1982 to 1993 time period has the expected sign.

challenges to established parties 565 These results suggest that the existence of a polarized party system does not necessarily benefit anti-political-establishment parties. This becomes especially evident when we examine the longitudinal relationship between the variables. Thus, the degree of polarization of a party system and changes over time in polarization appear to be largely unrelated to the electoral support for anti-political-establishment parties. So far, the effects of the two independent variables party system polarization and establishment party divergence on the electoral fortunes of anti-political-establishment parties have been viewed as essentially mutually exclusive. This is not necessarily the case. Ignazi (1996) maintains that the increasing polarization of party systems and the resulting growth in support for right-wing extremist parties in the late 1980s had been sparked by an earlier radicalization of the main establishment parties themselves. In other words, according to Ignazi, one would expect that countries in which the establishment parties were more divergent in 1982 and which had experienced an increase in party system polarization between 1982 and 1993 should demonstrate a higher level of support for anti-political-establishment parties in 1993 than other countries. This model shall be termed Model A for the purposes of the multivariate regression analyses. Table 4 presents the results of the multivariate analyses. The results do not seem to support the notion that Ignazi s hypothesis on the rise of parties of the extreme right can be extended to cover all antipolitical-establishment parties. Model A as a whole is not statistically significant and neither of the two coefficients has the expected sign. Moreover, the only strong indicator in this model is establishment party divergence in 1982. According to Ignazi s hypothesis one would have expected a positive relationship between establishment party divergence in 1982 and the combined anti-political-establishment party score in 1993. However, the standardized regression coefficient indicates a relatively strong negative relationship that is, the less divergent the establishment parties were on the left-right political scale in 1982, the higher the combined anti-political-establishment party score was in 1993. However, there are alternatives to Ignazi s model. One could argue that the party system features that are most conducive to anti-politicalestablishment parties are a very polarized party system in which the establishment parties are positioned rather close to each other on the left-right political scale. This arrangement would provide anti-political-establishment parties the greatest possible amount of space either to the left or to the right of the establishment parties. It would enable the anti-political-establishment parties strategically to position themselves rather close to the establishment

Table 4. Multivariate analyses Multivariate models Model A Model B Model C Model D 1982 to 1993 to 1945 1945 1993 1945 Party system features 1993 1974 1982 1993 1982 1993 1993 1974 1982 1974 Establishment party divergence 1945 1974-0.28 1982-0.64* -0.84** 1993-0.79** 1982 to 1945 1974 difference -0.40-0.64* 1993 1982 difference -0.23-0.26 1993 to 1945 1974 difference -0.54* -0.43* 566 amir abedi Party sustem polarization 1945 1974 0.46* 1982 0.63* 1993 0.38* 1982 to 1945 1974 difference 0.46-0.08 1993 1982 difference -0.13 0.27 0.41 1993 to 1945 1974 difference 0.25-0.35 R 2 0.36 0.24 0.47 0.53 0.30 0.05 0.25 0.20 0.11 0.43 Note: Table entries are standardized regression coefficients. *p < 0.05; ** p < 0.01.

challenges to established parties 567 parties and thus pick up additional, disaffected, voters closer to the centre without jeopardising their more radical followers unless an even more radical party pops up closer to the extreme position vacated by that party. However, since this study concerns itself with the combined vote for all antipolitical-establishment parties, the possible negative fate of any individual party does not affect the electoral fortunes of these parties in general. Rather, anti-political-establishment parties should generally profit from the increasing space available to them under that scenario. In its cross-sectional version, labelled Model B, this scenario can be tested for 1945 1974, 1982 and 1993, respectively. This model thus brings together theoretical expectations I and IV. In its longitudinal version, Model C will test theoretical expectations II and V, and Model D will test theoretical expectations III and VI. The results of the multivariate analyses for Model B are more encouraging than they were for Model A. The R 2 values for 1982 and 1993 are fairly strong. In both of these years, establishment party divergence emerges as the strongest indicator of the combined anti-political-establishment party votes in 1982 and 1993, respectively. In both cases, the standardized regression coefficients indicate a very strong negative relationship between the two variables. This supports the conclusion reached in the previous correlation analyses. The R 2 value for 1945 1974 is not as strong and the coefficients indicate that party system polarization is the stronger predictor of the anti-political-establishment party vote share. Overall, the results for Model B partially reinforce the findings of the bivariate analyses lower ideological divergence among the establishment parties is related to higher levels of electoral support for anti-politicalestablishment parties. However, it seems that the addition of party system polarization to the model has increased its explanatory power. The previous correlation analyses have suggested that the polarization variable by itself is not sufficient to explain the level of support for anti-political-establishment parties at a single point in time. The results of the multivariate analyses have revealed that a polarized party system can be conducive to the electoral success of parties that challenge the political establishment provided that the establishment parties are, at the same time, positioned closer to each other on the left-right continuum. Model C tries to explain the combined anti-political-establishment party scores in 1982 and 1993 with the changes over time in party system polarization and establishment party divergence. The R 2 values for the three manifestations of Model C are weaker than the ones for Model B. Nevertheless, the signs of the standardized regression coefficients indicate that antipolitical-establishment parties seem to be electorally more successful when the party system is becoming more polarized and when the establishment parties

568 amir abedi converge toward the centre of the political spectrum. Establishment party divergence again emerges as the stronger predictor of the combined antipolitical-establishment party vote share for two of the three manifestations of Model C. While the results for Model C appear to support the notion that the relationship between the variables is not only cross-sectional but also longitudinal, the same does not seem to hold for Model D. Model D explores the relationship between changes in party system polarization and establishment party divergence over time and the change in the combined anti-politicalestablishment party score over that same time span. While the standardized regression coefficients support the notion that there is a negative association between support for anti-political-establishment parties and divergence between the establishment parties on the left-right scale, they do not necessarily reinforce the conclusion that polarizing party systems are positively associated with an increase in the vote share of anti-political-establishment parties. Rather, there seems to be a weak negative relationship between changes in party system polarization over time and changes in the combined anti-political-establishment party vote for both the 1945 1974 to 1982 and the 1945 1974 to 1993 time periods. However, the results consistently indicate the existence of a negative relationship between changes in establishment party divergence and changes in the combined anti-political-establishment party vote share. Overall, the results of the multivariate analyses basically reinforce the conclusions reached in the bivariate analyses. 6 First, anti-political-establishment parties appear to do better in elections when the establishment parties are positioned relatively close to each other on the left-right political scale. The evidence in support of that hypothesis is fairly strong when one looks at that relationship across space and at a single point in time. One has to be more cautious, however, when making the claim that a change over time in the positioning of the establishment parties has an effect on the electoral fortunes of anti-political-establishment parties. Nevertheless, the consistently negative signs of the standardized regression coefficients suggest that the anti-politicalestablishment parties electoral fortunes generally improve when the establishment parties converge over time. Second, the results of the bivariate and multivariate analyses do not lend much support to the hypothesis that the degree of party system polarization has a strong independent impact on the electoral fortunes of anti-political-establishment parties. More specifically, there is no consistent indication that a polarizing party system is conducive to the success of anti-political-establishment parties. However, the results for Models B and C provide some evidence in support of the argument that these parties can profit from a polarized or, to a somewhat lesser degree, polarizing

challenges to established parties 569 party system whenever there is less divergence between the establishment parties that is, whenever the establishment parties have left more ideological space unoccupied. Thus, party system polarization reinforces the effect that establishment party divergence has on the level of electoral support for parties that challenge the political establishment. The results of the bivariate and multivariate analyses provide more support for the existence of a cross-sectional relationship between party system features and the combined anti-political-establishment party vote than for the existence of a longitudinal relationship. This affects party system polarization more so than establishment party divergence. While the coefficients provide consistent evidence in support of the existence of a negative relationship between changes in establishment party divergence and changes in the level of support for anti-political-establishment parties, there is no consistent indication that an increase in party system polarization is associated with an increase in anti-political-establishment party vote shares. What could explain this discrepancy in outcomes? First of all, the results appear to imply that it is the degree of polarization of a party system, as opposed to a change in polarization, that is more important to the level of success of anti-politicalestablishment parties, especially in connection with a close positioning of the establishment parties on the left-right scale. More precisely, a comparatively small change in the degree of polarization of a party system will not have a significant effect on the electoral fortunes of anti-political-establishment parties. For example, it does not make a big difference to a very polarized party system whether it experiences a small increase or decrease in polarization as long as it remains very polarized. However, in the case of establishment party divergence we are dealing with a relatively small range of space and a larger number of voters. Thus, even a comparatively minor change in the degree of divergence can be expected to have a noticeable effect on the level of support for anti-political-establishment parties. The results certainly indicate the need for further research in this area. It would, for example, be useful to include individual-level data in any future examination of the effects of party system features on the electoral fortunes of parties that challenge the political establishment. Conclusions The main goal of this article has been to explore the possible independent effects of party system features on the electoral fortunes of anti-politicalestablishment parties. Two hypotheses were tested. First, Hainsworth s (1992a), Kitschelt & McGann s (1995) and Mair s (1995) contention that anti-

570 amir abedi political-establishment parties thrive mainly in party systems in which the establishment parties are positioned rather close to each other on the left-right political scale. Second, Piero Ignazi s (1992, 1996, 1997) suggestion that antipolitical-establishment parties are more successful in polarizing party systems. The bivariate correlation and multivariate regression analyses lend some credence to the thesis that party system features do indeed influence the electoral fortunes of anti-political-establishment parties. However, the analyses provide more support for the establishment party divergence hypothesis than for the one that stresses the importance of party system polarization. Divergence has emerged as the stronger indicator, both when trying to predict crossnational differences in the electoral fortunes of anti-political-establishment parties at a single point in time and, more especially, when it comes to examining the relationship between changes in party system features over time and changes in the level of electoral support for anti-political-establishment parties. Furthermore, the effect of these party system features on the electoral fortunes of anti-political-establishment parties has appeared most clearly in the cross-sectional analyses. The results from the longitudinal analyses were more ambiguous. The results of the analyses conducted in this article provide evidence in support of Hainsworth s (1992a), Kitschelt & McGann s (1995) and Mair s (1995) hypothesis while at the same time challenging Ignazi s (1992, 1996, 1997) argument. More precisely, the results call into question whether the polarization hypothesis can be extended from applying specifically to rightwing extremist parties to covering all anti-political-establishment parties. Nevertheless, there is some indication that anti-political-establishment parties can profit from a polarized or, to a somewhat lesser extent, polarizing party system provided that it is accompanied by a less divergent positioning of the establishment parties on the left-right political scale. Ignazi (1997: 319) has argued that [a] growing political distance between the extremes and the collapse of the centre parties could favour institutionalised extreme-right parties and Green parties. However, the results of the analyses conducted in this study suggest that it is not necessarily only the collapse of the centre parties, but also the overcrowding of the centre that enables anti-politicalestablishment parties to gain an electoral foothold in the established party systems of many advanced industrial democracies. The results of this study suggest that a successful strategy against antipolitical-establishment parties must include increasing the ideological divergence between establishment parties. Voters must be offered a clearer choice. This is obviously easier said than done. Establishment parties in most advanced industrial democracies currently seem to feel that they need to move closer to the centre in order to remain, or become, electorally competitive. In