Identifying Europe ?

The role and dynamics of a european cleavage

Robert Andersen

Jocelyn A.J. Evans1

Introduction

According to Lipset and Rokkan (1967) social cleavages based on social class, religion and linguistic group are typically seen as the foundation for both party presence and electoral support of these parties. These relationships have a long history, beginning with universal suffrage and the consolidation of the democratic acquis in individual countries. Nonetheless, recent research suggests that traditional cleavage structures are in decline (Dalton, Flanagan and Beck, 1984; Franklin, Mackie and Valen, 1991); this is especially the case for the class/party relationship (Clark et al, 1998; Evans, 1999). Some argue that a decline in traditional cleavage voting is associated with a rise in post-industrial values, whereby voters have become decreasingly concerned with group interests and more concerned with individualistic issues (Inglehart, 1987,1990; Beck, 1992). There has been little empirical research, however, assessing whether new cleavages are also developing.

At the same time as the alleged rise in individualism the European Union has become progressively implanted. With the decline in traditional cleavages, it is of interest whether a new cleavage based on European sentiments is developing. As a challenge to the systemic status quo ante of nation-state governance, which mobilises parties on pro- and anti- stances, and which has both material and symbolic implications for members of the electorate, there is no reason to expect that such a cleavage could not co-exist with increasing political individualisation. In fact, a number of authors have posited the emergence of a pro-/anti-European dimension amongst political parties in contemporary EU democracies (e.g., Benoit, 1997; Taggart, 1998), as well as potential members and neighbours (Valen, 1976; Ringdal and Valen, 1998; Kopecky and Mudde, 2002). But do party differences in stances on the European issue indicate the presence of a European cleavage ?

There is ample evidence that suggests attitudes toward European integration—among both elites and masses—vary considerably according to country and time (Eichenberg and Dalton, 1993; Gabel, 1998; Anderson, 1998; Janssen, 1991). Concomitantly, there are also numerous differences across countries in terms of party stances on Europe (Flood, 2002; Szczerbiak and Taggart, 2002). Although attitudes toward Europe appear to cut-across traditional political orientations, such as the Left-Right continuum (Hix, 1994), there is little evidence that the European issue is realigning party systems and voters (Mair, 2000). According to Marks and Wilson (2000), voters use existing political cleavages as a ‘prism’ through which they view new issues, including Europe.

Nonetheless, previous research employing a cleavage approach has concentrated on the role of traditional cleavages—and the party groups associated with them—in orienting elites to the European question (Valen, 1976; Marks and Wilson, 2000). There has been little research on whether a European cleavage is developing separately in response to a dealignment or realignment of traditional cleavages. The present paper takes a political sociology approach to explore for evidence of a new political cleavage based on Europe in five countries : Belgium, France, Italy, Luxembourg and the Netherlands.2 Our analysis uses survey data from the Mannheim Eurobarometer Trend file. These data represent individuals over many years, allowing us to assess the extent to which parties that have explicitly mobilised in opposition to Europe have been successful, and whether this success is increasing.

Cleavage structure in European democracies

Lipset and Rokkan’s theory of party systems and voter alignments is predicated upon the nation-state building and democratization processes which led self-identifying social groups to form out of opposition to, and then integration with, an encroaching state that challenged territorial, cultural and economic loyalties (Lipset and Rokkan, 1967; see also Evans, 2004; Flora 1999). The development of these cleavages depended on the loyalties and thus identity—or, to employ the Parsonian term, ‘orientation’—of individuals within social sub-systems (Parsons and Smelser, 1956). In most continental European nations—including the five countries of interest in this article—the National and Industrial Revolutions are considered the ‘critical junctures’ for traditional political cleavage formation and party system alignment.

In simple terms, from the National Revolutions came the opposition between Catholic (later Christian Democrat) and secular Liberal parties, the former principally representing the haute and petty bourgeoisie and rural sectors, the latter urban and professional and proto-professional groups. The Industrial Revolutions, on the other hand, led to the incorporation of the working class, leading to the formation of Communist, Socialist and subsequently Social Democratic parties, which in the terms of Duverger’s ‘sinistrism’ (Duverger, 1951), assumed the Liberal parties’ left flank, and thus squeezed the Liberals into a centrist position. Each country has its own contextual specificities, but these three party types— Socialist/Social Democratic Left, Liberal Centre,3 Christian Democrat Right (or Conservative)—generally continue to dominate in most continental European countries, including those we examine.

The importance and longevity of these traditional cleavages has made the threshold high for the identification of any new cleavage (Marks and Wilson, 2000). According to Bartolini and Mair (1990 : 215), three criteria must be satisfied to define a political cleavage : (1) there must be a clearly identifiable group; (2) the group must be a self-perceived group; and (3) there must be explicit mobilisation of this group by a political party or actor. Potential contenders have been posited—such as the public-private division (Dunleavy, 1979; Manza and Brooks, 1999; Evans, 2005) and “new class” divisions developing among professionals (Brint, 1984)—but these are better seen as divisions within old cleavage groups rather than as new cleavages that threaten the old cleavage structure (see, for example Weakliem, 1991). Moreover, although so-called ‘post-material issues’ such as the environment have become more important (Inglehart, 1977), they are better seen as absorbed within the traditional Left-Right continuum, rather than as reflecting new political cleavages (Knutsen, 1995 : 86). Even radical right parties tend to have social bases of support that divide along traditional class and religious lines (see Andersen and Evans, 2003), and thus do not reflect new divisions in society.

Although Bartolini and Mair’s criteria are clearly applicable historically, much of this may be due to a retrospective historical clarity which no more pertained at the time than it does in the contemporary period—for example, delineating somewhat simplistically a homogeneous ‘working class’ or a distinct linguistic group. More importantly, from the traditional cleavage perspective, a key element missing for the establishment of cleavages based on most new issues is the presence of a critical juncture. Essentially, the Parsonian dilemmas of orientation can be stretched to delineate individuals’ loyalties to ‘old’and ‘new’ and their slow division and implantation according to such developed identities. Given the sovereign and institutional challenge of the EU to national polities—a concept which retrospectively could be subsumed within the notion of ‘critical juncture’—a potential divide exists between those with loyalty to the old nation-state and others looking to the supranational arena.

The European issue as nascent cleavage

Although there is no reason not to expect Europe to be the basis for a strong cleavage in the future, most commentators do not think it is so today. The following comment by Mair perhaps sums up conventional wisdom : ‘To the simple question of whether Europe has had a direct impact on the format of national party systems, the equally simply answer must be an unequivocal “no”’ (Mair, 2000 : 31). The reasons for this lack of apparent effect are numerous : lack of knowledge about European matters amongst the electorate; lack of interest in European politics; a democratic deficit that, despite the growing power of the European Parliament, distances policy-making from the electoral arena. In addition, anti-European parties also commonly oppose other salient elements of the national political system. In other words, these parties typically include Europe as one of a series of protests or objections to the system, rather than focus mostly on Europe (Taggart, 1998; Evans, 2003).

From the standpoint of electoral alignment, however, and the extent to which individuals no longer vote according to their social structural position, there is evidence of changing electoral dynamics that include the Europe issue (Belot and Cautrès, 2004 : 141). As policy-domains increasingly come under European as well as domestic auspices, together with governance more generally under the terms of such agreements as the Maastricht and Nice Treaties and the still vulnerable EU Constitution, the EU presents a challenge to the old nation-state political systems. The argument here is not over democratic deficit, or whether the EU wishes to subsume nations in a federal super-state, but rather that it is a challenge to existing national institutional frameworks, and is perceived as such by supporters and opponents alike.

Looking at the parties which adopt a consistently Eurosceptic position, the two main families which emerge are the Marxist-Communist Extreme Left and Extreme Right. In both cases, there is a rejection of the European Union although for very different ideological reasons (Evans, 2000). The Extreme Left rejects a capitalist supranational arena derived from a free trade area which disadvantages workers and those insecure socio-economic positions. The Extreme Right, similarly disenchanted by capitalist aspects, but precisely for their internationalist tenor, promotes the nation-state and notions of ethno-cultural separatism, if not superiority, as the basis for a successful and stable socio-political system.

While the ideological bases may differ, the notion of shifting loyalty is similar for both the Extreme Left and the Extreme Right. Pro-Europeans accept a new socio-political arena that partially or wholly replaces the nation-state. Anti-Europeans, whether of the Extreme Left or Extreme Right, reject this notion, promoting instead either an internationalist collective or a reactionary nationalist or sub-nationalist unit.4 It would be a mistake, however, to see Extreme Left Marxist-Communist Parties as completely rejecting the nation-state. In the post-war period, all such parties have identified the nation to a greater or less extent as a key source of promoting workers’ interests, and continue to do so in opposition to a capitalist European agenda. To the extent that commentators distinguish between pro- and anti-European parties, and the concomitant pro- and anti-system stances held by both, it is evident that the pro-European parties promote the European arena as a possible focus of identification as a bloc, rather than a single party.

For the cases of interest, namely Belgium, France, Italy, the Netherlands and Luxembourg, reaction to the European project differs widely between countries. In terms of critical juncture, unlike the National and Industrial Revolutions, the ‘European Revolution’, although in many ways as arbitrary a concept as its predecessors, has at least developed at an identical rate in all member nations to the extent that the Treaties of Rome, Maastricht and Nice have all been adopted and ratified at approximately similar times, with similar demands (if not outcomes) in all countries. Nonetheless, the precise nature of opposition to European integration varies according to traditional political structuration, national context, leadership and other political issues.

Of the five nations we examine, France is undoubtedly the country displaying the most wide-ranging level of Euroscepticism across time, both for the historically powerful Communist Party and more lately its Trotskyite micro-rivals Lutte Ouvrière and the Ligue Communiste Révolutionnaire on the Extreme Left, and on the Extreme Right for the Front National and Mouvement National Républicain. France also has elements of the Gaullist Right and reactionary Republican Left, most notably the Mouvement des Citoyens, which show anti-EU sentiment. Since the 1980s, there has been a growing tendency for such elements of the mainstream Right to separate on the grounds of retaining national sovereignty, for instance under Philippe De Villiers and Charles Pasqua’s Mouvement pour la France and Rassemblement pour la France, respectively. We might expect, then, a transfer of traditionally mainstream Right voters to these souverainiste anti-European parties. At the other end of the spectrum, Belgium and Luxembourg have displayed perhaps the lowest levels of opposition to Europe. In the former, Deschouwer and van Assche (2002) identify only Vlaams Blok as a Eurosceptic party. They note that, in comparison to other Eurosceptic parties across Europe, however, even Vlaams Blok should be seen as relatively soft in its opposition, not least because its regionalist stance on Flanders’ independence is seen as compatible with and even facilitated by the European project. This may be usefully contrasted with, say, the Lega Nord in recent years in Italy which, while regionalist, is outspoken in its condemnation of Europe integration (Conti, 2002). In the past, its regionalist aspirations led it to support the European integration process. Nevertheless, its increasingly extremist and anti-system position has led it—in line with other Extreme Right parties—to adopt an anti-European position.5 Leaving aside the Europe issue, there is evidence that traditional cleavages have some impact on anti-EU voting in most European countries. Perhaps most importantly, the importance of social class as a predictor of EU support is well-established. Individuals tend to favour EU integration when they are likely to derive economic benefit from it (Gabel and Palmer, 1995; Gabel, 1998). For example, individuals in high-level management positions tend to be more likely to favour integration than blue-collar workers who fear that they will fall prey to the lack of existing European welfare provisions or lose out to cheap immigrant labour. We might also expect individuals in higher strata in the private sector to be more positive towards integration than routine clerical staff in short-term contracted, non-autonomous ‘junk jobs’ (Esping-Andersen, 1990 : 215) with less incentive to welcome the vagaries of market competition.

Previous research also suggests that religion is related to support for European integration. Marks and Wilson (2000 : 438) argue that the principal religious divide on Europe lies between Catholics and Protestants, the former being pro-European given the Catholic Church’s own supranational agenda and the historical legacy of Catholic parties’ fights against the secular nation-state building process.6 Conversely, the Protestant faiths are more likely to oppose Europe given their more introspective, national independent status. On the other hand, research on conservative values and authoritarianism, which display strong association with pro- / anti-European feeling (Heath et al, 1999; Grunberg and Schweisguth, 2003), suggests that Catholicism is linked to more rejectionist stances on progressive political agendas than are secular voters who are characterised by more universalist values (Andersen and Evans, 2004).

The literature on the nature of Euroscepticism and any role it may play across time in voting alignments is remarkably thin. Consequently, the analysis that follows is largely exploratory, in that we have no strong preconceptions about how the countries will differ. We begin by assessing changes in attitudes towards Europe and European identity over time, and whether there has been any associated change in support for anti-EU parties over time. We then model support for anti-EU parties, with the goal of determining the impact of anti-EU sentiment and European identity. Finally, we test whether the impact of anti-EU sentiment and European identity on support for these parties has changed over time. In other words, we look for evidence of a new European cleavage that relates European sentiments to parties with anti-EU platforms.

Data and method

Our data are from the Mannheim Eurobarometer Trend File, which contains many of the most important trend questions of the Eurobarometer surveys conducted between 1970 and 1999 (Schmitt and Scholz, 2001). In total this amounts to survey data from 877,223 respondents living in the 15 EU countries and Norway. The data for each country are from multistage probability samples. For the present analyses, however, we focus only on respondents from Belgium, France, Italy, Luxembourg and the Netherlands. We limit our analysis to these five countries for two reasons : (1) they are all original members of the European Community, so if the European enterprise is to have had any influence on political cleavages, we might expect to find it most visibly here, (2) data on other countries are too incomplete in order to be able to carry out an extensive over time analysis.

Dependent Variable

The dependent variable is intended vote in the next national legislative election. We divide this variable into a binary variable : support for Eurosceptic parties is coded 1; support for all other parties and nonvoters are coded 0. Division of parties into these categories was based on Szczerbiak and Taggart’s Eurosceptic classifications (Szczerbiak and Taggart, 2002 : 18-19). Table 1 presents the parties coded as Eurosceptic for each country.

Independent Variables

We are primarily interested in two independent variables :
attitudes toward European integration and European identity. Attitudes toward European integration were tapped with the following question (codes in parentheses): “In general, are you for or against efforts being made to unify Western Europe ?” Responses to the question were : Very much for (coded 1); To some extent for (coded 2); To some extent against (coded 3); Very much against (coded 4). European identity was measured with the following question : “In the near future do you see yourself as : (1) <respondent’s nationality> only; (2) <respondent’s nationality> and European; (3) European and <respondent’s nationality>; (4) European only.” We recoded this variable into two categories : European (responses 2 - 4) versus not European (response 1).

Control Variables

In order to avoid making spurious inferences due to confounding variables, our main analyses control for several variables thought to be important to vote in these countries. Most importantly, we control for the traditional cleavages of social class and religion. We use a five-category social class variable adapted from Erikson and Goldthorpe (1992), with the following categories : (1) unskilled manual labourers; (2) skilled manual labourers; (3) routine nonmanual workers; (4) self-employed; (5) professionals and managers (see Andersen and Heath, 2002). Given that the five countries in our analysis are predominantly Catholic, with only the mixed case of the Netherlands, the Catholic/Protestant divide is of less interest than the religious/nonreligious divide. As a result, our models control for religious attendance, which is divided into three categories : (1) at least once a week, (2) occasionally, and (3) never. The models also control for age and gender. Finally, year is included in all of the models, both as a control and as a variable of interest with respect to how it interacts with attitudes towards Europe and European identity in its effect of support for the anti-EU parties.

Methods

We start with a descriptive analysis of the patterns in attitudes towards European integration, European identity and support for anti-EU parties over time for each country. In the case of attitudes towards European integration and support for anti-EU parties we have data spanning the complete range of the study, thus giving us enough time periods to smooth the trend over time using nonparametric regression.7 Since the European identity question was asked only since 1992, there are not enough time periods to effectively use nonparametric regression (nor is it necessary to comprehend the trend). As a result, for this variable we simply graph the proportion of respondents in each country who have a European identity at each survey year.

Our main analysis consists of two separate binary logit models for each country. The first set (Model 1) explores the effects attitudes towards Europe on support for anti-EU parties from 1974-1996. The second set of models (Model 2) assesses the effect of European identity on support for anti-EU parties from 1992-1998. We also include terms to test for an interaction in the effects of these predictor variables and time in both models. Preliminary analysis suggested that year should be included in both models as a categorical variable. In other words, the year-to-year variation in support for the anti-EU parties could not be adequately captured with a linear trend or a simple polynomial. Model 1 includes all of the control variables discussed earlier. For Model 2, religiosity was necessarily excluded because it was not present for all years, but all of the other control variables are included in the model. Due to the small number of respondents who supported anti-EU parties in Luxembourg, that country is necessarily excluded from the logit model analysis because the models would be extremely unstable and not trustworthy.

Results

Trends in EU attitudes, European identity, and anti-EU party support

Figure 1 shows the results from the simple nonparametric regressions of attitudes towards integration on time. Recall that high scores indicate negative attitudes towards European integration. In all countries, except Italy, there is a similar pattern whereby average attitudes remained fairly stable until about 1985—the year in which the principles of the Single European Act were established and the Iberian peninsula expansion ratified—and then become increasingly more pro-Europe afterwards. On the other hand, in Italy attitudes remained fairly stable throughout the period under study. It should also be noted that the Italian average is consistently higher (i.e., more anti-Europe) than it is for all other countries. Despite the unequivocally strong pro-European line adopted by the Italian pentapartito governments, this stable level of Euroscepticism betrays the strong Extreme Left and Extreme Right antipathy to the EC / EU across time for the ideological reasons outlined earlier. The decline post-1990 corresponds to the transformation of the PCI into the PDS and its adoption of social democratic rather than socialist / communist principles.

The findings in Figure 1 are mirrored in Figure 2, which shows the percent in each country claiming a European identity from 1992 to 1998. Although we would ideally explore responses to this variable over a longer period of time, the question was not asked in the surveys until 1992. It is difficult to see any trend in European identity over time in any of the countries, but differences across countries are apparent. For example, in both Figure 1 and Figure 2 we see that those in the Netherlands and Belgium tend to identify with the European community than those from other countries. It is tempting to see support for Europe in Belgium and the Netherlands as reflecting the disproportionate advantage such small countries receive from the EU, particularly at the international level. Luxembourg provides a counter-example, however, showing remarkably low levels of European identification despite its position at the heart of Europe.

With the exception of perhaps Italy, we have thus far seen that attitudes towards European integration are generally becoming more positive. But are these general trends reflected in a decline in support for anti-EU parties ? We begin to assess this question with Figure 3, which shows trends from the nonparametric regressions of anti-EU party support on time. Given the findings described above, one should expect a decrease in support for anti-EU parties, but that is generally not the case. In fact, only in Italy and Belgium did support for anti-EU parties decline over time, and recall that in the case of Italy increased support for European integration was less obvious than for any other country. In France support has remained quite stable, and in the Netherlands there is an apparent increase in support for anti-EU parties, mainly through the electoral breakthrough of the Green Left in the 1989 environment-oriented election. Although support is generally lower than in the other countries, it also remained relatively stable in Luxembourg except for a large drop from 1993 to 1994 which seems sensible to interpret as random variation than a real decline in the population. Before we analyse these findings in detail, however, we examine models predicting party support that control for possibly confounding factors.

Attitude towards Europe and EU party support

We now turn to the logit models regressing support for EU parties attitudes towards Europe. These models also include control variables and terms for the possible interaction between year (treated as a factor) and attitudes toward Europe. Recall that we do not fit these models to the Luxembourg data because the number of respondents supporting anti-EU parties was far too small.

Table 2 shows an analysis of deviance for the Europe and Year effects. The chi-square values are from type II tests and thus represent the difference in deviance between a saturated model (i.e., all terms are included), and a model that excludes only the term in question. In other words, they assess whether the model fit is improved by including the terms. It is clear from this table that there are indeed statistically significant differences in support for anti-EU parties over time in all four countries—in all cases the model fit is significantly better when differences in support for the anti-EU parties are allowed to vary over time.

More interesting in Table 2 are the differences in the impact of anti-EU sentiment over time. Firstly, in the case of Italy and the Netherlands, there is a significant effect of attitudes towards Europe, but the model fit does not improve by allowing the effect of attitudes towards Europe to change from one year to the next. This provides some evidence, then, of a consistent anti-EU vote mobilised on bases other than traditional cleavage structures. Secondly, the analysis of deviance indicates that attitude towards Europe affects party support in France, but that the importance of this attitude changed from one year to the next. On the other hand, for Belgium, attitudes towards Europe matter only when its impact over time is considered, suggesting that it is necessary to include the interaction in the final model. Of course, Table 2 tells us nothing about the direction and magnitude of any of the effects discussed above. For this we must have a closer look at the structure of the models.

Rather than display the numerous logit coefficients from these models in a table, we choose to display the fitted probabilities of voting for an anti-EU party in graphical form. This facilitates better interpretation of the effects in the model, especially given the complexity of the logit coefficients associated with interaction terms. Figure 4, then, plots the fitted probabilities of supporting an anti-EU party according to responses to the European Union question for each country. Recall that responses to this question range from a high of 4 (negative attitude towards Europe) to 1 (positive attitude towards Europe). In calculating these fitted probabilities, only attitudes towards Europe and year are allowed to vary—all other terms are set to their means (for quantitative variables) or proportions (for categories of categorical variables). In other words, the lines in the graphs represent fitted probabilities across time for “typical” respondents.

Only for France and the Netherlands are the lines clearly separated throughout the entire period, and thus showing a consistent impact of attitude towards EU on estimated probability of voting for an anti-EU party. Here we see the expected pattern—the more Eurosceptic the voter, the more likely s/he is to vote for an anti-EU party. There also appears to be a growing cleavage based on European attitudes in these two countries, though as we noted earlier, the differences across years is not statistically significant for the Netherlands. Moreover, the differences are generally much larger in France. There is a relatively clear trend in France towards higher probability of voting for an anti-EU party among strong anti-Europeans, while the probability of voting for an anti-EU party remains stable for others. In this case, then, we have possible evidence of a developing cleavage structure based on Europe : increasing polarisation of an electorate choosing a range of parties all characterised by a rejection of the new supranational institutional framework representing the critical juncture. This corresponds to the pro-/anti-European dimension and the more general pro/anti-system divide which other commentators have used to characterise the contemporary French system (Belot and Cautrès, 2004; Grunberg and Schweisguth, 2003).

The pattern is not the same for the other countries, however. More specifically, we see in Figure 4 that although there are significant changes in the effect of attitudes toward Europe across years, these differences do not follow any clear pattern. Moreover, the probability of voting for an anti-EU party is often unrelated to position on European integration—most notably, in Belgium where the strongest anti-European category is least likely to vote for one of these parties at the beginning of the period under study. More importantly, there is no evidence of an increasing polarization in anti-EU vote according to attitude towards Europe. Consistent with the analysis of deviance, there is no evidence of attitudes towards Europe having any effect whatsoever on support for anti-EU parties in Italy.

European Identity and EU Party Support

We now turn to Table 3, which shows an analysis of deviance for logit models predicting support for anti-EU parties from European identity from 1992-98. This table shows a number of noteworthy findings. Firstly, European identity has a significant effect on support of anti-EU parties in all countries except the Netherlands. We also notice that, even when holding European identity constant, there were significant differences in support for these parties from year to year in all countries except Italy. Finally, in none of the countries was there a statistically significant change in the effect of European identity on support for anti-EU parties.

All models control for age, sex and social class. Religion is not controlled for. Year is entered in the models as a categorical variable.

Figure 5 shows the fitted probabilities of support for anti-EU parties for those who have a European identity and those who do not. We see that in most countries there are clear differences in support according to identity—those with a European identity are far less likely to support the anti-EU parties. Only with respect to France, however, are the findings consistent with those regarding attitudes towards Europe—i.e., the more Eurosceptic and the more one identifies with Europe, the less likely one is to support anti-EU parties.

Perhaps the most striking finding in Figure 5 is the fact that, although attitudes towards Europe were able to distinguish vote for the anti-EU parties in the Netherlands, European identity does not. This implies, then, that there is not a direct equivalence between identity and EU support—those with national or other identities as their principal loyalty are still able to support the EU process.

Finally, Figure 5 shows a growing trend in Italy for support for anti-EU parties, regardless of identity. This is perhaps contrary to what might be expected given the state’s push for European integration—even the Eurosceptic aspects of the Berlusconi governments, with Forza Italia included in the pro-Europe reference category, are offset—and thus probably better reflects a change in the Lega Nord’s success at mobilizing on other issues aside from the European issue, rather than a change in public sentiment. More specifically, this trend probably reflects the Lega Nord’s increasing shift towards protest politics characteristic of Extreme Right populism.

Discussion

In summary, our results are quite mixed. The trendless fluctuation in European attitudes toward integration across time illustrates the shifting and hence non-structural importance of the issue. In contrast, the inertia of the European / non-European identities emphasises the latent presence of an identity, but at a socio-cultural rather than political level. Regardless of the level of support for European integration, however, when controlling for class and religious effects in models predicting anti-EU party support, we find that Europe is an issue par excellence. Still, European identity and attitudes towards Europe do not appear to form the basis of a cohesive social and political cleavage.

Although there have been changes in the effects of European attitudes on voting over time, the mark of an emerging cleavage— namely a monotonic trend marking the increasing salience of this effect—is largely absent. There are important country differences, however. In France, for example, the European issue increasingly divides pro- from anti-Europeans in terms of voting over time. Although there was no similar pattern with respect to the effects of European identity on anti-EU vote, we examined a very short period of time (1992-98) for this variable and thus the findings should be interpreted cautiously. There is some evidence, then, that a European cleavage is developing in France. Only time will tell whether this cleavage solidifies or whether it simply reflects other systemic facets of the platforms of the Extreme Right, Extreme Left and ‘ultra-Republican’ parties in France that make up the anti-EU parties as we defined them.

Aside from France, the evidence of an anti-EU political cleavage is weak at best. In fact, in Luxembourg support for the EU is so overwhelming that the parties espousing anti-EU sentiments are almost nonexistent. They are so weak, in fact, that in no years were the numbers in the surveys who supported them large enough to allow even fitting a logit model to predict vote. Even in the other countries where there is more—though still small—anti-EU support, there is no convincing evidence of change over time. In other words, voters appear to choose anti-European parties for reasons other than a simple orientation that is at odds with that of an increasing number of voters who identify with Europe. With this in mind, the nature of anti-EU parties is relevant. The period under study was characterized by a rise in radical right parties, which as time went on, tended to become more anti-European. In other words, although the parties we explore are clearly anti-EU, many of them also have other elements that may discourage voters from supporting them, regardless of whether they are strongly Eurosceptic.

Our conclusion that Europe has little structural effect on party systems and long-term voter alignments is consistent with the conjectures of Marks and Wilson (2000). One might argue that our findings are purely confirmatory, but this would be misguided. Unlike others, we have explicitly looked for evidence of a separate European cleavage. Previous research has either shown that existing cleavages dictate attitudes toward Europe, or that there is no rapid restructuring of alignments based upon Europe. In contrast, guided by Bartolini and Mair’s (2000) criteria for a political cleavage, we examined how EU sentiments are related to anti-EU party support. In other words, we looked for evidence of a particular group of voters choosing a party that explicitly represented those interests. The sustained stability of the effects of EU sentiments on anti-EU vote across time, rather than a visible watershed, suggests that there has not been a critical juncture leading to the development of a unique cleavage based on Europe.

Despite the absence of an identity cleavage, our analysis provides further evidence that Europe has mattered at certain points in time and in certain countries. France in 1980 is a case in point. The relatively strong showing of the Extreme Left, outstripping the Moderate Left in votes, if not in seats, in the face of defeat by the Moderate Right in the 1979 European elections undoubtedly strengthened anti-EC resolve among the French electorate and forged a salient issue out of Europe.8 The drop in 1985 to previous levels confirms the disappearance of this issue as a salient topic, however, probably reflecting the resurrection of traditional cleavage-related economic concerns during the first Socialist government’s period of incumbency.

In using a political sociology approach, we have assumed that political parties arise in response to the demands of voters. In other words, party structures and platforms reflect concerns of social cleavages. Institutionalists, on the other hand, argue that the electoral systems largely influence the development of parties. Others still would argue that cleavage structures and electoral systems both have consequences for the development of parties (see, for example, Ordeshook and Shvestova, 1994; Amorim Neto and Cox, 1997). If this is the case, some of the differences in anti-EU party support across countries are likely to reflect different institutional arrangements in the countries. It follows, then, that a social cleavage may not develop into a political cleavage simply because the parties do not adequately represent that cleavage in the face of an inauspicious institutional context.

We should also acknowledge limitations in the data. The fact that the question we use to tap European identity was not asked in any surveys before 1992 limited us from directly exploring the effects of identity over time. We attempt to overcome this problem by exploring the effects of attitudes toward European integration, in place of identity, on anti-EU vote. This allowed us to extend the analysis back to 1975. Although we found little in the way of a developing European cleavage using this measure, it is possible that one would be uncovered if there were a better measure of European identity. In fact, these two measures had quite different effects on vote for the short period when they were both present, indicating that attitudes towards the European integration cannot necessarily be assumed to equivalent to European identities. Nonetheless, it is impossible to retrospectively reconstruct a measure of European identity for those years for which it was not asked.

In conclusion, our findings reinforce those of existing research on the role of the European issue in electoral alignments and choice. To the extent that European identity exists, it remains latent and largely untapped in most countries. Europe emerge as an important issue from time to time, and even appears to be increasingly structurally implanted in one country at least—France, but the conditions for cleavage-emergence are still unmet. In other words, although Europe is increasingly impinging upon polities and the lives of those within them, its presence is not yet strong enough to justify the passing of a ‘critical juncture’. Whether future deepening, rather than widening, of Europe has such an effect remains to be seen.

Robert Andersen

Jocelyn A.J. Evans1

Notes

[ 1] Please address correspondence to Robert Andersen ((andersr@ mcmaster. ca)or Jocelyn Evans ((j. a. evans@ salford. ac. uk). Following our usual practice, authors are listed in alphabetical order though both had an equal role.

[ 2] We exclude the sixth original EC member – Germany – due to a lack of Eurosceptic parties of any electoral significance in this country.

[ 3] For a consideration of variants of Liberalism and its straddling Left and Right, see Kirchner (1988).

[ 4] We would regard most regionalist movements as essentially reactionary in their political stance, looking for a return to a pre-nation state status quo ante.

[ 5] Many authors now treat Alleanza Nazionale (AN) as a conservative and softly Eurosceptic, rather than Extreme Right and anti-European party (Ignazi, 2005). Given the time-frame of our analysis, we leave AN in the latter category, its moderation occurring over the last few years only.

[ 6] Although see Kalyvas (1998) for a more nuanced utilitarian view of the role of the Church and Catholic parties in democratization.

[ 7] We use locally weighted scatterplot smoothing with window spans of 60% of the data. These models essentially fit local polynomial regressions and join the fitted values from the regressions together to form the regression smooth (Fox, 2000). In other words, the nonparametric regression smooths some of the year-to-year variation so that we can more easily visualize long-term trends but still allows a more nuanced look at changes over time than would a parametric regression. All models control for age, sex, religion and social class. Year is entered in the models as a categorical variable. Attitude towards Europe is entered as a covariate, which explains the 1 df associated with it.

[ 8] hhttp :// www. ulb. ac. be/ soco/ cevipol/ documentation/ France/ Europeennes/ 1979. ht m