Electoral Waves: An Analysis of Trends, Spread and Swings of Votes Across 20 West European Countries,

Transcription

1 National Centre of Competence in Research (NCCR) Challenges to Democracy in the 21 st Century Working Paper No. 43 Electoral Waves: An Analysis of Trends, Spread and Swings of Votes Across 20 West European Countries, Daniele Caramani Institute of Political Science University of St. Gallen Bodanstrasse 8, CH-9000 St. Gallen Phone: Web: October 2010

2 Electoral Waves: An Analysis of Trends, Spread and Swings of Votes Across 20 West European Countries, Abstract The paper analyses the flux of votes towards and away from party families in Europe. This classical indicator of shifting votes is used to assess the degree to which there exist similar waves of electoral change across Europe which take place simultaneously in most countries. Theoretically, the paper distinguishes trends, spreads and swings. Based on electoral data for 20 West European countries since 1970, the analysis focuses on uniform swings for families of left and right, as well as for greens, regionalists and populists. The detection of waves is interpreted as an indicator of a Europeanised electorate referring to common supra-national issues. Results attest for swings as far as the green and populist families are concerned, but less so for left right parties and regionalists. Keywords: Electoral swings; Parliamentary elections; Europeanisation; Party families This paper is part of the project From National to Supra-National Democracy in the European Union funded by the National Centre for Competence in Research Challenges to Democracy in the 21st Century (NCCR Democracy, Swiss National Science Foundation).

3 Introduction Since the first electoral analyses carried out in the US and Europe, the commonality and simultaneity of electoral change has been considered an important indicator of the degree of a political system s integration. Common shifts of votes occurring between two elections in all territorial areas of a political system mean that its parts move together and that everywhere they respond similarly to common factors and issues. Such general simultaneous changes of votes point to an advanced degree of formation and integration of political systems. While in the past uniform change has been investigated exclusively at the national level, it can be used today to analyse the formation of a European-wide electorate. At the level of national electorates and party systems shift of votes were thought of being an important indicator of the degree of nationalisation of electoral politics. 1 This paper extends the use of this indicator to assess the degree of Europeanisation of electoral politics. Insofar as one can detect common waves of change that take place simultaneously everywhere in Europe or, at least, in most European political systems we can speak of a Europeanised or integrated dimension of electoral politics. 2 To what extent is electoral change common and simultaneous in Europe? To what extent do national electorates move together in waves of change? The paper addresses these questions and aims to assess the degree to which waves of change that wipe across Europe at the same time indicate the development of an incipient European-wide party system. This is the reason why the analysis 1 For a review of the other dimensions of the nationalisation of electoral politics see Caramani (2004: 32 43). Two further dimensions are particularly relevant also for the assessment of the Europeanisation of electoral politics: first, the homogeneity of voting behaviour across countries and, second, the vertical displacement of issues from the national to the European level. 2 The transposition of these indicators from the national to European level can be found in Caramani (2006 and forthcoming). For a similar approach see Jerôme et al. (2006) who analyse Europe as one nation. 2

4 starts in the early 1970s, the decade when European integration took an electoral dimension with the introduction of direct elections to the European Parliament. First, the analysis in this paper includes shifts between the major parties of the left and right (communists, socialists, liberals and conservatives). It aims to look systematically at the frequent but mostly impressionistic attempt among observers in the public sphere to identify continental-wide shifts in electorates mood following major electoral or cabinet changes in a few countries. Second, the analysis goes beyond the left right dimension and considers what the literature has identified has other main elements of electoral de-alignment, also mainly since the early 1970s: 3 The green wave of the 1980s: The parallel institutionalisation of new politics especially through environmental parties as the main partisan translation of the generational cleavage between materialist and post-materialist values (Inglehart, 1990; Poguntke, 2002). 4 The regionalist wave of the 1990s: The emergence of ethnic and regionalist parties in the spaces left void by the alleged weakening of the normative and identity role of the nation-state occurred through supra-national political integration and economic globalisation, as well as decentralisation processes (Keating, 1998 and 2004). The populist wave of the 2000s: The emergence of extreme right-wing populist parties from the cleavage between winners and losers of integration and globalisation, and the perceived threats to identities and labour markets engendered 3 On de- and re-alignment see the answer in terms of stabilisation of voting behaviour in Bartolini and Mair (1990). For assessment of recent volatility trends see Lane and Ersson (2007) and Tavits (2008). 4 The generational value cleavage at the origin of civil rights movements (especially in the US), pacifism and feminist did not lead to other stable party organisations. 3

5 by the opening up of barriers to competition from immigrants and external production centres (Kitschelt, 1995; Kriesi et al., 2008). In trying to assess the existence of European-wide tides of change across most or all countries, this paper wishes to contribute to a growing literature that transposes theories and concepts developed to explain the formation of national political spaces to the formation and integration of a European political space (Bartolini, 2005; Hix, Noury and Roland, 2007; Marks and Steenbergen, 2002). Within this literature, the theory of the nationalisation of politics occupies an important place as it focuses directly on the electoral dimension of system-formation and integration. 5 The transposition of nationalisation to Europeanisation is therefore an important step towards understanding the formation and integration from a mass electoral perspective (Caramani, 2006: 17 22; Jerôme et al., 2006). The paper departs from much of the literature on cross-country dynamics also insofar as it consider different types of general shifts. 6 The paper therefore starts with a theoretical discussion of different types of waves : uniform swings, trends and spread. It then moves to describe the methodological bases of the study, namely a broad cross-country longitudinal analysis. The main bulk of the paper is devoted to the presentation of empirical results for each type of wave. 5 Not surprisingly in the wake of two major comparative works on nationalisation (Caramani, 2004 and Chhibber and Kollman, 2004) the number of contributions on the topic has increased in recent years (see Alemán and Kellam, 2008; Bochsler, 2006; Hicken, 2008; Croissant and Schächter, 2008; Kasuya and Moenius, 2008). 6 This is a literature mainly concerned with economic voting. See Lewis-Beck and Paldam (2007) and Powell and Whitten (993). For an overview see Lewis-Beck and Stegmaier (2007). 4

6 Analysing Electoral Waves: Swings, Trends and Spread In this theoretical section, the distinction between different types of waves is presented. There are three main types of waves: (1) uniform swings, (2) trends, (3) spread. While the paper focuses on uniform swings, it is important to contrast them against other types of waves. The analysis of uniform swings of votes across territorial units can be defined a classical indicator of electoral studies going back to periods in which individual investigation techniques based on surveys did not exist and in which analyses relied on aggregate, that is territorial data. The golden age of electoral geography includes the earliest electoral studies as those of M. Hansen (Norsk Folkepsikologi, 1898), A. Siegfried (Tableau Politique de la France de l Ouest, 1913) and H. Tingsten (Political Behavior, 1937). More recently, two further classical works have used this indicator: E. Schattschneider s The Semisovereign People (1960) and D. Butler and D. Stokes Political Change in Britain (1974). It has since then been largely forgotten. 7 This paper argues that today it can fruitfully be applied to the analysis of European integration processes. The integration of electorates and party systems has been analysed primarily at the national level: nationalisation refers to a process of transformation of politics from local to national. Processes of Europeanisation refer on the other hand to the transformation of politics from national to European. Such processes of electoral integration, whether at the national or European level entail distinct dimensions. First, a horizontal dimension referring to the territorial homogenisation of electoral behaviour in the units composing a system. This is a 5

7 process of standardisation. Second, a vertical dimension referring to the dislocation of issues, organisations and competences from the sub-systemic to the systemic level. This is a process of centralisation. 8 The horizontal dimension includes both (1) the homogenisation of support for different parties across regions and (2) the uniformity of change between two elections, that is swings (Caramani, 2004: 36). Both measure territorial similarity of electoral behaviour: the first is static whereas the second is dynamic. Since Schattschneider s pioneering work, swings have been used to detect the location of the forces influencing voters behaviour. Stokes s studies in particular have been aimed at measuring where voters stimuli are located. 9 These studies attempted to establish the location of issues and factors that were relevant in the voters choice through variance component models based on the degree of uniformity of swings across regions or constituencies. They hoped to be able to assess whether the relevant factors where situated locally, regionally or nationally. However, as Claggett, Flanigan and Zingale demonstrated, there is no means in analysis of variance-based techniques to distinguish among the types of forces which may cause the nonuniform response by the local units (1984: 83). Whereas it is plausible that uniform swings in a country are caused by national influences (if one excludes the possibility of coincidences), in cases of non-uniform swings it is impossible to 7 On related issues such as electoral cycles see Kayser (2009), Kim and Fording (1998 and 2001), Lebo and Norpoth (2007) and Norpoth (2002 and 2004). 8 Already Schattschneider referred to these two dimensions observing in the US that, on the one hand, the two main parties spread in all states, receive increasingly homogeneous share of votes and with uniform swings across states from one election to the next and, on the other hand, that national issues started to dominate the political agenda replacing local ones (Schattschneider, 1960: 79 93). 9 See Stokes (1965 and 1967) and later publications by Sundquist (1973), Katz (1973) and Butler and Stokes (1974). 6

8 determine at what level the stimulus is located. Since then attempts at measuring the location of forces has been abandoned. 10 In this paper no attempt is made to operationalise the location of forces influencing voters choice: whether European, national or regional. As mentioned, the measurement of this dimension presupposes survey data. It can, however, be considered that parallel waves that take place simultaneously all-over Europe indicate similar concerns and issues. In Schattschneider s words we ought to suspect that something has happened to the political system when we observe that the Republican party gained ground in every state in 1952 and lost ground in forty-five states in 1954, gained ground throughout the country in 1956 and lost ground in nearly every state in 1958 (Schattschneider, 1960: 93). The indicator used here to operationalise uniform swings aims to assess the generality and simultaneity of similar change in voting behaviour. Swings entail a change that covers all or almost all territorial units of a political system (constituencies or regions at the national level; countries at the European level). 10 It is today generally considered that determining the location of forces individual survey data are needed. Studies based on aggregate data tend to underestimate national influences as non-uniform swings, too, can be produced by common national factors to which local units give a differentiated response (see Katz, 1973). In their analysis of British elections Butler and Stokes find results attesting to the presence of local factors modifying the uniforming effects of national influences (on the socalled Butler-Stokes hypothesis see Butler and Stokes, 1974: 143ff). 7

9 Three elements in this indicator are important: A similar change: It is not just measuring change, but change in the same direction; 11 A general change: The similar change must take place everywhere, that is, in all territorial units or almost all territorial units (be it constituencies at the national level or countries at the European level); A simultaneous change: A similar and general change must take place at the same time point in different territorial units. Furthermore, when one speaks of the similarity of change (first bullet point above, two aspects need to be taken into consideration. First, the uniformity in the direction of change in the vote for a party (or a party family) from one election to the next, that is, whether a party (or a party family) increases or decreases its share of votes. This refers to the signs + or. The more similar signs in the various territorial units, the more uniform change. Second, the uniformity in the amount, or size, of change in the vote for a party (or a party family) from one election to the next. One has to establish whether a party (or a party family) increases/decreases its support everywhere (in each territorial unit) with equivalent percentages. Such an analysis obviously presupposes a similar direction of change In this respect it is a different measure from volatility (for example, the Pedersen Index) in which the absolute change is measured disregarding + and signs. 12 A related difficulty which is mentioned further down related to the uniformity of the amount of change is the relationship between amount of change and size of party (or party family) in the different territorial units, be it constituencies at the national level or, as in the present case, countries at the European level. If the size of a party (or party family) is different across territorial units, the amount of change tends to be proportional to the party s share of votes. The uniformity of amount, therefore, manifests itself through a non-uniformity of values (see Butler and Stokes, 1974: ). 8

10 General (everywhere), simultaneous (at the same time) and similar change (same direction and size of vote shifts) must be distinguished from overall trends in which, for example, a party increases its vote nationally or a party family increases its votes at the European level. First, such an increase (or decrease) can be due to one or a few territorial units only (not necessarily all or most). Trends are therefore not general. Second, such an increase (or decrease) can take place through increases in all units but at different times. Trends are therefore not necessarily simultaneous. What swings measure, instead, is the degree to which a similar and simultaneous change takes place everywhere (in few, many or all countries). This is expressed in the percentage of territorial units in which the similar and simultaneous change takes place. If a party family increases its electoral support in the same year in all countries but one or two (say, 90%) one can speak of a uniform swing. The higher the percentage, the more general the swing. Furthermore, the higher the shift of votes, towards or away from, a given party family, the stronger the swing. Therefore, not only it is important to measure the uniformity of the direction of change, but also size. Finally, it should be mentioned that swings have mainly been analysed in the context of two-party systems (swings between Labour and Conservatives in Britain, or between Democrats and Republicans in the US). This paper extends the meaning of swings to the appearance and disappearance of parties, what here is called spread. If in all or most countries new parties of a same family appear simultaneously, this is considered a swing from 0 to a given percentage of votes (positive swing). Similarly, the disappearance is considered a negative swing if it takes place simultaneously in many countries. Armed with this simple but informative indicator the paper aims at establishing the levels of generality in electoral change across Europe and the existence of waves between left and right, and with the rise of green, regionalist and populist parties since

11 Research Design and Data The analysis considers 20 West European countries (including Cyprus and Malta). The time period covers all elections since the first election in the 1970s up to the most recent election in 2008 allowing for time series over almost 40 years. 13 Because of the need of longitudinal analysis to document common shifts of votes, the analysis excludes Central and East European countries that democratised after the break-down of the Berlin Wall in Countries that democratised later such as Greece, Portugal and Spain ( ) are included since those dates. All national legislative elections to the lower houses have been considered. All parties scoring at least 1% nationally at one specific election have been included in the analysis. Smaller parties have been grouped into an other parties category. In total, 211 elections have been analysed. 14 The analysis of the uniformity of electoral swings across countries faces additional problems with respect to swings across constituencies in national elections. First, parties across countries differ whereas across constituencies in one country the parties are the same. In 13 For Cyprus complete results for 1970 and 1976 are missing and have been excluded. In Cyprus the seats of the Turkish community are vacant (24 out of 80). Overseas seats for Denmark (Greenland and Faroe Islands), France, the Netherlands, Portugal have not been considered while for Italy (2006 and 2008) they are included. In Germany Zweitstimmen have been used. Similarly in Italy (1994, 1996 and 2001) votes for the PR-list distribution of seats have been taken. In France votes cast in the first ballot have been considered. Finally, in Ireland and Malta computations have been carried out on first preferences. 14 All data have been newly collected from official sources. Errors and inconsistencies in official data have been corrected with the help of statistical offices and ministries. As for all longitudinal analyses of voting patterns it is important to have information on the transformation of parties over elections. Parties change name and they merge and split. Yet the new units should not always be considered as new parties as this would artificially affect swing levels. A change of name or even ideology does not give rise to a new party. Information to assess such continuity is thus crucial and the dataset is based on careful information about name changes, mergers and splinters. In case of mergers, the new party has been considered the follower of the largest of the two or more previous parties. In case of splinters, the largest of the new parties has been considered the follower of the old party. 10

12 Europe there is no such thing as a pan-european party. For this reason the analysis that follows is based on party families rather than parties. 15 The use of families is necessary if one wishes to aggregate party votes at the supra-national level to see whether or not there is swing in Europe. To create party families, for each election the votes for parties of a same family (say, two liberal parties in the same country at the same election) have been added together. To this end all parties have been classified in 22 party families (plus others) have been created. To allow for ideological change, furthermore, this classification applies to parties at each election. The same party can therefore be classified in one family at election t and in another one at election t+1 (if there has been an important ideological transformation). 16 Second, at the European level there is no common time point. Whereas at the national level all constituencies vote at the same time in a general election, at the European level elections take place in different years in the different countries. Election points across countries are not simultaneous. Since countries hold elections in different years, the election year cannot be used as a time unit for comparative purposes. Periods of time of five years have been used instead as this is the maximum time of a legislature (some countries vote every three or four 15 For a general discussion on the classification of party families see Mair and Mudde (1998) and Zuckerman et al. (2007). The ideological analysis of the parties in this article is based on the definition parties give themselves of their position as collected in a new database on party families. These data have been cross-checked with other classifications such as those from the Comparative Manifestos Project (CMP, see Budge et al., 2001; Klingemann et al., 2006) and available data from various expert surveys such as the Chapel Hill project (Marks et al., 2006; Ray, 2006; Steenbergen and Marks, 2007) and the Benoit, Hunt and Laver project (Laver and Hunt, 1992; Benoit and Laver, 2006). The CMP and expert surveys do not classify all parties but only the major ones (see Data Appendix). 16 Party families are the following: communists, new left, socialists, greens, social liberals, liberals, conservatives, agrarians, nationalists, extreme right, ethnic parties, regionalists, people s parties, Catholics, Protestants, animalists, feminists, pro-alcohol, anti-alcohol, pensioners, personal, Orthodox, anti-europe, Europeans and others. The analysis in this paper, however, is limited to left right and to regionalist, green and populist families. 11

13 years, but none every six years or more). In tables and figures, therefore, the date 1970 refers to , 1975 to , and so on. 17 Evidence of Electoral Waves in Western Europe This section examines the commonalities of change across Europe of electoral behaviour and presents the evidence in three steps. First, the main trends since Second, patterns of spread of families across Europe. Third, the uniformity and simultaneity of change in support which represents the central part of the analysis. Trends The term trend denotes general temporal patterns of increase or decrease of votes for given party families. Contrary to uniform swings trends need not be patterns that take place simultaneously in European countries. Figure 1 displays the trends in voting support since 1970 for two groups of party families. The first group includes the three party families that in recent years have attracted the attention of scholars and media for their novelty and success. The clearest trend is that of the greens which pass from a support of 0% in the early 1970s (they existed nowhere) to 3% and then 4% on average in Europe after Somewhat similarly, the populist right-wing parties pass from roughly 2% to 4 5% on average since 17 While the choice of these time units may seem arbitrary, taking five-year periods withing full decades is intuitive. Results for different time units (for example, , , etc.) do not lead to significantly different results. 18 On average means that in countries in which a family does not exist it is counted as 0%. 12

14 1990. No trend can on the contrary be detected for regionalists (including regionalists strictly speaking, as well as ethnic parties) which remain around 2 3%. [Figure 1 about here] The second group of party families includes the main parties of the left (socialists and communists) and right (conservatives and liberals) as well as confessional and people s parties that, in a number of countries, occupy the ideological space of liberals and/or conservatives (as in Austria, Belgium, Germany or the Netherlands and, before 1992, in Italy). Here one observes a basic stability with an overall decrease of support for these parties from roughly 88% to 80%. This includes a very slight decline for social democrats, communists and liberals and a somewhat larger one for confessional parties. While trends tell us that some families increase or decrease their support, they do not tell us whether this happened everywhere and/or simultaneously, which is the information we need as an indicator of Europeanisation. For this we need to turn to other indicators. Spread Spread denote the appearance across countries of a specific party family. Spread must be distinguished from trends as the latter may be caused by the increase of votes in one or few national parties only, without the appearance of parties in other countries and without any spread. First, spread indicates the appearance of new party families. In Europe there are few new party families. The greens are the clearest case, alongside with very minor families such as feminists, animalists (for the protection of animals), pro- or anti-european parties and pensioners parties. In the past, this would have been the case of communist parties that 13

15 started forming (in most cases as splinters from socialists and social democrats) in the early 1920s almost simultaneously everywhere. Second, we also call spread the appearance across countries of party families that were confined to one or few political systems. In both cases, Europeanisation means the spatial diffusion of party families in political systems in which they previously did not exist. As for trends, however, this needs not be a simultaneous process (which would correspond, for example, to the appearance of a party family in 20 countries at the same time). 19 [Figure 2 about here] If one looks at Figure 2 depicting the number of countries in which party families exist from 1975 to the present we see that to different degrees all three party families are characterised by a process of spread or diffusion across European political systems. The clearest spread is that of green parties. In the five-year period there is only the French Green Party, but soon the number of countries in which a green party exists increases to 11 (in with Austria, Belgium, Denmark, Germany, Italy, Luxembourg, the Netherlands, Sweden and Switzerland) and then to 14 (in with Ireland, Malta, Cyprus and Iceland). 20 In the period there are such parties in 16 out of 20 West European countries (or 85% including also Greece and the UK). 21 In this sense, one can speak of a green wave (spread) and a Europeanisation process of this family in the 1980s. 19 Spread can be considered a type of swing insofar as, with the formation of a party family in a given country, there is a swing from 0% to something. Spread corresponds to measures of increasing territorial coverage (see Caramani, forthcoming). 20 It may be useful to recall that parties are operationalised as existing when they receive at least 1% of the vote nationally. 21 Greens disappear in Denmark, and Luxembourg is not counted for as no election took place before this paper was drafted. 14

16 The regionalist and populist spread take place in the period Both are less significant than the green one. First, both families existed since long. Second, both families already existed in a number of countries, though not many. In regard to extreme right-wing parties in the period there were seven such parties: Austria (Freiheitliche Partei Österreichs), Denmark (Fremskridspartiet), Finland (the Unity Party which however soon disappears), Greece (National Alignment, same as Finland), Italy (Movimento Sociale), Norway (Fremskrittspartiet) and Switzerland (Schweizer Demokraten). 22 Their presence in European countries increases sharply in when they appear in Belgium and France (Front National), Germany (Republikaner), Luxembourg (National Bewegong), the Netherlands (Centrumdemocraten) and Sweden (Ny Demokrati). Soon after, however, in a number of countries these parties disappear as in Luxembourg (where they had appeared in 1985), the Netherlands and Sweden, and later in Germany and Switzerland. Since 2005 such parties exist in nine countries out of 20 in Western Europe (less than 50%). Concerning regionalist parties during the same period, namely , we also have a number of parties that appear across Europe. Whereas such parties existed previously in Belgium, Iceland (sporadically), Spain and the UK, we see such parties appear also in Switzerland and, most notably, in Italy with the Lega Nord. Excluding Iceland and Switzerland, and including the Swedish party in Finland (which is an ethnic rather than a regional party), these are the five cases that over time have remained stable. In reality, however, it is difficult to speak of a regionalist wave in Europe in terms of spread across countries as their presence remains limited to Belgium, Finland, Italy, Spain and the UK (25% of the total number of countries). 22 The Swiss SVP (Schweizerische Volkspartei) is classified as a conservative party. 15

17 Uniform Swings A more specific acceptation of spread stresses the similarity, simultaneity and generality of change. Uniform swings differ from the spread or diffusion across countries as seen in the previous sub-section not so much because they take place between existing party families (as noted, technically spread means a swing from no votes to some votes), but rather because of the simultaneity of change in all territorial units. What swings indicate is similar change that takes place simultaneously in a more or less large number of territorial units. We start by looking at the uniform direction of swings. The Data Appendix at the end of the paper indicates the swings for each party in each country over the entire period of time. On the basis of these data two indicators have been calculated for each party family. First, the number of countries in which a swing is positive or negative (i.e., in which the family wins or loses) as a percentage of the number of countries in which this family exists. 23 The denominator in this case is not the total number of countries (except for the socialist family which exists in all countries). Second, the number of countries in which a swing is positive or negative as a percentage of all countries in which democratic elections took place. 24 In this second case, a family which does not exist and does not appear in the subsequent election is considered zero swing (from nothing to nothing). 23 The minus sign indicates a negative swing (a party family loses votes). 24 The number of countries is 20 with the exception of and (Cyprus missing) and (Luxembourg missing). In calculating percentages countries have not been weighted by size of their population. For a procedure along these lines applied to ideological shifts, see Kim and Fording (1998 and 2001). 16

18 For both indicators, a family that does not exist in a country and then appears is considered a positive swing; a family that exists and disappear in the next election is considered a negative swing. [Figure 3 about here] Figures 3 and 4 are based on the first indicator and give the percentages of countries in which either a positive or negative simultaneous swing exists for each time point since The further the curves are apart, the more general the simultaneity of change in Europe. The clearest examples of uniform direction of swing are those displayed by greens. Twice in and all green parties (100%) win votes with respect to the previous fiveyear period. The solid curve indicating positive swings is at the top of the graph, while the dashed line indicating negative swings is at 0%, meaning that for those two periods no green party in Europe lost votes. It is obviously rare that all parties of a same family win or lose simultaneously in all countries. For the greens, however, we see that this happens twice and that also in and the uniformity of swing is very high (around 85%), meaning that only the greens of few countries display a change against the main stream. On the contrary, in one sees that 50% of the greens lose votes and 50% win votes indicating a perfect non-uniformity in that no direction, either positive or negative, prevails in Europe. The uniform swing for the greens is clearly one that takes place in the 1980s, in total four clear time points with uniform swings. As far as right-wing populist parties are concerned, there are two moments of uniform swings since In and we have a first wave. Here in 75% of the countries these parties make a win as the solid line indicates, whereas in 25% they lose. While this is 17

19 not a perfectly uniform swing as in the case of the greens, it shows a favourable moment for these parties across Europe albeit, as mentioned, not everywhere. This is also the moment at which the number of such parties across Europe increase: they exist in 12 countries in as Figure 2 above shows. A third clear moment of uniformity towards positive swings for populist parties is : around 85% of these parties (where they exist) make a win, while 15% make a loss. As far as regionalist parties are concerned, there are two moments of clear negative uniform swings. First, in when all four parties lose votes simultaneously (in Belgium, Spain and the UK, as well as in Iceland where the party disappears, counting as negative swing). Second, in when three out of four families lose votes (only in Italy the regionalist parties slightly increase their votes with respect to the previous period). Consistent with what much of the literature has stressed is the rather uniform positive swing in when 75% of regionalist parties in Europe increase their electoral support. Four out of five parties win, most significantly in Italy but also in Switzerland where the Lega dei Ticinesi appears, and with the exception of Spanish regionalists that overall lose votes. If one takes the total number of time points for which there is a uniform swing (i.e., a similar direction of change in at least 75% of the countries) for the three families together, there are 10 such points out of 21 (seven time points for each of the three families). 25 Based on regionalists, greens and populists, therefore, it is difficult to say that Europe moves together with uniform and simultaneous changes of votes since the 1970s. These results would rather point to the fact that in most cases it is national specificities that predominate. 18

20 [Figure 4 about here] A similar conclusion is suggested by looking at the swings of the main parties of the left right dimension: communists, socialists, liberals and conservatives (Figure 4). Out of a total of 28 time points, at one point only there is a uniform swing (as operationalised in at least 75% of countries moving in the same direction), namely when conservatives lose almost everywhere in Results are the same if liberals and conservatives (as well as religious parties) are considered together as a right wing family and if socialists and communists are considered together as a left wing family. Looking at Figure 4 reveals that clear uniform swings in the same direction are rare for the major parties of the left right dimension. The family that displays the least of them is the liberal one. Between 14 and 18 such parties exist in the 20 countries considered in this paper depending on the period of time. The pattern is one of hardly more than half of them moving in the same direction. This means that no swing whatsoever can be seen at the European level as far as the liberal family is concerned. As mentioned, the conservative parties across Europe lose almost everywhere (in 10 out of the 13 countries in which they existed at the time) in the second half of the 1980s. Apart from that, one sees some degree of uniformity in and with 10 and 9 parties, respectively (out of 14) increasing their support (about 70% of uniformity). As far as the socialists are concerned, there are two negative periods in which about 70% of the parties in the 20 countries lose: and Other than that no clear swings are present. Finally, most of the communist parties (which exist on average in 14 countries over the different periods of time) lose votes between 1980 and 1994 (around 70% of them), win in and lose again in As mentioned, among these four families and for seven periods of time, in only one case there is a uniformity above 75% (that is, three quarters of 25 The threshold of 75% has been chosen as it represents the middle point between perfect uniformity 19

21 the parties in the various countries moving in the same direction) indicating the absence of common waves of change along the left right dimension since Figure 5 displays the same information based on the second indicator, namely the number of parties with a homogenous swing, but this time as a percentage of the total number of countries (including those in which the family does not exist). This information allows one to see that in a number of cases positive swings are in reality due to the appearance of parties in new countries: when positive swings (solid lines) correspond to the decline of the dotted line indicating in how many countries a family does not exist. This is the case of greens and populists. Similarly, several negative swings are due to the disappearance of parties in a number of countries, as in the case of liberals and communists (the dashed line increases as well as the dotted one). Only when the dotted line remains flat, are we in the presence of swings between existing parties. 26 As one sees, however, this is rarely the case. Among these events is the rather uniform decline of communists in the 1980s, and the swing against conservatives in the second half of the 1980s and then the positive one in For the rest it appears that, rather than swings we are in the presence of episodes of spread with parties of a specific family appearing and diffusing in new territories. [Figure 5 about here] (100%) and perfect non-uniformity (50%). 26 With respect of the first indicator, the denominator is here larger. While in Figure 4 the percentages of uniform negative and positive swings add up to 100%, here they do not as in a number of countries the family does not exist. A third curve has therefore been inserted (corresponding to those of Figure 3, but in percentages) indicating the percentage of countries in which nothing has changed from one period to the next. A zero swing corresponds to a family that exists nowhere and does not appear in the subsequent period. The only graph that corresponds to that in Figure 4 is the one for socialists, since this family exists in all countries. As before, if a party appears it is considered a positive swing (from 0 to something), if it disappears it is considered a negative swing (from something to 0). 20

22 After having seen the uniformity of the direction of swings for families across European countries, we now turn to the uniformity of the amount of the shift of votes and look at the uniformity of the size of swings. It makes sense, however, to do this only for families and time points in which a uniform direction has been detected. This limits the analysis to the following observations: 27 The green positive swings in , and The regionalist negative swings of and , as well as the positive swing of The populist positive swings in , and The conservative negative swings of In total these make up for 10 observations (excluding the green wave because of a small number of cases). Furthermore, to quantify the uniformity of the amount of swing the standard deviation of the swing for individual families across countries has been calculated: how large is the variation in the gains or losses of a family s votes across countries at a given time point? This measure is calculated for the countries in which the family actually exists. 28 [Figure 6 about here] 27 As before, the analysis includes only swings in which there is uniformity of direction in at least 75% of parties of a same family. 28 The appearance of a party of a given family in a given country is considered a positive swing (by the amount of votes in percentage the party gets at its first election) and the disappearance is considered a negative swing (by the amount of votes in percentages the party had at its last election before disappearing): 0 (at election t ) votes% (at election t 1 ). 21

23 Looking at Figure 6 and Table 1, two features of the size of swings can be stressed. The first feature is that with few exceptions swings are not large ones. Figure 6 indicates parties in single countries with larger swings with bold lines (families with small swings have been designated with very thin lines to avoid jamming the graphs). As seen above, the greens display the larger uniformity of direction (especially in ). From the graph in Figure 6 one sees, however, that in spite of a general growth all over Europe, this growth is very limited in size. It is an increase of 3.20% in , 2.70% in and 2.44% in The single maximum gain of votes for a single party is 7.81% in France in the period Also the regionalists display uniform directions of change in three occasions (100% in , 83% in and 80% in ). Besides being based on very few parties, these swings are again very small ones (two negative ones: 1.52% and 1.70% on average; and one positive one: 3.36% on average). The single major increase is that of the Lega Nord in Italy at the beginning of the 1990s with an increase of 8.65%. The largest swings are clearly those of the right-wing populist parties in , and Here we find on average increases of votes for these parties between 4.03% and 5.21%, the single largest increase being that of the Austrian populist right (FPÖ and BZÖ) in after a dramatic drop in the previous period. Other large increases are those of the French right (mainly the Front National) with 9.83% in , and of Italy s right in (8.10%). Finally, as one sees from both the figure and the table, the uniform negative swing of the conservatives in is also minimal with an average loss of votes of 1.25% (the maximum loss was 2.67% in Portugal). [Table 1 about here] 22

24 The second feature to be stressed besides the small size of swings is that these swings are not uniform as far as their size is concerned. The non-uniformity of the size of swings appears visually in Figure 6. Among the ten cases of uniform directions (operationalised as at least 75% of parties of a same family moving in the same direction), some increase/decrease their votes much more than others. The standard deviation (as well as the range between minimum and maximum swing) in Table 1 summarises these differences in the amount of swing between parties of a same family. 29 In most cases, the standard deviation is low because of the small size of swings in the first place (for the three populist swings it is larger). One sees large differences in the amount of change between populist parties in in Austria, Belgium, Denmark, Italy, Germany, the Netherlands, Norway and Sweden. 30 The three swings for the Greens (1980, 1985 and 1990) are much more homogenous as far as the size of change is concerned. The same applies to the two negative swings for regionalists in 1980 and The positive regionalist swing of 1990 is more diverse because of the big increase of the Lega Nord in Italy. To summarise, both the actual size and the degree of uniformity of swings strengthen the indications received from the analysis of the direction of swing, namely that common simultaneous swings across Europe are few and limited in size mainly those of greens and populists. Conclusion Electoral waves in general and the uniformity of electoral swings across territorial units more specifically be it constituencies at the national level or countries at the European level is an important indicator attesting for the degree of integration and formation of electorates and 29 For each of the ten uniform swings (direction) only parties moving in the same direction have been taken to calculate the standard deviation. 30 The only big drop in votes for a right-wing party takes place in France in this period. 23

25 party systems. This paper has explored for the first time evidence of common simultaneous swings across Europe from 1970 to the present for 20 countries, adapting concepts and methods that previous literature has applied to the national level. The previous analysis has shown that waves of change do take place, but that on the other hand only few uniform swings occurred in Europe since 1970 and that they have been marginal phenomena limited in size. First, no swings of a given uniformity (i.e., taking place in at least 75% of the 20 West European countries analysed) take place along the left right dimension that is for four main families: communists and socialists on the one hand, and liberals and conservatives on the other. Wins and losses of votes for these families vary from country to country and only in very few cases it is possible to detect change in at least two thirds of the countries at the same time and in the same direction. In most cases, change between elections is country-specific and does not give rise to Europeanised patterns which could attest for common shifting preferences or overarching issues. Second, a given number of swings have taken place mainly for party families that have developed as spread since The green family is the clearest example of change across Europe with time points (between 1980 and 1994) in which in all countries (where such parties exist) greens have spread and made a gain of votes simultaneously. The populist family provides another example, in and, above all, after As far as the regionalists are concerned, the evidence is much less robust. These three families are those that the literature had identified as the main cases of novelty and change of West European party systems. In spite of the evidence on the uniformity of direction, the size of such swings is not large (the largest is that for populist parties) and not uniform. Furthermore, the analysis of the number of countries in which parties of a same family exist shows that the uniformity of positive and negative swings is strongly associated with spread, that is the appearance and disappearance of parties. 24

26 Classical studies at the national level from Schattschneider s 1960 book on the US to Stokes comparisons of the US and the UK (1965 and 1967; see also Butler and Stokes, 1974) have demonstrated the importance of common national shifts of votes across territorial units for the formation of a national electorate that responds nationally to common themes and issues. These are indicators not only of electoral integration, but also of systemwide responsiveness and accountability. The indications that the previous analysis give, however, point out that a European-wide electorate responding Europeanly to common supra-national themes and issues has not yet developed. Responsiveness and accountability seem, on the contrary, to remain anchored in specific national contexts explaining the absence, in most cases, of European-wide patterns of swings across national electorates and party systems. To some extent, these results lead to amend the conclusions reached on the basis of different indicators, such as those of a progressive convergence and homogenisation of voting behaviour in Europe since 1970 (Caramani, forthcoming). Such a homogenisation does indeed take place, but the causes of it are unlikely to be Europeanwide supra-national factors but rather national ones. 25

27 References Alemán, E. and M. Kellam, The nationalization of electoral change in the Americas. Electoral Studies 27: Bartolini, S., Restructuring Europe: Centre Formation, System Building, and Political Structuring Between the Nation State and the European Union. Oxford: Oxford University Press. Bartolini, S. and P. Mair, Identity, Competition and Electoral Availability: The Stabilisation of European Electorates, Cambridge: Cambridge University Press. Bochsler, D., The nationalisation of political parties: A triangle model, applied on the Central and Eastern European countries. Central European University Political Science Journal 4: Butler, D. and D. Stokes, Political Change in Britain: The Evolution of Electoral Choice. London: Macmillan (second edition). Caramani, D., The Nationalization of Politics: The Formation of National Electorates and Party Systems in Western Europe. Cambridge: Cambridge University Press. Caramani, D., Is There a Euorpean Electorates and What Does It Look Like? Evidence from Electoral Volatility Measures, West European Politics 29: Caramani, D., forthcoming. The Europeanisation of Electoral Politics: An Analysis of Converging Voting Distribution sin 30 European Party Systems, Party Politics. Chhibber, P. and K. Kollman, The Formation of National Party Systems: Federalism and Party Competition in Canada, Great Britain, India, and the United States. Princeton, N.J.: Princeton University Press. Claggett, W., Flanigan, W. and N. Zingale, Nationalization of the American electorate. American Political Science Review 78: Croissant, A. and T. Schächter, Die Nationalisierung politischer Parteien und Paretiensysteme in asiatischen Neo-Demokratien. Politische Vierteljahresschrift 49: Hicken, A., Building Party Systems in Developing Democracies. Cambridge: Cambridge University Press. Hix, S., Noury, A. and G. Roland, Democratic Politics in the European Parliament. Cambridge: Cambridge University Press. Inglehart, R., Culture Shift in Advanced Industrial Societies. Princeton, N.J.: Princeton University Press. Jerôme, B., Jerôme-Speziari, V. and M. Lewis-Beck, Partisan Dynamics and the European Nation. Paper presented at the Midwest Political Science Association, Chicago. Kasuya, Y. and J. Moenius, The nationalization of party systems: Conceptual issues and alternative district-focused measures. Electoral Studies 27:

28 Katz, R., The attribution of variance in electoral returns: An alternative measurement technique. American Political Science Review 67: Keating, M., The New Regionalism in Western Europe: Territorial Restructuring and Political Change. Cheltenham: Elgar. Keating, M., Regions and Regionalism in Europe. Cheltenham: Elgar. Kim, H. and R. Fording, Voter Ideology in Western Democracies, European Journal of Political Research 33(1): Kim, H. and R. Fording, Voter Ideology, the economy, and the International Environment in Western Democracies, Political Behavior 23(1): Kitschelt, H., The Radical Right in Western Europe: A Comparative Analysis. Ann Arbor: University of Michigan Press. Kriesi, H. et al., West European Politics in the Age of Globalization. Cambridge: Cambridge University Press. Lane, J.-E. and S. Ersson, Party system instability in Europe: Persistent differences in volatility between West and East. Democratization 14: Lebo, M. and H. Norpoth, The PM and the Pendulum: Dynamic Forecasting of British Elections. British Journal of Political Science 37: Lewis-Beck, M. and M. Paldam, Economic Voting: an Introduction. Electoral Studies 19(2 3): Lewis-Beck, M. and M. Stegmaier, Economic Models of Voting. In Dalton, R. and H.-D. Klingemann (eds.), The Oxford Handbook of Political Behaviour. Oxford: Oxford University Press (pp ). Marks, G., and M. Steenbergen, Understanding political contestation in the European Union. Comparative Political Studies 35: Norpoth, H., On a Short Leash: Term Limits and the Economic Voter. In Dorussen, H. and M. Taylor (eds.), The Context of Economic Voting. London: Routledge (pp ). Norpoth, H., Forecasting British Elections: A Dynamic Perspective. Electoral Studies 23: Poguntke, T., Green Parties in National Governments. London: Cass. Powell, B. Jr. and G. Whitten, Cross-National Analysis of Economic Voting: Taking Account of the Political Context. American Journal of Political Science 37: Schattschneider, E., The Semisovereign People: A Realist s View of Democracy in America. New York: Holt, Rinehart, and Winston. Stokes, D., A variance component model of political effects. In Claunch, J. (ed.), Mathematics Applications in Political Science. Dallas: The Arnold Foundation, pp Stokes, D., Parties and the nationalization of electoral forces. In Chambers, W. and W. Burnham (eds.), The American Party System: Stages of Political Development. Oxford: Oxford University Press, pp Sundquist, J., Dynamics of the Party System: Alignment and Realignment of Political Parties in the United States. Washington, D.C.: The Brooking Institution. 27

29 Tavits, M., On the linkage between electoral volatility and party system instability in Central and Eastern Europe. European Journal of Political Research 47: Zuckerman, A., Dasovic, J. and J. Fitzgerald (2007). Partisan Families: The Social Logic of Bounded Partisanship in Germany and Britain. Cambridge: Cambridge University Press. 28

30 Data Appendix: Swings by Party Family in Europe Communists Countries Austria Belgium Cyprus Denmark Finland France Germany Greece Iceland Ireland Italy Luxembourg Netherlands Norway Portugal Spain Switzerland Conservatives Countries Cyprus Denmark Finland France Greece Iceland Ireland Italy Norway Portugal Spain Sweden Switzerland United Kingdom

31 Extreme right-wing parties Countries Austria Belgium Denmark Finland France Germany Greece Iceland Italy Luxembourg Netherlands Norway Sweden Switzerland Greens Countries Austria Belgium Cyprus Denmark Finland France Germany Greece Iceland Ireland Italy Luxembourg Malta Netherlands Sweden Switzerland United Kingdom

32 Liberals Countries Austria Belgium Cyprus Denmark Finland France Germany Greece Iceland Ireland Italy Luxembourg Netherlands Norway Portugal Spain Sweden Switzerland United Kingdom Regionalists Countries Belgium Iceland Italy Spain Switzerland United Kingdom

33 Socialists Countries Austria Belgium Cyprus Denmark Finland France Germany Greece Iceland Ireland Italy Luxembourg Malta Netherlands Norway Portugal Spain Sweden Switzerland United Kingdom

34 Tables and Figures 33

35 Figure 1. Percentage of votes for party families: (five-year periods averages) Note: The bottom-up rank order of families in the legend is the same of that of the patterned areas in the graphs. Regionalist are here depicted separate from ethnic parties even if in the analysis they are considered together.