Lakatos reloaded. A reply to Lister


Michael Lister has responded to my critique of his work with what Lakatos (1980) called a “problem shift”. In his original paper, Lister (2007) presented empirical evidence that selective welfare state institutions undermine norms of solidarity, which in turn depresses turnout in general elections. In my comment (Arzheimer 2008), I demonstrate that, while the hypothesis is certainly plausible, the evidence itself is inconclusive at best. From my reading of Lister’s reply, I take it that we are in agreement on this point at least. Our disagreements (or mutual misunderstandings?) relate to three different issues, which I will address in turn.

The Status of Social Norms and the Problem of Ecological Fallacy

Lister (2009) argues that I do “great violence” to his position by oversimplifying his presentation of the causal chain that links a) institutions, b) social norms, c) individual (internalised) norms, d) individual voting behaviour and e) turnout, and also by falsely accusing him of committing an ecological fallacy. In my bid to develop a stylised representation of his argument (Arzheimer 2008: 682), I do indeed simplify his sophisticated theoretical account and omit social norms, and I do so for two good reasons: first, in the reductionist spirit of Almond and Verba (1965), it is not clear how social norms are different from the distribution of individual attitudes; and second, even if we assume that social norms do have an independent ontological status at the macro level, we have no independent measure for them and cannot, therefore, model their relationship with institutions. As for my “false accusation”, Lister (2007) studies a correlation between two aggregate variables, inequality, as a proxy for institutions, and turnout, which he then interprets in terms of individual agency. That approach, by definition, constitutes an ecological fallacy.

A quantitative impasse?

In my comment on Lister’s original article, I point out that institutional variables tend to vary a lot between countries but are often fairly stable over time within countries. As a result, it is often extremely difficult to disentangle institutional and country effects. Responding to this point, Lister (2009) concludes that a “statistical evaluation of this [i.e. his] hypothesis is virtually impossible”, but this is not major problem because his original results “should be seen as ‘indicative’” anyway. I could not disagree more strongly.

First, new statistical methods have been developed that are better suited to dealing with “weak” datasets (see the references in my comment). Second, over the last three decades or so, waves of democratisation have given us many more (varied) observations, which most studies have so far ignored. Moreover, political scientists and welfare economists have begun to tap into the power of experimental surveys that allow us to confront citizens with a whole range of alternative hypothetical welfare-state arrangements. The quantitative analysis of such experiments at the micro level is a useful complement to macro-quantitative information.

Finally, and perhaps most importantly, we should remember that the whole purpose of quantitative analysis is to test competing hypotheses so that the most plausible can be identified, given the data at hand. If one is not interested in this type of reality-check, there is little point in conducting quantitative analysis.

Qualitative vs Quantitative Analysis of the Institutions-Attitudes link – a false dichotomy

In his response to my comment, Lister suggests that the alleged link between institutions and attitudes should be studied by applying “more qualitative techniques, such as case studies”. He then briefly compares the development of the welfare state and levels of social trust in Sweden and the United States.

Following King, Keohane and Verba (1994: 4), qualitative research does not rely on numerical measurement but aims at providing more detailed information on the phenomenon in question. However, the same logic of inference associated with quantitative research and the same general rules of scientific inquiry apply. In this instance, an appropriate qualitative research design might include a detailed analysis of welfare-state legislation (at the macro level), accompanied by some carefully selected in-depth interviews with Swedish and American citizens that illustrate their experience of the welfare state and their more-or-less-trusting relationship with other citizens.

What we get instead is the information that some benefits introduced under President Johnson were effectively abolished under the Reagan and Clinton administrations so that the system remained highly selective, that cuts were made in Sweden during the 1990s without changing the universal nature of the system, that social trust has fallen in the US from 53% in 1963 to 35% in 1999 and that  levels of social trust in Sweden are extremely high at present and were presumably slightly lower in the past.

How is this different from the original quantitative analysis? The institutional information could easily be translated into an ordinal or quasi-interval measure for universalism (from low to extremely low in the United States, and from extremely high to quite high in Sweden), while the information on social trust comes directly from quantitative population surveys such as the World Values Survey.

The dependent variable has thus changed from turnout to social trust, yet we are still left with an opaque aggregate correlation that could have been caused by any number of processes at the micro level. The biggest difference between this new information and the original analysis is that Lister has dropped all control variables and reduced the number of observations from about 130 to 4.

Lakatos famously differentiated between “progressive” and “degenerative” problem shifts. Whether the switch from a time-series cross-sectional design to a case-study design falls into the former or the latter category is a judgement best left to the reader.


Almond, Gabriel A. and Verba, Sidney, 1965: The Civic Culture. Political Attitudes and Democracy in Five Nations. Boston, Little, Brown and Company.

Arzheimer, Kai, 2008: ‘Something old, something new, something borrowed, something true? A comment on Lister’s ‘Institutions, Inequality and Social Norms: Explaining Variations in Participation’’, British Journal of Politics and International Relations, 10,  681-697

King, Gary, Keohane, Robert O. and Verba, Sidney, 1994: Designing Social Inquiry. Scientic Inference in Qualitative Research. Princeton, Princeton University Press.

Lakatos, Imre, 1980: The Methodology of Scientific Research Programs. Cambridge, Cambridge University Press.

Lister, Michael, 2007: ‘Institutions, inequality and social norms: Explaining variations in participation’, British Journal of Politics and International Relations, 9, 20–35.

Contextual Factors and the Extreme Right Vote in Western Europe, 1980-2002


After the Second World War, the extreme right (ER) in Western Europe was associated with the atrocities of the Nazis and their puppet regimes (Rydgren 2005) and was therefore politically isolated and insignificant in most countries of the region. But from the early 1980s on, an unexpected third wave of right-wing extremist party activity swept over the continent. All of a sudden, parties that were dubbed as “extreme”, “radical”, “populist” or “new” right proved highly successful at the polls in countries such as Austria, Belgium, Denmark, France, Italy, Norway, Sweden, and Switzerland.

Problems of terminology and idiosyncratic features notwithstanding, a consensus1 emerged that these parties should be grouped into a single party family. While this group of extreme right parties (ERP) is arguably more heterogeneous than other party families (Mudde1996), its members are reasonably distinct from the mainstream or established right and share a number of ideological features, in particular their concern about immigration, which swiftly became the single most important issue for these parties (van der Brug and Fennema2003).

By the 1990s, scholars of electoral behavior had also identified a set of core features of the ERP electorates. While the most successful of these parties have managed to attract some votes from virtually all social groups, the bulk of the extreme right’s support comes from non-traditional segments of the working and lower-middle classes who are worried about the presence of Non-West European immigrants in their respective societies. There is generally a much greater propensity to vote for the ER amongst men, voters who are either young or rather old, those with a low level of formal education and amongst the manual workers, the petty bourgeoisie, and those in routine non-manual employment (see the review in  Arzheimer and Carter2006, 421-422). This sharp social profile is matched by an equally clear attitudinal profile: as a number of studies have demonstrated, the voters of the West European ER are to a large degree motivated by xenophobic feelings and beliefs (see e.g. van der Brug and Fennema2003).

A whole host of national and a smaller number of comparative studies have replicated these findings time and again. However, surprisingly little attention has been paid to the more intriguing twin question of why the ER’s support is so unstable within many countries over time, and why these parties are so weak in many West European countries. Only a handful of contributions have looked into this question at all, and each of the existing studies has its shortcomings. Moreover, the findings from different studies often contradict each other. The aim of this article is therefore to employ fresh data and a more adequate modeling strategy in order to provide a more comprehensive and satisfactory answer to the question of why the support for the ER in Western Europe varies so much over time and across countries.

The remainder of this article will proceed as follows. After a brief introduction to the main theories of ER voting, the existing longitudinal and comparative studies on the ER vote in Western Europe will be reviewed. Following that, a multi-level model of the ER vote in Western Europe will be presented. The article ends with a discussion of the main findings and their implications for future research.

Theoretical Accounts for Extreme Right Support in Western Europe

Starting with then contemporary attempts to explain the rise of the Nazi party and the Italian Fascists, social scientist have developed a multitude of theoretical accounts of support for the extreme right. The complexity of these accounts notwithstanding, they can be grouped in four broad strands (Winkler1996).

A first group of authors focuses on largely stable and very general attributes of the ER’s supporters, namely personality traits and value orientations, that make them more receptive for the ER’s appeals than their compatriots. The most prominent example of this line of research is arguably Adorno et al. (1950).

A second strand of the literature is chiefly concerned with social disintegration, which is characterized by a (perceived) break-down of social norms (“anomia”) and intense feelings of anxiety, anger and isolation brought about by social change. Allegedly, this mental state inspires a longing for strong leadership and rigid ideologies that are provided by the ER. A classic example of this approach is Parsons (1942).

According to a third class of accounts which draw heavily on theories from the field of social psychology, group conflicts are the root cause of the ER’s successes. This strand of research is, however, rather heterogeneous. At one end of the spectrum, there are classical theories of scapegoating (e.g. Dollard et al.1939). They argue that (ethnic) minorities provide convenient targets for the aggression of those members of the majority who are frustrated by their lack of status and other resources because these minorities tend to be both different from one’s own reference group and powerless. Otherwise, the choice of the victimized group is largely random and purely driven by emotions.

At the other end of the range, theories of Realistic Group Conflict beginning with Sherif and Sherif (1953) emphasize that ethnic conflicts can be driven by a bounded yet instrumental rationality. If xenophobia is the result of a conflict between immigrants and lower class natives over scarce resources (low-paid jobs, welfare benefits), discrimination against immigrants, proliferation of racist stereotypes and support for the ER can be interpreted not as an emotive reaction but rather as part of an instrumental strategy. This idea is especially prominent in more recent accounts (e.g. Esses, Jackson and Armstrong1998).

Finally, theories of ethnic competition (Bélanger and Pinard1991), “status politics” (Lipset and Bendix1951) , “subtle”, “modern” or “symbolic” racism (Kinder and Sears1981), and social identity (Tajfel et al.1971) all cover a middle ground between these two poles. While the various labels obviously highlight different aspects of group conflicts, recent research (Pettigrew2002) usefully suggests that most if not all of these approaches could be subsumed under the concept of relative deprivation: members of one social group feel that in comparison with another social group, they are not getting what they feel they are entitled to, even if they know that they get more than the other group.

While all three approaches have a lineage that spans more than five decades, most recent comparative research explicitly or implicitly combines theories of group conflict with elements of a fourth perspective that complements and expands on the three major approaches. In Winkler’s (1996) original survey of the literature, this emerging perspective was presented under the label of a “political culture” that constrains the effects posited by the other approaches. However since the mid-1990s, interest in a whole host of other, more tangible contextual factors has grown tremendously, and it is now widely believed that the interplay between group conflicts and system-level variables can help explain the striking differences in support for the ER over time and across countries.2

Building on previous work by Kriesi et al. (1992) and Tarrow (1996), Arzheimer and Carter (2006) have argued that these contextual factors should be subsumed under the concept of a “political opportunity structure” . Such a structure consists of short-, medium- and long-term contextual variables, which capture the “openness or accessibility of a political system for would-be political entrepreneurs” (Arzheimer and Carter2006, 422) and affect the chances (and thereby presumably the motivation) of politicians to create and maintain an electorally viable ERP.

However, while the concept of opportunity structures is certainly useful, it is also notoriously vague. As the review in the next section will demonstrate, there is no consensus (yet) on what variables are part of an opportunity structure. On the other hand and somewhat paradoxically, the notion of “opportunity” has implications that might be to restrictive: many context factors like unemployment or immigration will not only provide the political elite with an incentive to mobilize as entailed by the concept but will also have a direct and possibly more important impact on voters’ preferences. Given that comparative data on the perceptions and strategic decisions of (would-be) members of the political elite are unavailable, it is empirically impossible to separate the two potential effects of a contextual variable.

More generally, social psychological theories of group conflict were developed and tested in the context of small group research, where psychological and sometimes even physiological processes can be closely monitored, often in an experimental or quasi-experimental setting. Datasets that are available for longitudinal and comparative analyses, on the other hand, are restricted to a handful of attitudinal measures and a set of simple socio-demographic variables that were consistently replicated over the years. This problem is mitigated, however, by the fact that all theoretical accounts of ER support tend to identity a similar set of socio-demographic groups that should be most susceptible to the appeals of the ERPs. Moreover, both national and cross-sectional comparative studies of ER support have confirmed strong and consistent links between these socio-demographic indicators of group membership and more nuanced attitudinal measures.

Therefore, although data limitations make is impossible to unpack the details of the underlying psychological process, it is clear that the impact of micro- and macro-level variables on support for the extreme right in Western Europe should be modeled jointly. Only such a multi-level model provides one with unbiased estimates of the contextual effects, because the differences in the socio-demographic and attitudinal composition of the European electorates are controlled for. A multi-level model therefore represents a significant improvement over the existing empirical accounts for the ER’s support, which will be reviewed in the next section.

Previous Findings on Contextual Determinants of the Extreme Right’s Electoral Support

Jackman and Volpert (1996) conducted the first large scale3 quantitative comparative analysis of the ER’s electoral support by estimating a Tobit model of the ERPs’ vote share. Their main findings were that(1) the ER benefits from high unemployment,(2) higher electoral thresholds reduce the support for the ER, and(3) multi-partyism in combination with a proportional electoral system is associated with higher levels of ER voting.

Some technical issues notwithstanding (see Golder2003a), the analysis by Jackman and Volpert was ground-breaking both in terms of its spatial and its temporal coverage, and yet, there are some obvious substantive problems with it. First, Jackman’s and Volpert’s selection of cases is problematic in at least one instance: They include the Alianza Popular/Partido Popular, which became the major party of the established right in Spain in the 1980s and is not normally considered an ERP (Ignazi2003, 190-191). On a related account, their time-frame is problematic since there is now a wide agreement that the “third wave” did not come to itself before the early 1980s, when immigration became the core issue of the ER and many ERP tried adopted new strategies and communication frames that had proved successful in France (Rydgren2005). This leads to an obvious problem with the Scandinavian Progress Parties that started out as anti-tax parties in the 1970s and only moved into the ER camp during the early 1980s (Svåsand1998).

Second, Jackman and Volpert (1996) analyze the impact of a (somewhat limited number of) polity-level variables on aggregate support for the ER but completely ignore the micro-level, which is at the center of all theoretical explanations of ER voting. Finally, by modeling electoral returns, Jackman and Volpert restrict their analysis to a handful of (very important) snapshots in the political histories of the 16 countries. But while election results are decisive for the creation, composition and survival of governments, the ongoing level of support for the ER can have a tremendous impact on proposed and actually implemented policy via the strategic calculations of the established parties, even if the ER is not (yet) represented in parliament (Minkenberg2001).

For these reasons, Knigge (1998) models aggregate support for ERPs as measured by the bi-annual Eurobarometer surveys in Belgium, France, the Netherlands, West Germany, Denmark, and Italy between 1984 and 1993 in a Time-Series Cross-Sectional design and concludes that immigration and political dissatisfaction correlate with higher levels of support for the ER. Conversely, the effect of unemployment is negative.

While Knigge’s contribution is an improvement over the analysis by Jackman and Volpert because she analyses time-series with more and uniformly spaced data-points, it clearly falls short in terms of country-coverage. Moreover, like Jackman and Volpert, Knigge confines herself to the macro-level while a comprehensive model of extreme right support should clearly include both micro- and macro-level factors.

Precisely this is the aim of the useful study by Lubbers, Gijsberts and Scheepers (2002), who merge surveys from 16 West European countries with a host of aggregate variables. From a series of multi-level models they conclude that after controlling for individual anti-immigrant attitudes and political dissatisfaction, the number of non-Western residents as well as characteristics of the ERP themselves have a substantial impact on the likelihood of an extreme right vote, whereas the unemployment rate has no significant effect.

However, their contribution is problematic, too, in a number of ways. First, the merging of data from six national election studies with data sets from three different supra-national projects obviously raises problems of validity and reliability. Second, the number of N = 17 level-two units is too low for multi-level modeling by any conventional standard, especially given the authors’ interest in estimating variance components (Hox2002, 173-179). Finally, they focus on a rather brief time period, thereby excluding the 1980s (when the ER became a relevant political actor for the first time in decades) and much of the 1990s. Moreover, unlike Jackman and Volpert and Knigge, Lubbers, Gijsberts and Scheepers discard any cross-temporal variation in the ER’s support by pooling surveys from different years.

This is certainly not a problem in the analysis presented by Golder (2003b), which proceeds along similar lines like that of Jackman and Volpert but covers 19 West European countries including many “failed cases” like Iceland, Ireland, or Malta, and 165 elections between 1970 and 2000. From his findings, Golder concludes that(1) the ER benefits from both high levels of unemployment and high levels of immigration, and that(2) there is an additional positive interaction between unemployment and immigration..

Although Golder’s results are suggestive, like the other aggregate analyses they do not allow one to draw conclusions about micro-level processes, e.g. about the propensity and the motivation of the unemployed to vote for the ER. Technical sophistication notwithstanding, most of the problems discussed in regard with Jackman and Volpert and Knigge therefore apply to Golder’s study as well.

While all studies discussed so far have addressed unemployment as one potential determinant of ER support, Swank and Betz (2003) were the first who empirically analyzed the the mediating effects of welfare state institutions on the ER vote. In yet another macro model, they regress the electoral returns of the ER in 83 elections that were held between 1981 and 1998 in 16 West European countries on trade openness, capital mobility and foreign immigration as well as on the level of social protection and a number of other contextual variables. From their findings, they conclude that the number of asylum seekers is positively related to ERP success, whereas a high level of welfare state protection reduces the appeal of the ER.

However, although the impact of unemployment is at the center of the debate, Swank and Betz (2003, 228) use a fairly general index of welfare state benefits. While their approach is innovative, this variable is clearly not ideal for their purpose. Moreover, all concerns regarding the aggregate analyses by Jackman and Volpert, Knigge, and Golder obviously apply here, too.

Finally, in the most recent contribution to the field, Arzheimer and Carter (2006) have tried to overcome some of the limitations of the existing research by merging data from 24 national election surveys (conducted in seven countries between 1984 and 2001) with a host of aggregate-level information such as party-positions, proportionality of the electoral system, unemployment, and immigration figures. Like Knigge, they find a negative effect of aggregate unemployment. Moreover, they conclude that established right-wing parties that take a very tough stand on immigration may actually legitimize the policies of the ER, and that grand coalition government prior to elections raises the odds of an ER vote being cast.

However, Arzheimer’s and Carter’s paper is not without problems, too. First, while their mode of analysis is less demanding on the data than multi-level modeling, the parsimony of their model comes at a price as they have to assume that no unit (=country) effects remain after controlling for the impact of the aggregate variables. If this assumption does not hold, bias will result. Second, unlike Knigge and Lubbers, Gijsberts and Scheepers, they do not measure support for the ER between elections. Third, since they rely on national election studies that vary wildly in terms of attitudes questions asked, their range of individual-level variables is restricted to “objective” features like age, class, and education. Fourth, the number of level-2 units (elections) is rather low in relation to the large number of contextual variables in which they are interested.

To summarize, the existing research has demonstrated that contextual factors (and most prominently immigration and unemployment) have a systematic effect on support for the ER in Western Europe. However, while models of ethnic competition (that are consistent with micro-level theories of support for the ER) strongly suggest that immigration, unemployment, and their interaction should all have positive effects, it is unclear whether and under what conditions this is true. Moreover, it is by no means obvious that unemployment and immigration are truly more important than other contextual factors.

Support for the Extreme Right in Western Europe, 1980-2002


While a lot of progress has been made since the paper by Jackman and Volpert was published, the previous section has shown that none of the existing studies on the contextual determinants of the extreme right’s vote is entirely satisfactory. The analysis presented here tries to overcome these limitations by(1) combining a relatively large number of relevant system-level variables with individual socio-demographic and attitudinal data that are measured in a comparable fashion,(2) covering the whole time-span between 1980 and 2002, and(3) not excluding contexts where the ER is very weak.

At the micro-level, the model includes information on the respondents’ gender, age, level of education, and social class. The gender gap in support for the ER is well known, even if its causes are controversial (Gidengil et al.2005). The other socio-demographic indicators reflect the theoretical and empirical links between group membership, attitudes and the likelihood of a vote for the ER. For instance, it is well known that voters with high levels of educational attainment are more likely to embrace liberal values (Weakliem2002) and have little reason to feel threatened by low-skilled immigrants. Younger voters, members of Europe’s declining “petty bourgeoisie”, manual workers, the unemployed and maybe pensioners, on the other hand, should be highly susceptible to the appeal of the ER because they compete with immigrants for scarce resources.

Three measures model the impact of ideologies and more specific political preferences. First, while longitudinal data on immigration attitudes are largely unavailable, in most countries the ER has taken a negative stance on European Integration and has tried to link this theme to its core issues of immigration, national sovereignty, and law and order. Therefore, the model contains a control for Euroscepticism.

Second, the notion of a “protest vote” features prominently in some of the earlier accounts of the “third wave”. It is, however, unclear what a protest vote should entail. On the one hand, some authors suggest that “protest” is something irrational and emotional that is unconnected to values and ideologies and primarily “a vote against things” (Mayer and Perrineau1992, 134). But on the other hand, it is obvious that much of this “protest” is not un-ideological at all but clearly directed “against the policy or the absence of policy in this respect [migrants and law and order]” (Swyngedouw2001, 218-219).

To account for these “protest motives”, the model contains both an indicator for general political dissatisfaction as well as a control for political ideology (the familiar left-right self-placement scale). This makes it possible to separate the alleged “pure protest” from ideology- and policy-based considerations. Moreover, controlling for ideology accounts for the fact that the political left can benefit from Euroscepticism, too.

Most national and comparative studies of the ER vote have demonstrated rather stable and uniform effects for these individual-level variables. The crucial question here is whether these regularities do still hold once contextual variables are included and the spatial and temporal coverage is extended beyond that of previous analyses.

At the macro-level, the model aims at bringing together the most relevant variables from the contributions discussed above without taking a shotgun approach that would render both estimation and interpretation infeasible. Given their prominence in the literature and the fact that ethnic competition theories provide a clear rationale for the interpretation of their effects, the inclusion of unemployment and immigration plus an interaction between both figures is a matter of course.

The “protective” effect of welfare state benefits found by Swank and Betz (2003) deserves closer inspection,too, not least because it has clear implications for public policy. Moreover, the findings by Swank and Betz contravene one particularly influential early account of the ER’s support in Western Europe: Kitschelt’s (1995) hypothesis that a combination of authoritarian and market-liberal stands would guarantee electoral success for the ER. However, in line with the argument about ethnic competition in the labor market, instead of the general benefit data a more specific measure of benefits for the unemployed will be used.

Following Arzheimer’s and Carter’s approach, two institutional features that most clearly embody the concept of a (durable) opportunity structure, namely political decentralization and the degree of disproportionality of the electoral system are included, too. In the case of decentralization, Arzheimer and Carter present arguments both for a positive and a negative relationship with the ER vote. On the one hand, subnational elections can work as a “safety valve” for dissatisfied citizens that would ceteris paribus reduce support for the ER in national elections. On the other hand, these second-order elections provide the ER with opportunities for acquiring political experience, access to the media, and credibility. While neither of these two effects is borne out in their original analysis, disproportionality is of particular interest because the existing research seems to disproves the common wisdom that less proportional systems help to “keep the rascals out”.

Finally, Lubbers, Gijsberts and Scheepers as well as Arzheimer and Carter have argued that a comprehensive model of ER voting should also reflect the impact of genuinely political short-term factors such as the political agenda, the general tenor of the political debate on the issues of the ER, and the ideological positions of the political parties in a given country at a given time. While both groups of authors use a somewhat idiosyncratic terminology, their two competing hypotheses can easily be re-expressed within the framework of well-established theories of political behavior.

On the one hand, classical theories of spatial competition (cf. Enelow and Hinich1984) which treat the distribution of preferences in the electorate as exogenous and fixed in the medium run suggest that the rise of ER in the 1980s can be explained by the mainstream parties’ persistent reluctance to cater to the existing demand for strict immigration and asylum policies. Consequentially, support for the ER must decline if the established parties (with their track record of past performance in government and their much broader appeal) take a tougher stand on immigration and multi-culturalism, thereby “stealing” the ER’s issues. As Bale (2003, 76) observes, this “conspiracy of silence” theory of ER electoral successes is rather popular with political pundits in many West European countries.

On the other hand, a more subtle claim is often made in the literature on the extreme right: if a mainstream party takes a radical position on the extreme right’s issues, the public can interpret this as a signal that these policies are relevant, and that the contents and style of extreme right politics are no longer taboo (see e.g. Thränhardt 1995). As a consequence, at least some of the voters who support the ER’s policies but shy away from voting for a stigmatized party will now, in Jean-Marie Le Pen’s words, “prefer the original to the copy”. Moreover, other voters who were previously not aware of these issues may now be induced to evaluate the parties with respect to this policy dimension.

While authors like Thränhardt and Bale interpret this chiefly as a Machiavellian gambit by the established right (who have less to lose and more to win than the established left if immigration moves up the political agenda), an increase in the importance of immigration, asylum and race will without doubt benefit the extreme right, too. Although the connection is rarely made in the literature, these effects can easily be interpreted in terms of agenda setting and priming

To gauge the potential effects of party competition, Lubbers, Gijsberts and Scheepers (2002) rely on an expert survey, from which the derive two measures that capture the “immigration restriction climate” and the available “space for the ER”. Arzheimer and Carter (2006) draw on the Comparative Manifesto Dataset. Using party statements on internationalism, multi-culturalism, national lifestyle, and law and order, they construct two variables, namely the ideological position of the major party of the ER and the ideological distance between the two major mainstream parties. The latter approach seems preferable, because unlike the expert survey, the manifesto data are inherently dynamic, based on a well-defined and reliable coding procedure, easily available for replication, and cover the whole period under study.

However, to further improve on Arzheimer and Carter and to link the empirical analysis more closely to the underlying theories, the construction of both variables was slightly modified. First, considering only the ideological position of the major moderate right party seems overly restrictive. Often, the margin of what is politically acceptable will in fact be defined by the position of a smaller party of the right (or left, see Thränhardt 1995, 328 on anti-immigrant measures taken by Communist mayors in France).4 Consequentially, the most radical position on the ER’s issues taken by an electorally relevant party that is not part of the ER (cf note 11) is used as an indicator for electoral competition. This approach has the additional benefit of avoiding somewhat arbitrary decisions about what constitutes the “major” party amongst a whole group of more or less equal-sized political groupings.

Second, Arzheimer’s and Carter’s indicator of convergence between the two major parties was replaced by two separate measures for the variance and the salience of statements by all established parties on the issues of the ER. The salience measure ignores the direction of these statements and focuses solely on the space devoted to these issues. More salience is equivalent to a more prominent position of these issues on the agenda, which presumably benefits the ER. The variance measure, on the other hand, reflects Zaller’s (1992, chapter 6) more subtle proposition that the public will often follow the views of the elites if there is consensus amongst them, whereas visible disagreement amongst the elites conducive to polarization.5

To summarize, if standard spatial theories of voting apply, support for the ER will ceteris paribus be lower where the established parties position themselves further to the right. If, however, theories of agenda setting and priming prevail, the extreme right should benefit if their issues(1) feature prominently in elite discourses directed at voters and(2) if there is little consensus about what should be done. On the other hand, if elites downplay these issues and if there is little conflict amongst elites, this should reduce support for the ER. Consequentially, the “conspiracy of silence” could be a viable political strategy.

Finally, two caveats are in order. First, the phrase “no longer taboo” suggest that the timing of manifesto statements can have a crucial effect: once a taboo is broken, it could be difficult if not impossible to restore it. In principle, one could classify countries according to(1) whether the established parties have ever adopted the issues of the ER and(2) if so, whether they have returned to their original position, and introduce this classification as an additional variable. But since the number of West European countries is low, in reality it is infeasible to model this effect of timing.

Second, the model does not contain any measures for another important class of short-term factors, namely the content of the mainstream-media. However, while the media will most probably haven some effect even when party positions are controlled for (see Boomgaarden and Vliegenthart 2007 for a single-country study that tests this proposition), relying solely on party manifestos can actually be an advantage since it could be argued that political messages sent by other parties could to a degree reflect anticipations about future and reactions to previous successes of the ER, which would in turn lead to endogeneity bias. While this argument may apply to the statements that parties and politicians issue on a daily basis, endogeneity is less likely to be a problem with party manifestos, manifestos are the outcome of a lengthy deliberation process within the respective party. Moreover, unlike individual statements, the commit they represent a public policy commitment. Therefore, manifestos are probably the best and most reliable measure for a party’s position and political message.


The analysis covers the member states of the European Union (EU) as it existed before the Eastern enlargement rounds plus Norway.6 Individual level data come from the European Commission’s bi-annual series of Eurobarometer surveys.7 The number of missing values in the Eurobarometer for the variables under study is rather low, yet listwise deletion of cases with missing information reduces the sample size by about one third and can lead to overly optimistic standard errors and biased estimates. As a safeguard, the Multivariate Imputation by Chained Equations procedure devised by van Buuren and Oudshoorn (1999) was used to create eleven imputed data sets. All analyses were carried out both on the original and the completed data sets, but the results are almost identical.

Contextual information was drawn from official election results, OECD databases and printed reports (OECD1992, 1999, 2001200220032004), the datasets produced by the Comparative Manifesto Project (Klingemann et al.2006), the UNHCR statistical yearbook (UNHCR2002), and Lijphart’s (1999) seminal study of institutional arrangements in Western democracies.

The analysis spans the years from 1980 to 2002. During these 23 years, 1,065 individual Eurobarometer surveys were conducted in 18 countries.8 Each survey where at least one respondent voiced an intention to vote for an ERP and where all the required individual-level information was available was retained, yielding a total of about 175.000 respondents nested in 267 individual surveys. Surveys without any supporters of the ER were excluded, which effectively removed the United Kingdom and the Republic of Ireland from the analysis.9

While Golder (2003b) is right that excluding these “failed cases” is likely to lead to biased estimates in studies where aggregated electoral support is modeled, the case is less straightforward in a setup where individual voting intentions are analyzed. In countries where the ER is very weak, strong effects of social desirability are likely to bias the measurement of ER support. Moreover, supporters of the ER are often prevented from voting for their preferred party because the ERPs will not field candidates in most constituencies.10 Finally, due to financial constraints of the pollsters, the supporters of tiny ERPs are often coded as voting for “other” parties. As a result, support for the ER will be underestimated in contexts where those parties are already very weak, which will lead to a different kind of bias. While there is no perfect solution for this dilemma, restricting the analysis to contexts where it is at all possible to trace support for the ER by means of mass opinion surveys is a reasonable compromise.

The remaining surveys provide an exceptionally good coverage of the “third wave”, including the early successes and failures in the 1980s. The only major gaps are Norway in the 1980s and late 1990s (when the country was not yet/not any longer covered by the Eurobarometer) and Austria between 1986 (when Haider became chairman of the FP”O) and 1994 (when Eurobarometer polling started).


The dependent variable is vote intention for an ERP,11 calling for logit or probit multi-level analysis because the observations in the dataset are obviously nested. However, the way in which this nesting should be modeled is less obvious. Observations could be conceived of as (1) persons nested in countries nested in time, (2)  persons nested in time nested in countries, or (3) persons, cross-classified by time and countries. Of these, (2) is the most appropriate variant for a number of reasons. First, a cross-classification would be structurally incomplete, because a number of country-years are not covered by the Eurobarometer or were excluded because there were no ER voters. Second, persistent effects unit (country) effects are quite strong, whereas there is no indication of any effects of time that would be uniform across countries. Finally, while time-points are random in the sense that they can be conceived of as a sample from a large universe of days/weeks on which a survey could have been conducted, countries are not sampled from a population but are essentially “fixed” (Berk, Western and Weiss1995).

For these reasons, countries are represented by a series of dummy variables, which are common to all surveys from a given country.12 This modeling strategy effectively reduces the number of levels to two (see Duch and Stevenson 2005 for an application), namely the individual and the particular context of the survey wave in which she was interviewed. Thus, the model can be written as

Please see the PDF-Version

where i is an index at the person-level and j is an index at the context level. Hence, yij is the individual vote intention for an ERP, which is assumed to be binomially distributed (??). The logit of the probability to vote for the extreme right (??) depends on a linear combination of k individual variables (x1ij⋅⋅⋅xkij), l contextual variables (z1j⋅⋅⋅zlj), 14 fixed country effects (c1⋅⋅⋅c14), and a random disturbance at the context level (u0j).13 The latter is assumed to be normally, identically and independently distributed (??).14 Since the structure of the model is logistic, the binomial distribution of the vote intentions is assumed to adequately account for randomness at the individual level. All models were estimated in MLwiN 2.02 using the Penalized Quasi-Likelihood method based on a second-order Taylor expansion (PQL2).


Estimates for the four components of the model are presented in table 1. The rows in the lower third of the table contain the unit effects for the 14 countries under study, i.e. the logit of an ER vote when all individual and contextual variables are set to zero.15 While the coefficients themselves are of little intrinsic interest, their huge variation implies that even if a whole host of individual and contextual variables are controlled for, there are persistent differences between these countries that must be due to other, durable factors.

At the bottom of the table, σu02 represents the residual variance at the contextual level, i.e. the normally distributed random shocks that affect all voters in a given country at a given time. This figure is about 40 per cent lower than in a null model that contains only unit effects and a random term (not shown as a table), suggesting that the combination of contextual and individual variables goes a long way in understanding the puzzle of ER support. Nonetheless, a normal distribution with a variance of 0.3 will still produce a considerable number of rather large random shocks.16

The effects of the individual-level variables can be discerned from the topmost panel of table 1. As it turns out, the expected patterns re-appear even when one controls for contextual variables, unit effects and contextual variance. In line with theoretical expectations, groups who compete with immigrants for scarce resources and who have exhibited the highest level of xenophobia in the past – manual workers, younger voters, and the unemployed – show significantly more support for the ER than other groups. The gender gap is equally prominent. While the logistic link implies that effects are not linear-additive and that the proportion of ER support depends on the level of all independent variables, membership in either of these socio-demographic groups roughly doubles the probability of an ER vote.17 Again in line with previous findings, holding a university degree massively reduces the probability of a vote for an ERP, whereas being a pensioner has no significant effect on the vote.

Turning to the attitudinal variables, being a eurosceptic18 more than doubles the probability of an extreme right vote, but political dissatisfaction and ideology have even stronger effects. Dissatisfaction is operationalized through a four-point rating scale, therefore its maximal impact on the logit is 1.7 points. Left-right self-placement was measured on a ten-point rating scale, so its maximal effect on the logit is 4.7 points. Maximal effects paint a somewhat unrealistic picture since few voters hold extreme attitudes, but even if one considers the more conservative interquartile range of 1 point (dissatisfaction) and 3 points (ideology), it is obvious that political dissatisfaction and political leanings have significant and rather dramatic effects on the propensity to vote for the ER even when they are mutually controlled for.

Jointly, the coefficients in the upper panel of table 1 provide fresh evidence from a large number of political contexts that the ER vote is not based on protest alone, and that the ER is by no means a “catch all” that mobilizes all social groups in a similar way (Mayer and Perrineau 1992; see van der Brug and Fennema 2003 for a broader discussion). Rather, the ER’s success is based on its appeal to a constituency that has a distinct social and attitudinal profile.

This picture is complemented by th coefficients for the contextual variables, which are presented in the middle panel of table 1. “Disproportionality” and “decentralization” refer to the Gallagher-Index for the most recent election and the index devised by Lijphart (1999, 189), respectively. “Asylumseekers” reflects the number of new applications for asylum status per capita,19 “unemployment” refers to the standardized unemployment rates which are supplied by the OECD, whereas “unemployment benefits” reflects the impact of the OECD’s “Gross Unemployment Benefit Replacement Rates”.20 Three multiplicative interaction terms were created to reflect the hypotheses that the effects of unemployment and immigration reinforce each other (Golder2003b), while unemployment benefits can mitigate these effects (Swank and Betz2003).

“Toughness” (the most radical position on these issues taken by any party that is not considered to be part of the ER), “salience”, “variance” and the interaction of the latter two reflect information on party competition and political elite messages in a given context and were constructed as outlined above.21 To ease the interpretation and to reduce the likelihood of numerical problems, the unemployment, asylum and benefit rates as well as the measures for salience and variance were centered at their respective grand mean.

[Table 1 about here.]

In line with the findings by Arzheimer and Carter, decentralization and disproportionality of the electoral system do not have a statistically significant impact on support for the ER when other contextual variables are held constant. Again, the data provide no evidence that less proportional systems can curb support for the ER. It should, however, be borne in mind that (with the notable exception of France) the variation in the degree of proportionality is limited.22

The number of asylum seekers, unemployment rates,23 and the salience of the ER’s issues, however, all seem to have substantial and statistically significant effects on support for the ER. But in the presence of interaction terms, the size and statistical significance of these main effects must be calculated conditionally. The estimated effect for the number of asylum seekers, for instance, refers to a situation where both the centered unemployment rate and the centered level of unemployment benefits are held constant at a level of 0, i.e. at the mean of the original variables. Similarly, the positive effect of the unemployment rate is conditional on average levels of immigration and unemployment benefits.

[Figure 1 about here.]

Contrary to predictions derived from ethnic competition theory, and contrary to Golder’s (2003b) findings that are based on a different specification and on macro data alone, the interaction between levels of unemployment and immigration is negative: at higher-than-average levels of asylum applications and unemployment, the effects of both variables do not reinforce each other. Rather, a ceiling effect is observed that limits the impact of both contextual variables on the ER’s support. More specifically, if immigration is very high, unemployment does not matter any more. This is illustrated in Figure 1: at levels of immigration that are slightly higher than average (> 0.7), the effect of unemployment is not longer significantly different from zero (upper panel). Since the distribution of asylum application rates is right-skewed, this applies to roughly 20 per cent of all contexts. Note, however, that the effect of unemployment is rather weak even where immigration is at its empirical minimum of -0.98. Immigration, on the other hand, has a significantly positive effect even when unemployment is up to five points above its average (lower panel). This threshold is only exceeded in ten per cent of all contexts.

Similarly, more generous income replacement rates will reduce the effect of unemployment by a small and the effect of immigration by a considerable amount, as indicated by the two negative interaction effects. Moreover, unemployment benefits have an additional effect of their own, but this effect is small and statistically insignificant at average levels of immigration and unemployment. Even if unemployment and immigration rates are at their empirical extrema, this main effect will hardly effect the outcome.

[Figure 2 about here.]

The interpretation of the effects that political messages sent by other parties have on support for the ER is more straightforward. In line with Arzheimer and Carter, “toughness”, i.e. the ideological position of the most radical amongst the established parties, has no significant effect. This constitutes prima facie evidence against the “conspiracy of silence” hypothesis derived from spatial models of voting.

However, the two variables that reflect ideas of agenda setting and priming do have an effect. At an average level of variation in the party positions, a greater salience of the ER’s issues in the party manifestos is ceteris paribus related to a higher level of ER support. This effect (which prevails though objective factors like immigration and unemployment are controlled for) is quite pronounced: The interquartile range of the salience variable is 6.3, which translates into a change of 0.77 points on the logistic scale. This is roughly equivalent to the individual-level effect of being dissatisfied with European Integration.

The effect of salience is somewhat muted at higher levels of variation in the party statements as indicated by the negative sign of the interaction term, but even where the (centered) variance is higher than 50, the coefficient is significantly positive, as can bee seen from the lower panel of Figure 2. This applies to more than 95 per cent of all contexts. In fact, for 90 per cent of all contexts, the variance falls into the interval [-14.1; 36.6], implying that very often the negative interaction has no substantive consequences at all and the effect of salience prevails. The effect of variance, on the other hand, is neither statistically nor substantively significant, regardless of the level of salience.

[Figure 3 about here.]

While logit coefficients convey the direction of the effects and provide a means for testing their statistical significance, they are less useful for assessing the political relevance of a given variable. Here, the most relevant quantity is the predicted effect on the ER’s share of the votes, which depends on the respective level of all independent variables. A convenient tool for illustrating this impact are figures which plot the predicted probability as a function of one to three focal independent variables while all other independent variables are held constant at pre-specified levels that represent theoretically interesting “scenarios” (King, Tomz and Wittenberg2000). More recently, Mitchell and Chen (2005) have suggested that for complicated models, one could aggregate the average individual effects of all independent variables that are not varied in the graph into a single quantity which they dub “covariate contribution”. A relatively small number of covariate contributions (say three) could then be employed to cover a whole host of different “scenarios”.

This method is used in Figure 3 to illustrate the joint effect of unemployment rates, immigration, and unemployment benefits on the probability of a vote for the extreme right. Covariate components were obtained by calculating the logit for each of the 174,452 respondents from the fixed effects in the leftmost column of Table 1 and subtracting the joint impact of the three contextual variables and their interactions from this quantity. Then, the covariate contribution was set to the fifth, seventh, and ninth decile of its distribution. Levels of immigration and levels of unemployment benefits were set to the first, the fourth, the sixth, and to the ninth decile of their respective empirical distributions, while the unemployment rate was varied from its first to its ninth decile.

Figure 3 clearly demonstrates that while the effects of those three contextual variables that feature most prominently in previous research are statistically significant, their political relevance will often be negligible. The short-dashed lines that represent the lower half of all the individually calculated covariate contributions are basically flat, which indicates a trivial relationship between unemployment figures and the probability of a vote for the ER. Moreover, unemployment benefits and immigration levels hardly affect support for the ER, although weak effects are discernible if one compares the graphs within the same row or same column: support is minimal where benefits and immigration figures are close to their minimum (upper left corner), but increases very slightly (by less than two percentage points) where either of the two variables comes closer to its maximum.

If the contribution of other covariates is set to a somewhat higher level as represented by the solid lines, unemployment rates, unemployment benefits and immigration have a slightly stronger but rather complex effect on the predicted support for the extreme right. Comparing the graphs within each column reveals that higher levels of immigration are related to higher levels of ER support, although the differences are still small. Higher levels of benefits are related to higher levels of support, too, but this effect is restricted to contexts with below-average levels of immigration (the first two rows of graphs). As regards the effect of unemployment rates, a positive effect becomes visible, but only in contexts where either levels of immigration or unemployment benefits are very low. Most interestingly and in line with the findings by Swank and Betz, at high levels of immigration, unemployment benefits reduce the impact of unemployment, i.e. the line that represents this relationship is flat.

Finally, if the contribution of other covariates is set to a very high level (that could be due to individual-level effects, a strong unit-effect, the impact of other contextual variables or a large and positive random shock at the contextual level), the three contextual variables will have a strong and intertwining effect on the probability of a vote for the ER. Basically, a comparison within the columns of Figure 3 reveals that higher levels of immigration are associated with higher support for the extreme right. This effect, however, is much stronger where the level of welfare state protection is average or below average. At higher levels of unemployment benefits, the impact of immigration is much reduced. On the other hand, unemployment benefits are positively related to higher levels of support where immigration is low (c.f. the first two rows of Figure 3). Finally, unemployment figures are often strongly related to support for the ER, but where immigration is high, this relationship effectively vanishes, which reflects the lack of the positive interaction posited by Golder (2003b).

[Figure 4 about here.]

Analyzing the impact of party positions involves only two variables and their interaction and is therefore simpler. Figure 4 graphs the relationship between the salience of the ER’s issues in other parties’ manifestos and the expected vote share of the ER for four levels of variance in party statements and three levels of covariate contributions. As one would expect from the coefficients in Table 1 and the graphical analysis in Figure 2, the variance of party statements has very little impact on the success of the ER. The salience of the statements, however, is highly relevant for the ER’s electoral success, provided that the contribution of other covariates is large. For the lower 50 per cent of all covariate contributions, even very high levels of salience hardly increase the ER’s electoral fortunes, and at the seventh decile, the differences between high- and low-salience contexts are still small. If the contribution of other covariates falls into the upper third of its distribution, however, the political effect of salience is huge.


While the relationship between support for the ER and the contextual variables is much more complex than suggested by previous research, the basic results of the graphical analysis can be easily summarized. First, in line with theories of ethnic competition, the ER will benefit from high levels of immigration and unemployment, but this effect is moderated by the institutions of the welfare state. Generous unemployment benefits seem to curb the additional impact of unemployment where immigration levels are high. On the other hand, if immigration levels are very low, generous unemployment benefits increase the probability of an ER vote. Accordingly, the lowest levels of ER support are predicted for a system with minimal benefits, low unemployment rates, and minimal immigration. Extreme right mobilization would be most facilitated by high unemployment and high levels of either immigration or unemployment benefits (but not both). Independent of these objective social and economic conditions, political factors, i.e. the salience of the ER’s issues in the manifestos of other parties have a remarkable effect on the ER’s prospects.

Second, while these findings are of both political and theoretical interest, they apply only to a situation where the probability of an ER vote is already rather high due to other factors. For roughly 70 per cent of all covariate constellations, the contextual variables will have a small impact on the probability of a ER vote whereas for the remaining 30 per cent, contextual factors can tip the balance and can make an ER vote much more (or much less) likely.

Third, a considerable portion of the covariate contributions is due to country-specific intercepts and context-specific random effects. Consequentially, the political relevance of the contextual variables will be more pronounced (1) within sub-groups that exhibit disproportionate levels of support for the extreme right (e.g. politically dissatisfied right-leaning workers) but (2) also more generally in countries where the propensity of an ER is rather high across the board (Austria, Belgium, France, Denmark, Norway), and (3) in contexts that are affected by a substantial random shock (e.g. a media scare). The substantive implications of this latter possibility should not be underestimated: from the distributional assumption in Equation ?? and the parameter estimate of 0.3 in Table 1, it follows that roughly 35 per cent of all random shocks will shift the logit of an ER vote for all citizens in a given context by at least 0.5 points upwards or downwards. If a case is near the median of the covariate contributions, a difference of that size can render contextual variables politically relevant or irrelevant.


This paper set out from the twin question of why support for the ER varies so much across time and political systems. More specifically, its aim was to assess the impact of contextual variables on the support for the parties of the ER in Western Europe. The analyses presented here differ from previous accounts in two crucial ways:(1) The effects of individual and contextual variables are modeled jointly and(2) all relevant and available Eurobarometer data sets were included, resulting in maximal spatial and temporal coverage.

The findings on the individual level largely confirm previous results from national studies: The ER’s electorate has a clear social and attitudinal profile. These results rule out explanations that link the ER’s electoral appeals solely or chiefly on “protest”.

The picture at the contextual level is more complex. First, there is no empirical support for the “conspiracy of silence” hypothesis. On the contrary: in line with theories of agenda setting and priming, the salience of the ER’s issues (immigration and national identity) in the manifestos of the established parties has a strong positive impact, whereas the “toughness” of the established parties has no significant effect.

Second, both unemployment rates and immigration have generally a positive impact on the ER vote, but their respective effects do not reinforce each other. Rather, a ceiling effect is observed. Moreover, unemployment benefits can reduce support for the ER in certain constellations. Third, the political relevance of these effects crucially depends on the contribution of other covariates. Often, even constellations of contextual variables that clearly favor the ER will be of little political consequence.

Finally, even after differences in the composition of societies (via individual-level variables), features of the context, and random variation at the contextual level are taken into account, there are striking differences between countries as revealed by the estimates for the unit effects ranging from -8.7 to -3.2 points on the logistic scale. Put differently, given the levels of the variables included in the model, in Austria, Italy and Denmark the ER is persistently much stronger and in Spain, Sweden, and Finland, it is much weaker than one would expect it to be. Future research should focus on factors such as access to the media, organisational strength of the ER and links with other actors, political culture, and elite cues other than those in manifestos to come up with an explanation for these differences.


Adorno, Theodor W., Else Frenkel-Brunswik, Daniel J Levinson and R. Nevitt Sanford. 1950. The Authoritarian Personality. New York: Harper.

Arzheimer, Kai and Elisabeth Carter. 2006. “Political Opportunity Structures and Right-Wing Extremist Party Success.” European Journal of Political Research 45:419–443.

Bale, Tim. 2003. “Cinderella and Her Ugly Sisters. The Mainstream and Extreme Right in Europe’s Bipolarising Party Systems.” West European Politics 26:67–90.

Bélanger, Sarah and Maurice Pinard. 1991. “Ethnic Movements and the Competition Model. Some Missing Links.” American Sociological Review 56:446–457.

Berk, Richard A., Bruce Western and Robert E. Weiss. 1995. “Statistical Inference for Apparent Populations.” Sociological Methodology 25:421–458.

Boomgaarden, Hajo G. and Rens Vliegenthart. 2007. “Explaining the rise of anti-immigrant parties: The role of news media content.” Electoral Studies 26:404–417.

Dollard, John, Leonard W. Doob, Neal E. Miller, O.H. Mowrer and Robert R. Sears. 1939. Frustration and Aggression. New Haven, London: Yale University Press.

Duch, Raymond M. and Randy Stevenson. 2005. “Context and the Economic Vote. A Multilevel Analysis.” Political Analysis 13:387–409.

Eatwell, Roger. 2004. The Extreme Right in Britain. The Long Road to ’Modernization’. In Western Democracies and the New Extreme Right Challenge, ed. Roger Eatwell and Cas Mudde. Routledge pp. 62–79.

Enelow, James M. and Melvin J. Hinich. 1984. The Spatial Theory of Voting. An Introduction. Cambridge, London, Melbourne: Cambridge University Press.

Esses, Victoria M., Lynne M. Jackson and Tamara L. Armstrong. 1998. “Intergroup Competition and Attitudes Toward Immigrants and Immigration. An Instrumental Model of Group Conflict.” Journal of Social Issues 54:699–724.

Freeman, Gary P. 1998. The Decline of Sovereignty? Politics and Immigration Restriction in Liberal States. In Challenge to the nation-state: Immigration in Western Europe and the United States, ed. Christian Joppke. Oxford University Press pp. 86–109.

Gidengil, Elisabeth, Matthew Hennigar, AndrE Blais and Neil Nevitte. 2005. “Explaining The Gender Gap in Support for the New Right: The Case of Canada.” Comparative Political Studies 38(10):1171–1195.

Golder, Matt. 2003a. “Electoral Institutions, Unemployment and Extreme Right Parties. A Correction.” British Journal of Political Science 33:525–534.

Golder, Matt. 2003b. “Explaining Variation in the Success of Extreme Right Parties in Western Europe.” Comparative Political Studies 36:432–466.

Hox, Joop. 2002. Multilevel Analysis. Techniques and Applications. Mahwah: Lawrence Erlbaum.

Ignazi, Piero. 2003. Extreme Right Parties in Western Europe. Oxford u. a.: Oxford University Press.

Jackman, Robert W. and Karin Volpert. 1996. “Conditions Favouring Parties of the Extreme Right in Western Europe.” British Journal of Political Science 26:501–521.

Kinder, Donald R. and David O. Sears. 1981. “Prejudice and Politics. Symbolic Racism vs. Racial Threats to the Good Life.” Journal of Personality and Social Psychology 40:414–431.

King, Gary, Michael Tomz and Jason Wittenberg. 2000. “Making the Most of Statistical Analysis. Improving Interpretation and Presentation.” American Journal of Political Science 44:341–355.

Kitschelt, Herbert. 1995. The Radical Right in Western Europe. A Comparative Analysis. Ann Arbor: The University of Michigan Press.

Klingemann, Hans-Dieter, Andrea Volkens, Judith Bara, Ian Budge and Michael D. McDonald. 2006. Mapping Policy Preferences II. Comparing 24 OECD and 24 CEE Countries, 1990-2003. Oxford u.a.: Oxford University Press.

Knigge, Pia. 1998. “The Ecological Correlates of Right-Wing Extremism in Western Europe.” European Journal of Political Research 34:249–279.

Kriesi, Hanspeter, Ruud Koopmans, Jan-Willem Duyvendak and Marco G. Giugni. 1992. “New Social Movements and Political Opportunities in Western Europe.” European Journal of Political Research 22:219–244.

Lijphart, Arend. 1999. Patterns of Democracy. Government Forms and Performance in Thirty-Six Countries. New Haven: Yale University Press.

Lipset, Seymour M. and Reinhard Bendix. 1951. “Social Status and Social Structure. A Re-Examination of Data and Interpretations. II.” British Journal of Sociology 2(3):230–254.

Lubbers, Marcel, Mérove Gijsberts and Peer Scheepers. 2002. “Extreme Right-Wing Voting in Western Europe.” European Journal of Political Research 41:345–378.

Mayer, Nonna and Pascal Perrineau. 1992. “Why Do They Vote for Le Pen?” European Journal of Political Research 22:123–141.

Minkenberg, Michael. 2001. “The Radical Right in Public Office. Agenda-Setting and Policy Effects.” West European Politics 24(4):1–21.

Mitchell, Michael N. and Xiao Chen. 2005. “Visualizing Main Effects and Interactions for Binary Logit Models.” The Stata Journal 5:64–82.

Mudde, Cas. 1996. “The War of Words. Defining the Extreme Right Party Family.” West European Politics 19:225–248.

Norris, Pippa and Christopher Wlezien. 2005. “Introduction: The Third Blair Victory: How and Why?” Parliamentary Affairs 58(4):657–683.

OECD. 1992, 1999, 2001. Trends in International Migration. SOPEMI. Paris: OECD.

OECD. 2002. Benefits and Wages. OECD Indicators. Paris: OECD.

OECD. 2003. OECD Employment Outlook. Towards More and Bettter Jobs. Paris: OECD.

OECD. 2004. Quarterly Labour Force Statistics (no. 4). Paris: OECD.

Parsons, Talcott. 1942. “Some Sociological Aspects of the Fascist Movements.” Social Forces 21:138–147.

Pettigrew, Thomas F. 2002. Summing Up: Relative Deprivation as a Key Social Psychological Concept. In Relative Deprivation. Specification, Development, and Integration, ed. Iain Walker and Heather J. Smith. Cambridge: Cambridge University Press pp. 351–373.

Rydgren, Jens. 2005. “Is Extreme Right-Wing Populism Contagious? Explaining the Emergence of a New Party Family.” European Journal of Political Research 44:413–437.

Schmitt, Hermann, Evi Scholz, Iris Leim and Meinhard Moschner. 2002. The Mannheim Eurobarometer Trendfile 1970-2002. Data Set Edition 2.00. Codebook and Unweighted Frequency Distributions. Updated by Iris Leim and Meinhard Morschner, ZA Cologne. Zentralarchiv für Empirische Sozialforschung.

Sherif, Muzafer and Carolyn W. Sherif. 1953. Groups in Harmony and Tension. An Integration of Studies on Intergroup Relation. New York: Harper and Brothers.

Svåsand, Lars. 1998. Scandinavian Right-Wing Radicalism. In The New Politics of the Right. Neo-Populist Parties and Movements in Established Democracies, ed. Hans-Georg Betz and Stefan Immerfall. Houndmills, London: Macmillan pp. 77–93.

Swank, Duane and Hans-Georg Betz. 2003. “Globalization, the Welfare State and Right-Wing Populism in Western Europe.” Socio-Economic Review 1:215–245.

Swyngedouw, Marc. 2001. “The Subjective Cognitive and Affective Map of Extreme Right Voters: Using Open–ended Questions in Exit Polls.” Electoral Studies 20:217–241.

Tajfel, Henri, M.G. Billig, R.P. Bundy and Claude Flament. 1971. “Social Categorization and Intergroup Behaviour.” European Journal of Social Psychology 1:149–178.

Tarrow, Sidney. 1996. Power in Movement. Social Movements, Collective Action, And Politics. Cambridge: Cambridge University Press.

Thränhardt, Dietrich. 1995. “The Political Uses of Xenophobia in England, France and Germany.” Party Politics 1:323–345.

UNHCR. 2002. 2002 UNHCR Statistical Yearbook. United Nations High Commissioner for Refugees.

van Buuren, Stef and Karin (CGM) Oudshoorn. 1999. Flexible Multivariate Imputation by MICE. (07.10.2005): TNO Prevention and Health.

van der Brug, Wouter and Meindert Fennema. 2003. “Protest or Mainstream? How the European Anti-Immigrant Parties Developed into two Separate Groups by 1999.” European Journal of Political Research 42:55–76.

Weakliem, David L. 2002. “The Effects of Education on Political Opinions. An International Study.” International Journal of Public Opinion Research 14:141–157.

Winkler, Jürgen R. 1996. Bausteine einer allgemeinen Theorie des Rechtsextremismus. In Rechtsextremismus. Ergebnisse und Perspektiven der Forschung. PVS Sonderheft 27, ed. Jürgen W. Falter, Hans-Gerd Jaschke and Jürgen R. Winkler. Opladen: Westdeutscher Verlag pp. 25–48.

Yonwin, Jessica. 2004. Electoral performance of far–right parties in the UK: Standard Note: SN/SG/1982. House of Commons Library.

Zaller, John R. 1992. The Nature and Origin of Mass Opinion. Cambridge, New York, Oakleigh: Cambridge University Press.


PIC(a) Effect of unemployment
PIC(b) Effect of immigrationGraphs are based on estimates in Table 1. Both variables are centered. 95% confidence interval calculated with Variance-covariance matrix estimated under listwise deletion. The level of unemployment benefits is set to its mean of zero.

Figure 1:The conditional effects of unemployment and immigration


PIC(a) Effect of variance
PIC(b) Effect of salienceGraphs are based on estimates in Table 1. Both variables are centered. 95% confidence interval calculated with Variance-covariance matrix estimated under listwise deletion. The level of unemployment benefits is set to its mean of zero.

Figure 2:The conditional effects of salience and variance


PICGraphs are based on estimates in Table 1 (listwise deletion). All variables are centered. Covariate components set to -4.7 (short dash), -3.8 (solid line), and -2.3 (long dash).

Figure 3:The joint impact of unemployment, asylum seekers, and unemployment benefits on the probability of a vote for the extreme right


PICGraphs are based on estimates in Table 1. Both variables are centered. Covariate components set to -5 (short dash), -4 (solid line), and -2.7 (long dash).

Figure 4:The effect of the extreme right’s issues on the extreme right vote


Listwise DeletionMultiple Imputation
18-29 years0.437(0.041)0.419(0.042)
30-45 years0.194(0.039)0.172(0.035)
>65 years-0.162(0.054)-0.160(0.047)
Education: middle/high0.056(0.036)0.069(0.034)
Education: university-0.324(0.043)-0.251(0.036)
Petty Bourgeoisie0.043(0.048)0.088(0.042)
Dissatisfied: EU0.751(0.036)0.729(0.038)
Dissatisfied: Democracy0.607(0.018)0.551(0.018)
Asylumseekers ×Unemployment-0.024(0.014)-0.031(0.014)
Unemployment benefits0.013(0.010)0.009(0.010)
Unemployment benefits ×Unemployment-0.002(0.002)-0.001(0.002)
Unemployment benefits ×Asylumseekers-0.010(0.005)-0.009(0.005)
Variance ×Salience-0.001(0.000)-0.001(0.000)

Logistic multi-level model. PQL2 estimates, model-based standard errors in parentheses. MI results based on eleven separate imputations.

Table 1:Support for the extreme right: Socio-demographics, attitudes, country effects, and contextual variables

*I wish to thank Paul Whiteley for advice and comments. I am also indebted to the participants of staff seminars at Essex and Mainz and to Elisabeth Carter, Jocelyn Evans and Chris Wendt for helpful comments at various stages of this research project.

1Much of the early literature is devoted to the perhaps not entirely fruitful twin debates on the “correct” label and on criteria for membership in this party family. However, at least the latter question is more or less settled since the mid-1990s: “we know who they are, even though we do not know exactly what they are” (Mudde 1996, 233, see also note 11). As regards terminology, this article refers to the “extreme right” because this seems to be the most commonly used label in recent research.

2On the other hand, theories of authoritarianism and anomia provide very limited analytical leverage because they focus on largely stable psychological states. Therefore, it is difficult to see how they could explain short-term fluctuations of ER support within a given country or persistent differences between otherwise largely similar countries.

3Jackman and Volpert analyze 103 elections that were held in 16 countries between 1970 and 1990.

4The relatively small Christian Social Union in Germany is a point in case: while they are clearly positioned in the political mainstream and have always formed an alliance with the much larger Christian Democratic Union at the national level, they take a tougher stand on immigration than the “Post-Fascist” Italian National Alliance (Lubbers, Gijsberts and Scheepers2002).

5According to Zaller, this effect is moderated by the political awareness of the respondents. In principle, the role of political awareness could be modeled by a cross-level interaction. However, since data on political awareness is rather limited, and since this moderating effect is not central to the argument presented in this paper, this route was not pursued.

6Switzerland is excluded from the analysis both for substantial reasons as well as for a lack of data.

7The partial cumulation of the Eurobarometer produced by a team led by Hermann Schmitt (Schmitt et al.2002) greatly facilitated the construction of the data set. An appendix containing details of the coding and imputation procedures as well as scripts for Stata and Mlwin that can be used to replicate the findings, additional tables and an assessment of the robustness of the findings are available through the author’s dataverse at .

8Because of the economic, social, and political-cultural differences, there are separate surveys for East and West Germany. Norway did not accede the EU in 1994 but did participate in the Eurobarometer between 1990 and 1996 and then again in 2002/2003.

9Luxembourg had to be excluded because the OECD does not calculate standardized benefit rates for this country. Estimates for a number of more parsimonious models that are based on a larger number of respondents and contexts including Luxembourg are presented in the online appendix.

10Britain is a case in point. In 2005, the British National Party, now the most important party of the ER in the United Kingdom (Eatwell2004), contested 119 of the 646 Westminster constituencies, i.e. less than 20 per cent (Norris and Wlezien2005, 678). With 57 and 33 candidates, the numbers were even lower in 1997 and 2001 (Yonwin2004, 7).

11The variable is coded as 1 if an respondent intends to vote for the Freedom Party in Austria, the Front National or the Vlaams Blok in Belgium, the Freedom Party or the Danish People’s Party in Denmark, the Rural Party or the True Fins in Finland, the National Front in France, the German People’s Union, Republikaner or National Democrats in Germany, the EPEN, the National Front and Political Spring in Greece, the National Alliance and the Northern League in Italy, the Center Parties and the Lijst Pim Fortuyn/Leefbar Nederland in the Netherlands, the Freedom Party in Norway, the “Christian Democrats” in Portugal, the various Falange Parties in Spain, and New Democracy in Sweden. Voters of other parties and self-declared non-voters are coded as 0.

12To simplify the presentation, the model contains no constant but rather one unit-dummy for each country.

13Note that a double index indicates variation both across persons and contexts, while variables with a single index vary across contexts but are constant over persons within the same context.

14More explicitly, the Variance-Covariance Matrix Ωu that governs the distribution of u0j is assumed to be a diagonal matrix whose elements are identical.

15AT = Austria, BE = Belgium, DE-E = East Germany, DE-W = West Germany, DK = Denmark, ES = Spain, FI = Finland, FR = France, GR = Greece, IT = Italy, NL = Netherlands, NO = Norway, PT = Portugal, SE = Sweden.

16If one considers a shock of 0.7 points on the logistic scale (which is equivalent to the effect of Euroscepticism) as “large”, this threshold will be exceeded in about 20 per cent of all realizations.

17All socio-demographic variables enter the model as dummy indicators.

18Euroscepticism is measured by a dummy variable.

19Other figures such as the number of non-white residents or the share of foreign-born residents could have been employed, too, but the data on asylum seekers and refugees are preferable for at least three reasons: first, asylum applications (and family reunification claims, which are often related to previous applications) have provided the main route for new legal immigration into Wester Europe since the 1970s (Freeman1998, 94), second, unlike other measures they are comparable across time and countries, and third, asylumseekers and refugees have become the main focus of the ER’s propaganda. To prevent numerical problems, the numbers were entered as applications per 1,000 residents.

20The OECD calculates “Gross Unemployment Benefit Replacement Rates” by averaging over several types of households, durations of unemployment, and income levels before unemployment.

21Parties of the ER were excluded from all calculations. For salience and variance, the figures were weighted according to the relative size (vote share) of the respective parties. Since the party manifestos are usually published only when an election is imminent, party positions between publication dates were interpolated.

22Removing France from the sample does not substantively affect the results.

23The individual employment is controlled for.


Christian Religiosity and Voting for West European Radical Right Parties


The academic literature on parties and voters of the extreme, radical or populist right is vast, and from this work we know that some voters are more likely than others to vote for these parties. The effects of certain socio-demographic characteristics on the radical right vote have been very well documented and there is a consensus in this literature that male voters, young voters, voters with low or middle levels of education and voters from certain social classes are more likely to vote for radical right parties than are other electors (see for example Arzheimer and Carter 2006; Betz 1994; Lubbers et al. 2002). Studies also agree that the attitudes of voters impact on their likelihood of casting a vote for these parties and that negative attitudes towards immigrants are particularly powerful in predicting a vote for a radical right party (Billiet and De Witte 1995; Lubbers et al. 2002; van der Brug et al. 2000).


Within this body of literature the impact of a voter’s religious attachment, involvement and attitudes on his or her propensity to vote for a party of the radical right has received relatively little attention, at least as compared to the effects of gender, age, education or class and the influence of certain attitudes. This is not wholly surprising given the importance of these other predictors. Furthermore, models of radical right voting are likely to have omitted variables that relate to religion for practical reasons: reliable, comparative data on religious behaviours and beliefs are hard to come by.


We believe, however, that there are valuable reasons for investigating the link between a voter’s religious attachments and beliefs and his or her likelihood of voting for a radical right party. And this is not because of the ever-present academic desire to ‘fill a gap in the literature’, although a gap does clearly exist (Mudde 2007: 296). Rather, in the first instance, our desire to explore this relationship rests on the widespread acknowledgement that, despite their decline (Crewe 1983; Crewe and Särlvik 1983; Dalton et al. 1984), traditional social cleavages continue to be important in structuring partisan alignments and electoral choice (Mair et al. 2004), and that the divide between religious and secular voters is still a relatively strong predictor of vote (Dalton 1996). To begin with, therefore, we are guided by research such as Girvin’s, which argues that ‘although electoral behaviour is affected by other factors such as gender and class, church attendance in a number of cases is the single most important variable in explaining voting decisions’ (Girvin 2000: 13; see also Norris and Inglehart 2004).


Secondly, we would argue that it is useful to concentrate on the impact of religion on a specific electoral choice – namely the likelihood of a vote for the radical right – because such a focus will ultimately tell us more about the role of religiosity in electoral choice. As we shall see, there are a number of good reasons to suggest that religiosity will reduce the likelihood of a vote for the radical right, and yet there are also good reasons to suggest that it might increase this likelihood. By disentangling the various influences of religiosity on the radical right vote, and by assessing their strength, we may gain a better understanding of the ways in which religiosity does or does not affect electoral choice in general.


In this article we therefore propose to investigate the impact of religiosity on the radical right vote because this endeavour serves a dual purpose: from the religiosity end of the telescope we seek to learn more about the impact of religiosity on electoral choice, while from the radical right end of it, we aim to gain an understanding of the predictive strength of religiosity on the radical right vote.


It also transpires that we have chosen to point our telescope into the sky at a rather interesting time. To be sure, traditional social cleavages have weakened and levels of church membership and religious participation have declined (Girvin 2000), yet religion has also rather unexpectedly assumed a greater centrality in the political life of West European societies in recent years. Its return to the global political agenda – as evidenced most pronouncedly by the war between Al Qaeda and ‘the West’ – has had considerable domestic implications in Western Europe, aggravating tensions between Christian or agnostic majorities and a host of minority groups that are increasingly defined (by themselves and the outside world) not in ethnic, but in religious terms. Conflicts over the symbolism of headscarves worn in public institutions in France, rows about veils in the UK, death-threats aimed at female politicians from Islamic backgrounds in the Netherlands and in Germany, and the crisis over the Danish cartoons are just some examples of such tensions. While it is too early to gauge the precise impact of such developments on long-term electoral choices, this context does make our decision to revisit the link between religiosity and electoral choice rather timely.


The rest of this article follows a conventional structure: the next section outlines our conceptualization of religiosity and our favoured terminology, and sets out our theoretical framework and hypotheses. We then explain our model and our variables, and describe our data and methodology. Having done this, we present our results and discuss our findings. We close with an assessment of the importance of religiosity in predicting electoral choice both for radical right parties and indeed more generally.



Religiosity and voting for the radical right: conceptualization and theoretical framework


As mentioned above, few studies have explored the impact of a voter’s religious attachment, involvement and attitudes on his or her likelihood of voting for a party of the radical right. What is more, those that have devoted attention to this question have, in the main, been single-country studies (e.g. Billiet 1995; Billiet and De Witte 1995; Lubbers and Scheepers 2000; Mayer 1998; Mayer and Perrineau 1992; van der Brug 2003; Westle and Niedermayer 1992). There are just four cross-national studies of radical right voting that have included an examination of the effect of religiosity, and the findings of these were rather mixed in that two found that religiosity had weak and inconsistent effects on party preference (van der Brug et al. 2000; van der Brug and Fennema 2003), while the other two concluded and that less religious (Norris 2005: 138-9) or non-religious (Lubbers et al. 2002: 348) people were over-represented in the radical right electorate.


Crucially, and in stark contrast to the more recent studies that examine the relationship between church involvement and ethnocentrism or prejudice (e.g. Billiet et al. 1995; Eisinga et al. 1990, 1999), these comparative analyses conceptualize and operationalize religiosity in a rather simple way: van der Brug et al. (2000) and van der Brug and Fennema (2003) include a composite variable in their models, which is made up of religious denomination and church attendance, Norris (2005) makes use of a measure of religious self-identification, and Lubbers et al. (2002) distinguish between non-religious people, religious people belonging to non-Christian denominations, and Christian people. We would argue that these conceptualizations and operationalizations are problematic because they are too blunt to untangle the different effects that religiosity may have on the likelihood of radical right vote and, as a result, they are likely to underestimate the total effect of religiosity (Bartle 1998). Research on religiosity and ethnocentrism (discussed below) suggests that religious affiliation, involvement and belief structures can be linked to the radical right vote in different ways and so it is crucial to conceptualize religiosity in a manner that captures its different aspects or dimensions, and the ways in which these might interact. To this end we conceptualize religiosity as a combination of religious affiliation, church attendance, private religious practice and self-stated religiosity. Precisely because our conceptualization captures the different aspects of religious activity and beliefs, we favour the term ‘religiosity’ over ‘religiousness’ or simply ‘religion’.


As indicated in the introduction, there are reasons to believe that religiosity may reduce the likelihood of a radical right vote, and yet there are also reasons to believe it may increase it. Focusing first on why religiosity might reduce the likelihood of such a vote, to begin with there is plenty of evidence to suggest that religious affiliation and involvement will lead to a greater likelihood of a voter voting for a party of the mainstream right, such as a Christian, Christian Democratic or conservative party that has traditionally defended religious interests, than any other type of party, including a party of the radical right. Of course Christian and Christian Democratic parties differ from conservative parties in terms of their origins and ideologies, with the former traditionally defending Christian values and the latter having no links with organized religion, but both long-standing research and more contemporary studies have shown that religious voters have tended to favour parties of the mainstream right, irrespective of whether these parties are of the Christian Democratic or the conservative type.


Many analyses of voting in Weimar Germany report that the Catholic electorate was less permeable to the NSDAP than other sections of society, and attribute this to Catholic voters’ attachment to the Zentrum party, as well as to the integrating role played by Catholic networks and organizations (e.g. Childers 1983: 188-9; Falter 1991; Grunberger 1971: 552; Lipset 1971: 147-9; Mommsen 1996: 353). And despite widespread secularization, the attachment of religious voters to Christian Democratic or conservative parties continues to be observed today. Norris and Inglehart, for example, argue that ‘in industrial and postindustrial societies […] religious participation remains a significant positive predictor of Right orientations’, even after controlling for a whole range of other socio-demographic, economic and contextual factors. Indeed, they conclude that ‘religious participation emerges as the single strongest predictor of Right ideology in the model, showing far more impact than any of the indicators of social class’ (2004: 204-7. See also Girvin 2000: 21). Given these findings, we believe it is therefore reasonable to expect a certain degree of ‘encapsulation’ of religious voters by Christian, Christian Democratic or conservative parties (see Hypothesis H1 below).


Secondly, we also expect religious voters to be less likely to vote for a party of the radical right than other voters for the simple reason that radical right parties will not appeal to them (see Hypothesis H2a below). On the one hand, radical right parties do nothing to attract religious voters since they do not discuss religion in their ideologies and programmes. Instead, these parties have only addressed the subject for purposes of political advantage and mobilization and/or because it fits in with their world-view. For example, the parties are much more concerned about non-Western religions (particularly Islam) that are said to be a threat to Western culture and society than they are about any of the moral substance of religious teachings, or about what adhering to a faith might actually mean and entail. In some specific cases the radical right’s failure to appeal to religious voters is also explained by anti-clerical traditions (as in Austria and Germany), or by the fact that the parties have libertarian roots (like in Norway and Denmark). On the other hand, the issues that the parties do discuss and the views they have on these issues are often very much at odds with the beliefs and values of religious voters. After all, the values, beliefs, and traditions associated with most contemporary versions of the Christian faith are those of tolerance, compassion and altruism, and these find little in common with the authoritarian, xenophobic and even racist ideologies and appeals of the parties of the radical right, and the practice of targeting some of the most vulnerable groups in society such as refugees and immigrants.


For a number of different reasons, therefore, it is wholly reasonable to suggest that religiosity might ‘insulate’ voters from the appeals of a party of the radical right. However, for a variety of other reasons, it also makes sense to hypothesize the contrary, and to expect religious affiliation, religious involvement and the intensity of religious beliefs to be linked with a greater support for a party of the radical right. As regards affiliation, a number of studies, starting with that by Allport and Kramer (1946), have concluded that people with no religious affiliation show lower levels of ethnocentrism than people who describe themselves as Catholic or Protestant (see also Pettigrew 1959). As for religious involvement, dozens of analyses have pointed to the existence of a relationship between church attendance and levels of prejudice. The seminal work by Adorno et al. (1950) was one of the first to report a curvilinear relationship between church attendance and prejudice. While, in general, it found higher levels of ethnocentrism among churchgoers than among non-attenders, more specifically it found that regular churchgoers and non-attenders were both less prejudiced than those who attended church on a less frequent or an irregular basis. A number of subsequent analyses, carried out both in the US and in Europe, have reached similar conclusions (e.g. Allport and Ross 1967; Eisinga et al. 1990; Gorsuch and Aleshire 1974; Petersen and Takayama 1984; Pettigrew 1959; Studlar 1978). Other studies have proposed that prejudice also depends on the nature of particular religious convictions or belief structures and that people with strong religious beliefs are prone to developing a ‘closed belief-system’, which has often been linked to ethnocentrism and authoritarianism (Glock and Stark 1966; Rokeach 1960; but see also Middleton 1973; Ploch 1974; Roof 1974 for a critique of this argument). While many of the studies just mentioned may reflect a climate specific to the United States of the 1950s and 1960s, a link between closed religious belief-systems and ethnocentrism has also been uncovered in a more recent analysis of religion and prejudice (Altemeyer 2003) as well as in a recent pan-European youth survey (Ziebertz et al. forthcoming).


To be sure, many of the early studies on religiosity and prejudice have been criticized on theoretical, conceptual and methodological grounds (see Eisinga et al. 1999 for a useful summary). Many failed to ascertain whether religious doctrines act as a trigger for prejudice, or whether, conversely, they legitimate existing prejudices. In addition, these early studies have been attacked for failing to adequately specify both dependent and independent variables, and in particular for muddling up different dimensions or aspects of religiosity, such as affiliation, church attendance, and belief structures (Scheepers et al. 2002). Finally, many of these early works also tended to examine bivariate relationships only, and did not control for other social variables such as age, educational level, class, or localism.


Despite the shortcomings of these studies, however, there is still good reason to hypothesize that religiosity may be linked with a greater propensity to vote for a radical right party because the literature cited above clearly points to a link between religiosity and ethnocentrism. And since negative attitudes towards immigrants – which are closely related to ethnocentrism – are one of the most powerful predictors of a vote for a party of the radical right (as discussed above), it makes sense to hypothesize a two-step link between religiosity, anti-immigrant sentiment and voting for a radical right party, with religious people showing a greater likelihood of voting for the radical right than other people (see Hypothesis H2b below).


On the basis of these arguments, a number of hypotheses – which bring together different strands of theory that have not been considered in combination before – may be advanced as to the impact of religiosity on the likelihood of a radical right vote. Of course, despite these arguments, it could well be that religiosity is not a cause of radical right thinking, but is instead a correlate, since religious people are not only older (Argue et al. 1999), but also tend to have lower levels of education (see Johnson 1997) and therefore are less likely to embrace liberal-democratic values than their compatriots. We therefore also advance a hypothesis that proposes that religiosity has no direct effect on the likelihood of a radical right vote, and that instead, any effect is due to socio-demographic characteristics alone (Hypothesis H3 below). Our (competing) hypotheses are as follows:

  •  H1: Religious people are less likely to vote for the radical right because they are firmly attached to Christian Democratic or conservative parties;
  • H2a: Religious people are less likely to vote for the radical right because they are less likely to adopt negative attitudes towards immigrants;
  • H2b: Religious people are more likely to vote for the radical right because they are more likely to adopt negative attitudes towards immigrants;
  • H3: All direct relationships between religiosity and the vote are spurious (i.e. once radical right-wing attitudes and party identification are controlled for, the remaining effects of religiosity are due to the socio-demographic profile of religious people and will disappear completely if group memberships are taken into consideration.)


In principle, these mechanisms can reinforce or counterbalance each other. In addition, the extent to which these hypotheses may be borne out in practice will clearly depend on differences in national contexts and on features of each political system. It is well beyond the scope of this study to examine these differing national contexts (see for example Broughton and ten Napel 2000; Hanley 1994; van Hecke and Gerard 2004), but as a starting point we may point to the importance of differences in the strength of the religious cleavage. In the Lutheran countries of Scandinavia the religious cleavage is relatively weak (Madeley 2004), and so encapsulation by Christian Democratic parties is likely to be moderate at best. By contrast, in denominationally mixed countries (such as the Netherlands and Switzerland), where this cleavage is stronger, greater encapsulation is to be expected. Secondly, any traditional links between the church and specific political forces are likely to be relevant. In France, for instance, there has historically been a close connection between fundamentalist streams within the Catholic Church and anti-modern and illiberal political forces (Minkenberg 2003; Veugelers 2000). In this context, religiosity is likely to have a quite different connotation than in countries that lack such a tradition.


The characteristics of individual parties will also have an effect on our findings. Most obvious is whether the parties of the mainstream right are Christian Democratic or, as in France, are conservative parties. Even where Christian Democracy prevails significant differences exist between the parties: while some parties, such as the Austrian ÖVP, are catch-all parties that have attempted to integrate a host of different ideological tendencies (Fallend 2004), others, like the Belgian CVP and PSC remain confessional parties (Lucardie and ten Napel 1994). Different still are the Scandinavian Christian Democratic parties, which emerged much later and which grew ‘out of traditions of religious dissent representing various shades of dissatisfaction with the religious establishment among activist minorities’ (Madeley 2004: 218). On a more specific, policy-level, some Christian Democratic parties have tended to stress the Christian values of compassion and tolerance and are therefore inclined to support the rights of immigrants (see della Porta 2002 on the case of Italy, where a strong, Catholic pro-immigrant movement exists), whereas others– like the German CSU– have taken a tough stand on immigration (Lubbers et al. 2002: 356).


Radical right parties also differ in their ideological profiles (Betz 1994; Carter 2005; Ignazi 1992; Kitschelt 1995; Taggart 1995) and these differences are likely to have implications for our findings since the parties will attract different socio-economic segments of the electorate, and will entice voters with different attitudes. While most parties of the radical right have no specific interest in religion, the French Front National has always (not least through its stand on abortion) tried to appeal to conservative Catholics, and the Italian Alleanza Nazionale is actively trying to develop a more Christian/conservative profile. The latter party is also unusual insofar as it places much less emphasis on the issue of immigration and is much less xenophobic than most other parties of the radical right (della Porta 2002). For all these reasons, therefore, we certainly expect country differences. That said, it is not our intention here (especially with only eight cases) to test explanations for these differences, even though we can engage in some speculation as regards our results.



Modelling the links between background variables, religiosity and the radical right vote


Although our model is a little complicated, its basic structure (see Figure 1) is of the simple block-recursive type that has been fruitfully applied in electoral research before (Bartle 1998; Miller and Shanks 1996) and that helps us establish the direction of the flow of causality. Located at the very beginning of the causal chain are several socio-demographic variables that are exogenous: although these socio-demographics will often affect the level of religiosity as well as the development of political attitudes and the vote, it is inconceivable that religiosity will cause gender, age, education or class. Religiosity in turn can have a causal effect both on political attitudes and on behaviour, but it is implausible to assume the reverse. Finally, the vote itself depends (amongst other things) on attitudes, religiosity and socio-demographic features but does itself not alter these variables.


In contrast to the comparative studies mentioned above, which included religiosity as an independent variable, our model incorporates religiosity as a variable that appears before political attitudes in the causal chain. This allows us to consider the different ways in which religiosity may affect the likelihood of a radical right vote. In particular we can examine whether its effects are direct, indirect, or are due to background variables (i.e. whether they are spurious).




The actual model on which our analysis is based is represented in Figure 2. The dependent variable in the analysis is vote for a party of the radical right, as depicted on the right hand side of the diagram (Block IV). This, we argue, is likely to be influenced by three sets of independent variables: religiosity (Block II); radical right attitudes (Block III); and socio-demographics (Block I). In addition, it is likely to be influenced by an intervening variable, namely an individual’s party identification with a Christian Democratic or conservative party (labelled ‘CD-PID’). This is also located in Block III.




We begin by considering the impact of the three sets of independent variables independently of each other. The variable ‘Religiosity’ is a latent variable constructed from four observable variables (rel1-rel4) that tap the different aspects of religiosity that previous research has identified, namely religious affiliation, church attendance, private religious practice and self-stated religiosity (see below for further details on the data). We treat these variables as indicators of a single latent variable because they are highly correlated in all countries under study. This allows us to deal with one variable only and yet to continue to benefit from the advantages that multi-indicator variables bring in terms of enhanced reliability and validity of results. As alluded to above in Hypothesis 3, independent of any identification with conservative or Christian Democratic parties and independent of an individual’s radical right attitudes we expect to see no direct relationship between religiosity and the radical right vote because the parties of the radical right pay little attention to religious issues.


The early studies discussed above examined the link between religiosity and ethnocentrism – i.e. a tendency to regard one’s own ethnic and cultural group as superior and to treat other groups with contempt (Sumner, 1906). We would argue that, since (non-Western) immigrants make up the most prominent ‘out-group’ in West European societies, it makes sense to operationalize this concept by including variables that capture an individual’s attitudes towards immigrants. ‘Radical Right Attitudes’ are therefore measured by 21 observable attitudinal variables (labelled rra1, rra2 etc in Figure 2) that relate to views on immigrants and refugees. Empirically, these 21 variables show a very high degree of intercorrelation and are thus treated as indicators of a single latent variable. Clearly, since previous research has shown that anti-immigrant sentiment is one of the strongest predictors of a radical right vote, we expect to see a positive relationship between this variable and the radical right vote.


Our third set of independent variables is composed of socio-demographic variables. These include age, gender, class and education. In line with the findings of previous studies, we expect a greater propensity to vote for the radical right among younger voters as compared to older voters, among male voters as compared to female voters, among voters with lower levels of education compared to those with high levels of education; and among working-class voters, farmers and the ‘petty bourgeoisie’.


As regards our intervening variable (‘CD-PID’) that refers to voters’ identification with a Christian Democratic or conservative party, clearly, we expect voters who identify with such parties to be less likely to vote for a party of the radical right than voters who display no such identification.


Of course, the three independent variables just discussed are not expected to exert an effect on the propensity of a radical right vote in isolation only. Rather, socio-demographic variables are likely to have an impact on an individual’s religiosity, and on his or her attitudes. This is shown in Figure 2 by arrows that flow from ‘Socio-Demographics’ to ‘Religiosity’, and from ‘Socio-Demographics’ to ‘Radical Right Attitudes’. In addition, socio-demographics are likely to have an impact on the likelihood of an individual’s identification with a Christian Democratic or conservative party, hence the further arrow that runs from ‘Socio-Demographics’ to ‘CD-PID’. We also cannot rule out the possibility that the socio-demographics have a direct impact on the vote after controlling for religiosity, radical right attitudes, and ‘CD-PID’, and there is therefore an arrow connecting ‘Socio-Demographics’ and ‘Radical Right Vote’ directly, capturing any residual effects of group membership on the vote that might remain after controlling for attitudes. These include any spurious effects of religiosity (Hypothesis H3).


Religiosity, for the theoretical reasons discussed above, is likely to have either a negative or a positive impact on radical right attitudes (Hypotheses H2a and H2b). This is shown by the arrow in Figure 2 that runs from ‘Religiosity’ to ‘Radical Right Attitudes’. In addition, we expect religiosity to have an effect on identification with a Christian Democratic or conservative party.


Radical right attitudes are very likely to have a direct effect on the vote for the radical right. Yet we cannot rule out that they might additionally be correlated with ‘CD-PID’ because people who identify with established, mainstream right-wing parties may be more likely to hold radical right attitudes than other citizens. That said, we can make no assumption as to the direction of this relationship, and so our model depicts a mere correlation, as represented by a double-headed arrow running between ‘Radical Right Attitudes’ and ‘CD-PID’.


This model enables us to test whether religiosity influences the radical right vote in any way whatsoever. If religiosity does affect the radical right vote, the model allows us to test whether it does so directly, or indirectly (through radical right attitudes and/or an identification with a Christian Democratic or conservative party), or whether the effect of religiosity is spurious (i.e. related to socio-background variables). The model thus allows us to test a number of alternative ‘routes’ that have so far largely been neglected or conflated in the literature on religiosity and on the radical right.



Data and Methodology

Our data come from the first round of the European Social Survey (EES), the fieldwork of which was conducted in 2002. This database is particularly attractive because it includes a whole host of measures of radical right attitudes as well as of religious views and behaviours. From the 22 countries covered in this survey we selected eight West European systems that have witnessed a substantial and persistent support for the radical right: Austria, Belgium, Denmark, France, Italy, Netherlands, Norway and Switzerland. While countries in which the radical right has been unsuccessful should be included in macro-level explanations of party success so as to avoid selection bias, it makes no sense to include them in micro-level models. If not a single respondent reports the intention to vote for the radical right (as in Spain, Sweden, or the UK), there is simply nothing to model. By much the same token we excluded Germany as the number of self-declared radical right voters here was tiny (n=10), making conventional logit or probit modelling unfeasible.


Respondents under the age of 18, non-citizens, and members of non-Christian faiths were excluded. In six of the eight countries included in this study there was little variation in the denomination of respondents who indicated they were of a Christian faith. Only in the Netherlands and Switzerland were there significant numbers of both Catholics and Protestants. The impact of different religious doctrines can therefore only be examined in these two countries, and this is confined to noting differences between Catholic and Protestant voters only, since the ESS does not disaggregate between different strands of Protestantism.


All respondents who stated that, in the last election, they had voted for the Austrian Freiheitliche Partei (FPÖ), the Flemish Vlaams Blok (VB) or the Belgian Front National (FNb), the Danish Dansk Folkeparti (DF) or Fremskridtspartiet (FRPd), the French Front National (FN) or Mouvement National Républicain (MNR), the Italian Alleanza Nazionale (AN), Lega Nord (LN) or Movimento Sociale-Fiamma Tricolore (Ms-Ft), the Dutch Lijst Pim Fortuyn (LPF), the Norwegian Fremskrittspartiet (FRPn), or the Swiss Freiheitspartei der Schweiz (FPS), Lega dei Ticinesi (LdT), Schweizer Demokraten (SD) or Schweizerische Volkspartei (SVP) were given a code of 1. All remaining respondents were given a code of 0. There was an average of 1,700 respondents per country.


As regards the socio-demographic variables we coded male respondents as 1 and female respondents as 0, and we recoded age into three categories that reflect the findings of previous studies on its effects on the radical right vote (18-29; 30-65; older than 65). For social class, data was first mapped onto the familiar Goldthorpe-Scheme. Then, to keep things as simple as possible, we created a dummy variable that takes the value 1 for those classes that have shown the greatest support for the radical right in the past – workers, farmers, and the petty bourgeoisie – and 0 for all others. For education we used the ESS’s seven-point scale of achievement that ranges from ‘no primary education’ (1) to ‘second stage of tertiary education’ (7).


We made use of the four measures contained in the ESS that capture different aspects of religious activity and beliefs. The first two concern the regularity with which an individual prays outside of religious services and the regularity with which he or she attends religious services (other than on occasions such as weddings, funerals etc.). These were each measured on a seven-point scale ranging from 1 (‘every day’) to 7 (‘never’). We reversed both scales to facilitate interpretation. The third measure taps religious affiliation and simply asks whether the respondent belongs to a Christian church or considers him or herself to be a Christian. Respondents who replied in the affirmative were coded as 1 and all others were coded as 0. The final measure of religiosity asks the respondent for a self-assessment of religiosity and is measured on a scale that ranges from 0 (‘not at all religious’) to 10 (‘very religious’). As is clear from its wording, this question is not about formal religious membership. It can thus be interpreted as a measure of the intensity of non-institutionalized Christian beliefs.


Identification with a Christian Democratic or conservative party in the sense of the Ann-Arbor model was operationalized as a simply dummy variable. Respondents who identified with the ÖVP in Austria; the CVP (now CD&V) or PSC (now CDH) in Belgium; the KF or KD in Denmark; the RPF, UMP or UDF in France; the CCD-CDU (now UDC), Forza Italia or NPSI in Italy; the CDA, CU or SGP in the Netherlands; the KRF or Høyre in Norway; and the CVP or EVP in Switzerland were coded as 1, while all others were coded as 0.


Finally, as mentioned above, we selected 21 observable attitudinal variables from the ESS to construct our latent variable ‘Radical Right Attitudes’. These cover a number of subdimensions of radical rightist thinking including attitudes towards the economic, social and cultural impact of immigrants, attitudes towards race and ethnicity, and attitudes towards immigrant and refugee rights. These variables were measured on a variety of scales. The full details of all 21 variables, as well as the full datasets for each country, can be found in the replication archive at


Since we have a significant number of variables in our model we did not use listwise deletion. Rather, we employed Multiple Imputation by Chained Equations (MICE), a very versatile imputation method that fills the gaps in the data set with a range of ‘plausible’ values. As our core dependent variable, one intervening variable and several of our indicator variables are dichotomous, we estimated the models with an extension of the Structural Equation Modelling (SEM) framework, implemented through the program MPlus, which allows for transparent handling of categorical variables (see Muthén 2004 for an overview).


To identify our model, the scales of the two latent variables (religiosity and radical rightist attitudes) had to be fixed. We did this by setting the coefficients for the paths from the latent constructs to an arbitrary indicator (praying and wages respectively) to one. Since we expect the basic structure outlined in our model to apply in all countries but the actual strength of the relationships to vary across systems, we estimated our models on a per-country basis with no equality constraints. Most parameters presented in the tables below are unstandardized regression coefficients. Exceptions are the effects on the dichotomous variables (identification with a Christian Democratic or conservative party, belonging to a Christian church/considering oneself a Christian, and radical right vote), which are represented by unstandardized probit coefficients. While all the relationships between variables were estimated simultaneously, we will discuss our findings from each regression in turn, so as to make interpretation easier.



Religiosity and radical right voting: findings and discussion

The overall fit between our model and our data is good. The Root Mean Square Error of Approximation is well below the conventional threshold of 0.1 in all countries and comes close to 0.05 in most countries, which indicates a ‘very good’ fit. The measurement models for religiosity and radical right attitudes also perform very well: all coefficients are significant (throughout this article we use the conventional 5 per cent threshold) and positive. Moreover, all are, by and large, within the same range. Full details of these measurement models can be found at


Turning now to the substantial relationships, Table 1 shows the regression of religiosity on the socio-demographics and enables us to see which of the different groups in the eight societies are, on average, more (or less) religious. The findings again point to a largely uniform pattern across the countries: holding other socio-demographic variables constant, men are considerably less religious than women and older citizens are more religious than younger people. Importantly, since the age groups 30-65 and 66+ have large positive coefficients, Table 1 also indicates that young men – who make up the social group that shows a disproportionally high level of support for the radical right in all West European countries – are also the group least likely to be religious. By contrast to gender and age, education (with the exception of Switzerland and Italy) and class have no significant effects once the other variables are controlled for.




Next, since previous research has shown that radical right-wing attitudes are an excellent predictor of the radical right vote, we turn our attention to the antecedents of these attitudes. As can be seen in Table 2, we find that education has a significant and strong negative effect on radical-right attitudes in all eight societies under study even when the other socio-demographic variables and religiosity are held constant. This result is in line with existing research that found that higher levels of education are usually associated with more liberal views (Coenders and Scheepers 2003; Weakliem 2002). Class has the expected significant positive effect on radical-right attitudes: working-class voters, farmers and voters categorized as belonging to the petty bourgeoisie show a greater propensity of holding radical right-wing attitudes than other class groups even after controlling for education. The only exception here is the Netherlands, where the effect of class is still positive but is somewhat weaker and is not statistically significant. The effect of age on radical right-wing attitudes is mostly positive – i.e. older people have, on average, and after controlling for the other factors, slightly more radical right-wing attitudes than their younger compatriots. The two exceptions here are Italy, where age effects are reversed, and the Netherlands, where they are insignificant. By contrast, gender has no discernible effect on radical right-wing attitudes, with the exception of Norway, where men have somewhat more radical right-wing attitudes than women.


Finally, with respect to religiosity, we find that this variable has hardly any effect at all on people’s attitudes towards radical right issues: in five of the eight countries (including the two denominationally mixed ones), the coefficients are not significantly different from zero, and in the three remaining societies, the effect is very weak. From the findings presented in Table 2, we can conclude that both hypotheses H2a and H2b are falsified: in the eight West European societies under study, religious people are neither more nor less likely to adopt negative attitudes towards immigrants than their agnostic compatriots once the background variables are controlled for.



From Table 2 alone, one might be tempted to conclude that religiosity has no political consequences in Western Europe’s secularised societies. However, Table 3, which shows the probit regression of Christian Democratic / conservative party identification on religiosity as well as on the set of socio-demographic variables, indicates that this assertion would be incorrect: religiosity continues to have a huge impact on one’s likelihood of identifying with a Christian Democratic or conservative party even if the effects of socio-demographic variables are controlled for. The coefficients are substantial and significant in all countries, although it is interesting to note that the effect is a little weak in Italy and is unusually strong in the Netherlands. In the Netherlands the effect is substantially stronger for Catholics than it is for Protestants, while in Switzerland it is marginally stronger for Catholics than it is for Protestants (not shown as a table). Of course, with reference to our hypotheses, the strong impact of religiosity on party identification is a necessary but not a sufficient condition for the validity of Hypothesis H1, which suggested that religious people are less likely to vote for the radical right because they are firmly attached to Christian Democratic or conservative parties.




Table 3 also shows that men are more likely to identify with a Christian Democratic or conservative party than women. That said, since men are less religious in all countries, the direct positive effect of gender on party identification will often be effectively neutralised (in Belgium, France, and Norway) or even outweighed (in Denmark) by a negative indirect effect of gender via religiosity. The effect of class is negative throughout Western Europe, but is only significant in the two Scandinavian countries, where it most likely reflects the strength of the labour/capital cleavage. As for education, its effect is significantly positive in Austria, France, and Norway, but insignificant in all other countries. Finally, the effect of age is significant only in France, where it is huge. Again, this is after controlling for religiosity, which is already positively related to age, meaning that the direct and indirect effects of age will reinforce each other.


The (residual) correlation between identification with a Christian Democratic / conservative party and radical right-wing attitudes is negligible in all countries (see Table 4). This implies that supporters of these parties are neither more nor less likely to adopt negative attitudes towards immigrants than other voters once religiosity and socio-demographics are held constant.




Table 5 shows the probit regression of a vote for a party of the radical right on radical right-wing attitudes, religiosity, party identification, and the standard set of socio-demographic variables. A first observation is that the well-known effects of gender, age, class, and education are not significantly different from zero in most countries. The obvious explanation for this finding is that the strong effects of these socio-demographic attributes often found in studies of the radical right vote basically reflect the group differences in the strength of right wing attitudes that can be discerned from Table 2. That is, while education, for example, has a massive impact on attitudes, which in turn substantially affects the vote, the correlation between education and the vote disappears once attitudes are controlled for.




The explanatory power of attitudes is all the more evident in Table 5 if we look at the coefficients of radical right-wing attitudes. These are significant, large, and within the same range in seven of the eight countries. Table 5 therefore confirms that radical right-wing attitudes are a powerful predictor of the radical right vote, and that support for these parties should not be interpreted as a non-ideological, protest vote (van der Brug et al. 2000; van der Brug and Fennema 2003). The only exception here is Italy, where the effect is rather weak and is insignificant. This can be explained in part by the fact that the vast majority of Italian radical right-wing voters voted for the Alleanza Nazionale – a party has moderated its profile in recent years and that historically displayed limited hostility to foreigners in its ideology anyway (Carter 2005; Newell 2000).


The direct effect of religiosity on the probability of voting for a radical right party is less uniform across our countries. In Italy, religiosity has a borderline significant negative impact, while in Switzerland (where the effect is virtually identical for Catholics and Protestants) and France being religious clearly raises the probability of a radical right vote. Put differently, this indicates that in Switzerland and France the radical right appeals to religious voters net of them being encapsulated by Christian Democratic or conservative parties and of them being more or less anti-immigrant than other people. While there is no obvious explanation for this in the case of the Swiss SVP, the findings for France are in line with the FN’s appeals to a small but distinct fundamentalist Catholic constituency. In the five other countries, religiosity has no significant direct effect on the likelihood of voting for a radical right party – a finding that lends support to Hypothesis H3.


Finally, Table 5 indicates that the effects of identifying with a Christian Democratic or conservative party on the likelihood of voting for a party of the radical right are negative and often very large, although they are not significant in three of the eight countries under study. Combined with the results shown in Table 3, this provides further evidence for the validity of Hypothesis H1: in many cases, religious people are less likely to vote for the radical right because they are firmly attached to Christian Democratic or conservative parties.




From our model we can conclude that religiosity does play a significant role in explaining the radical right vote in Western Europe but that the picture is somewhat more complex than the (early) psychological research would suggest. In a bid to disentangle the various mechanisms, Table 6 illustrates the direct, indirect and total effects of religiosity on the likelihood of casting a vote for a party of the radical right in all eight countries under study. The first row of the table shows that the effect of religiosity via party identification is (often strongly) negative in all countries and significantly so in five of eight. By contrast, the second row illustrates that the effect of religiosity via radical right-wing attitudes is mostly weak and insignificant. The sum of these indirect effects (reported in the third row) is negative in all countries and significantly so in five of them. The direct effect of religiosity on the likelihood of casting a vote for a party of the radical right is reported in the fourth row of the table, which repeats the information from Table 5 above. The direct effect of religiosity is not uniform across the countries: in five of the eight societies it is not significant, whereas in France and Switzerland it raises the probability of a radical right vote, and in Italy it lowers this probability. These findings clearly highlight the importance of national contexts, and underline just how much religiosity, and indeed what it means to be religious, are shaped by distinct national influences. The final row of Table 6 reports the total effect of religiosity (indirect and direct). This is negative and significant in five countries, is negative and borderline-significant in Austria, and is not significantly different from zero in France and Switzerland.





The question that this article set out to investigate was whether religiosity influences the likelihood of an individual casting a vote for a party of the radical right in Western Europe. Our interest in this issue was guided by existing bodies of literature that led us to believe that a link between religiosity and radical right voting might well exist and by the fact that very few comparative studies have examined the subject. In an attempt to answer our question, we specified four separate hypotheses regarding the relationship between religiosity and voting for a radical right party. These enabled us to untangle the different effects that religiosity has on the radical right vote. In the first instance we suggested that religiosity might prevent people from voting for the radical right because religious people tend to develop an identification with a Christian Democratic or conservative party, and are thus simply not available to the parties of the radical right (Hypothesis H1). We also proposed that religiosity might have an effect on the support for the parties of the radical right via attitudes, and that this effect could either be negative (Hypothesis H2a) or positive (Hypothesis H2b). Lastly, we suggested that once attitudes and socio-demographic attributes are controlled for, there would be no substantial relationship between religiosity and the radical right vote (Hypothesis H3).


Somewhat surprisingly, this last hypothesis is not born out in practice in three of the eight countries, where there are significant direct effects of religiosity. There is no obvious explanation for the moderate negative direct effect of religiosity on the likelihood of a radical right vote in Italy, or its clearly stronger positive effect in Switzerland. By contrast, however, the positive effect of religiosity on the likelihood of a vote for the radical right in France is more easily accounted for. Not only has the Front National always taken a tough stand on issues such as abortion, homosexuality and the role of the church, but the party also has links with ultra-Catholic groups opposed to the church’s alleged ‘liberalism’ (Minkenberg 2003; Veugelers 2000). While studies of the Front National’s electorate demonstrate that most of its voters are overwhelmingly attracted by the party’s stance on immigration and are unconcerned about issues related to the church and its traditional teachings, and while the official church has become a leading critic of the FN’s anti-minority policies (Mayer and Perrineau 1992; Veugelers 2000), it is quite possible that these elements of the party’s appeal are attractive to a small segment of Catholic fundamentalists.


It also transpires that neither Hypothesis H2a nor Hypothesis H2b is born out in practice. We found no evidence that religious people are less likely to vote for the radical right because they are more altruistic, tolerant and compassionate and thus less likely to espouse negative attitudes towards immigrant; and nor did we find evidence to support the contrary suggestion that such people are more likely to vote for these parties because their religiosity is linked to higher levels of prejudice. While the second link in this causal chain (that anti-immigrant attitudes are very strong predictors of radical right voting) is confirmed in our findings (except in Italy, where, it has been argued, the AN is substantively different from other radical right parties), the first link is not: we found no relation between religiosity and anti-immigrant attitudes. All the effects were either statistically insignificant or irrelevant in substantial terms.


Of course, whether the absence of an overall relationship between religiosity and anti-immigrant sentiment is due to different mechanisms that counter-balance each other or to a true non-relationship cannot be ascertained with the data at hand. Yet, if we accept the absence of a link between religiosity and anti-immigrant attitudes at face value, this is clearly at odds with the findings of the earlier literature, and thus raises interesting questions. Setting aside concerns over the conceptual and methodological rigour of the early studies, one possible explanation for this contradiction would be that religiosity and ethnocentrism may well have been linked when these previous analyses were carried out (mainly in the 1950s and 1960s), but that this relationship has since waned and disappeared. Indeed, religious teachings, values and convictions are unlikely to have remained unaffected by social change, secularization and globalization, and it is thus very likely that belief systems are today less ‘closed’ than they used to be, and religious outlooks less ‘particularistic’. Yet the problem with this line of reasoning is that, everything else being equal, we would expect to have seen greater support for parties of the radical right in the 1950s and 1960s as compared to today. And this is clearly not the case: the radical right has been electorally more successful in the last two decades than at any point since World War Two.


Perhaps then the explanation is not temporal but geographical. Indeed, the vast majority of the studies that pointed to a link between religiosity and ethnocentrism were carried out in the US and it may well simply be that, while there was a relationship between religiosity and ethnocentrism among these respondents, that same relationship does not exist within West European electorates. This of course, once again, points to the importance of national contexts, both in terms of what religion means and entails in different societies and in terms of its manifestation and representation in the political system.


Clearly we can only speculate about the reasons why we found no link between religiosity and anti-immigrant attitudes and, as we noted above, it could be that there are different relationships between religiosity and anti-immigrant sentiment that actually counter-balance each other. From our more narrow perspective, however, regardless of this relationship, we can confidently conclude that in the societies under study, religiosity does not affect the vote for the radical right because of any influence religiosity might have on anti-immigrant attitudes.


Attitudes, however, remain crucial. Indeed, while the first link in our suggested causal chain (that religious people have either higher or lower levels of anti-immigrant sentiment) was falsified by our findings, the second was not. Like others (van der Brug et al. 2000; van der Brug and Fennema 2003), we found that negative attitudes towards immigrants are very strong predictors of radical right voting. Our analyses thus provide further evidence that voters who vote for parties of the radical right are doing so because they agree with the policies of these parties, and in particular with their anti-immigration appeals.


In contrast to H3, H2a and H2b, Hypothesis H1 is borne out in practice: in all countries religiosity has a substantial and statistically positive effect on the likelihood of a voter identifying with a Christian Democratic or conservative party. This in turn massively reduces the likelihood of casting a vote for a party of the radical right in many countries. We therefore conclude that ‘good Christians’ are neither especially tolerant towards ethnic minorities nor attracted by the radical right’s anti-immigrant rhetoric. Rather, to a large degree, they are simply still attached to Christian Democratic or conservative parties, and although they do not necessarily vote for these parties, this attachment ‘vaccinates’ them against voting for a party of the radical right (see Scarbrough 1984 on this idea of ‘vaccination’ in an electoral context).


This demonstrates that religiosity continues to be an important predictor of electoral choice. Yet, this ‘vaccine effect’ is likely to become weaker with time due to general de-alignment trends induced by social modernization and value change. Just as the parties of the mainstream left can no longer count on a traditional base of working class voters, Christian Democratic and conservative parties are today faced with fewer religious voters than they once were. Thus, in spite of still being able to ‘encapsulate’ religious voters, this natural reservoir of support is shrinking. All other things being equal, therefore, this points to an increase in the potential of radical right parties.




We would like to thank John Bartle, Thomas Poguntke, Elinor Scarbrough and Jack Veugelers for their valuable comments and suggestions on an earlier version of this article. We are also grateful to two anonymous reviewers and the editor of this journal for their helpful comments. Of course, the usual disclaimer applies.






Adorno, Theodor W., Else Frenkel-Brunswik, Daniel J. Levinson, and R. Nevitt Sanford (1950). The Authoritarian Personality. New York: Harper and Row.


Allport, Gordon W., and Bernard M. Kramer (1946). ‘Some Roots of Prejudice’, The Journal of Psychology, 22, 9-39.


Allport, Gordon W., and Michael J. Ross (1967). ‘Personal Religious Orientation and Prejudice’, Journal of Personality and Social Psychology, 5:4, 432-43.


Altemeyer, Bob (2003). ‘Why do Religious Fundamentalists Tend to be Prejudiced?’, International Journal for the Psychology of Religion, 13:1, 17-28.


Arendt, Hannah (1951). The Origins of Totalitarianism. New York: Harcourt, Brace and Co.


Argue, Amy, David R. Johnson, and Lynn K. White (1999). ‘Age and Religiosity: Evidence from a Three-Wave Panel Analysis’, Journal for the Scientific Study of Religion, 38:3, 423-35.


Arzheimer, Kai, and Elisabeth Carter (2006). ‘Political Opportunity Structures and Right-Wing Extremist Party Success’, European Journal of Political Research, 45:3, 419-43.


Bartle, John (1998). ‘Left-Right Position Matters, but does Social Class? Causal Models of the 1992 British General Election’, British Journal of Political Science, 28:3, 501-29.


Berman, Sheri (1997). ‘Civil Society and the Collapse of the Weimar Republic’, World Politics, 49:3, 401-29.


Betz, Hans-Georg (1994). Radical Right-Wing Populism in Western Europe. London: Macmillan.


Billiet, Jaak B. (1995). ‘Church Involvement, Ethnocentrism, and Voting for a Radical Right-Wing Party: Diverging Behavioral Outcomes of Equal Attitudinal Dispositions’, Sociology of Religion, 56:3, 303-26.


Billiet, Jaak, and Hans De Witte (1995). ‘Attitudinal Dispositions to Vote for a “New” Extreme Right-Wing Party: The Case of “Vlaams Blok”’, European Journal of Political Research, 27:2, 181-202.


Billiet, Jaak, Ann Carton, and Rob Eisinga (1995). ‘Contrasting Effects of Church Involvement on the Dimensions of Ethnocentrism: An Empirical Study among Flemish Catholics’, Social Compass, 42:1, 97-107.


Broughton, David, and Hans-Martien ten Napel, eds. (2000). Religion and Mass Electoral Behaviour in Europe. London: Routledge.


Carter, Elisabeth (2005). The Extreme Right in Western Europe: Success or Failure? Manchester: Manchester University Press.


Childers, Thomas (1983). The Nazi Voter: The Social Foundations of Fascism in Germany, 1919-1933. Chapel Hill, NC: The University of North Carolina Press.


Coenders, Marcel, and Peer Scheepers (2003). ‘The Effects of Education on Nationalism and Ethnic Exclusionism: an International Comparison’, Political Psychology, 24:2, 313-43.


Crewe, Ivor (1983). ‘The Electorate: Partisan Dealignment Ten Years On’, West European Politics, 6:4, 183-215.


Crewe, Ivor, and Bo Särlvik (1983). Decade of Dealignment. Cambridge: Cambridge University Press.


Dalton, Russell J., Scott C. Flanagan, and Paul Allan Beck, eds. (1984). Electoral Change in Advanced Industrial Democracies: Realignment or Dealignment? Princeton, NJ: Princeton University Press.


Dalton, Russell J. (1996). ‘Political Cleavages, Issues and Electoral Change’, in Lawrence LeDuc, Richard G. Niemi, and Pippa Norris (eds.), Comparing Democracies: Elections and Voting in Global Perspective. Thousand Oaks, CA: Sage, 319-42.


della Porta, Donatella (2002). ‘Immigration and Protest: New Challenges for Italian Democracy’, South European Society and Politics, 5:3, 108-32.


Eisinga, Rob, Jaak Billiet, and Albert Felling (1999). ‘Christian Religion and Ethnic Prejudice in Cross-National Perspective: A Comparative Analysis of the Netherlands and Flanders (Belgium)’, International Journal of Comparative Sociology, 40:3, 375-93.


Eisinga, Rob, Albert Felling, and Jan Peters (1990). ‘Religious Belief, Church Involvement, and Ethnocentrism in the Netherlands’, Journal for the Scientific Study of Religion, 29:1, 54-75.


Fallend, Franz (2004). ‘The Rejuvenation of an “Old Party”? Christian Democracy in Austria’, in Steven van Hecke and Emmanuel Gerard (eds.), Christian Democratic Parties in Europe since the End of the Cold War. Leuven: Leuven University Press, 79-104.


Falter, Jürgen W. (1991). Hitlers Wähler. Munich: Beck.


Girvin, Brian (2000). ‘The Political Culture of Secularization: European Trends and Comparative Perspectives’, in David Broughton and Hans-Martien ten Napel (eds.), Religion and Mass Electoral Behaviour in Europe. London: Routledge, 7-27.


Glock, Charles Y., and Rodney Stark (1966). Christian Beliefs and Anti-Semitism. New York: Harper and Row.


Gorsuch, Richard L., and Daniel Aleshire (1974). ‘Christian Faith and Ethnic Prejudice: A Review and Interpretation of Research’, Journal for the Scientific Study of Religion, 13:3, 281-307.


Grunberger, Richard (1971). A Social History of the Third Reich. London: Penguin


Hanley, David, ed. (1994). Christian Democracy in Europe: A Comparaive Perspective. London: Pinter.


Ignazi, Piero (1992). ‘The Silent Counter-Revolution: Hypotheses on the Emergence of Extreme Right-Wing Parties in Europe’, European Journal of Political Research, 22:1, 3-34.


Johnson, Daniel C. (1997). ‘Formal Education vs. Religious Belief: Soliciting New Evidence with Multinomial Logit Modeling’, Journal for the Scientific Study of Religion, 36:2, 231-46.


Kitschelt, Herbert (1995). The Radical Right in Western Europe: A Comparative Analysis. Ann Arbor, MI: The University of Michigan Press.


Kornhauser, William (1959). The Politics of Mass Society. London: The Free Press.


Lipset, Syemour Martin (1971). Political Man. London: Heinemann.


Lubbers, Marcel, and Peer Scheepers (2000). ‘Individual and Contextual Characteristics of the German Extreme Right-Wing Vote in the 1990s: A Test of Complementary Theories’, European Journal of Political Research, 38:1, 63-94.


Lubbers, Marcel, Mérove Gijsberts, and Peer Scheepers (2002). ‘Extreme Right-Wing Voting in Western Europe’,  European Journal of Political Research, 41:3, 345-78.


Lucardie, Paul, and Hans-Martien ten Napel (1994). ‘Between Confessionalism and Liberal Conservatism: the Christian Democratic Parties of Belgium and the Netherlands’, in David Hanley (ed.), Christian Democracy in Europe: A Comparaive Perspective. London: Pinter, 51-70.


Madeley, John T. S. (2004). ‘Life at the Northern Margin: Christian Democracy in Scandinavia’, in Steven van Hecke and Emmanuel Gerard (eds.), Christian Democratic Parties in Europe since the End of the Cold War. Leuven: Leuven University Press, 217-41.


Mair, Peter, Wolfgang C. Müller, and Fritz Plasser (2004). ‘Introduction: Electoral Challenges and Party Responses’, in Peter Mair, Wolfgang C. Müller, and Fritz Plasser (eds.), Political Parties and Electoral Change. London: Sage, 1-19.


Mayer, Nonna (1998). ‘The Front National in the Plural’,  Patterns of Prejudice, 32:1, 4-24.


Mayer, Nonna, and Pascal Perrineau (1992). ‘Why do they Vote for Le Pen?’, European Journal of Political Research, 22:1, 127-41.


McGann, Anthony J., and Herbert Kitschelt (2005). ‘The Radical Right in the Alps: Evolution of Support for the Swiss SVP and Austrian FPÖ’,  Party Politics, 11:2, 147-71.


Middleton, Russell (1973). ‘Do Christian Beliefs Cause Anti-Semitism?’, American Sociological Review, 38:1, 33-52.


Miller, Alan S., and Rodney Stark (2002). ‘Gender and Religiousness: Can Socialization Explanations be Saved?’, American Journal of Sociology, 107:6, 1399-423.


Miller, Warren E., and J. Merrill Shanks (1996). The New American Voter. Cambridge, MA: Harvard University Press.


Minkenberg, Michael (2003). ‘The Policy Impact of Church-State Relations: Family Policy and Abortion in Britain, France, and Germany’, West European Politics, 26:1, 195-217.


Mommsen, Hans (1996). The Rise and Fall of Weimar Democracy. Chapel Hill, NC: The University of North Carolina Press.


Mudde, Cas (2007). Populist Radical Right Parties in Europe. Cambridge: Cambridge University Press.


Muthén, Bengt O. (2004). Mplus Technical Appendices. Los Angeles, CA: Muthén and Muthén, (retrieved 29/11/2005).


Newell, James L. (2000). ‘Italy: The Extreme Right comes in from the Cold’, Parliamentary Affairs, 53:3, 469-85.


Norris, Pippa (2005). Radical Right: Voters and Parties in the Electoral Market. New York: Cambridge University Press.


Norris, Pippa, and Ronald Inglehart (2004). Sacred and Secular: Religion and Politics Worldwide. New York: Cambridge University Press.


Petersen, Larry R., and K. Peter Takayama (1984). ‘Religious Commitment and Conservatism: Toward Understanding an Elusive Relationship’, Sociological Analysis, 45, 355-71.


Pettigrew, Thomas F. (1959). ‘Regional Differences in Anti-Negro Prejudice’, Journal of Abnormal and Social Psychology, 59:1, 28-36.


Ploch, Donald R. (1974). ‘Religion as an Independent Variable: A Critique of Some Major Research’, in Allan W. Eister (ed.), Changing Perspectives in the Scientific Study of Religion. New York: Wiley and Sons, 275-94.


Rokeach, Milton (1960). The Open and Closed Mind: Investigations into the Nature of Belief Systems and Personality Systems. New York: Basic Books.


Roof, Wade Clark (1974). ‘Religious Orthodoxy and Minority Prejudice: Causal Relationship or Reflection of Localistic World View?’, American Journal of Sociology, 80, 643-64.


Royston, Patrick (2005). ‘Multiple Imputation of Missing Values: Update of Ice’, The Stata Journal, 5: 188–201.


Rubin, Donald B. (1987). Multiple Imputation for Nonresponse in Surveys. New York: Wiley.


Scarbrough, Elinor (1984). Political Ideology and Voting: An Exploratory Study. Oxford: Clarendon Press.

Scheepers, Peer, Mérove Gijsberts, and Evelyn Hello (2002). ‘Religiosity and Prejudice against Ethnic Minorities in Europe: Cross-National Tests on a Controversial Relationship’, Review of Religious Research, 43:3, 242-65.


Studlar, Donley T. (1978). ‘Religion and White Racial Attitudes in Britain’, Ethnic and Racial Studies, 1:3, 306-15.


Sumner, William Graham (1906). Folkways: A Study of the Sociological Importance of Usages, Manners, Customs, Mores and Morals. Boston, MA: Ginn.


Taggart, Paul (1995). ‘New Populist Parties in Western Europe’, West European Politics, 18:1, 34-51.


van Buuren, Stef, and Karin Oudshoorn (1999). Flexible Multivariate Imputation by MICE. Leiden: TNO Prevention and Health, (retrieved 7/10/2005).


van der Brug, Wouter (2003). ‘How the LPF Fuelled Discontent: Empirical Tests of Explanations of LPF Support’, Acta Politica, 38:1, 89-106.


van der Brug, Wouter, and Meindert Fennema (2003). ‘Protest or Mainstream? How the European Anti-Immigrant Parties Developed into Two Separate Groups by 1999’, European Journal of Political Research, 42:1, 55-76.


van der Brug, Wouter, Meindert Fennema, and Jean Tillie (2000). ‘Anti-Immigrant Parties in Europe: Ideological or Protest Vote?’, European Journal of Political Research, 37:1, 77-102.


van Hecke, Steven, and Emmanuel Gerard Eds. (2004). Christian Democratic Parties in Europe since the End of the Cold War. Leuven: Leuven University Press.


Veugelers, John W. P. (2000). ‘Right-Wing Extremism in Contemporary France: A “Silent Counterrevolution”?’,  Sociological Quarterly, 41:1, 19-40.


Walter, Tony, and Grace Davie (1998). ‘The Religiosity of Women in the Modern West’, British Journal of Sociology, 49:4, 640-60.


Weakliem, David L. (2002). ‘The Effects of Education on Political Opinions: An International Study’, International Journal of Public Opinion Research, 14:2, 141-57.


Westle, Bettina, and Oskar Niedermayer (1992). ‘Contemporary Right-Wing Extremism in West Germany: “The Republicans” and their Electorate’, European Journal of Political Research, 22:1, 83-100.


Ziebertz, Hans-Georg, William K. Kay, and Ulrich Riegel, eds. (forthcoming). Youth in Europe III. Münster, Germany: LIT.





Table 1:        Determinants of religiosity


Religiosity on…AustriaBelgiumDenmarkFranceItalyNeths.NorwaySwitz.
Age 30-650.58*0.43*0.52*0.33*0.32*0.21*0.37*0.66*
Age over 650.79*1.31*0.98*1.08*0.67*0.65*0.90*1.00*

Notes: Entries are unstandardized coefficients; standard errors are in brackets, *: p<.05





Table 2:        Determinants of radical right attitudes


Radical right attitudes on…AustriaBelgiumDenmarkFranceItalyNeths.NorwaySwitz.
Age 30-650.26*0.19*0.170.25*-0.23*-
Age over 650.60*0.30*0.56*0.34*-0.32*0.190.43*0.30*

Notes: Entries are unstandardized coefficients; standard errors are in brackets, *: p<.05





Table 3:        Determinants of Christian Democratic / conservative party identification


CD PID on…AustriaBelgiumDenmarkFranceItalyNeths.NorwaySwitz.
Age 30-650.*
Age over 650.150.330.290.80*0.070.14-0.01-0.11

Notes: Entries are unstandardized probit coefficients; standard errors are in brackets, *: p<.05

Table 4: Correlation of Christian Democratic / conservative party identification and radical right attitudes


Correlation with…AustriaBelgiumDenmarkFranceItalyNeths.NorwaySwitz.
Rad right att0.13*-0.080.13*0.19*0.13-0.01-0.04-0.04

Notes: Entries are correlations (Pearson); standard errors in brackets, *: p<.05.





Table 5:        Determinants of radical right voting


Radical right voting on…AustriaBelgiumDenmarkFranceItalyNeths.NorwaySwitz.
Rad right att0.72*0.59*0.60*0.65*0.190.62*0.59*0.50*
Age 30-65-0.26-0.09-0.220.720.15-0.04-0.30*-0.16
Age over 65-0.15-0.45-0.170.350.47-0.28-0.52*0.07
CD PID-0.92*-0.40-0.17-0.83*-0.26-0.50*-0.61*-0.69*

Notes: Entries are unstandardized probit coefficients; standard errors are in brackets, *: p<.05.
Test for CD PID is one-tailed.





Table 6:        Decomposition of the effect of religiosity


Religiosity on radical right votingAustriaBelgiumDenmarkFranceItalyNeths.NorwaySwitz.
Via CD PID-0.48*-0.26-0.06-0.30*-0.07-0.51*-0.29*-0.41*
Via Rad right att0.03-0.01-0.04*0.040.00-0.020.04*0.01
Total indirect-0.46-0.27-0.11*-0.26*-0.06-0.53*-0.26*-0.40*

Notes: Entries are unstandardized coefficients; standard errors are in brackets, *: p<.05.

Test for effect via CD PID is one-tailed.



Protest, Neo-Liberalism or Anti-Immigrant Sentiment: What Motivates the Voters of the Extreme Right in Western Europe?


1 The Research Problem: What Motivates the Voters of the Extreme Right?

The so-called “third wave” of post-war right-wing extremism (Beyme 1988) in Western Europe caught comparative political science by surprise. After the Second World War, the Extreme Right in Western Europe had been associated with the atrocities of the Nazis and their puppet regimes (Rydgren 2005) and was therefore politically isolated and insignificant in most countries of the region. But from the early 1980s on, parties that were dubbed as “Extremist”, “Radical”, “Populist” or “New” Right or any combination thereof1 and had been located at the margins of the political systems suddenly proved highly successful at the polls in countries such as Austria, Belgium, Denmark, France, Italy, Norway, and Sweden.

The diversity of these parties looked somewhat bewildering. Some had rather obvious connections with the old inter-war right while others qualified as “modern” (Ignazi 2002). Two of them pursued a separatist agenda (the Vlaams Blok and the Lega Nord) whereas the majority was firmly committed to national unity, and some of them had been founded as early as in the 1940s (the Italian MSI) whereas others (most notably New Democracy in Sweden) were only formed shortly before their first electoral successes.

But problems of terminology and idiosyncratic features notwithstanding, it soon became clear that these parties had some important commonalities and should be grouped into a single party family (Mudde 1996). While the members of this family may lack a common party label, and while there are few institutionalised structures to facilitate their transnational co-operation – two criteria that are frequently employed for party family membership, see Mair and Mudde 1998: 214-215 – the sociological profiles of their respective electorates turned out to be very similar. Moreover, the parties of the Extreme Right share a number of ideological features, in particular their concern about immigration, which became the single most important issue for these parties from the late 1980s on (Hainsworth 1992Kriesi 1999van der Brug and Fennema 2003).2

There is less agreement, however, as to what motivates the voters of the Extreme Right. In many of the earlier accounts, the notion of a (pure) “protest vote” features prominently. While there is no universally accepted definition of what constitutes a protest vote (but see Kang 2004), this literature suggests that protest reflects an unsatisfactory performance of the political system. Protest is therefore disconnected from ideology and should primarily be understood as “a vote against things” (Mayer and Perrineau 1992: 134). In a similar fashion, van der Brug and Fennema (2007: 478-479) argue that “the prime motive behind a protest vote is to show discontent with the political elite”, whereas political attitudes would be of less importance. This interpretation fits neatly into the discourse on anti-party sentiment that gained prominence in the early 1990s (Poguntke and Scarrow 1996), and there can be little doubt that at least some of the Extreme Right parties could benefit from widespread feelings of distrust and disaffection with the established parties.

The more recent literature, however, acknowledges that quite often, protest is not un-ideological at all but clearly directed “against the policy or the absence of policy in this respect [imigration and safety]” (Swyngedouw 2001: 218-219). Consequentially, the vast majority of comparative studies of the Extreme Right vote now adopt a theoretical framework that is based on the notion of a conflict between non-Western immigrants and the indigenous population over scarce resources (jobs, welfare benefits). Prominent examples of this approach include Jackman and Volpert (1996), Knigge (1998), Lubbers et al. (2002), Golder (2003a), and Arzheimer and Carter (2006), who all analyse the joint impact of immigration and unemployment on the electoral returns for the Extreme Right. More recently, Swank and Betz (2003) and Arzheimer (2008) have introduced the level of welfare benefits as an additional mediating variable.

Given the findings from this literature and the importance that the issue of immigration has gained for the parties of the Extreme Right, it makes obvious sense to assume that the voters of the Extreme Right are primarily motivated by concerns about immigration. Due to data restrictions, however, there is surprisingly little empirical evidence to support this view. The four studies by Jackman and Volpert, Knigge, Golder, and Swank and Betz are based on polity-level data alone and therefore have to take the anti-immigrant sentiment of the Extreme Right voters for granted. But even those comparative analyses that employ micro-data either assume a link between anti-immigrant sentiment and the Extreme Right vote or do have to rely on sub-optimal indicators.

Arzheimer and Carter (2006), for instance, present a hybrid model of Extreme Right voting that combines variables measured at the micro-level with information on the polity-level to capture the effects of “Political Opportunity Structures” on the individual vote. But this model does not include any items on individual political attitudes because the national election surveys on which their analysis is based “do not provide adequate data on attitudes” (Arzheimer and Carter 2006: 425). Rather, Arzheimer and Carter treat socio-demographic indicators like age, gender, and formal education as proxies for political preferences and values that might or might not dispose a respondent to vote for the Extreme Right (Arzheimer and Carter 2006: 421-422).

In a similar fashion, Lubbers et al. (2002: 357) estimate a complex multi-level model of Extreme Right voting. But because they use various data sources, they have to rely on single measure for anti-immigrant sentiment that is common to these data-sets, namely the question whether the respondents feels that “there are too many immigrants” in the country. While this is obviously a much more direct approach to the alleged link between anti-immigrant sentiment and the Extreme Right vote, operationalising a complex phenomenon like anti-immigrant sentiment with a single indicator is risky because this variable will be subject to both systematic and random measurement error. Likewise, even the useful study by van der Brug and Fennema (2003) that focuses exclusively on the question of whether the vote for the Extreme Right should be considered a “protest vote” relies on a single indicator to assess the impact of anti-immigrant sentiment on the Extreme Right vote, namely the subjective importance and satisfaction with immigration policies.

While anti-immigrant sentiment and (to a lesser degree) notions of “pure protest” dominate the recent discussion, two early but very influential accounts of the the “third wave” provided a rather different explanation for the success of the Extreme Right. Both Betz (1994) and Kitschelt (1995) claim that economic (neo-)liberalism is the key ingredient in the Extreme Right’s electoral “winning formula” (Kitschelt 1995: viii). According to them, “modern” parties like the Freedom Party in Austria or the Front National in France are enormously successful because they mix xenophobic statements with an attack on high taxation, the welfare state and its bureaucracy. Such a program would appeal to working class and lower middle class voters who feel that they do not benefit from “big government” but are likely to suffer from comparative disadvantages in a globalising labour market. More “traditional” Extreme Right parties like the German DVU and Republikaner, however, would never attract a similarly large constituency because they were wedded to the welfare policies of the inter- and postwar Extreme Right.

With the benefit of hindsight, the Extreme Right’s involvement with neo-liberal policies during the early 1990s now looks more like a brief fling. Consequentially, Betz (2003) has altoghether abandoned the idea that the Extreme Right does seriously pursue a “neo-liberal” agenda or has done so in the past, while Kitschelt has modified his original ideas considerably (McGann and Kitschelt 2005). Given the professional stature of both authors and the impact their respective monographs had on the field, an empirical test of the “winning formula” thesis is, however, overdue.

To summarise, while anti-immigrant sentiment has emerged as the most prominent motivation behind the Extreme Right vote in Western Europe, alternative accounts do exist and adequate tests of the respective causal links have by and large been restricted to a host of national studies (e.g. Billiet and Witte 1995Clark and Legge 1997Mughan and Paxton 2006). This is obviously problematic because each of these studies uses a different set of indicators, thereby rendering comparisons over time and countries invalid.

Fortunately, comparable data on attitudes towards immigrant as well as on electoral behaviour have recently become available with the first round of the European Social Survey (ESS). The aim of this article is therefore to make use of these data for modelling the effect of protest, immigrant sentiment and economic liberalism on the Extreme Right vote while at the same time controlling for a larger number of background variables than previous studies.

2 Data, Model and Methodology


Figure 1: A Simplified Model of the Extreme Right Vote in Western Europe


Data for the present study were collected in 2002/2003 under the auspices of the European Social Survey project. Of the 22 countries covered by this survey, seven were selected that have witnessed substantial support for the Extreme Right in recent years: Austria, Belgium, Denmark, France, Italy, the Netherlands, and Norway.3 While Golder (2003b) has argued that polity-level studies of the Extreme Right should also look at the “failed cases” (e.g. Spain or the UK) to avoid selection bias, it makes no sense to include them in micro-level analysis. If no one reports support for the Extreme Right, then there is simply nothing to model.4

Figure 1 represents the basic structure of the model. “Vote” is a dummy variable that takes the value of 1 if a respondent has voted for a party of the Extreme Right (see footnote 1) and the value of 0 if he or she has abstained or voted for another party. Vote is regressed on a a number of control variables as well as on the standard indicator (Arzheimer 2002) for non-ideological protest motives (“And on the whole, how satisfied are you with the way democracy works in this country?”), on sentiment towards immigrants, and on economic liberalism, thereby providing a direct test for the three most popular hypotheses about the motives or Extreme Right voters.


CultureHow important should it be for immigrants to be committed to the way of life in [country]
WagesAverage wages and salaries are generally brought down by people coming to live and work here
Skilled LabourPeople who come to live and work here help to fill jobs where there are shortages of workers
JobsWould you say that people who come to live here generally take jobs away from workers in [country], or generally help to create new jobs?
Social SecurityMost people who come to live here work and pay taxes. They also use health and welfare services. On balance, do you think people who come here take out more than they put in or put in more than they take out?
EconomyWould you say it is generally bad or good for [country]’s economy that people come to live here from other countries?
Cultural ThreatWould you say that [country]’s cultural life is generally undermined or enriched by people coming to live here from other countries?
Quality of LifeIs [country] made a worse or a better place to live by people coming to live here from other countries?
CrimeAre [country]’s crime problems made worse or better by people coming to live here from other countries?
Labour MigrationAll countries benefit if people can move to countries where their skills are most needed
Multi-CulturalismIt is better for a country if almost everyone shares the same customs and traditions
Religious DiversityIt is better for a country if there are a variety of different religions
Linguistic DiversityIt is better for a country if almost everyone is able to speak at least one common language
ImmigrationIf a country wants to reduce tensions it should stop immigration
Fair Share[Country] has more than its fair share of people applying for refugee status

Scales for “Wages”, “Skilled Labour”, “Labour Migration”, “Multi-Culturalism”, “Religious Diversity”, “Linguistic Diversity”, “Immigration” and “Fair Share’ run from 1 (“Agree Strongly”) to 5 (“Disagree Strongly”). All other scales run from 1 to 11.

Where necessary, scales were reversed so that high values refer to the pro-immigrant position.

Table 1: Indicators for Immigrant Sentiment

In the literature, immigrant sentiment is often portrayed as a complex phenomenon (Mughan and Paxton 2006). Moreover, given the different levels and patterns of immigration in Wester Europe, one cannot take for granted that interviewees from different countries respond to any given indicator in exactly the same way. Therefore, immigrant sentiment is conceptualised as a “latent” (not directly observable) variable in the model.5 Having a separate measurement model for this attitude makes the overall model more robust and allows one to assess the reliability of the indicators in comparative perspective.

The 15 indicators selected for the measurement of anti-immigrant attitudes (see Table 1) reflect two major components of anti-immigrant sentiment, namely perceptions of material and cultural threats. However, while these two sub-dimensions are conceptually separable (Mughan and Paxton 2006: 342-343), the respective items display a very high degree of intercorrelation in all countries under study and are therefore interpreted as indicators for a single latent variable. To ease the interpretation, items were rescaled so that positive values of the latent variable correspondent to pro-immigrant sentiment whereas negative values stand for anti-immigrant sentiment.


Income EqualisationThe government should take measures to reduce differences in income levels
Government InterventionThe less that government intervenes in the economy, the better it is for [country]
Trade UnionsEmployees need strong trade unions to protect their working conditions and wages

Scales run from 1 (“Agree Strongly”) to 5 (“Disagree Strongly”)

For “Government Intervention”, the scale was reversed so that high values refer to the economically liberal position, too.

Table 2: Indicators for Economic Liberalism

To test Betz’s and Kitschelt’s early hypothesis about the importance of pro-market attitudes for the Extreme Right vote, the model contains a second latent variable dubbed “Economic Liberalism”. It is constructed from three indicators that capture resistance against equalisation of incomes, against trade unions and against state intervention in the economy (see Table 2).

As outlined above, socio-demographic variables often play an important role as proxy variables for attitudes in the existing research on the Extreme Right because theory suggests various causal links between both groups of variables. For instance, ethnic competition theory suggests that higher levels of formal education should be associated with lower levels of anti-immigrant sentiment (because most non-Western migrants are unskilled) and therefore with a lower propensity to vote for the Extreme Right. Moreover, there is ample evidence that formal education promotes liberal values (e.g. Weakliem 2002), whose adoption should also reduce levels of anti-immigrant sentiment. Either way, once anti-immigrant sentiment is controlled for, socio-demographic variables should have only minimal direct effects on the vote.

Again, most existing research simply assumes that socio-demographics can be used as a proxy variables for anti-immigrant sentiment, precisely because good indicators for attitudes are not generally available. To test this assertion as well as to control for residual effects, the model contains a large selection of socio-demographic variables (gender, age, union membership, church attendance, class6, employment status, education7, household size, and relationship status8) that have been shown to have an effect on the Extreme Right vote in previous research. Both direct and indirect (via anti-immigrant sentiment and economic liberalism) effects link these variables and the vote.

Finally, it has been noted that the literature on the voters of the Extreme Right is empirically and analytically not well connected to the very large body of research on mainstream electoral behaviour (Arzheimer 2008). In a bid to overcome this unfortunate divide, two standard attitudinal measures were included in the model as additional controls: according to the Michigan school, party identification9 is the single most important predictor of electoral behaviour, whereas ideology (left-right self placement) plays a prominent role in spatial approaches to electoral behaviour that build on the work of Hotelling (1929) and Downs (1957). Not controlling for these important predictors could lead to significant bias in the results.

The presence of latent variables and the (block-causal)10 structure of the model call for Structural Equation Modelling (Kaplan 2000), a statistical technique that allows one to estimate the parameters for the relationships between several variables simultaneously. Estimation was carried out on a per-country basis so that the sign and strength of effects can be compared across polities.

While Structural Equation Modelling (SEM) is now a well-established technique, three complications remain. First, the vote for the extreme right is a dichotomous variable whereas SEM was originally developed for continuous variables. However, modern software allows one to specify a nonlinear link between a dichotomous variable and its antecedents to deal with this problem.11

Second, while levels of item non-response are generally very low in the ESS, even a small proportion of missing values adds up in a model with so many variables. To avoid bias, over-optimistic standard errors and the massive reduction of the sample size that would result from listwise deletion (i.e. complete case analysis), Multiple Imputation by Chained Equations (MICE, see van Buuren and Oudshoorn 1999) was applied to fill the gaps in the data with a range of plausible values. For each country, 21 separate imputations of the original data were created using Royston’s (2005) implementation of MICE in Stata. Since MICE is a stochastic procedure, the differences between these imputations reflect the uncertainty about the missing values. Results from the 21 separate analyses were then combined in Mplus according to the rules outlined in Rubin (1987). This somewhat complex procedure yields approximately unbiased parameter estimates and conservative standard errors that take the amount of missing data into account, thereby providing an additional margin of safety.

Finally, an (arbitrary) scale for the two latent variables (immigrant sentiment and economic liberalism) must be established to identify the model. This was done by assuming that these variables are standardised, i.e. that they have a mean of zero and unit variance.

3 Findings

3.1 Overall Fit and Measurement Models



Root Mean Squared Errors of Approximation (RMSEA) and number of observations (N), averaged over 21 imputations

Table 3: Fit of the Model

Amongst the many fit indices for Structural Equation Models that have been proposed in the literature, the Root Mean Squared Error of Approximation (RMSEA) is arguably the most popular at the moment because it has a well-known distribution and is less sensitive to the size of the sample than other measures (Garson 2008). Table 3 shows that in all seven countries, the RMSEA is well below the conventional threshold of 0.1 and actually comes very close to the value of 0.05, which indicates a very good fit.

Estimates for the measurement model for immigrant sentiment are equally encouraging (Table 4). All coefficients are statistically significant12 and have the correct sign. Moreover, for most indicators the parameters are roughly within the same range, implying that the respective indicators are more or less equivalent. Since the entries in Table 4 are really just unstandardised regression coefficients, their interpretation is straightforward. For instance, the last of the 15 items asks whether the respondent agrees that the country gets a “fair share” of refugees. If one now compares two respondents with average (0) and rather positive (1) feelings towards immigrants, this difference of one standard deviation on the latent variable results in a substantively higher (about 0.5 points on a five-point rating scale) level of agreement with the pro-refugee statement, with the strongest effect (0.558 points) in Austria and the weakest (0.316 points) in Italy.


Skilled Labour−0.173(0.023)−0.232(0.025)−0.207(0.025)0.023(0.026)
Social Security1.489(0.048)1.096(0.046)1.084(0.052)1.291(0.051)
Cultural Threat1.618(0.049)1.156(0.048)1.566(0.055)1.723(0.058)
Quality of Life1.455(0.039)1.265(0.040)1.469(0.046)1.537(0.043)
Labour Migration−0.196(0.025)−0.191(0.023)−0.112(0.027)−0.079(0.021)
Religious Diversity−0.507(0.026)−0.302(0.026)−0.466(0.030)−0.383(0.027)
Linguistic Diversity0.243(0.017)0.071(0.017)0.107(0.015)0.107(0.014)
Fair Share0.558(0.026)0.541(0.024)0.455(0.031)0.465(0.027)
Skilled Labour−0.352(0.028)−0.214(0.020)−0.125(0.017)
Social Security0.998(0.061)1.108(0.041)0.991(0.041)
Cultural Threat1.404(0.071)1.069(0.037)1.328(0.043)
Quality of Life1.199(0.062)1.141(0.037)1.157(0.034)
Labour Migration−0.183(0.026)−0.088(0.019)−0.101(0.019)
Religious Diversity−0.310(0.031)−0.214(0.019)−0.369(0.019)
Linguistic Diversity0.066(0.021)0.139(0.013)0.108(0.012)
Fair Share0.316(0.035)0.474(0.019)0.388(0.019)

Entries are unstandardised regression coefficients (WLSMV) based on 21 imputations. Standard errors in parentheses. See Table 1 in the appendix for the full text of the items

Table 4: Immigrant Sentiment: Measurement Model

The measurement model for economic liberalism (Table 5), however, works less well. More specifically, the item on government interventions is only loosely connected to the latent variable13 whereas the two other items perform well. This is probably due to an extremely skewed distribution of the responses: in all countries but Austria, majorities in excess of 70 per cent either support government interventions or are at least indifferent. However, since scepticism about government interventions obviously reflects the theoretical content of economic liberalism, the item was retained.


Income Equalisation0.470(0.054)0.288(0.037)0.277(0.072)0.430(0.053)
Government Intervention0.161(0.036)−0.187(0.040)0.036(0.069)−0.086(0.043)
Trade Unions0.748(0.082)0.388(0.050)0.648(0.206)0.425(0.051)
Income Equalisation0.333(0.066)0.354(0.036)0.361(0.031)
Government Intervention−0.033(0.039)0.015(0.023)0.010(0.026)
Trade Unions0.747(0.146)0.573(0.056)0.425(0.036)

Entries are unstandardised regression coefficients (WLSMV) based on 21 imputations. Standard errors in parentheses.

Table 5: Economic Liberalism: Measurement Model

3.2 Antecedents of Economic Liberalism and Immigrant Sentiment

Table 6 presents the coefficients for the regression of economic liberalism on a range of socio-demographic control variables. Rather unsurprisingly, the unemployed, trade unionists, and members of the working class show substantively lower levels of economic liberalism than other respondents. Conversely, interviewees holding university degrees show disproportionate support for market capitalism. However, the relative and absolute strength of these effects varies considerably. In France, for instance, education is the key factor, whereas class and trade union membership are dominant in Denmark or Austria. Moreover, gender makes a significant difference in all countries but Italy: even if a whole host of other socio-demographics is controlled for, men tend to support market mechanisms more strongly than women.


Age: 18-29−0.157(0.113)−0.103(0.181)−0.292(0.174)0.061(0.159)
Age: 30-450.041(0.082)0.403(0.141)0.244(0.119)0.062(0.138)
Age: over 65−0.012(0.112)−0.795(0.232)−0.123(0.195)−0.019(0.183)
Religion: none−0.005(0.081)−0.082(0.128)0.061(0.108)−0.093(0.119)
Church Attendance0.038(0.023)−0.076(0.047)0.066(0.050)0.038(0.043)
Trade Union Member−0.348(0.085)−0.551(0.132)−0.423(0.143)−0.068(0.220)
Petty Bourgeoisie0.385(0.158)0.728(0.234)0.388(0.224)0.324(0.232)
Working Class−0.427(0.103)−0.822(0.173)−0.485(0.157)−0.042(0.175)
University Degree0.103(0.093)0.990(0.170)0.261(0.140)0.907(0.131)
Household Size = 10.056(0.109)−0.108(0.202)0.222(0.206)0.136(0.194)
Age: 18-29−0.249(0.153)−0.077(0.134)0.038(0.127)
Age: 30-45−0.280(0.110)0.102(0.078)0.107(0.092)
Age: over 65−0.135(0.165)−0.219(0.139)−0.332(0.235)
Religion: none0.067(0.109)−0.002(0.083)−0.062(0.082)
Church Attendance0.012(0.028)−0.037(0.026)0.004(0.034)
Trade Union Member−0.229(0.119)−0.580(0.091)−0.645(0.086)
Petty Bourgeoisie0.016(0.134)0.282(0.235)−0.041(0.216)
Working Class−0.464(0.142)−0.245(0.104)−0.160(0.113)
University Degree0.228(0.146)0.551(0.083)0.604(0.094)
Household Size = 10.024(0.169)−0.042(0.148)0.019(0.157)

Entries are unstandardised regression coefficients (WLSMV) based on 21 imputations. Standard errors in parentheses.

Table 6: Regression of Economic Liberalism on Socio-Demographics

As discussed in sections 1 and 2, most comparative analyses of the extreme right vote in Western Europe rely on a putative link between socio-demographic indicators of group-membership on the one hand and anti-immigrant sentiment on the other. Table 7 demonstrates that this practice is justified, at least up to a degree: formal education emerges clearly as the single most important predictor of sentiment towards immigrants. In all seven countries studied here, respondents with a university degree report much more positive feelings towards immigrants than other interviewees. The difference is roughly equivalent to half a standard deviation of the latent variable and varies from 0.48 points in the Netherlands to 0.69 points in Austria. Moreover, even though education is controlled for, class has an effect, too: in most countries, members of the working class and the “petty bourgeoisie” display a much more negative attitude towards immigrants than other respondents. Other variables (unemployment in particular) have smaller and more erratic effects.


Age: 18-290.318(0.087)0.193(0.095)0.205(0.107)0.513(0.098)
Age: 30-450.229(0.065)0.137(0.072)0.057(0.076)0.417(0.083)
Age: over 65−0.301(0.086)−0.160(0.107)−0.148(0.107)0.044(0.109)
Religion: none0.280(0.064)−0.020(0.068)0.102(0.066)0.019(0.073)
Church Attendance0.057(0.018)0.039(0.022)0.085(0.029)0.041(0.025)
Trade Union Member−0.017(0.060)−0.031(0.064)−0.025(0.076)0.268(0.114)
Petty Bourgeoisie−0.306(0.118)−0.318(0.123)−0.143(0.153)0.031(0.164)
Working Class−0.454(0.077)−0.314(0.082)−0.206(0.083)−0.293(0.099)
University Degree0.694(0.072)0.593(0.090)0.645(0.086)0.669(0.077)
Household Size = 1−0.157(0.085)−0.222(0.105)−0.052(0.128)−0.116(0.106)
Age: 18-29−0.046(0.130)0.195(0.091)0.135(0.086)
Age: 30-450.154(0.087)0.242(0.059)0.107(0.066)
Age: over 65−0.166(0.129)−0.315(0.095)−0.346(0.134)
Religion: none0.300(0.087)0.078(0.067)0.117(0.056)
Church Attendance0.014(0.023)0.033(0.021)−0.039(0.023)
Trade Union Member−0.083(0.100)0.154(0.060)0.063(0.054)
Petty Bourgeoisie−0.192(0.107)−0.151(0.122)−0.353(0.166)
Working Class−0.444(0.106)−0.221(0.082)−0.256(0.074)
University Degree0.487(0.152)0.476(0.060)0.679(0.064)
Household Size = 1−0.092(0.141)0.178(0.102)−0.045(0.104)

Entries are unstandardised regression coefficients (WLSMV) based on 21 imputations. Standard errors in parentheses.

Table 7: Regression of Immigrant Sentiment on Socio-Demographics

3.3 The Extreme Right Vote in Comparative Perspective

Finally, Tables 8 and 9 presents the (probit) regression of the Extreme Right vote on socio-demographic and attitudinal variables. A first important finding is that once attitudinal variables are controlled for, in all of the seven countries socio-demographic variables have no significant effect on the vote whatsoever. Put differently, the large and persistent differences as regards the propensity of various social groups to vote for the Extreme Right that have been observed in national and comparative studies are entirely due to differences between these groups with respect to the five attitudinal variables included in the model.

As regards the protest motives, the effect of satisfaction with the way democracy works in one’s country is statistically insignificant in most countries. Only in Belgium and the Netherlands, there is a link between (dis)satisfaction and the Extreme Right vote. While the absolute value of the coefficient is small (-0.14 and -0.12 respectively), its potential impact is large because satisfaction was measured on a ten-point rating scale. However, the interquartile range for satisfaction in Belgium and the Netherlands (loosely speaking the difference between those who are fairly dissatisfied and those who are fairly satisfied) amounts to only 3 and 2 points respectively, which would result in a rather moderate impact on the likelihood of an Extreme Right vote. On balance, these results suggest that the role of “pure protest” motives is very limited.

Similarly, economic liberalism is obviously not a key ingredient in the electoral winning formula for the Extreme Right: its effects are insignificant in all countries. Crucially, the effect is negative (though statistically insignificant) for the voters of the French Front National, Kitschelt’s 1995 “master case” of a “new” rightist party.

On the other hand, positive sentiment towards immigrants generally exerts a significant negative effect on the vote. Put differently, concerns about immigrants and immigration policies emerge as major motivation for the voters of the Extreme Right in six out of seven countries. The single exception is Italy, where the effect is not significantly different from zero. This specific finding sheds an interesting light on the Alleanza Nazionale, whose supporters make up the vast majority14 of the Italian Extreme Right voters in the data set: first, even the Alleanza’s neo-fascist predecessor party MSI displayed only very limited hostility to foreigners (Newell 2000), and second, the party has moderated its profile so much in recent years that some scholars do not longer consider it as part of the Extreme Right. While one can obviously not judge a party by its voters, the results demonstrate that the Alleanza’s supporters are different in so far as they are apparently not particularly attracted by anti-immigrant rhetoric and policies. Rather, they seem to be motivated by their general left-right preferences and their identification with the party.

As regards ideology, the findings are similarly clear-cut: more right-leaning respondents are far more likely to vote for the extreme right even after immigrant sentiment is controlled for in all countries but Denmark and France, where the effect does not pass the conventional threshold of statistical significance. Again, this speaks against the idea that the voters of the Extreme Right are motivated by pure protest motives which are unrelated to policy considerations.

Finally, party identifications have a very strong and highly plausible effect on the Extreme Right vote: respondents who identify with these parties display a very high propensity to vote for them, whereas an identification with any other party acts as an effective deterrent. While this may seem fairly obvious (if not tautological), almost all existing analyses neglects the role of party identification. This is problematic precisely because party identification has such a strong effect on the vote. If this force is ignored, severe bias can result. Also, like with other parties, the match between party identification and voting behaviour is by no means perfect. The share of identifiers amongst the voters of the Extreme Right varies between 25 (Netherlands) and 67 (Italy) per cent, whereas between 54 (France) and 85 (Italy) per cent of the identifiers vote for the respective party.

This important role of party identification provides additional evidence against the pure protest hypothesis. Moreover, only when this strong yet imperfect link is controlled for, one can truly appreciate the importance the influence of immigrant sentiment and ideology: although the single most important predictor of the Extreme Right vote is statistically held constant, policy-related attitudes still exert a very strong influence.


Immigrant Sentiment−0.196(0.059)−0.189(0.057)−0.363(0.051)−0.201(0.063)
Economic Liberalism0.144(0.080)−0.025(0.124)−0.046(0.107)−0.147(0.109)
PID: Extreme Right2.251(0.215)1.820(0.216)1.928(0.244)1.410(0.277)
PID: other−0.761(0.212)−0.841(0.287)−0.646(0.153)−0.642(0.218)
Left-Right Placement0.157(0.052)0.110(0.042)0.068(0.037)0.088(0.052)
Satisfied: democracy−0.050(0.033)−0.139(0.039)−0.030(0.036)−0.075(0.045)
Age: 18-29−0.065(0.341)−0.033(0.251)0.127(0.216)−0.336(0.410)
Age: 30-45−0.050(0.220)0.013(0.210)−0.082(0.194)0.090(0.217)
Age: over 65−0.018(0.233)−0.249(0.338)0.303(0.242)0.404(0.358)
Religion: none0.411(0.199)0.121(0.179)0.137(0.152)−0.173(0.224)
Church Attendance0.080(0.060)−0.065(0.088)−0.117(0.084)−0.015(0.089)
Trade Union Member−0.044(0.205)0.039(0.185)0.221(0.196)0.250(0.328)
Petty Bourgeoisie−0.080(0.389)−0.001(0.401)−0.153(0.357)−0.381(0.527)
Working Class0.424(0.225)0.056(0.218)0.253(0.196)−0.035(0.244)
University Degree0.238(0.288)−0.329(0.445)−0.727(0.416)−0.037(0.297)
Household Size = 10.108(0.337)0.347(0.276)−0.322(0.266)−0.052(0.404)

Entries are unstandardised Probit regression coefficients (WLSMV) based on 21 imputations. Standard errors in parentheses.

Table 8: Regression of the Extreme Right Vote on Socio-Demographics and Attitudes I


Immigrant Sentiment−0.035(0.066)−0.254(0.041)−0.259(0.044)
Economic Liberalism0.148(0.075)0.028(0.048)0.087(0.070)
PID: Extreme Right2.498(0.302)1.434(0.159)1.363(0.113)
PID: other−1.171(0.467)−0.577(0.091)−1.052(0.150)
Left-Right Placement0.173(0.070)0.150(0.022)0.152(0.031)
Satisfied: democracy−0.023(0.054)−0.117(0.023)−0.018(0.025)
Age: 18-290.003(0.511)−0.240(0.175)−0.233(0.148)
Age: 30-450.012(0.324)−0.060(0.101)−0.208(0.130)
Age: over 650.288(0.433)−0.202(0.197)−0.294(0.249)
Religion: none0.240(0.334)0.156(0.105)−0.079(0.101)
Church Attendance−0.066(0.087)−0.050(0.039)−0.114(0.041)
Trade Union Member−0.028(0.514)−0.214(0.116)0.087(0.110)
Petty Bourgeoisie0.231(0.336)0.007(0.150)−0.215(0.252)
Working Class−0.300(0.440)0.081(0.127)−0.014(0.130)
University Degree0.161(0.431)−0.001(0.121)−0.226(0.160)
Household Size = 10.046(0.523)0.109(0.184)−0.008(0.178)

Entries are unstandardised Probit regression coefficients (WLSMV) based on 21 imputations. Standard errors in parentheses.

Table 9: Regression of the Extreme Right Vote on Socio-Demographics and Attitudes II


Figure 2: The Effect of Ideology, Immigrant Sentiment, and Party Identification on the Extreme Right Vote (No Party ID)


Figure 3: The Effect of Ideology, Immigrant Sentiment, and Party Identification on the Extreme Right Vote (Extreme Right Party ID)


Figure 4: The Effect of Ideology, Immigrant Sentiment, and Party Identification on the Extreme Right Vote (Other Party ID)

This is most readily seen when the findings are converted from the probit scale back to the “quantity of interest” (King et al. 2000), i.e. the probability of a vote for the Extreme Right. For instance, for a right-leaning (say 7 on the ideology scale) respondent from Norway who identifies with the Freedom Party (1) and has rather negative (-1) attitudes towards immigrants, one would simply multiply these values with their respective coefficients, add the constant and plug the result (-1.699 + 7 × 0.152 + 1 × 1.363 + -1 × -0.259 = 0.987) into the standard cumulative density function Φ to obtain the probability of an Extreme Right vote (0.838).15 Figures 24 show how this probability varies with ideology (the solid, short-dashed and long dashed lines), immigrant sentiment, and party identification.16

From Figures 2, it is readily apparent that both ideology and immigrant sentiment have a sizeable impact amongst non-partisans: for a right-leaning voter who dislikes immigrants, the probability of a vote for the Freedom Party quickly approaches 40 per cent, while this probability is 20 per cent or less for left-leaning voters, especially if they are favourably disposed towards immigrants.17 But even amongst those respondents who identify with the Freedom Party (see Figure 3), the probability of an Extreme Right vote is clearly less than 100 per cent and varies considerably with ideology and immigrant sentiment (Figure 3).18 Perhaps the most interesting constellation is depicted in Figure 4. Here, one can see that even respondents who identify with another party have a sizeable probability of voting for the Freedom Party, provided that they are right-leaning and strongly oppose immigration (cf. the upper-left corner of the graph). While such a vote would still be a rather rare event, the probability of an Extreme Right vote is considerably (i.e. roughly ten times) higher in this group than amongst those respondents who have a more favourable attitude towards immigrants (cf. the lower-right corner of the graph).



Entries are ratios of the expected vote shares amongst anti-immigrant (-1) and pro-immigrant (+1) centrist (5) citizens with no party identification. The (mostly insignificant) effects of all other variables were set to zero.

Table 10: The impact of immigrant sentiment amongst independents in comparative perspective

Further graphical comparisons between countries are hampered by the fact that the base level of Extreme Right support (as reflected by the constant in Tables 8 and 9) varies considerably, resulting in essentially flat lines for countries with low levels of Extreme Right support. Therefore, ratios of predicted vote shares were calculated to put the impact of immigrant sentiment into comparative perspective. These calculations focus on a group that is of particular interest for political strategists in all West European countries: centrist citizens who are not attached to a particular party. In Norway, for instance, members of this group who display a rather positive (+1)19 attitude towards immigrants have a probability of roughly 12 per cent to vote for the Freedom Party. But for members of the same group who clearly dislike immigrants (-1 on the scale), the probability of a Freedom Party more than doubles. As can be gleaned from Table 10, for most countries this ratio is roughly in the same range.

The obvious exception is Italy, were immigrant sentiment makes virtually no difference as regards the electoral prospects of the Extreme Right.20 On the other hand, in Austria and Denmark support for the Extreme Right roughly triples with increasing anti-immigrant sentiment. While this figure might be slightly misleading in the case of Austria because support for the Freedom Party was generally very low amongst centrist independents so that the political impact of immigrant sentiment must remain limited within this group, attitudes vis-a-vis immigrants make all the difference in Denmark. Here, even those independent centrists who hold favourable views of immigrants have a seven per cent probability of voting for the Extreme Right. Consequentially, the model predicts that about one in five independents who strongly dislike immigrants but have otherwise centrist political preferences will vote for the Extreme Right.

4 Conclusion

Parties of the “Extreme”, “Radical” or ‘Populist” Right have become a permanent feature of West European politics, and since the mid-1980s, immigration has been the most important issue for them. Recent research has linked the levels of support these parties receive to polity-level variables such us unemployment and immigration. However, comparative micro-level evidence on the motives of their voters is still scarce.

Using recent survey data and a more appropriate measurement model than previous research, this article has demonstrated that Kitschelt’s 1995 hypothesis about the importance of neo-liberal policy preferences is not borne out in practice, and that the role of “pure protest” motives is very limited. Rather, the Extreme Right vote is driven by intense feelings of anti-immigrant sentiment in all countries but Italy. In line with theories of ethnic group conflict, these feelings are particularly strong within those segments of the electorate that compete with immigrants for scarce resources (low paid jobs and welfare benefits).

While the effects of anti-immigrant sentiment are strong, they are, however, moderated by general ideological preferences and party identification. On the basis of a new data set and a richer statistical model, these findings therefore confirm earlier claims that the Extreme Right vote can be explained by general causal mechanisms that apply to other parties, too (van der Brug and Fennema 2003). More specifically, the Extreme Right vote can be understood as the result of long-term political preferences and affiliations on the one hand and (immigration) policy-related attitudes on the other.21 Once these standard variables are measured adequately, it seems largely unnecessary to consider static22 and idiosyncratic factors like personality traits (Adorno et al. 1950) or alienation in today’s mass society (Kornhauser 1960). Rather, comparative electoral research should focus on the specific circumstances under which immigration is politicised and perceived as a problem that can move votes.


Adorno, Theodor W., Else Frenkel-Brunswik, Daniel J Levinson and R. Nevitt Sanford. 1950. The Authoritarian Personality. New York: Harper.

Arzheimer, Kai. 2002. Politikverdrossenheit. Bedeutung, Verwendung und empirische Relevanz eines politikwissenschaftlichen Begriffes. Wiesbaden: Westdeutscher Verlag.

Arzheimer, Kai. 2008. Die Wähler der extremen Rechten 1980-2002. Wiebaden: VS Verlag für Sozialwissenschaften.

Arzheimer, Kai and Elisabeth Carter. 2006. Political Opportunity Structures and Right-Wing Extremist Party Success. In: European Journal of Political Research, 45: 419–443.

Betz, Hans-Georg. 1994. Radical Right-Wing Populism in Western Europe. Houndmills, London: Macmillan.

Betz, Hans-Georg. 2003. The Growing Threat of the Radical Right. In: Merkl, Peter H. and Leonard Weinberg (eds.), Right-Wing Extremism in the Twenty-First Century, London: Frank Cass. 74–93.

Beyme, Klaus von. 1988. Right-Wing Extremism in Post-War Europe. In: Beyme, Klaus von (ed.), Right-Wing Extremism in Western Europe, London: Frank Cass. 1–18.

Billiet, Jaak and Hans de Witte. 1995. Attitudinal Dispositions to Vote for a ‘New’ Extreme Right–Wing Party: The Case of ‘Vlaams Blok’. In: European Journal of Political Research, 27: 181–202.

Clark, John A. and Jerome S. Legge. 1997. Economics, Racism, and Attitudes toward Immigration in the New Germany. In: Political Research Quarterly, 50: 901–917.

Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper.

Garson, David. 2008. Structural Equation Modeling. College of Humanities and Social Sciences, North Carolina State University.

Golder, Matt. 2003a. Electoral Institutions, Unemployment and Extreme Right Parties. A Correction. In: British Journal of Political Science, 33: 525–534.

Golder, Matt. 2003b. Explaining Variation in the Success of Extreme Right Parties in Western Europe. In: Comparative Political Studies, 36: 432–466.

Hainsworth, Paul. 1992. Introduction. The Cutting Edge. The Extreme Right in Post-War Western Europe and the USA. In: Hainsworth, Paul (ed.), The Extreme Right in Europe and the USA, London: Pinter. 1–28.

Hotelling, Harold. 1929. Stability in Competition. In: The Economic Journal, 39: 41–57.

Ignazi, Piero. 2002. The Extreme Right. Defining the Object and Assessing the Causes. In: Schain, Martin, Aristide Zolberg and Patrick Hossay (eds.), Shadows Over Europe. The Development and Impact of the Extreme Right in Western Europe, New York: Palgrave. 21–37.

Jackman, Robert W. and Karin Volpert. 1996. Conditions Favouring Parties of the Extreme Right in Western Europe. In: British Journal of Political Science, 26: 501–521.

Kang, Won-Taek. 2004. Protest Voting and Abstention Under Plurality Rule Elections: An Alternative Public Choice Approach. In: Journal of Theoretical Politics, 16: 79–102.

Kaplan, David. 2000. Structural Equation Modeling. Foundations and Extensions. Thousand Oaks: Sage.

King, Gary, Michael Tomz and Jason Wittenberg. 2000. Making the Most of Statistical Analysis. Improving Interpretation and Presentation. In: American Journal of Political Science, 44: 341–355.

Kitschelt, Herbert. 1995. The Radical Right in Western Europe. A Comparative Analysis. Ann Arbor: The University of Michigan Press.

Knigge, Pia. 1998. The Ecological Correlates of Right-Wing Extremism in Western Europe. In: European Journal of Political Research, 34: 249–279.

Kornhauser, William. 1960. The Politics of Mass Society. London: Routledge and Kegan Paul.

Kriesi, Hanspeter. 1999. Movements of the Left, Movements of the Right: Putting the Mobilization of Two New Types of Social Movements into Political Context. In: Kitschelt, Herbert, Peter Lange, Gary Marks and John D. Stephens (eds.), Continuity and Change in Contemporary Capitalism, Cambridge: Cambridge University Press. 398–423.

Lipset, Seymour Martin and Stein Rokkan. 1967. Cleavage Structures, Party Systems, and Voter Alignments. An Introduction. In: Lipset, Seymour Martin and Stein Rokkan (eds.), Party Systems and Voter Alignments. Cross-National Perspectives, New York/London: Collier-Macmillan. 1–64.

Lubbers, Marcel, Mérove Gijsberts and Peer Scheepers. 2002. Extreme Right-Wing Voting in Western Europe. In: European Journal of Political Research, 41: 345–378.

Mair, Peter and Cas Mudde. 1998. The Party Family and Its Study. In: Annual Review of Political Science, 1: 211–229.

Mayer, Nonna and Pascal Perrineau. 1992. Why Do They Vote for Le Pen? In: European Journal of Political Research, 22: 123–141.

McGann, Anthony J. and Herbert Kitschelt. 2005. The Radical Right in The Alps. Evolution of Support for the Swiss SVP and Austrian FPO. In: Party Politics, 11.

Mudde, Cas. 1996. The War of Words. Defining the Extreme Right Party Family. In: West European Politics, 19: 225–248.

Mughan, Anthony and Pamela Paxton. 2006. Anti-immigrant sentiment, policy preferences and populist party voting in Australia. In: British Journal of Political Science, 36: 341–358.

Newell, James L. 2000. Italy. The Extreme Right Comes in from the Cold. In: Parliamentary Affairs, 53: 469–485.

Poguntke, Thomas and Susan E. Scarrow. 1996. The Politics of Anti-Party Sentiment: Introduction. In: European Journal of Political Research, 29: 257–262.

Royston, Patrick. 2005. Multiple Imputation of Missing Values. Update of ice. In: The Stata Journal, 5: 188–201.

Rubin, Donald B. 1987. Multiple Imputation for Nonresponse in Surveys. New York: John Wiley & Sons.

Rydgren, Jens. 2005. Is Extreme Right-Wing Populism Contagious? Explaining the Emergence of a New Party Family. In: European Journal of Political Research, 44: 413–437.

Swank, Duane and Hans-Georg Betz. 2003. Globalization, the Welfare State and Right-Wing Populism in Western Europe. In: Socio-Economic Review, 1: 215–245.

Swyngedouw, Marc. 2001. The Subjective Cognitive and Affective Map of Extreme Right Voters: Using Open–ended Questions in Exit Polls. In: Electoral Studies, 20: 217–241.

van Buuren, Stef and Karin (CGM) Oudshoorn. 1999. Flexible Multivariate Imputation by MICE. (07.10.2005).

van der Brug, Wouter and Meindert Fennema. 2003. Protest or Mainstream? How the European Anti-Immigrant Parties Developed into two Separate Groups by 1999. In: European Journal of Political Research, 42: 55–76.

van der Brug, Wouter and Meindert Fennema. 2007. Causes of Voting for the Radical Right. In: International Journal of Public Opinion Research, 19: 474–487.

Weakliem, David L. 2002. The Effects of Education on Political Opinions. An International Study. In: International Journal of Public Opinion Research, 14: 141–157.

1Endless debates not withstanding, there is still no agreement as to what is the most appropriate terminology. In practice, however, this has not hampered scientific progress. As Mudde (1996: 233) observes, “we know who they are, even though we do not know exactly what they are”. In the remainder of this paper, I shall use “Extreme Right” as a shorthand for the Austrian Freedom Party, the Flemish Vlaams Blok/Vlaams Belang, the French-speaking Belgian Front National, the Danish People’s Party and the Danish Progress Party, the French Front National and the Mouvement National Républicain (MNR), the Italian Alleanza Nazionale, Lega Nord and Movimento Sociale-Fiamma Tricolore, the Dutch Lijst Pim Fortuyn (LPF), and the Norwegian Progress Party, simply because it is the most common label for these parties.

2An attempt at a slightly stricter definition of the Extreme Right would involve three elements: i) while their economic policies are quite flexible and of lesser importance, parties of the Extreme Right take a tough stand on immigration and do often (though not always) take a “right” position with respect to many other issues that form the authoritarian-libertarian dimension of political conflict, ii) in terms of political style and patterns of co-operation with other parties within their respective political system, they are usually not well integrated and present themselves as outsiders or radical alternative to the established parties and elites, and iii) although they may be “extreme” in these respects, they are not necessarily “extremist”, i.e. beyond the liberal-democratic pale (see Arzheimer 2008 for a more elaborate discussion of these issues). While this definition still leaves considerable room for interpretation, in reality there is hardly any disagreement amongst scholars as to which parties belong to the Extreme Right family (Mudde 1996).

3While the Swiss SVP is often considered as a party of the Extreme Right, Switzerland was excluded because its institutional structure is vastly different from other West European countries and because until recently, the transformation of the SVP was confined to the so-called “Zurich wing” of the party.

4While the Extreme Right in Germany is slightly stronger than in Spain or the UK, Germany had to be excluded from this analysis because of the very low number of self-confessed supporters of the Extreme Right in the German part of the ESS.

5Following a well-established convention, latent variables are represented by ovals in Figure 1. Observable variables are represented by rectangles.

6From the information in the ESS, a simplified version of the Goldthorpe scheme (which is widely used in comparative research) was derived.

7The ESS team provides a scale of educational attainment that greatly facilitates international comparisons.

8The latter two variables – single person households and having/not having a partner – reflect notions of social isolation that are prominent in the older literature on right-wing extremism. Church attendance and union membership are primarily included as controls for the effects of traditional West European cleavages (Lipset and Rokkan 1967) but can also be interpreted as indicators for social integration.

9The variable was coded as trichotomous: identification with a party of the Extreme Right vs. identification with some other party vs. no identification at all (the reference category).

10Assertions about causality in non-experimental settings are always problematic. However, while variables in block I (socio-demographics) can clearly have a causal effect on the attitudes in block II (via socialisation and other processes of attitude formation), it is difficult to conceive of a process through which attitudes would affect socio-demographics. Similarly, the vote cannot possibly have a causal effect on socio-demographics. A causal effect of past behaviour on present attitudes via some sort of cognitive rationalisation process cannot be ruled out completely, though it seems unlikely that this would be a huge problem here.

11All models were estimated with MPlus 4.0, which provides estimators for both logit and probit links. Here, the latter was chosen because it is computationally much more attractive.

12Throughout this paper, the conventional five percent threshold is used.

13In Austria, the sign is correct but the effect is rather weak (though statistically significant). In Denmark, Italy, the Netherlands and in Norway, the parameter is not significantly different from zero. In Belgium and France, there is a weak but statistically significant effect that has the wrong sign.

14Because the number of Fiamma Tricolore and Lega Nord voters is very small (13), it is not possible to differentiate between them and the Alleanza voters.

15For simplicity’s sake, the other independent variables can be ignored since their effects are not significantly different from zero.

16The results refer to Norway but would be broadly similar for other countries. The values of 4, 7 and 5 for ideology reflect the lower quartile, upper quartile and median of the empirical distribution.

17The overall probability of a Freedom Party vote is rather high. This reflects the fact that the Freedom Party attracted more than 20 per cent of the vote in the Storting election of 2005.

18Empirically, the number of left-leaning, pro-immigrant Freedom Party identifiers is of course rather limited.

19The two latent variables are scaled so that a value of 0 is equivalent to the national average (see section 2). A value of +/-1 is one standard deviation above/below the national average.

20The calculations for Table 10 are based on the estimate for the respective coefficient in Table 9 (-0.035). However, the t-test test indicates that there is insufficient evidence (at the five per cent level) to reject the hypothesis that the coefficient is exactly zero. If one is willing to take the result of the test at face value, the ratio in Table 10 would be exactly 1.

21Presumably, candidate orientations are important, too, but these can not be measured with the data at hand.

22The (somewhat crude) indicators for alienation/social integration that are included in the model – household size, marital status, church attendance and union memberships – display few substantial effects in Tables 7, 8 and 9. The ESS questionnaire (like most other surveys) contains no indicators for personality traits, but the very notion of a disposition that is stable over decades is difficult to reconcile with the fluctuations of Extreme Right support in Western Europe. For a more comprehensive discussion and test of the traditional explanations of right-wing support in Western Europe see Arzheimer 2008.

Isoliert oder gut vernetzt? Eine vergleichende Exploration der Publikationspraxis in der PVS


1. Einleitung1

Wie andere wissenschaftliche Disziplinen befasst sich die Politikwissenschaft auch mit ihrer eigenen Entwicklung (siehe zu Deutschland etwa Bleek/Lietzmann 1999; Bleek 2001; Falter/Wurm 2003; Arendes 2004). Die disziplinäre Selbstvergewisserung lenkt die Aufmerksamkeit nicht zuletzt auf wissenschaftliche Publikationen als für moderne Wissenschaft gleichsam konstituierende Faktoren (Hyland 2004: 1). Gerade begutachtete Zeitschriften gelten in der Politikwissenschaft – ähnlich wie in anderen Sozialwissenschaften – zunehmend als Ort wissenschaftlichen Fortschritts. Publikationen und Zitationen in möglichst renommierten Zeitschriften2 versprechen daher einzelnen Forschern und ihren Institutionen Reputation (Merton 1968) und werden als ein Gradmesser wissenschaftlicher Leistungsfähigkeit (Wissenschaftsrat 2008: 8-9) betrachtet, der auch Rekrutierungs- und Finanzierungsentscheidungen zugrunde gelegt wird. In diesem Sinn wurden Publikationen und Zitationen in begutachteten Zeitschriften genutzt, um die Qualität und Leistungsfähigkeit politikwissenschaftlicher Institute in verschiedenen Ländern, darunter in der Bundesrepublik, zu bestimmen (Klingemann 1986; Crewe 1988; Klingemann et al. 1989; Plümper 2003; Hix 2004a, b; Dale/Goldfinch 2005). Ebenso wurde die Bedeutung von Zeitschriftenveröffentlichungen für die Karrierechancen von Politikwissenschaftlern untersucht (Plümper/Schimmelfennig 2007).

Bei allen Unterschieden eint diese Arbeiten eine Grundposition: Zeitschriftenveröffentlichungen werden weitgehend isoliert voneinander betrachtet, gleiches gilt für die Autoren (siehe aber Goodin/Klingemann 1996; Carter/Spirling 2008). Diese Perspektive ist für bestimmte Fragestellungen gut geeignet, verstellt jedoch den Blick auf den Charakter von Wissenschaft als kollektiver Unternehmung.3 Essentiell für wissenschaftlichen Fortschritt ist ein barrierefreier und unvoreingenommener Austausch von Ideen und Erkenntnissen (etwa Merton 1942). Aus dieser Sicht wäre es optimal, wenn eine wissenschaftliche Disziplin sich als sprichwörtliche „kleine Welt“ (etwa Milgram 1967) erwiese: Jedes Mitglied der scientific community kennt die Erkenntnisse und Methoden jedes anderen Mitglieds unmittelbar oder zumindest mittelbar, so dass die Möglichkeit einer wechselseitigen theoretischen, methodologischen und substantiellen Befruchtung besteht. Als dysfunktional müsste es hingegen etwa gelten, wenn Forscher nur die Arbeiten einer bestimmten „Schule“ oder der Mitglieder einer hochspezialisierten Teildisziplin wahrnähmen (etwa Jervis 2002: 188). In diesem Sinne wurde in der deutschen Politikwissenschaft wiederholt die Sorge geäußert, für die als Schritte zum Erkenntnisfortschritt notwendige Differenzierung und Spezialisierung werde mit der „Aufbröselung“ und „Zerfledderung“ der Disziplin ein zu hoher Preis bezahlt (Veen 1982: 7; siehe auch Dogan 1996). Im Vergleich dazu zeigten sich Goodin/Klingemann (1996) in bezug auf die Politikwissenschaft in weltweiter Perspektive eher optimistisch.

In diesem Aufsatz wollen wir einen Beitrag zur Diskussion über die Integration der deutschen Politikwissenschaft leisten. Diesen Gegenstand, der u.a. die Publikationspraxis, den Austausch auf wissenschaftlichen Tagungen und die Nachwuchsrekrutierung einschließt, können wir an dieser Stelle nicht umfassend untersuchen. Wir konzentrieren uns vielmehr auf die kooperative und kommunikative Integration, wie sie in den Beiträgen in der Politischen Vierteljahresschrift (PVS) zu beobachten ist. Es handelt sich dabei um eine Zeitschrift, die in ihrem Selbstverständnis politikwissenschaftlichen Beiträgen aller theoretischen und methodologischen Richtungen und zu allen Teildisziplinen offensteht. Insoweit bietet sie sich als Forum des unvoreingenommenen wissenschaftlichen Austausches und der kommunikativen Integration an. Damit ist jedoch nicht gesagt, dass sie auch tatsächlich als solches genutzt wird (siehe zur Wahrnehmung seitens der DVPW-Mitglieder Faas/Schmitt-Beck 2008). In der Publikationspraxis könnten Forscher durchaus in der PVS nebeneinander publizieren, ohne miteinander zu kommunizieren oder gar zu kooperieren. Im Ergebnis würden die Möglichkeiten, die die PVS als eine general interest-Zeitschrift bietet, nicht genutzt. Um diese Fragen zu klären, untersuchen wir Häufigkeit und Struktur von Ko-Autorschaften und Zitationen in der PVS zwischen 1966 und 2007. Damit wir die Befunde zur PVS besser einordnen und beurteilen können, haben wir, wie im nächsten Abschnitt dargelegt, zusätzlich drei weitere internationale Zeitschriften untersucht.

2. Untersuchungsgegenstand, Daten und Methoden4

Unsere explorative Analyse verfolgt das Ziel, die kooperative und kommunikative Integration in den Beiträgen zur PVS untersuchen. Dazu betrachten wir zwei verschiedene, aber verwandte Phänomene: Zitationen und Ko-Publikationen.5 Zitationen indizieren generell eine intellektuelle Beziehung zwischen zitierender und zitierter Quelle (Lin/Kaid 2000: 145). Aus einer Zitation als solcher lässt sich jedoch wenig mehr ablesen als die schiere Kenntnisnahme der zitierten Quelle, was nicht zuletzt mit den verschiedenen Funktionen von Zitationen in der wissenschaftlichen Kommunikation zusammenhängt (etwa Dubois 1988). Eine Quelle kann zustimmend oder kritisch zitiert werden; auf sie kann um der intellektuellen Redlichkeit willen oder aus strategischem Kalkül Bezug genommen werden, sie kann wegen ihrer herausragenden wissenschaftlichen Qualität, wegen der Reputation des Autors, wegen persönlicher Beziehungen zum Verfasser oder auch zum Zwecke der Eigenwerbung erwähnt werden (z.B. Merton 1968; Brooks 1988). Die Aussagekraft von Zitationen ist also begrenzt. Trotz der aufgezeigten Ambiguitäten erscheint es sinnvoll, Zitationen als einen Indikator für das Fehlen unüberwindbarer Hürden zwischen der zitierten und der zitierenden Quelle aufzufassen.

Ko-Autorschaften implizieren persönlichen Kontakt und aktive Kooperation zwischen Wissenschaftlern. Sie setzen eine gewisse gegenseitige Wertschätzung voraus und gehen in der Regel einher mit einer intensiven intellektuellen Auseinandersetzung. Ko-Autorschaften können daher als ein besserer Indikator für wissenschaftlichen Austausch gelten als Zitationen (Peters/van Raan 1991; Norris 1993). Allerdings sind auch der Aussagekraft dieses Indikators Grenzen gesetzt. So ist nicht gesichert, dass substantielle Beiträge zu Publikationen mit einer Ko-Autorschaft honoriert werden (Heffner 1979; Chandra et al. 2006). Überdies können Kooperationen unterschiedlichen Motiven entspringen und verschiedenen Mustern folgen (Melin 2000; Hara et al. 2003; MLA Task Force 2006: 56).

Eine Kooperation ist – wie unsere knappe Skizze verdeutlicht hat – wesentlich voraussetzungsreicher als eine Zitation. Daher nehmen wir an, dass Zitationen erheblich häufiger vorkommen als Ko-Autorschaften. Zugleich ist es eine offene Frage, inwieweit die Kooperations- und Zitationsnetzwerke einander ähneln oder sich unterscheiden. Es könnten in beiden Hinsichten ähnliche Muster eines auf enge Zirkel beschränkten Austausches auftreten. Sollten diese Zirkel auf beiden Ebenen sogar dieselben Personen umfassen, könnte man nicht nur von Hemmnissen für den innerdisziplinären Austausch sprechen, sondern es läge auch nahe, von Zitierkartellen zu sprechen. Ebensogut kann die empirische Analyse in bezug auf Zitationen und Ko-Autorschaften unterschiedliche Muster zutage fördern. Beispielsweise könnten Wissenschaftler eng über Zitationen verknüpft sein, aber praktisch nicht kooperieren.

Die Daten zum Zitations- und Kooperationsverhalten in der PVS basieren auf einer Auswertung des von Thomson/Reuters vertriebenen Social Science Citation Index (SSCI), in dem die PVS seit 1966 erfaßt wird. Von vergleichbaren Projekten wie etwa Solis unterscheidet sich der SSCI unter anderem dadurch, dass er nicht nur Daten zu einzelnen Artikeln, sondern auch die Zitationsbeziehungen zwischen diesen Artikeln erfaßt, sofern die zitierende und die zitierte Zeitschrift beide im Index enthalten sind. Mit Hilfe der von uns entwickelten Software lassen sich aus diesem Beziehungsgeflecht Aussagen zum Kooperations- und Zitationsverhalten in der PVS ableiten, die dann mit Hilfe netzwerkanalytischer Methoden untersucht werden können.

Dabei ist festzuhalten, dass der SSCI keine unproblematische Datenquelle darstellt. Zum einen werden Monographien sowie Beiträge in Sammelbänden und einer Vielzahl kleinerer Zeitschriften vollständig ausgeblendet. Zum anderen sind die Informationen im SSCI bei näherer Betrachtung häufig unvollständig, inkorrekt oder zumindest inkonsistent. Beispielsweise fehlen (insbesondere natürlich bei Zitationen von Artikeln, die im Erscheinen begriffen sind) oft Angaben zu Seitenzahlen und Jahrgängen. In anderen, weitaus häufigeren Fällen werden für die Namen von Zeitschriften variierende Abkürzungen verwendet und Vornamen von Autoren wahlweise ausgeschrieben oder in kreativer Weise abgekürzt.

Das erste dieser beiden Probleme ist grundsätzlicher Natur: Im sozialwissenschaftlichen Bereich gibt es momentan trotz einiger vielversprechender Ansätze keine Alternative zum SSCI. Die Auswirkungen des zweiten Problems versuchen wir durch eine pragmatische Analysestrategie abzumildern, indem wir uns auf die Auswertung möglichst eindeutiger Informationen aus dem SSCI konzentrieren. Konkret heißt dies u.a., dass wir Angaben zu Seitenzahlen, Heftnummern und Jahrgängen ignorieren und die Namen der Autoren normalisieren, indem wir nur den Familiennamen sowie den Anfangsbuchstaben des ersten Vornamens erfassen.6 Im Sinne unserer Forschungsfrage ist dies eine konservative Strategie, die in seltenen Fällen dazu führen könnte, dass verschiedene Artikel bzw. Personen mit identischem Familiennamen und ähnlichen Vornamen irrtümlich zusammengefaßt werden, was zu einer (sehr moderaten) Überschätzung der Kohärenz politikwissenschaftlicher Forschung führen würde.

Um die Befunde zur PVS besser beurteilen zu können, betrachten wir zusätzlich Kooperations- und Zitationsmuster in drei weiteren vergleichbaren Zeitschriften. Wir untersuchen zum einen die Österreichische Zeitschrift für Politikwissenschaft (ÖZP), die seit 1971 von der Österreichischen Gesellschaft für Politikwissenschaft herausgegeben wird. Wie die PVS ist sie die Zeitschrift einer nationalen Vereinigung von Politikwissenschaftlern im deutschsprachigen Raum. Als zweiten Referenzpunkt haben wir Political Studies (PS) gewählt. PS erscheint seit 1950 und wird von der britischen Political Studies Association, der mitgliederstärksten britischen Vereinigung von Politikwissenschaftlern herausgegeben. Ein wesentlicher Unterschied zwischen PS auf der einen und PVS und ÖZP auf der anderen Seite liegt im Verbreitungsgebiet, da PS auf Englisch und damit in der zur lingua franca der Politikwissenschaft avancierten Sprache erscheint. Allerdings gilt PS – nach unterschiedlichen Maßstäben – nicht als eine der weltweit führenden politikwissenschaftlichen Zeitschriften (Garand/Giles 2007: 296; Plümper 2007). Um zu prüfen, ob die Muster des wissenschaftlichen Austauschs in Spitzenzeitschriften anders aussehen als in PVS, ÖZP und PS, haben wir drittens zum Vergleich auch die Publikationspraxis im British Journal of Political Science (BJPS) untersucht. BJPS erscheint seit 1970 und hat sich als eines von wenigen nicht-amerikanischen Periodika unter den weltweit bedeutendsten politikwissenschaftlichen Zeitschriften etabliert. Da es sich wie bei PS um eine britische Zeitschrift handelt, sind die Ergebnisse leichter vergleichbar, als wenn wir eine amerikanische Spitzenzeitschrift als Vergleichsmaßstab heranzögen.7

Zur Analyse von Kooperations- und Zitationsbeziehungen bedienen wir uns der Methoden, die für die Untersuchung sozialer Netzwerke entwickelt wurden (zur Einführung siehe etwa Scott 2000, De Nooy et al. 2005, Knoke/Yang 2008). Die Autoren, die in den vier Zeitschriften publizieren, bilden die Knoten der von uns analysierten Netzwerke und Subnetzwerke. Zitationen stellen eine gerichtete, Kollaborationen eine ungerichtete Verbindung zwischen zwei Knoten her. Um die Analyse übersichtlich zu halten, betrachten wir in der quantitativen Analyse nicht jede individuelle Zitation (etwa Hyland 2004) bzw. Kollaboration, sondern lediglich das Faktum der Erwähnung bzw. der Zusammenarbeit zwischen zwei Autoren. Ob ein Autor einmal oder mehrmals, zustimmend oder kritisch erwähnt wird, unterscheiden wir nicht; ebensowenig untersuchen wir die stilistische Gestaltung von Zitationen (etwa Dubois 1988) oder die Häufigkeit der Zusammenarbeit.8

-Abbildung 1 etwa hier-

Kleinere Netzwerke lassen sich vergleichsweise leicht visualisieren und per Augenschein analysieren. Bei den von uns betrachteten Publikationsnetzwerken mit jeweils mehreren hundert Knoten ist dies jedoch nicht mehr möglich. Wir konzentrieren uns deshalb auf die quantitative Analyse und auf die Identifikation von Subnetzwerken. Das wichtigste Konzept, dessen wir uns dabei bedienen, ist das der „Komponente“ (De Nooy et al. 2005: 66-70). Bei einer Komponente handelt es sich schlicht um einen Teil des Netzwerkes, in dem es möglich ist, von jedem beliebigen Knoten zu jedem anderen beliebigen Knoten zu gelangen. Abbildung 1 illustriert die Bedeutung des Konzeptes: Das fiktive Netzwerk von sieben Autoren zerfällt in drei Komponenten. Im Fall der ersten Komponente ist dies sehr leicht zu erkennen, da zwischen den Autoren 1, 2 und 3 alle theoretisch denkbaren Verbindungen realisiert sind. In Komponente 2 sind die Autoren 5 und 6 jeweils mit Autor 4 verbunden, so dass diese drei Personen ebenfalls eine Komponente bilden, obwohl zwischen 5 und 6 keine direkte Verbindung besteht. Autor 7 schließlich ist von allen anderen Personen isoliert und bildet deshalb eine Komponente für sich.

In Abbildung 1 handelt es sich um ein Kooperationsnetzwerk, d.h. die Verbindungen zwischen den Knoten sind ungerichtet. In Zitationsnetzwerken hingegen spielt die Verbindungsrichtung eine Rolle. Deshalb lassen sich hier „schwache“ von „starken“ Komponenten unterscheiden. Letztere betrachten nur solche Subnetzwerke als verbunden, in denen es unter Beachtung der Verbindungsrichtung möglich ist, von jedem beliebigen Knoten aus jeden anderen Knoten zu erreichen. Starke Komponenten setzen folglich eine sehr enge wechselseitige Verknüpfung voraus, die in Zitationswerken nur selten zu beobachten ist. In unseren Auswertungen konzentrieren wir uns deshalb auf schwache Komponenten, für deren Konstituierung die Richtung der Verbindung keine Rolle spielt.

-Abbildung 2 etwa hier-

Die Aufteilung eines Netzwerkes in Komponenten stellt in der Regel nur den ersten Schritt der Analyse dar. Zur Identifikation zentraler Knoten innerhalb einer Komponente stehen im wesentlichen drei Maße zur Verfügung. Die einfachste dieser Kennziffern ist der Grad ((in-)degree) eines Knotens, d.h. die Zahl der Autoren, mit denen eine Person zusammenarbeitet bzw. von denen er oder sie zitiert wird. Für unsere Fragestellung wichtiger sind jedoch zwei alternative Maße, nämlich die „closeness centrality“ und vor allem die „betweenness centrality“ (de Nooy et al. 2005: 123-132). Letztere vermittelt einen Eindruck davon, wieviele der Verbindungen innerhalb einer Komponente über eine bestimmte Autorin bzw. einen Autor vermittelt sind.

Darüber hinaus ist es sinnvoll, große Komponenten in kleinere Subgruppen („Cliquen“) zu zerlegen. Die Zahl der in der Literatur vorgeschlagenen Methoden zur Definition und Identifikation solcher Subgruppen (N-Cliquen und –Clans, K-plexes und –Cores, F-Groups etc.) ist kaum überschaubar. Die meisten dieser Definitionen setzen eine relativ hohe Netzwerkdichte voraus, die in den von uns analysierten Daten nicht zu erwarten (und auch nicht zu finden) ist. Im folgenden konzentrieren wir uns deshalb auf das Konzept der Dreieckskonnektivität, das sich in jüngster Zeit für die Analyse großer Netzwerke als sehr nützlich erwiesen hat und in der Lage ist, große Subgruppen („Schulen“, „Ansätze“ etc.), die im Vergleich zu ihrer Umgebung relativ stark integriert sind, zu aufzuspüren.

Zur Anwendung dieses Verfahrens werden zunächst alle sogenannten „3-Ringe“ identifiziert (vgl. für das folgende Achmed et al. 2007). Ein solcher Ring besteht aus drei Personen, die alle direkt miteinander verbunden sind, d.h. es handelt sich um eine vollständige ungerichtete Triade, die ein „Dreieck“ bildet. Dieser Verbindungstyp repräsentiert eine intensive intellektuelle Austauschbeziehung innerhalb einer Subgruppe.9

Anschließend wird jeder Verbindungslinie im Netzwerk die Zahl der Dreiecke, denen sie angehört, als Wert zugewiesen, um besonders wichtige Verbindungen im Netzwerk zu identifizieren, die eine kohärente Subgruppe von Personen zusammenhalten. Im nächsten Schritt wird dann jeder Person der maximale Wert der Verbindungen zugeordnet, an denen sie partizipiert. Auf diese Weise erhält man auf der Individualebene einen Indikator für die (lokale) Integration in das Wissensnetzwerk. Abschließend wird dann nach zusammenhängenden Gruppen von Personen gesucht, die alle ein Mindestmaß an Integration aufweisen. Hierzu verwenden wir den sogenannten Island-Algorithmus.10

Abbildung 2 illustriert das Verfahren für ein fiktives Netzwerk von sieben Autoren, die durch insgesamt 10 Beziehungen miteinander verbunden sind. Sechs der Autoren sind durch Dreiecksbeziehungen miteinander verknüpft, während der siebte (G) eine Randstellung einnimmt, da er nicht in einen Ring eingebunden ist. Unter den neun verbleibenden Beziehungen sind drei von besonderer Bedeutung, da sie zwei Ringen angehören: DE, CE und BC. Dementsprechend bilden die vier Autoren B, C, D und E eine besonders dicht verknüpfte Subgruppe innerhalb des Netzwerkes, während A, F und insbesondere G an der Peripherie des Netzwerkes liegen. Im Beispiel läßt sich dies auch per Augenschein erkennen. Für die Analyse von Komponenten mit mehr als einem Dutzend Mitgliedern ist ein stärker formalisierter Zugriff jedoch unverzichtbar.

3. Empirische Befunde

3.1 Ko-Autorschaften

Wie Tabelle 1 zeigt, wurden im Beobachtungszeitraum 647 Aufsätze in der PVS veröffentlicht. Rund vier von fünf Aufsätzen wurden von einzelnen Autoren verfasst, während zu 17 Prozent der Beiträge mindestens zwei Verfasser beisteuerten. Gemeinsam verfasste Beiträge sind somit beileibe nicht das vorherrschende Muster im wichtigsten politikwissenschaftlichen Periodikum in Deutschland. Darin ähnelt die PVS sehr deutlich ihrer österreichischen Schwesterpublikation, in der sich 18 Prozent in Ko-Autorschaft verfasste Beiträge finden, wie auch der britischen PS mit 19 Prozent Gemeinschaftsproduktionen. Eine Ausnahmestellung nimmt in dieser Hinsicht das BJPS ein, in dem mit 44 Prozent beinahe die Hälfte aller Beiträge von mindestens zwei Autoren verfasst wurde. Aus der umgekehrten Perspektive bestätigt sich dieser Befund: rund zwei Drittel der Autoren im BJPS arbeiteten an Ko-Produktionen mit, während in den drei anderen Publikationen etwa 40 Prozent der Autoren zu solchen Gemeinschaftswerken beitrugen. Gemessen an den Werten in den Naturwissenschaften ist freilich selbst der Anteil im BJPS eher niedrig (Newman 2001, Glänzel 2002).

Der Blick auf die gesamte Untersuchungsperiode verstellt den Blick auf interessante Entwicklungen zwischen 1970 und dem Beginn des 21. Jahrhunderts. 1970 lag die durchschnittliche Zahl von Autoren pro Beitrag in der PVS und den beiden britischen Zeitschriften unter 1,2, in der ÖZP mit 1,3 etwas darüber. Bei allen betrachteten Periodika stieg dieser Wert bis ins Jahr 2007 an, allerdings in unterschiedlichem Maße. In der ÖZP hat er sich kaum merklich auf 1,4 erhöht. Damit rückte sie, die 1970 noch das Feld anführte, an dessen Ende. Bei PVS und PS stieg die durchschnittliche Autorenzahl auf knapp unter 1,6, bei der BJPS sogar auf etwa 1,8. Damit ähnelt das BJPS den drei in dieser Hinsicht führenden amerikanischen politikwissenschaftlichen Zeitschriften American Political Science Review, American Journal of Political Science und Journal of Politics. Dagegen ähneln PVS und PS diesen Publikationen in deren Ausgaben aus den achtziger und neunziger Jahren, während die ÖZP eher an die drei amerikanischen Zeitschriften in den siebziger und achtziger Jahre erinnert (Fisher et al. 1998: 851; Chandra et al. 2006: 3).

Über die Gründe für die unterschiedlichen Entwicklungen können wir an dieser Stelle nur spekulieren. Zu einem gewissen Teil könnte die wachsende Bedeutung von Ko-Publikationen mit der steigenden Zahl potentieller Ko-Autoren im Zuge der Etablierung und Expansion des Faches, also der Einrichtung zusätzlicher Professuren, der wachsenden Zahl von Absolventen und Promovierten, zusammenhängen (Arendes 2004: 193). Neben veränderten disziplinären Normen und institutionellen Publikationsanreizen könnte zudem die Notwendigkeit wissenschaftlicher Arbeitsteilung eine Rolle spielen. So könnte etwa das Vordringen zunehmend anspruchsvollerer empirischer Analysen dazu beigetragen haben, weil diese spezielle Kenntnisse, Fähigkeiten sowie entsprechende Hard- oder Software erfordern (Fisher et al. 1998; Melin 2000).

– Tabelle 1 etwa hier –

Betrachten wir die Struktur der Ko-Publikationen, wird deutlich, dass in der PVS Ko-Autorschaften nicht nur selten, sondern auch auf kleine Autorenteams und wenige Beiträge beschränkt sind. Die identifizierten 213 Ko-Autoren gehören zu 78 Ko-Autorennetzwerken (=Komponenten, vgl. Abschnitt 2), die im Median aus zwei Personen bestehen. Ko-Autoren publizieren in den meisten Fällen nur einen Beitrag in dieser Zeitschrift gemeinsam; lediglich fünf von 183 Autordyaden publizierten einen zweiten Beitrag in der PVS.

-Abbildung 3 etwa hier-

Wählt man als Kriterium zur Abgrenzung relevanter Kooperationsnetzwerke eine Zahl von mindestens sechs Mitgliedern, finden sich in der PVS drei, vergleichsweise kleine Komponenten.11 Eine davon enthält ein Dreieck um Markus Klein, der zwei Beiträge mit Bürklin und Ruß und daneben noch drei weitere Aufsätze mit unterschiedlichen Ko-Autoren in der PVS veröffentlicht hat. Die zweite Komponente umfasst wiederum eine Autorentriade, die einen gemeinsamen Aufsatz veröffentlicht hat, und drei weitere Verfasser (Castles, Leibfried, Obinger), die drei Aufsätze mit zwei Mitgliedern des Dreiecks publiziert haben. Im Zentrum12 der größten, aber nicht sehr dichten Komponente schließlich steht Franz Urban Pappi, der etliche Beiträge in der PVS veröffentlicht hat. Interessanterweise lassen sich die Autoren in diesen vergleichsweise umfangreichen Komponenten allesamt der empirischen Forschung zuordnen, während Forscher auf dem Gebiet der politischen Theorie nicht in größere Netzwerke eingebunden sind.

Ähnlich wie in der PVS sind in der ÖZP 83 Komponenten vorzufinden, und auch hier bilden Kooperationspartnerschaften, die sich auf mehr als einen Artikel erstrecken, die große Ausnahme. Allerdings finden sich hier immerhin acht Komponenten aus mindestens sechs Autoren (vgl. Abbildung O-1 im Online-Anhang zu diesem Beitrag). Dies erklärt sich partiell daraus, dass in der ÖZP generell mehr Aufsätze mit mindestens vier Autoren veröffentlicht werden, die somit Kerne größerer Strukturen bilden können. Bestes Beispiel dafür ist ein Beitrag von Daldos et al. mit sechs Autoren. Daneben fällt ein Netzwerk ins Auge, das dadurch entstand, dass drei Autoren von Lachnit et al. mit anderen Autoren bzw. Mitgliedern anderer Netzwerke gemeinsam publizierten. Im Zentrum13 zweier weiterer Netze stehen mit Hans Heinz Fabris und Emmerich Tálos zwei Autoren, die 11 bzw. 12 Artikel in der ÖZP veröffentlicht haben. Diese Beispiele stehen auch stellvertretend für die immerhin 59 Autoren, die in der ÖZP zwischen drei und zwölf Aufsätze veröffentlichten. Für die PVS ist es hingegen höchst ungewöhnlich, wenn ein Autor mehr als einen einzigen Aufsatz in der Zeitschrift publiziert. In der individuellen Produktivität innerhalb dieses Periodikums unterscheidet sich die ÖZP somit deutlich von ihrer deutschen Schwester.14

Die beiden Zeitschriften trennt jedoch nicht nur dieses Merkmal. Vielmehr zeigt sich bei einer simultanen Analyse der ÖZP- und PVS-Daten, dass zwischen den Ko-Publikationsnetzwerken der jeweils führenden Fachzeitschriften der Nachbarländer praktisch keine Verbindung besteht. Obwohl Österreich und Deutschland keine Sprachgrenze trennt, scheint zwischen Politikwissenschaftlern beider Länder eine Kooperationsgrenze zu liegen. Damit werden Potentiale, die länderübergreifende Zusammenarbeit eröffnen könnte, nicht genutzt. Dieses Muster steht in Einklang mit der vergleichsweise geringen Offenheit der österreichischen und deutschen Arbeitsmärkte für Politikwissenschaftler (Armingeon 1997).

Noch deutlicher unterscheidet sich die PVS allerdings von den beiden britischen Zeitschriften, in denen mehr, größere und dichtere Kooperationsnetzwerke zu beobachten sind. In PS und BJPS finden sich 161 bzw. 165 Kooperationsnetzwerke. Ähnlich wie in der PVS und der ÖZP umfassen sieben Netzwerke in PS mindestens sechs Mitglieder, in BJPS sind es mit 17 deutlich mehr. Die drei größten Netzwerke in PS sind mit 16, 19 bzw. 21 Autoren aber wesentlich umfangreicher und zugleich dichter als ihre deutschen und österreichischen Pendants (vgl. Abbildung O-2 im Online Anhang). Besonders augenfällig ist die intensive Kooperation der empirischen Forscher Ron Johnston, Charles Pattie, Patrick Seyd und Paul Whiteley. Allein Johnston und Pattie haben fünf gemeinsame Artikel in PS veröffentlicht. In BJPS lassen sich sogar sieben Netzwerke mit 15 bis 48 Mitgliedern entdecken (vgl. Abbildung O-3 im Online-Anhang). Im Zentrum eines dieser produktiven Netzwerke stehen wiederum Johnston und Pattie, die im BJPS 11 gemeinsame Artikel veröffentlicht haben.

– Abbildung 4 etwa hier –

BJPS und PS können als die beiden führenden britischen general interest Zeitschriften gelten. Es liegt deshalb nahe, die Daten beider Zeitschriften zusammenzuführen, um einen Gesamteindruck von der Publikationspraxis in der britischen Politikwissenschaft zu gewinnen. In dieser neuen Betrachtungsweise ergeben sich 257 Ko-Autorennetzwerke. Die meisten davon sind sehr klein, aber immerhin 272 der 964 Ko-Autoren sind in vier größere Komponenten eingebunden. Zwei davon sind mit 21 bzw. 29 Mitgliedern mittelgroß, während die beiden anderen mit 77 bzw. 145 Personen deutlich umfangreicher als alle bisher betrachteten Strukturen sind. Wie Abbildung 4 deutlich macht, zeichnet sich das größte Netzwerk nicht durch eine sehr dichte Struktur aus. Vielmehr besteht es aus vier Subnetzwerken, die von einer schmalen Kette zusammengehalten werden. Würde man aus der von Helen Margetts bis Warren Miller reichenden Kette (in der Abbildung dunkel hervorgehoben) auch nur einen Autor entfernen, blieben zwei oder drei mittelgroße Komponenten übrig. Es fällt zudem auf, dass mit Johnston und Pattie die beiden produktivsten Autoren nicht zur größten Struktur gehören, sondern im Zentrum eines von mehreren lose verkoppelten Subnetzwerken der zweitgrößten stehen. Trotz der im Vergleich höheren Kooperationsdicht scheint es also auch in der britischen Politikwissenschaft Spielraum für eine weitere Integration zu geben.

Zusammenfassend läßt sich festhalten, dass in den beiden britischen Zeitschriften Ko-Autorschaften eine wesentlich gewichtigere Rolle spielen als in der PVS und der ÖZP. Zugleich sind selbst diese politikwissenschaftlichen Kooperationsnetzwerke klein und fragil im Vergleich zu jenen in Naturwissenschaften. So weist Newman (2001) in einer Analyse von Ko-Autorschaften in der Physik, der Computerwissenschaft, biomedizinischer Forschung und anderen Naturwissenschaften ‘Riesenkomponenten’ nach, die zwischen 57 und 93 Prozent aller Autoren umfassen. Im Fall der beiden britischen Zeitschriften waren es lediglich 15 Prozent. In den Naturwissenschaften ist die größte Komponente zudem typischerweise über 200mal größer als die zweitgrößte, in den von uns betrachteten Fällen war sie höchstens rund zweimal so groß. Auch wenn man berücksichtigt, dass in den Analysen zu den Naturwissenschaften weltweit alle relevanten Zeitschriften einflossen, spricht der Vergleich dafür, dass in den politikwissenschaftlichen Publikationen umfangreiche, dichte und dauerhafte Kooperationspartnerschaften eine verschwindend kleine Rolle spielen. Gemessen an den Kooperationsnetzwerken in den Naturwissenschaften, kann die in Zeitschriften publizierte politikwissenschaftliche Forschung kaum als wohlintegriert gelten, sondern erscheint vielmehr hochgradig fragmentiert. Dies gilt für die PVS und die ÖZP in noch stärkerem Maße als für ihre beiden britischen Pendants.

Dieses Resultat dürfte kein Artefakt der auf wenige nationale Zeitschriften beschränkten Auswahl sein, da eine (vorläufige) Analyse aller im SSCI enthaltenen politikwissenschaftlichen Zeitschriften für den Zeitraum zwischen 2000 und 2007 zu sehr ähnlichen Folgerungen führt (nicht ausgewiesen). Dieser Befund spricht auch dafür, dass weder sprachliche Barrieren noch geographische Distanzen die für die PVS wie die anderen nationalen Zeitschriften vorgefundenen Ergebnisse erklären können. Eher scheint es sich um ein generelles Charakteristikum politikwissenschaftlicher Forschung zu handeln.

3.2 Zitationen

In den betrachteten Zeitschriften finden wir durchschnittlich 38 Zitationen pro Beitrag, wobei sich über die Zeit ein steigender Trend erkennen lässt. Gleichzeitig nehmen nur sehr wenige Zitationen auf Artikel in der jeweiligen Zeitschrift Bezug. In der PVS beziehen sich 2,3% der Zitationen auf Aufsätze in dieser Zeitschrift. Damit liegt die PVS zwischen ÖZP mit 0,4% und PS mit 1% auf der einen Seite und dem BJPS mit 3,6% – und einem deutlich steigenden Trend hin zu 5% in den neunziger Jahren – auf der anderen Seite. Interne Zitationen spielen somit eine deutlich nachrangige Rolle. Dies kann als ein Hinweis darauf gewertet werden, dass die ausgewählten Periodika als general interest journals nur in begrenztem Maße als Foren des intensiven intellektuellen Austausches dienen.

Betrachten wir nun die Zitationen ein wenig genauer. Von den 594 Autoren, die in der PVS publizierten, zitierten 240, also rund 40%, zumindest einen anderen PVS-Aufsatz. In rund 10% (26) der Fälle handelt es sich dabei um eine Selbst-Zitation. Aus der umgekehrten Perspektive zeigt sich, dass 46% und damit knapp die Hälfte der Autoren (mit ihren PVS-Arbeiten) von niemandem zitiert wurden. 22% wurden von einem anderen Autor zitiert, 18% von zwei bis vier PVS-Verfassern. Die übrigen 24 Autoren wurden von mindestens fünf und maximal 21 PVS-Autoren zitiert. An der Spitze des Feldes finden sich Fritz W. Scharpf, Markus Klein, Hans-Dieter Klingemann, Max Kaase und Franz Urban Pappi mit Zitationen von 13, 14, 17, 18 bzw. 21 anderen PVS-Autoren.

– Abbildung 5 etwa hier –

Lassen die Zitationen Strukturen erkennen? Dazu betrachten wir zunächst die Netzwerke, die sich ergeben, wenn man wechselseitige Zitationen als Aufnahmekriterium verwendet. Es finden sich drei Autorenpaare, die ihre Arbeiten wechselseitig zitieren (Armingeon und Merkel, Gehring und Zürn, Stoiber und Thurner). Daneben ist eine umfangreichere Struktur aus elf Autoren zu erkennen (Abbildung 5). Ein Teil dieser Struktur resultiert aus einer in der PVS ausgetragenen Kontroverse zwischen Bürklin, Klein und Ruß auf der einen Seite und Klingemann und Inglehart auf der anderen Seite. Interessanter ist der Rest dieser Struktur. Im Zentrum stehen mit Hans-Dieter Klingemann, Max Kaase und Franz Urban Pappi drei Pioniere der Politischen Soziologie in der Bundesrepublik. Bemerkenswerterweise waren wie sie auch andere Mitglieder dieses Netzwerkes an der Universität Mannheim beschäftigt. Das gilt für Manfred Berger, Wolfgang Gibowski und Dieter Roth genauso wie in jüngerer Zeit für Susumu Shikano. In den Zitationen in der PVS erscheint Mannheim somit als ein Zentrum der politikwissenschaftlichen Forschung in Deutschland, dessen Mitglieder die Arbeiten ihrer lokalen Kollegen aufmerksam zur Kenntnis nehmen und öffentlich diskutieren.15

-Abbildung 6 etwa hier-

Betrachtet man nicht wechselseitige, sondern einseitige Zitationen als Einschlusskriterium, ergibt sich ein differenzierteres Bild (vgl. Abbildung 6, die Größe der Knoten ist hier proportional zur Zahl der eingehenden Zitationen). Neben einer ganzen Reihe kleiner und relativ uninteressanter Komponenten lässt sich ein größeres Netzwerk erkennen, das mit 171 Mitgliedern rund 70 Prozent der Autoren umfaßt. Diese Komponente repräsentiert die wichtigsten Akteure in der wissenschaftlichen Debatte, die innerhalb der PVS ausgetragen wird. Innerhalb dieses Netzwerkes lassen sich wiederum durch die Anwendung des oben vorgestellten Island-Algorithmus drei interessante Subgruppen identifizieren, deren Mitglieder untereinander durch aktive oder passive Zitation nach dem Prinzip der Dreiecks-Konnektivität (vgl. Abschnitt 2) verbunden sind. Das größte und dichteste Subnetz besteht aus 47 Wissenschaftlern, die sich im weiteren Sinne der empirischen Politischen Soziologie zuordnen lassen. Dessen Kern bilden wiederum jene Autoren, die das oben beschriebene Netz aus wechselseitigen Zitationen konstituieren. Eine zweite Gruppe besteht aus Wissenschaftlern aus dem Bereich der Internationalen Beziehungen mit einem gemeinsamen Interesse an der Europäischen Integration. Eine dritte Gruppe aus elf Autoren ist weniger homogen. Ein genauerer Blick zeigt, dass diese Gruppe sich allein der Tatsache verdankt, dass Thomas Plümper fünf Aufsätze zu drei unterschiedlichen Teilgebieten (Methoden, politische Ökonomie, Stand der Disziplin) in der PVS publiziert hat, die auf andere Beiträge in der PVS Bezug nehmen oder von diesen zitiert werden. Bemerkenswert ist darüber hinaus, dass mit Fritz Scharpf einer der am häufigsten zitierten PVS-Autoren keiner der drei genannten Gruppen zuzuordnen ist.

Aus der Definition der „Inseln“ ergibt sich, dass ein einzelner Autor stets nur einer bzw. keiner dieser stark verdichteten Großgruppen angehören kann. Um zu differenzierteren Aussagen über die individuelle Integrationsfähigkeit der PVS-Autoren zu gelangen, betrachten wir deshalb die Zahl der Triaden bzw. 3-Ringe, denen ein Verfasser angehört.16 Hinter dieser Vorgehensweise steckt die altbekannte Vorstellung, dass besonders einflußreiche Personen einer Vielzahl von „Cliquen“ (in diesem Falle Triaden) angehören (Kappelhoff 1986: 46). An der Spitze der Hierarchie stehen nach diesem Kriterium wiederum Franz-Urban Pappi und Hans-Dieter Klingemann, gefolgt von Markus Klein.

Abschließend haben wir die Autorenkomponente in Bi-Komponenten zerlegt, um potentielle intellektuelle „Gatekeeper“ zu identifizieren. Eine Bi-Komponente ist ein Subnetzwerk mit mindestens drei Mitgliedern, innerhalb dessen es keine Person gibt, die den Informationsfluß monopolisieren könnte. Anders gewendet existieren innerhalb von Bi-Komponenten eine Vielzahl von Verbindungen, die den Ausfall einer einzelnen Person kompensieren könnten (de Nooy et al. 2005: 141). Trotz unserer konservativen Definition von „Informationsfluß“ – auch hier ignorieren wir die Richtung der Zitationsbeziehungen – zeigt sich wiederum eine relativ starke Fragmentierung des Zitationsgeflechtes in der PVS. Rund 38% (64) der Autoren innerhalb der großen Komponente gehören keiner Bi-Komponente an. Diese Autoren befinden sich in den kettenartigen Außenbereichen von Abbildung 6 und sind kaum in die wissenschaftliche Diskussion innerhalb der PVS integriert. Daneben existieren eine große (55%) sowie drei kleine Bi-Komponenten, denen jeweils nur eine Handvoll Personen angehört. Letztere sind durch Scharpf, Kaase und Rattinger mit der größeren Bi-Komponente verbunden. Angesichts der sehr überschaubaren Fallzahlen scheint es jedoch übertrieben, die drei genannten Forscher allein auf Grund dieser Tatsache als „Gatekeeper“ zu betrachten.

Die ÖZP unterscheidet sich deutlich von der PVS. Nur 70 Autoren, was etwa 10% entspricht, werden zitiert oder zitieren andere ÖZP-Aufsätze. Unter diesen zitiert mit 21% ein doppelt so hoher Anteil wie unter den PVS-Autoren eigene Arbeiten. Schließt man diese Selbstzitate aus, finden sich unter den 70 Verfassern lediglich 4%, die nicht von einem anderen Aufsatz in der ÖZP zitiert werden. 74% werden ein- oder zweimal zitiert, 6% dreimal, und 14% vier- oder fünfmal. An der Spitze stehen in diesem Fall Wolfgang Müller, Peter Pernthaler und Christian Laireiter. Anders als in der PVS erbringt eine weiterführende Analyse keine interessanten Aufschlüsse über Zitationsstrukturen. Betrachtet man schwache Komponenten, ist ein etwas größeres Netz aus 16 Personen erkennbar, das jedoch zu einem großen Teil darauf beruht, dass Mitglieder des Autorenteams um Laireiter eigene Arbeiten zitierten. Legt man das härtere Kriterium wechselseitiger Zitation an, ist praktisch überhaupt keine Struktur erkennbar. Ins Auge fallen lediglich die wechselseitigen Zitationen von Fritz Plasser und Peter Ulram, die häufig auf frühere (gemeinsame) Arbeiten verweisen.

In der PVS werden durchaus Aufsätze österreichischer Autoren zitiert, wie umgekehrt PVS-Beiträge in der ÖZP erwähnt werden. Allerdings überlappen die Zitationsnetzwerke beider Zeitschriften nur minimal. Führt man die Daten von PVS und ÖZP zusammen, wächst die größte PVS-Komponente von 171 auf 195 Mitglieder, das größte ÖZP-Netzwerk von 16 auf 20. An den jeweiligen Strukturen in den Zitationen ändert die wechselseitige Ergänzung der Daten nichts. Das bedeutet, dass beide Zeitschriften weitgehend unverbunden nebeneinander existieren und die Autoren die Arbeiten im jeweils anderen Periodikum praktisch nicht per Zitation rezipieren. Zwischen beiden deutschsprachigen Zeitschriften liegt somit eine Kooperations- und Zitationsgrenze.

Zum Vergleich wollen wir nun wieder die beiden britischen Zeitschriften untersuchen. 359 Autoren gehören zum Zitationsnetz in der PS. Wechselseitige Zitationen sind seltener als in der PVS (17 Dyaden) und scheinen vor allem durch Bezugnahmen auf gemeinsame frühere Arbeiten zustande zu kommen. Daher kann es nicht erstaunen, dass – anders als in der PVS – keine große Komponente wechselseitiger Zitationen erkennbar ist. Legen wir auch hier das schwächere Kriterium an, erhalten wir eine 250 Autoren umfassende Komponente. Darin werden 40% der Autoren überhaupt nicht zitiert, 22% einmal, während 24% zwischen zwei- und viermal zitiert werden. Die übrigen Autoren erhielten bis zu 15 Zitationen; an der Spitze stehen Gordon Smith, Paul Whiteley, Michael Marinetto, Charles Pattie, Peter Hall und Paul Taylor. Wir finden in PS somit eine ähnlich stark ausgeprägte Zitationshierarchie wie in der PVS, während ein derart deutliches Gefälle in der ÖZP nicht zu beobachten ist.

Betrachten wir wieder Dreieckszitationen, finden sich zwei vergleichsweise dichte Subnetzwerke mit je rund 30 Mitgliedern. Ein erstes Netz um Autoren wie Dowding, Marsh, Rhodes und Marinetto bezieht sich auf Rational-Choice-Analysen von Governance, Institutionen und Policies. Wesentlich interessanter ist eine zweite Gruppe. Denn sie umfasst sowohl empirische Politische Soziologen wie Whiteley, Johnston und Pattie als auch Autoren, die sich für positive politische Theorie interessieren. Zur kommunikativen Integration beider Teilgebiete mögen relativ breite Konzepte wie Sozialkapital, Deliberation und Demokratie beitragen. Zugleich stehen dieses und das Rational-Choice-Netz weitgehend unverbunden nebeneinander. Die Integration kennt also durchaus Grenzen. Das wird auch deutlich, wenn man sich die kleineren Gruppen im unteren Teil von Abbildung O-4 im Online-Anhang vor Augen führt, die sich mit institutionellen Aspekten von Wahlen bzw. liberaler politischer Theorie befassen und von der übrigen zitationsgestützten Kommunikation in PS praktisch abgeschnitten sind.

Zum Zitationsnetzwerk der BJPS gehören immerhin 422 Autoren. Allerdings sind wechselseitige Zitationen mit 26 Fällen auch in dieser Zeitschrift eher die Ausnahme. Soweit sie auftreten, ergeben sie sich wiederum häufig daraus, dass Verfasser von Gemeinschaftswerken später darauf Bezug nehmen. Legt man das weichere Kriterium einseitiger Zitation an, lassen sich immerhin 368 Verfasser, also 87%, zu einer einzigen großen Komponente zusammenfassen. 52% dieser Autoren zitieren aus dem BJPS, ohne selbst darin zitiert zu werden. 12% werden einmal zitiert, 19% zwischen zwei- und viermal, während die restlichen 17% von mindestens fünf bis hin zu 38 Autoren zitiert werden. Gemessen an der Zitationszahl, sind James E. Alt, Ivor Crewe, Warren Miller, Roderick Kiewiet, Donald Kinder und Bo Särlvik die zentralen Autoren im BJPS.

Wesentlich wichtiger ist ein anderer Befund: Analysiert man wiederum die Dreieckskonnektivität, so ergibt sich – anders als in den drei anderen Zeitschriften – ein großer und dichter Kern, der 45% aller Autoren umfasst. Er vereint Politikwissenschaftler verschiedener Teilgebiete wie der Politischen Soziologie, der Rational-Choice-Theorie und der politischen Theorie. Dieses Ergebnis spricht für eine vergleichsweise stark integrierte Diskussion innerhalb dieses Periodikums. Gleichwohl gehören nicht alle häufig zitierten Autoren wie etwa John Helliwell und Keith Krehbiel zu diesem Kern. Besonders augenfällig sind die Positionen von Andrew Gelman, Gary King und George Tsebelis. Die Artikel dieser drei prominenten amerikanischen Autoren werden häufig im BJPS zitiert, nehmen aber keinerlei Bezug auf frühere Beiträge in dieser Zeitschrift. Dieser zunächst etwas überraschende Befund dürfte sich aus der Doppelfunktion des BJPS als britische und internationale bzw. quasi-amerikanische Zeitschrift erklären.

Führt man die Zitationsnetzwerke der beiden britischen Zeitschriften zusammen, resultiert ein 847 Autoren umfassendes Netz. Darin finden sich 68 wechselseitige Zitationen, und es ergibt sich eine recht große Gruppe von 24 Personen, die durch wechselseitige Zitationen verbunden sind. Im Zentrum dieses Subnetzes steht David Sanders, der mit immerhin zehn Autoren über wechselseitige Zitationsbeziehungen verbunden ist. 87% der Autoren gehören zu einer großen Komponente, in der 47% zitieren, ohne zitiert zu werden. 15% werden einmal zitiert, 22% zwischen zwei- und viermal. Die übrigen 16% erzielen bis zu 51 Erwähnungen, wobei mit Alt, Crewe, Miller und Särlvik vier vom BJPS bekannte und mit Marsh ein von PS bekannter Autor an der Spitze stehen. Ebenfalls aus der Analyse von PS kennen wir bereits die Gruppe liberaler Theoretiker um Canover und Moore. Das zweite, mit 299 Mitgliedern wesentlich größere Netzwerk ist weniger stark auf eine Zeitschrift oder Subdisziplin konzentriert und bildet gewissermaßen den Kern der Diskussion in britischen politikwissenschaftlichen Zeitschriften. Es lassen sich also – anders als im Fall von PVS und ÖZP – zwischen PS und BJPS deutliche Überlappungen erkennen, die eine zeitschriftenübergreifende Diskussion signalisieren.

4. Schlussfolgerungen

Ziel des vorliegenden Beitrages war es, die Strukturen wissenschaftlicher Kommunikation in der PVS in vergleichender Perspektive zu untersuchen. Wir konnten zeigen, dass gemeinsame Publikationen in der PVS eher selten vorkommen. Jedoch zeichnet sich über die Zeit ein aufsteigender Trend ab. Diese Befunde gelten ähnlich für die drei anderen betrachteten Zeitschriften, wobei allerdings der zunehmende Trend im BJPS am stärksten ausgeprägt ist. Soweit Kooperationsnetzwerke zu beobachten waren, erwiesen sie sich als klein und nicht sehr dicht. Im Vergleich zu Naturwissenschaften erscheint die Politikwissenschaft hinsichtlich gemeinsamer Zeitschriftenpublikationen somit eher als eine Ansammlung lose verkoppelter Inseln denn als eine „kleine Welt“, in der jeder mit jedem direkt oder indirekt zusammenarbeitet. Es scheint, als seien Politikwissenschaftler durch Unterschiede in den Untersuchungsgegenständen, theoretischen Positionen oder methodologischen Zugängen so weit voneinander getrennt, dass sie nicht zusammen publizieren (können). Legt man das harte Kriterium der Ko-Publikation an, kann man somit kaum von einer Integration der Politikwissenschaft sprechen, wie auch andere Sozialwissenschaften eher desintegriert wirken (Leahey/Reikowsky 2008).

Ein etwas anderes Bild zeichnet die Zitationsanalyse. Sie hat in der PVS wie in den anderen betrachteten Zeitschriften keine Anhaltspunkte für verbreitete Zitierkartelle an den Tag gebracht. Das schließt freilich nicht aus, dass solche in der Politikwissenschaft existieren, aber an anderen Orten – etwa in Sammelbänden und Monographien – ihren Niederschlag finden. Soweit das zuträfe, könnte man unsere Ergebnisse als Indiz dafür werten, dass die Regeln, denen der Publikationsprozess in den ausgewählten Zeitschriften unterliegt, diese Form kollusiven Verhaltens eher zu erschweren scheinen.

Darüber hinaus hat die Zitationsanalyse gezeigt, dass Autoren in den betrachteten Zeitschriften die dort vorher publizierten Beiträge durchaus zur Kenntnis nehmen und durch Erwähnung würdigen. Man kann also nicht davon sprechen, Politikwissenschaftler publizierten in wechselseitiger Ignoranz. Allerdings gilt dieser Befund für die betrachteten Zeitschriften in unterschiedlichem Maße. In der PVS konnten wir Kommunikationsnetzwerke identifizieren, die sich jedoch als nicht sehr umfangreich und dicht gewirkt erwiesen. Von einem echten Netz kann man ehesten im Hinblick auf Arbeiten zur Politischen Soziologie sprechen. Den Kommunikationsbeziehungen in der PVS relativ ähnlich sind jene in PS, wo ein Subnetzwerk ebenfalls dem Teilgebiet der Politischen Soziologie zuzurechnen ist. Zu den beiden anderen Zeitschriften überwiegen eher die Unterschiede. In der ÖZP ist deutlich weniger Kommunikation zwischen Autoren in Form von (wechselseitigen) Bezugnahmen zu erkennen. Gerade durch zahlreiche (einseitige) Zitationen von in diesem Periodikum veröffentlichten Beiträgen zeichnen sich die Publikationen im BJPS aus. Hier sind wesentlich dichtere und weniger auf eine Subdisziplin beschränkte Zitationsnetzwerke zu erkennen. Diese Konstellation dürfte dem Austausch und der Verbreitung neuer Ideen innerhalb einer (Teil-)Disziplin und damit dem wissenschaftlichen Fortschritt wesentlich zuträglicher sein als die Kommunikationsmuster in den anderen Zeitschriften. Überspitzt formuliert, erscheint das BJPS somit als ein echtes Forum wissenschaftlichen Austausches, während in der PVS, ÖZP und PS Autoren eher aneinander vorbei zu publizieren scheinen.17

Worin sind die Ursachen für diese Unterschiede zu suchen? Da sie sich in der Zahl der Referenzen sehr ähneln, können wir ausschließen, dass in den Zeitschriften unterschiedlich großer Wert auf die Würdigung früherer Forschung gelegt wird. Daher dürften sich die relativ losen Zitationsnetzwerke in PVS, ÖZP und PS eher daraus ergeben, dass sich Autoren darin bevorzugt auf Monographien oder Sammelbände beziehen oder aber auf Aufsätze aus anderen Zeitschriften als derjenigen, in der sie selbst publizieren. Tatsächlich zitierten im Untersuchungszeitraum Autoren der BJPS zu 40 Prozent Zeitschriftenartikel, und damit merklich häufiger als Verfasser von PVS, PS (27 Prozent) und ÖZP (17 Prozent).18 Diese Unterschiede können die dargestellten Muster zum Teil, aber nicht vollständig erklären. Darüber hinaus scheinen Autoren von PVS, PS und ÖZP, sofern sie Zeitschriften zitieren, seltener auf Artikel desjenigen Periodikums Bezug, in dem sie selbst veröffentlichen, als dies für BJPS-Autoren gilt. Über die Gründe dafür können wir an dieser Stelle nur spekulieren. Die Qualität der Beiträge könnte ebenso eine Rolle spielen wie die Reputation der Zeitschriften und disziplinäre Normen (siehe dazu etwa Schmitter 2002; Goodin/Klingemann 2002). Im Falle von PVS und ÖZP könnte es auch damit zusammenhängen, dass sie sprachlich bedingt nur für eine relativ kleine, arbeitsteilige scientific community zugänglich sind und daher manche Autoren in diesen Zeitschriften zu ihrem Spezialgebiet kaum Artikel (auf dem von ihnen gewünschten Niveau) finden können, auf die sie sich beziehen könnten.

Mit unserer komparativen Exploration haben wir einen wesentlichen Aspekt der Publikationspraxis in vier politikwissenschaftlichen Zeitschriften beleuchtet. Die Analyse weist jedoch auf mindestens ebensoviele Fragen hin, die in künftigen Arbeiten untersucht werden sollten. Es gilt, die Analyse auf Kooperations- und Zitationsnetzwerke auszudehnen, die andere Publikationen – wie Monographien, Sammelbände und andere Zeitschriften – umfassen. Derartige Analysen erlauben es nicht nur, die Publikationspraxis in den hier betrachteten Zeitschriften genauer zu untersuchen. Vielmehr ist sie auch geeignet, generellere Aussagen über die publikationsbasierte Kommunikation und Integration in der Politikwissenschaft zu treffen. Mit Blick auf die deutsche Forschung könnte sich möglicherweise herausstellen, dass die Kommunikation in bestimmten Zeitschriften vergleichsweise stark auf internationale Diskussionen ausgerichtet ist, während andere Periodika stärker in nationale Kommunikationsnetzwerke eingebunden sind. Auch scheinen zeitliche Veränderungen dieser Muster nicht ausgeschlossen, wie auch Unterschiede zwischen der politikwissenschaftlichen Diskussion in verschiedenen Ländern denkbar sind. Zudem bleibt zu klären, wie sich die Integration der Politikwissenschaft abseits wissenschaftlicher Veröffentlichungen, also etwa in bezug auf den Austausch auf Tagungen, Förderanträge, Forschungsprojekte oder persönliche Kontakte, darstellt. Es scheint somit, als hätten wir mit unserer Exploration ein fruchtbares Forschungsfeld betreten, das zu erschließen sich für eine selbstreflexive Politikwissenschaft lohnen sollte.


Ahmed, Adel, Vladimir Batagelj, Xiaoyan Fu, Seok-Hee Hong, Damian Merrick und Andrej Mrvar, 2007: Visualisation and analysis of the internet movie database, in: Hong, Seok-Hee und Kwan-Liu Ma (Hrsg.), Asia-Pacific Symposium on Visualisation 2007 (APVIS2007), Washington: IEEE Computer Society, 17-24.

Altman, David, 2006: From Fukuoka to Santiago: Institutionalization of Political Science in Latin America, in: PS: Political Science & Politics 39, 196–203.

Arendes, Cord, 2004: Politikwissenschaft in Deutschland. Standorte, Studiengänge und Professorenschaft 1949-1999, Wiesbaden: VS Verlag für Sozialwissenschaften.

Armingeon, Klaus, 1997: Karrierewege der Professoren und Professorinnen der Politikwissenschaft in der Schweiz, Österreich und Deutschland, in: Swiss Political Science Review 3, 1-15.

Bleek, Wilhelm, 2001: Geschichte der Politikwissenschaft in Deutschland, München: C.H.Beck.

Bleek, Wilhelm/Lietzmann, Hans J. (Hg.), 1999: Schulen in der deutschen Politikwissenschaft, Opladen: Leske und Budrich.

Brooks, Terrence A., 1988: Citer motivations, in: Kent, Allen (Hg.), Encyclopedia of Library and Information Science (Vol. 43, supp. 8). New York: Marcel Dekker, 48-59.

Calhoun, Craig, 1992: Sociology, Other Disciplines, and the Project of a General Understanding of Social Life, Chicago: University of Chicago Press.

Carter, David/Spirling, Arthur, 2008: Under the influence? Intellectual exchange in political science, in: PS: Political Science & Politics 41, 375-378.

Chandra, Kanchan/Gandhi, Jennifer/King, Gary/Lupia, Arthur/Mansfield, Edward, 2006: Report of APSA Working Group on Collaboration.

Crewe, Ivor, 1988: Reputation, Research and Reality: The Publication Records of UK Departments of Politics, 1978-1984, in: Scientometrics 14, 235-250.

Dale, Tom/Goldfinch, Shaun, 2005. Article citation rates and productivity of Australasian political science units 1995-2002, in: Australian Journal of Political Science 40, 425–434.

De Haan, J. 1997: Authorship patterns in Dutch sociology, in: Scientometrics 39, 197-208.

De Nooy, Wouter, Andrej Mrvar und Vladimir Batagelj, 2005: Exploratory Social Network Analysis with Pajek, Cambridge u.a.: Cambridge University Press.

Dogan, Mattei, 1996: Political Science and the Other Disciplines, in: Goodin, Robert E./Klingemann, Hans-Dieter (Hg.), A New Handbook of Political Science, Oxford: Oxford University Press, 97-130.

Dubois, Betty L., 1988: Citations in biomedical journal articles, in: English for Specific Purposes 13(1), 47-59.

Faas, Thorsten/Schmitt-Beck, Rüdiger, 2008: Die PVS und die deutsche Politikwissenschaft: Kurzbericht zur Umfrage unter den Mitgliedern der DVPW, in: Politikwissenschaft. Rundbrief der Deutschen Vereinigung für Politische Wissenschaft 139 (Herbst 2008), 33-35.

Falter, Jürgen W./Wurm, Felix W. (Hg.), 2003: Politikwissenschaft in der Bundesrepublik Deutschland. 50 Jahre DVPW, Wiesbaden: Westdeutscher Verlag.

Fisher, Bonnie S./Cobane, Craig T./Vander Ven, Thomas M./Cullen, Francis T., 1998: How Many Authors Does It Take to Publish an Article? Trends and Patterns in Political Science, in: PS: Political Science & Politics 31, 847-856.

Fowler, James H./Grofman, Bernard/Masouka, Natalie, 2007: Social networks in Political Science: Hiring and Placement of Ph.D.s, 1960-2002, in: PS: Political Science & Politics 40, 729-739.

Garand, James C./Giles, Michael W., 2003: Journals in the discipline: A report on a new survey of American political scientists, in: PS: Political Science & Politics 36, 293-308.

Giles, Michael W./Garand, James C., 2007. Ranking Political Science Journals: Reputational and Citational Approaches, in: PS: Political Science and Politics 40, 741–751.

Glänzel, Wolfgang, 2002: Coauthorship patterns and trends in the sciences (1980-1998): A bibliometric study with implications for database indexing and search strategies, in: Library Trends 50, 461-473.

Goodin, Robert E./Klingemann, Hans-Dieter, 1996: Political Science: The Discipline, in: Goodin, Robert E./Klingemann, Hans-Dieter (Hrsg.), A New Handbook of Political Science, Oxford: Oxford University Press, 3-49.

Goodin, Robert E./Klingemann, Hans-Dieter, 2002: In Defence of the New Handbook: A Comment on Criticism by Schmitter and Gunnell, in: European Political Science 1, 41-44.

Hara, Noriko/Solomon, Paul/Kim, Seung-Iye/Sonnenwald, Diane H., 2003: An emerging view of scientific collaboration, in: Journal of the American Society for Information Science and Technology 54, 952-965.

Heffner, Alan G., 1979: Authorship recognition of subordinates in collaborative research, in: Social Studies of Science 9, 377-384.

Hix, Simon, 2004: A Global Ranking of Political Science Departments, in: Political Studies Review 2, 293–313.

Hyland, Ken, 2004: Disciplinary Discourses. Social Interactions in Academic Writing, Ann Arbor: University of Michigan Press.

Jervis, Robert, 2002: Politics, Political Science, and Specialization, in: PS: Political Science and Politics 35, 187-189.

Kappelhoff, Peter, 1986: Cliquenanalyse. Die Bestimmung von intern verbundenen Teilgruppen in Netzwerken, in: Pappi, Franz-Urban (Hrsg.), Methoden der Netzwerkanalyse, München: Oldenbourg, 39-63.

Klingemann, Hans-Dieter, 1986: Ranking the Graduate Departments in the 1980s: Toward Objective Qualitative Indicators, in: PS: Political Science and Politics 19, 651-661.

Klingemann, Hans-Dieter/Grofman, Bernard/Campagna, Janet, 1989: The Political Science 400: Citations by Ph.D. Cohort and by Ph.D. Granting Institution, in: PS: Political Science and Politics 22, 258-270.

Knoke, David und Song Yang, 2008: Social Network Analysis, 2nd edition. Thousand Oaks u.a.: Sage.

Leahey, Erin/Reikowsky, Ryan C., 2008: Research Specialization and Collaboration Patterns in Sociology, in: Social Studies of Science 38, 425-440.

Lin, Yang/Kaid, Linda Lee, 2000: Fragmentation of the intellectual structure of political communication study: Some empirical evidence, in: Scientometrics 47, 143-164.

Melin, Göran, 2000: Pragmatism and Self-organization: Research Collaboration at the Individual Level, in: Research Policy 29, 31-40.

Merton, Robert K., 1942: Science and Technology in a Democratic Order, in: Journal of Legal and Political Sociology 1, 115-126.

Merton, Robert K., 1968: The Matthew Effect in Science, in: Science 159 (3810), 56-63.

Milgram, Stanley, 1967: The Small World Problem, in: Psychology Today 2, 60-67.

MLA Task Force on Evaluating Scholarship for Tenure and Promotion, 2006: Report of the MLA Task Force on Evaluating Scholarship for Tenure and Promotion, MLA.

Newman, M. E. J., 2001: The structure of scientific collaboration networks, in: Proceedings of the National Academy of the Sciences of the United States of America 16, 404-409.

Norris, R. P., 1993: Authorship patterns in CJNR: 1970-1991, in: Scientometrics 28, 151-158.

Peters, H. P. F./van Raan, A. F. J., 1991: Structuring scientific activities by co-author analysis: an exercise on a university faculty level, in: Scientometrics 20, 235-255.

Plümper, Thomas, 2007. Academic Heavy-Weights: The Relevance of Political Science Journals, in: European Political Science, 6, 41–50.

Plümper, Thomas/Schimmelfennig, Frank, 2007: Wer wird Prof – und wann? Berufungsdeterminanten in der deutschen Politikwissenschaft, in: Politische Vierteljahresschrift 48, 97-117.

Schmitter, Philippe C., 2002: Seven (Disputable) Theses Concerning the Future of ‘Transatlanticised’ of ‘Globalised’ Political Science, in: European Political Science 1, 23-40.

Scott, John, 2000: Social Network Analysis. A Handbook. 2. Aufl. London, Thousand Oaks, New Delhi: Sage.

Veen, Hans-Joachim, 1982: Einführung: Politikwissenschaft zwischen Selbstliquidation und politischer Integration, in: Konrad-Adenauer-Stiftung (Hrsg.), Entwicklungslinien der Politikwissenschaft in der Bundesrepublik Deutschland, Melle: Knoth, 7-11.

Wissenschaftsrat, 2008: Aufgaben, Kriterien und Verfahren des Evaluationsausschusses des Wissenschaftsrates, Berlin (, Zugriff am 29.9.2008)

Tabelle 1: Häufigkeit von Ko-Publikationen in den vier betrachteten Zeitschriften




















Anteil von Autoren mit einem einzigen Artikel





Artikel mit mehr als einem Autor

109 (17%)

144 (18%)

246 (19%)

364 (44%)


213 (36%)

281 (42%)

439 (36%)

607 (65%)

1Für wertvolle Hinweise und Anregungen bedanken wir uns bei den Gutachtern.

2Siehe zur Diskussion über die Messung von Qualität und Renommee politikwissenschaftlicher Zeitschriften Garand/Giles 2003; Giles/Garand 2007; Plümper 2007.

3Natürlich wird dieser Aspekt insofern berücksichtigt, als das Renommee einzelner Zeitschriften wie auch der Publikationen in referierten Zeitschriften zugeschriebene Stellenwert Ergebnisse kollektiver sozialer Konstruktion sind.

4Alle empirischen Analysen wurden mit dem von Batagelj und Mrvar entwickelten Programm Pajek ( durchgeführt. Datensätze, mit denen sich unsere Analysen replizieren lassen, stehen unter zur Verfügung.

5Mit dieser Auswahl blenden wir andere Aspekte forschungsbezogener wissenschaftlicher Tätigkeit, etwa andere Publikationsformen, Förderungsanträge, Forschungsprojekte, wissenschaftliche Vorträge und die Kommunikation auf wissenschaftlichen Tagungen und Konferenzen, aus. Dadurch geht uns zweifelsohne wertvolle Information verloren, und unsere Analyse kann nicht als repräsentativ für sämtliche Forschungsaktivitäten gelten. Allerdings betrachten wir einen besonders wichtigen Aspekt der Forschung. Denn Publikationen sind Ergebnisse erfolgreicher Forschungsprozesse, solche in begutachteten Zeitschriften enthalten Resultate, die dem Urteil kritischer Gutachter standgehalten haben. Unsere Analyse gibt also Aufschluss über die kooperative und kommunikative Integration politikwissenschaftlicher Forschung, die zu an renommierter Stelle abgedruckten Ergebnissen führte.

6Darüber hinaus berücksichtigen wir uns bekannte Änderungen des Familiennamens durch Heirat oder Scheidung.

7Weitere deutsche bzw. (partiell) deutschsprachige Zeitschriften wie beispielsweise die Zeitschrift für Parlamentsfragen, die Zeitschrift für Politik, die Schweizerische Zeitschrift für Politikwissenschaft oder die Zeitschrift für Internationale Beziehungen müssen außer Betracht bleiben, da sie vom SSCI nicht bzw. nicht über einen längeren Zeitraum erfaßt werden.

8 Die Zitationsanalyse schränken wir zeitlich nicht ein, berücksichtigen also etwa auch Zitationen eines Aufsatzes nach 30 Jahren. Dadurch erfassen wir nicht nur im unmittelbaren zeitlichen Zusammenhang geführte wissenschaftliche Diskussionen, sondern auch wiederaufgenommene Diskussionen und Fälle, in denen Argumente oder Befunde erst mit einiger Verspätung in Zeitschriftenpublikationen rezipiert wurden. Im Ergebnis zeichnet diese Methode, die gleichsam das disziplinäre Gedächtnis berücksichtigt, ein etwas optimistischeres Bild von der kommunikativen Integration als eine zeitlich beschränkte Analyse, die naturgemäß zu (noch) kleinteiligeren Strukturen führen würde..

9Im Falle gerichteter Beziehungen könnte dabei zwischen zwei verschiedenen Typen von 3-Ringen (transitiven und zyklischen) unterschieden werden. Wie oben dargelegt ignorieren wir jedoch die Richtung der Beziehung und fassen deshalb beide Typen zusammen.

10Vereinfacht dargestellt überflutet dieser Algorithmus das Netzwerk mit virtuellem Wasser, wobei der Wert eines Knotens als Geländehöhe interpretiert wird. Durch sukzessives Absenken des Wasserstandes lassen sich dann “Inseln” (=Regionen mit hoher Dichte) identifizieren. Um triviale Lösungen auszuschließen, geben wir als Mindestgröße für die Inseln eine Zahl von fünf Autoren vor. Die maximale Größe der Inseln entspricht der Gesamtzahl der Autoren minus eins. In der von uns verwendeten Variante des Algorithmus sind Inseln mit mehreren „Gipfeln“ (=Verdichtungszonen) zulässig, was wiederum eine konservative Strategie darstellt.

11Der Schwellenwert von sechs Mitgliedern wurde auf Grund der Tatsache gewählt, dass die meisten Kooperationspublikationen zwei oder drei Autoren haben. Bei einem niedrigeren Schwellenwert würden zusätzlich eine Reihe uninteressanter Komponenten erfaßt, die auf eine einzige gemeinsame Publikation mit einer vergleichsweise großen Zahl von Verfassern zurückgehen (z.B. Albert et al. im Heft 1/1996). Bei einem Schwellenwert von fünf Mitgliedern ergeben sich fünf Komponenten, bei einem Schwellenwert von vier steigt diese Zahl auf 39 Komponenten.

12Innerhalb dieser Komponente hat Pappi mit 0,64 die höchste (normalisierte) “betweenness centrality”. Intuitiv bedeutet dies, dass ein relative großer Anteil von Verbindungen zwischen Autoren durch Pappi vermittelt ist. Ein alternatives Kriterium für die Zentralität eines Autors ist die “closeness centrality”. Diese entspricht intuitiv der mittleren Entfernung eines Autors von allen anderen Autoren. Nach diesem Kriterium sind Pappi (0,47), Thurner (0,48) und Stoiber (0,46) in dieser Komponente die zentralen Akteure. Legt man als Maßstab den Degree, d.h. die Zahl der Verbindungen an, so ist Stoiber (gefolgt von seinen sechs Koautoren) die zentrale Person in der Komponente. Da dieser Befund ausschließlich auf einen einzigen (atypischen) Artikel zurückgeht, ist die betweenness centrality hier aussagekräftiger. Die betweenness centralisation der Komponente liegt bei rund 55 Prozent des theoretischen Maximums (das in einem sternförmigen Netzwerk mit einer einzigen Zentralgestalt erreicht würde).

13Kriterium ist auch hier wieder die betweenness centrality von 0,7 bzw. 0,75. Im folgenden verzichten wir auf die Ausweisung der entsprechenden Werte.

14Über die Gründe für diesen Unterschied können wir an dieser Stelle nur spekulieren. Eine gewisse Rolle könnte eine aus der kleineren Zahl österreichischer Wissenschaftler resultierende schwächer ausgeprägte Konkurrenz um den begrenzten Publikationsraum in der ÖZP spielen.

15Nur am Rande sei darauf verwiesen, dass Mannheim nach den Ergebnissen von Plümper (2003) das nach der Publikationsleistung in begutachteten Zeitschriften produktivste politikwissenschaftliche Institut in Deutschland ist.

16 Bei der Konstruktion der Inseln wurde die Zahl der Triaden betrachtet, denen eine Verbindungslinie zwischen zwei Autoren angehört.

17 Die hier beschriebenen Muster sind seit etwa Anfang der 1990er Jahre stabil. Eine weitergehende zeitliche Disaggreation wäre wenig sinnvoll, da Wissensnetzwerke per definitionem über die Zeit anwachsen. Im Sinne unserer Fragestellung nach der Integration der deutschen Politikwissenschaft ist die durchgängige Betrachtung des vollständigen Analysezeitraumes deshalb eine konservative Strategie: Die von uns beschriebenen Komponenten und Verdichtungsbereiche werden tendenziell (noch) kleiner, wenn etwa die wegweisenden Beiträge zur Politischen Soziologie aus den 1970er Jahren oder die Artikel zur EU aus den 1990er Jahren aus der Betrachtung ausgeschlossen würden. Analog dazu setzt die Untersuchung von vollständigen Zitationsnetzwerken voraus, daß nicht nur die rezenten Referenzen, sondern auch deren zeitlich potentiell weit zurückliegenden Quellen in den Blick genommen werden, da Zitationsnetzwerke in der Politikwissenschaft häufig mehrere Jahrzehnte überspannen. Angesichts der relativ geringen Größe und Dichte der von uns untersuchten Netzwerke wäre eine separate Analyse der Literatur aus einzelnen Dekaden wenig sinnvoll.

18Einige wenige Beiträge, die keine Referenzen enthalten, wurden aus der Berechnung ausgeschlossen.

How (not) to Operationalise Subnational Political Opportunity Structures


Analysing the influence that features of the national political context exert on the vote for Western Europe’s ‘extreme’ or ‘radical’ right[1] parties is now a minor industry (see for example Arzheimer & Carter 2006; Golder 2003; Jackman & Volpert 1996; Knigge 1998; Lubbers et al. 2002; Swank & Betz 2003), but in a recent contribution to this journal, Kestilä and Söderlund (2007) argue that research should give more consideration to what they call ‘subnational political opportunity structures’. According [k1] to Kestilä and Söderlund, focusing on the subnational context within one country mitigates against three problems that trouble the existing contextual analyses: i) at the subnational level, the number of contexts is large in comparison to the number of relevant variables; ii) unique features of the party system are obviously held constant; and iii) the heterogeneity of the radical right party family need not be of concern (2007: 774-775). [k2] This theoretical claim is backed up by an ecological analysis of the French regional elections of 2004. In a straightforward linear regression at the level of the département, Kestilä and Söderlund relate the electoral support for the Front National (henceforth FN) in the first round of those elections, as well as an index of the FN’s electoral success (that assesses the FN’s success in relation to the leading contender), to five aggregate variables: turnout in the first round of the 2004 election, the logged district magnitude in the previous regional elections of 1998, the effective number of party lists (Laakso-Taagepera index) in 1998, the share of immigrants born outside the European Union in 1999, and the unemployment rate in 1999. They find that turnout and district magnitude have significant negative effects on the FN’s electoral support,[2] whereas the effects of the number of party lists and unemployment are positive and significant. Most interestingly, the effect of immigration is not significantly different from zero in Kestilä and Söderlund’s model of FN aggregate support. From these results, Kestilä and Söderlund conclude that the radical right benefits from low turnout levels, and that greater proportionality of the electoral system does not increase support for the radical right but is actually related to substantially lowerlevels of support. They also conclude that the FN benefited when the effective number of party lists in 1998 was high, and when unemployment levels were high. By contrast, the share of immigrants present in each département did not affect the FN’s electoral score.


It should be noted that, while the results that relate to the district magnitude are in line with the findings of some other studies (e.g. Arzheimer & Carter 2006), those that pertain to unemployment and immigration are not: in country-level analyses the effect of unemployment is subject to on-going discussion, and the effect of immigration has been consistently found to be strong and positive.


From their results, and the findings of other studies notwithstanding, Kestilä and Söderlund argue that the ‘subnational political opportunity structure has been of great importance for the FN’ and more generally, that the subnational approach ‘is able to control a wider range of factors pertaining to the political system and tends to provide more reliable results’ (2007: 790).


Kestilä and Söderlund have invited other scholars working in the field to engage in a discussion of their unexpected findings. We have taken up this invitation because, while we concur that features of the subnational context are potentially relevant for the radical right vote and should be incorporated into more comprehensive accounts of support for these parties, we are not convinced by Kestilä and Söderlund’s conceptualization of what constitutes a subnational political opportunity structure, and nor are we persuaded by the empirical evidence they present.


We begin this article by highlighting some of the theoretical, conceptual and methodological problems present in Kestilä and Söderlund’s study. Then, since Kestilä and Söderlund were extremely forthcoming in providing us with their data, we engage in some re-analysis. In doing this we discuss the difficulties of estimating and interpreting the coefficients in Kestilä and Söderlund’s model and, using an indicator for the FN’s regional entrenchment that is independent of district magnitude and of the effective number of party lists, we demonstrate that features of the subnational political opportunity structure included in Kestilä and Söderlund’s model are essentially spurious. We close by offering an alternative operationalization of one of the key variables contained in Kestilä and Söderlund’s model.



Subnational Political Opportunity Structures: Conceptual Difficulties and Operationalization


The concept of political opportunity structures is notoriously vague, but at its core is the idea that certain variables can capture the degree of ‘openness or accessibility of a political system for would-be political entrepreneurs’ (Arzheimer & Carter 2006: 422). If one accepts this as a working definition, it follows that subnational political opportunity structures refer to a set of regional or local conditions that would either facilitate or hamper the attempts of the radical right to mobilize voters. Precisely because the concept of political opportunity structures is so vague, identifying these conditions and operationalizing these variables is a tricky task, and unfortunately, there are problems with the way in which Kestilä and Söderlund have gone about this in their study.


Our first misgiving concerns the inclusion of (regional) party system fragmentation in Kestilä and Söderlund’s model, operationalized by the variable ‘effective number of party lists’. As Kestilä and Söderlund suggest, the level of party system fragmentation might indeed be important to small or new parties either because a high level of fragmentation at the previous election might indicate that the system is open and therefore more favourable to such parties or, conversely, because a high level of fragmentation might indicate that a wide variety of alternatives already exists, rendering it difficult for a small or new party to make a breakthrough (2007: 784).


However, using this variable in relation to the FN is problematic because, as Kestilä and Söderlund themselves point out, the FN is neither small nor new: it is a well established political competitor that has acquired the status of third political force in many parts of France (2007: 775). Therefore, given that it is not knocking on the door of the French party system but is already clearly inside it, the issue for the FN is not how accessible the party system is (which is what party system fragmentation measures), but is how much political space the party has or, put differently, how much competition it faces at the right end of the political spectrum. In short, we would argue that party system fragmentation is not an appropriate variable with which to measure party competition in this instance and that a much more relevant variable would be one that taps the ideological space available to the FN (see below).


Kestilä and Söderlund do briefly discuss the ideological aspect of party competition when they note that cross-national studies that have analysed the impact of political opportunity structures have had to operationalize and measure ideological convergence or divergence and issue adoption. However, although they recognize ‘that the local party organisations may have an agenda somewhat deviant from the national one’ (2007: 783), they do not include the ideological dimension of party competition in their model on the basis that a subnational analysis such as theirs benefits from being able to ‘hold the ideological differences constant in each subunit due to the national character of the campaign in the elections of 2004’ (2007: 775).


Now, it might indeed be the case that the campaign for the regional elections of 2004 took on a national character, but that does not mean that the contest was ideologically similar in all départements. A quick glance at the first round of the 2004 regional election results reveals that in some regions competition on the right of the political spectrum was played out simply between the FN and one mainstream right party list (for example in Picardie). Elsewhere, however, there was more than one mainstream right party list and/or more than one list from the radical right. In Aquitaine, for instance, there were two mainstream right party lists, while in the Rhône-Alpes region not only was there an FN list, but there was also another ‘extreme right’ list. Similar patterns can be found with respect to the 1998 regional elections – the results of which Kestilä and Söderlund use to calculate their ‘effective number of party lists’.


The ideological nature of party competition has therefore varied by region and voters have been faced with a different choice of party lists according to where they live. This is likely to be important in an explanation of the FN’s success as the party may well be hindered by the presence of multiple mainstream right lists, and may also experience lower vote shares in départements where an alternative radical right list exists. By only including the party system fragmentation variable in their model, Kestilä and Söderlund fail to account for these trends.


Including the effective number of parties or party lists in analyses of the radical right party vote is also problematic for methodological reasons. This is because the vote share of the very party whose electoral success is being explained (in this case the FN) is included in the calculation of the effective number of parties: (1 / S pi2) for N parties, where p is the vote share of party i). This means that the two variables cannot possibly be independent of each other, and that, given the construction of the index, there must be a non-linear relationship between them, the exact shape of which depends on both the number and relative strength of other parties.


The co-variance of these two variables is most easily illustrated by simulation. To do this, we selected the results from four random départements in the 1998 regional elections (since départements were treated as districts in 1998) and let the FN’s vote share vary around its empirical value while holding the relative support within the mainstream right bloc and the absolute support for all other party lists constant.[3] Figure 1 illustrates the results of this simulation and clearly shows that, at least in these four départements, a change in the fortunes of the FN within the département would ceteris paribus have a strong, positive and almost linear impact on the effective number of parties.[4]



[Figure 1 about here]



By including the effective number of party lists in 1998 in their model, Kestilä and Söderlund therefore effectively regress the FN’s success in 2004 on a variable that already encompasses previous levels of support for the party. This is problematic for theoretical reasons, and our simulation shows that it also has very real implications in terms of interpreting the effect of this variable.


Turning to district magnitude, we take no issue with Kestilä and Söderlund’s decision to include this variable in their model. For theoretical reasons it makes sense to do so: it will be harder for the FN – a medium-sized party – to win votes in districts with a small magnitude than it will be for it to be successful in districts with a greater magnitude. Moreover, this effect will be exacerbated if the number of potential voters is small to begin with and if the party decides to invest fewer resources in districts with a small magnitude than it does in those with a larger magnitude.


What we are unhappy with, however, is the way in which this variable has been operationalized in Kestilä and Söderlund’s study. They regress the FN’s vote in each département in 2004 on the (logged) district magnitude in the 1998 regional election, and we would argue this is troublesome for two reasons.


Firstly, the inclusion of the (logged) district magnitude ignores the effect of legal thresholds which can and do override the effects of district magnitude. In the case of the electoral system of 1998, the five per cent legal threshold in place at the département level effectively cancelled out the effects of district magnitude in départements with a magnitude of 14 or more. Given that the district magnitude was 14 or more in 50 of the 94 départements, this has large implications for the model, and, as such, it would have made greater sense to include the effective magnitude or the effective threshold rather than simply the (logged) district magnitude.[5]


The second reservation we have about Kestilä and Söderlund’s operationalization of district magnitude concerns their decision to use the (logged) district magnitude in the 1998 regional election – i.e. in the previous regional election. Kestilä and Söderlund do this because they maintain that ‘changes in electoral laws may not necessarily have an immediate effect’ and that the psychological effect of an electoral system may take a while to manifest itself. They also note that ‘the district magnitude of 1998 and seats allocated to departments in 2004 have a very strong correlation’ (2007: 792, note 7).


We certainly do not dispute the fact that psychological effects of electoral systems may take a while to register with voters, and as such, other than keeping an eye on our comments above about legal thresholds, we would have no criticisms of the use of this variable had the electoral system of 1998 been identical to that used in 2004. The problem, however, is that it was not: the electoral system used in 1998 was fundamentally different to that used in 2004.


The system used in the 1998 regional elections (in use since the first regional elections of 1986) was a one-round proportional electoral system. Between 3 and 72 seats were distributed at the level of the département and, as mentioned above, there was a five per cent legal threshold in place. In 2004 a new two-round electoral system came into operation, however. Under this system, even though seats were eventually divided up between departmental sections, it was at the level of the region that lists were presented, votes were aggregated and seats were distributed. The district magnitude of the regions ranged from 43 in the Limousin and Franche-Comté regions to 209 in the Ile-de-France region but the existence of legal thresholds meant that the effect of district magnitude was effectively cancelled everywhere.[6]


Given that the electoral system changed so fundamentally between 1998 and 2004, we believe that it is unrealistic to argue that the 1998 system still exerted a psychological effect on voters and political elites in 2004. If voters are well-informed and rational enough to react to the mediating effects of electoral systems in the first place, then they are hardly likely, on the one hand, to take the effects of the 1998 electoral system into account, and yet, on the other, to fail to notice that the system has been changed thoroughly in the interim. And as concerns political elites, the effects of the 1998 system will not have entered into their calculations in 2004. Rather they will have taken the new electoral system into account when they decided on their campaigning strategies and on the resources they would invest in each district for the 2004 contest.


For these two reasons, therefore, we would argue that it does not make sense to use the (logged) district magnitude in the 1998 regional election as an indicator of the openness or accessibility of the political system in 2004. And the fact that the district magnitude of 1998 correlates very strongly with the seats allocated to départements in 2004 does not allay our fears because this correlation does not take account of the effects of legal thresholds and because the number of seats distributed to départements in 2004 is irrelevant since the allocation of seats took place at the level of the region, not at that of the département.


We are also uneasy about the inclusion of turnout in Kestilä and Söderlund’s model. The issue here is not how this variable has been operationalized, but rather why it is included in the model at all.


Of course, turnout is commonly included in national and comparative election studies, and a number of these works have observed a negative correlation between turnout and support for parties that are not fully integrated into the party system (Reif et al. 1997; van der Eijk et al. 1996), including the FN, whose vote share has been found to be highly correlated with turnout in both presidential and legislative elections (Auberger 2008). It makes good sense to include turnout in studies of this kind since they are in the business of explaining patterns in individual voter behaviour. In this instance, they are able account for the negative correlation they find by arguing that, while politically dissatisfied supporters of the established parties may refrain from voting altogether, politically dissatisfied supporters of non-established parties can express their dissatisfaction with their vote.


The purpose of an analysis that seeks to assess the impact of political opportunity structures on political parties is altogether different, however. Here the aim is to investigate the opportunities and incentives that a given (subnational) context affords parties and politicians, and crucially, we would contend that turnout is not part of that context. Since the (local) turnout is clearly not known to anyone before the evening of the election day, we would argue it is neither part of an opportunity structure nor a general contextual variable that could somehow affect the probability of a radical right vote.


We certainly recognize that the level of turnout might reflect the attitudinal atmosphere or the intensity of political competition in a particular locality, and that this in turn, may indeed be important in explaining the success of a political party. However, we have concerns about using turnout as an (ex-post facto) indicator of attitudinal atmosphere or political competition because turnout will be affected by a whole host of other factors including the political tradition of an area, the specific local issues, the personalities involved in the campaign, and even the weather. As such, interpreting the cause of differing levels of turnout is highly problematic.


With their last two variables, Kestilä and Söderlund assess the effect of immigration and unemployment on the FN vote. Their model includes the share of the population of each départment that is made up of immigrants born outside of the EU-15 and of unemployed people.[7] This, however, is problematic because the coefficients for immigration and unemployment pick up at least three different things: i) they may represent a true contextual effect whereby immigration and unemployment provide the FN with an incentive to mobilize voters and whereby voters who feel strongly about these issues have an opportunity to vote for a party that campaigns on them; ii) they pick up the effect of the composition of the départements; and iii) they reflect cross-level interactions between features of the context and features of individuals.[8]


This becomes clear if we consider départements with a high share of immigrants. If we assume that the presence of immigrants facilitates mobilization by the FN, then people living in such départements should, ceteris paribus, have a higher propensity to vote for the radical right. This contextual effect should therefore result in a positive aggregate correlation that reflects a subnational political opportunity structure.


However, things are not that simple because we also need to bear in mind the composition of départements and their immigrant population. In 1999 (the year of the census on which Kestilä and Söderlund rely), there were 4.0 million people living in France who had been born outside the EU-15.[9] However, the majority of these immigrants (57 per cent) were French citizens and, as such, had the right to vote. Presumably, they and any of their children born in France, as well as children born in France to non-naturalized immigrants, and many of these people’s friends will have a probability of voting for the FN that is close to zero. Everything else being equal, therefore, these individual effects will result in a substantial negative aggregate correlation that counteracts the positive relationship resulting from the contextual effect.


Given that in roughly one third of all départements immigrants born outside of the EU make up more than five per cent of the population, and that in some départements of the Ile-de-France and the Provence-Alpes-Côte d’Azur regions they comprise up to 20 per cent of the population, the individual effect is not negligible, something which is reflected in the bad model fit for the banlieues of Paris (see below). Finally, such a scenario implies a cross-level interaction too, in that while the FN will be able to mobilize more voters because of high number of immigrants, it will only be able to do so among non-immigrants.


The same logic applies to the effect of unemployment, too. Observed aggregate correlations between unemployment and the FN vote are the result of a contextual effect (voters respond to regional unemployment levels), a compositional/individual effect (the unemployed are presumably more likely to vote for the FN), and possibly also a cross-level interaction effect: after all, it seems reasonable to assume that the strength of the individual effect of unemployment varies with the prevalence of unemployment in one’s environment.


The aggregate correlations Kestilä and Söderlund present therefore conflate three conceptually different effects, the nature and size of which are impossible to separate without micro-data.[10] In addition, because these coefficients reflect the highly aggregated net result of different processes, they hide any co-variation that is likely to exist both between and among individual and contextual variables.[11] For these reasons, the coefficients for unemployment and immigration in Kestilä and Söderlund’s model do not provide reliable information on the role of unemployment and immigration within a political opportunity structure.


Estimating and interpreting the coefficients of Kestilä and Söderlund’s model


As the discussion above has demonstrated, Kestilä and Söderlund’s model is problematic because all of the independent variables included in it raise theoretical, conceptual and/or methodological concerns. In addition to the problems that beset individual variables, the number of cases in Kestilä and Söderlund’s model is not very large (N=94) and the units (départements) vary enormously in terms of their population. The standard deviation for this variable is .48 million people, and the distribution is substantially right-skewed. The number of inhabitants for the ten smallest départements varies between 77,000 and 190,000, whereas each of the ten largest départements has a population of between 1.3 and 2.6 million people. The implication is that a lot of information on individual behaviour is lost, and that the behaviour of citizens in large départements will, ceteris paribus, have a smaller impact on the aggregate correlations.[12]


However, even leaving aside these concerns, the effects of the different independent variables are either difficult to interpret or trivial. As mentioned already, the effects of immigration and unemployment cannot be unambiguously interpreted because both variables aggregate the individual characteristics of the voters of the 94 départements of mainland France and, in the process, conflate contextual and individual effects and cross-level interactions. As for the other variables contained in the model, even though they capture features of the départements that exist independently of the individuals living in them, as we will demonstrate, their political impact is small – a fact that is not apparent in Kestilä and Söderlund’s reading of their findings. Moreover, as we will also show, the estimates in Kestilä and Söderlund’s model are highly sensitive to the selection of cases.


Kestilä and Söderlund interpret their results mainly with reference to the relative size of the t-values, and this is problematic for three reasons. Firstly, the jury is still out on the question of whether it makes sense to calculate (classical) standard errors for data that are a population rather than a sample (Berk & Freedman 2003). Secondly, if significance tests are to be carried out, the calculation of the standard errors should take into account the spatial correlations that exist between départements. Ignoring these dependences violates the standard assumption that disturbances are identical and independently distributed. And thirdly, and most importantly, the size of a t-value (i.e. statistical significance) is not a criterion for substantive relevance, and so it assists little to an interpretation of the effects of the variable.


For these reasons, rather than focussing on the t-values, we would suggest that the effects of the variables are best interpreted by examining the expected change in the FN’s vote share for a given change of the independent variables. At the same time as examining this, it is also important to consider the distribution of the independent variables if we are going to be able to say anything about political realities.


Kestilä and Söderlund do show us what the expected change in the FN’s vote share is for changes in the independent variables. Indeed, although they do not discuss these expected changes in the text, in Table 3 of their article we can see that a unit increase in the logged district magnitude would reduce support for the FN by 3.45 percentage points, whereas a unit increase in the effective number of parties would increase the FN’s share by 1.14 points. However, what these results do not do is take into account the distribution of the district magnitude and the effective number of party lists across départements and as such, they tell us little about the true impact of these variables on particular départements.


Let us first consider the logged district magnitude in 1998 and its distribution across départements. By identifying the second and the third quartile we can ascertain that in half the départements the total number of seats to be filled was between 10 and 22. And we can work out that by increasing the district magnitude from 10 to 22 while holding all other independent variables constant the expected support for the FN would be reduced by a mere 2.7 points.[13] This suggests that the effect of the district magnitude is fairly small across these départements. What is more, if we were to examine the middle 90 per cent of the distribution instead, we would find that the expected difference between the smallest district of eight seats and the biggest district of 31 seats would be 4.7 points. This is still not very large, and here we are considering the vast majority of cases. Thus, even though the effect of the logged district magnitude is statistically significant, it seems that it is only really relevant when we consider very small and very large départements.[14]


When we repeat this exercise for the effective number of party lists, we see that increasing the effective number of party lists from 2.6 (the second quartile) to 3.5 (the third) would increase support for the FN by just one percentage point. And if we consider the middle 90 per cent of the distribution, where the effective number of party lists ranges from 2.2 to 4.2, we see that a change from 2.2 to 4.2 would give rise to a 2.3 percentage point increase in the FN’s vote share.


As well as taking into account the distribution of the independent variables across départements we also need to bear in mind that their effect can be conditional on the levels of the other four independent variables in each département. This is the case for district magnitude: the bivariate correlation between district magnitude and FN support is essentially nil in Kestilä and Söderlund’s model, and only becomes negative once both turnout and unemployment are included in the model. Yet if we move away from the overall model and consider different subgroups of départements, we see that the relationship is actually positive for départements with below-average unemployment and turnout levels, whereas it is negative if either turnout or unemployment or both are above average. This rings alarm bells because there is no obvious theoretical reason for this finding. As such, and particularly because the number of units is low, it points to the possibility that the negative effect of district magnitude (as well as the positive effect of the number of party lists) may be spurious and driven by outlying and otherwise unusual observations.


To investigate this suspicion further, we calculated a number of diagnostics (studentized residuals, Cook’s distance, and leverage values) that can be used to identify problems with the model fit. We found that one département – Seine-Saint-Denis, which, with the neighbouring départements of the Hauts-de-Seine and the Val-de-Marne, forms the infamous banlieues of Paris – clearly stood out. Seine-Saint-Denis has the highest share of immigrants born outside the EU and the second-largest population of that group in absolute terms, and yet levels of FN support here are far below what Kestilä and Söderlund’s model predicts: while they predicted that the FN would poll 25.5 per cent in this département, the actual result in 2004 was 15.8 per cent. This just goes to show what happens when the contextual and compositional effects of the immigration rate are conflated. Furthermore, the impact of this département on the model is large as it is an influential data point in terms of the independent variables and is the largest negative outlier. If it is excluded from the estimation the coefficient for the immigration variable almost doubles and becomes statistically significant.


The largest positive outlier, by contrast, is the Vaucluse in the Provence-Alpes-Côte d’Azur region. This département had an average district magnitude in 1998 and slightly above-average figures for all other independent variables. While Kestilä and Söderlund’s model predicts a vote share of 18.7 per cent for the FN in the Vaucluse, the actual result was a staggering 28.5 per cent. The high support for the FN in this département reflects a political tradition that dates back to the 1980s. In the legislative elections of 1986, Jacques Bompard, a founding member of the FN, polled 18 per cent for the party in this département – one of the best results for the party in that election. Bompard (who left the party in 2005) was also instrumental in the FN’s successes at the local and the regional level, and in 1995 he became mayor of Orange (a historical town in the Vaucluse), being one of the first members of the FN to hold such an office.[15]


Since the Vaucluse does not have much leverage (as regards the independent variables, it is pretty average in almost every way), excluding it from the estimation does not greatly affect the coefficients. However, with just 94 départements, the joint leverage of a small group of three or four cases can easily be a problem (Fox 1997: 281). Indeed, it is possible to manipulate the coefficients considerably by excluding a tiny fraction of the départements. For instance, excluding not only the Vaucluse but also Paris (i.e. département 75) and the Territoire de Belfort in the Franche-Comté region reduces the absolute value of the coefficient for the logged district magnitude from -3.4 to -2.8. By contrast, excluding Seine-Saint-Denis and two rural départements with low unemployment and immigration rates – the Cantal in the Auvergne and the Haute-Vienne in the Limousin – increases the coefficient to -4.3. Similarly, excluding Seine-Saint-Denis together with the Haut-Rhin in Alsace (a FN stronghold) and the Lot-et-Garonne in Aquitaine halves the coefficient for the effective number of party lists.


The most striking effect is observed if we consider the share of immigrants born outside the EU. This coefficient is rather small (.15) and statistically insignificant in Kestilä and Söderlund’s model. Excluding the Vaucluse and two other départements where the FN is very successful – the Ain in the Rhône-Alpes region and the Alpes-Maritime in Provence-Alpes-Côte d’Azur – further reduces the effect of immigration to .03. However, excluding Paris, Seine-Saint-Denis and either of the other banlieues départements (Val-de-Marne or Hauts-de-Seine) almost triples the coefficient and turns immigration in a powerful (and statistically highly significant) predictor of FN success. Not only does this illustrate just how sensitive the estimates in Kestilä and Söderlund’s model are to the selection of cases, but it also highlights once again that the model conflates contextual (i.e. opportunity structure) and compositional (i.e. individual) effects.


Clearly, one could question this practice of excluding individual départements for diagnostic purposes given that Kestilä and Söderlund are examining the population of French départements rather than a sample. That said, doing this does enable us to assess just how accurate an instrument Kestilä and Söderlund’s model is for examining the impact of subnational political opportunity structures on the FN vote in the regional elections of 2004, and indeed for making generalizations beyond this particular electoral contest.


To further investigate our concerns about the spurious nature of the effects of the variables included in Kestilä and Söderlund’s model, we introduced an alternative predictor into their model: the vote won by Jean-Marie Le Pen in each département in the first round of the 2002 presidential election. The theoretical relevance of this variable in the context of the regional elections of 2004 is clearly only modest. That is, while we do expect Le Pen’s 2002 vote score to be a strong predictor of the FN’s vote in the 2004 regional elections because this would demonstrate that FN support at the departmental level is stable over time, the main purpose of introducing this additional variable into the model is to observe what happens to the effects of the other independent variables.


We chose this particular variable because it allows us to control for the fact that, over decades, the FN has been much more successful in some parts of France than in others (Bréchon & Mitra 1992), something which is due to the stabilizing effect of local party organizations (Lipset & Rokkan 1967: 53) and to the compositional effects and structural factors that benefit the party. Furthermore, the vote won by Le Pen in 2002 is an attractive measure of FN entrenchment because it cannot possibly have been affected by the district magnitude and the effective number of party lists in the regional elections in 1998 as the 2002 election was held under a completely different electoral system. As such, including this new variable in the regression should yield unbiased results for the relationship between FN support in 2004 and district magnitude/party system fragmentation in 1998, net of any other (stable) factors that are related to FN success at the departmental level.



[Table 1 about here]



Table 1 presents the coefficients of Kestilä and Söderlund’s original model as well as those for the augmented model (column 2). As we expected, the vote for Le Pen in the presidential election of 2002 turns out to be a strong predictor of FN success in the regional elections of 2004: each percentage point increase in support in 2002 translates into an increase of .98 percentage points in the party’s vote in 2004.


More important, however, is the fact that, once the lepeniste vote is controlled for, all other factors except unemployment are of very minor importance, with estimates that are very close to zero.[16] The lack of relevance of the five original independent variables is confirmed by a further model (column 3 of Table 1) in which all of the five original predictors are dropped and which shows FN support in the 2004 regional elections to be essentially identical to Le Pen’s vote in 2002 minus a constant of 2.7 percentage points.[17]


The discussion above suggests that, in the first instance, the lack of robustness of Kestilä and Söderlund’s model means it is unable to provide a compact description of the FN’s success at the departmental level in the 2004 regional elections. However, from both of the alternative models presented in Table 1 we have to conclude that the features of the subnational political opportunity structure included in Kestilä and Söderlund’s original model were largely irrelevant in explaining the FN’s vote in the 2004 regional elections anyway. As such, Kestilä and Söderlund’s model does not enable reliable inferences to be made about the impact of contextual factors on the radical right vote in Western Europe more generally, let alone allow for inferences that are more reliable than those made in the existing cross-national studies.



Towards an alternative model of FN success in the 2004 regional elections?


While we have demonstrated that Kestilä and Söderlund’s analysis suffers from a whole host of conceptual and methodological problems, we are still convinced that subnational political opportunity structures can, in principle, be very useful in accounting for the electoral success of radical right parties (and indeed any other type of party) provided this concept is operationalized in a more stringent way.


Given the data at hand, and especially given the lack of micro-level data on immigration and unemployment, the most obvious way the model may be improved is by replacing the ‘effective number of party lists’ variable with a variable that captures the ideological nature of party competition in the regional elections of 2004. As we argued earlier, the effective number of party lists reflects the accessibility of the regional party system, which is rather irrelevant in the case of the FN. Moreover, this variable has an element of tautology to it as it is not independent of previous levels of support for of the party. We therefore suggest replacing the effective number of party lists with two very simple variables: i) the presence of a second ‘extreme right’ list presented by the Mouvement National Républicain (MNR), and ii) the number of lists submitted by parties of the moderate right. Information pertaining to the lists presented in each region is readily available from the French government’s website (


We would expect the presence of an MNR list to reduce support for the FN, albeit only slightly. Given that the MNR broke away from the FN in January 1999 and is led by Le Pen’s former deputy, Bruno Mégret, one would expect many voters to see this party as a substitute for the FN.[18] As such, the presence of an MNR list should, ceteris paribus, reduce support for the FN because the political space available to the FN is more crowded. That said, we anticipate that the effect will only be modest because the MNR’s challenge to the FN effectively collapsed with the 2002 presidential election (Kuhn 2005: 102), when Mégret picked up only 2.3 per cent of the vote in the first round while Le Pen won 16.7 per cent and went on to contest the second round against the incumbent president, Chirac.


The number of mainstream competitors should also have a negative effect on the FN’s vote. Since this effect will not necessarily be linear, we will distinguish between three different scenarios: the presence of a single mainstream right list; the presence of two such lists; and the presence of three or more.



[Table 2 about here]



As a point of reference, column 1 of Table 2 shows the regression of the FN’s vote share in 2004 on the effective number of party lists in 1998 – i.e. the indicator favoured by Kestilä and Söderlund. The effect of this variable is slightly stronger in this bivariate model than it was in Kestilä and Söderlund’s complete model, but the very low R2 shows that it explains only a tiny fraction of the variation in the FN’s support. What is more, as is evident in column 2, the effect of the effective number of party lists disappears completely if we control for entrenched FN support by once again introducing our alternative predictor (the vote won by Le Pen in the first round of the presidential elections of 2002) into the model.


Column 3 of Table 2 presents the results of a model based on our alternative operationalization of party competition. It includes a dummy variable which takes a value of 1 in each département where the MNR presented a list, and dummy variables for the presence of two mainstream right lists, and three or more mainstream right party lists. This alternative model clearly fits the data much better than the effective number of party lists model. It explains a larger share of the variance in the FN’s vote, and the lower Bayesian Information Criterion (BIC) indicates that even though it includes more independent variables (and hence loses two degrees of freedom), this alternative model is preferable to the effective number of party lists one.


In this model the coefficients for competition from the moderate right have a straightforward interpretation and confirm our expectations: the FN’s vote in 2004 is reduced in départements where there were multiple moderate right lists. Compared to départements where the moderate right presented just one list, the FN vote is substantially (by over 6 percentage points) reduced where the party faced two mainstream right party lists. Where there were more than two mainstream right lists, the FN’s vote is reduced by over 3 percentage points.


Contrary to our initial expectation, however, we see that the presence of an MNR list in the 2004 elections does not reduce support for the FN. Rather, the presence of an MNR list has a substantial positive effect on the vote of the FN in 2004: after controlling for party competition from the mainstream right, the FN is on average 3.2 percentage points stronger in départements where the MNR fields candidates. We might explain this unexpected positive effect by pointing to the strategic choices made by the MNR’s leadership. While the FN presented candidates in all regions (and all départements), the MNR, stretched for money and staff, focussed its efforts on regions where the radical right had done well in the past – i.e. in areas where it might expect to do well. It contested 11 of the 14 regions where then FN had won above-average results in 1998 but chose to fight in only 2 of the 7 regions where the FN’s performance was below its average in 1998 (Cramér’s V=.49). The coefficient therefore picks up both the negative impact of competition from the MNR as well as the positive effect of previous FN support.


Our tentative explanation is confirmed by the findings in column 4 of Table 2: once we introduce the now familiar indicator for entrenched FN support, the effect of a competing MNR list becomes negative, as expected. Moreover, the effects of competition from the moderate right remain negative, though they are substantially reduced. This latter finding might again reflect strategic considerations of the FN’s competitors. After all, there are clear incentives for the moderate right to present a unified list in FN strongholds, something which is evidenced by a substantial correlation of r=0.42 between the FN support in the preceding regional election and the presence of a single mainstream right list.


The most important point about the model presented in column 4 of Table 2 is that the presence of an MNR list and the fragmentation of the mainstream right continue to have a theoretically meaningful effect even if previous FN support is controlled for. Moreover, the BIC indicates that this is an improvement over both the model that combined Kestilä and Söderlund’s effective number of party lists and the lepeniste vote (column 2) and the ‘pure’ model of entrenched FN support from Table 1. We take this as evidence that local/regional ideological competition matters and that it should be included in a subnational political opportunity structure model for radical right parties. The same cannot be said for the effective number of party lists.


As regards the other variables in Kestilä and Söderlund’s model, there is, unfortunately, no ‘easy fix’. We believe that some of them – namely turnout and district magnitude in 1998 – should not be included in the model at all because their conceptual status is dubious. Replacing district magnitude in 1998 with district magnitude in 2004 is also not an option because there is effectively no variation in this variable due to the legal thresholds in operation. And as for immigration and unemployment, although these are clearly part of the subnational political opportunity structure, in the absence of micro-level data it is simply not possible to investigate their impact and to untangle their contextual, compositional and cross-level effects.


Given this situation it appears that a major data collection effort is required if subnational political opportunity structures are to be operationalized rigorously and analysed fully and we would argue that such an endeavour should really go beyond merging survey data with subnational immigration and unemployment figures because Lubbers and Scheepers (2002) have already conducted an analysis of this kind for France. Rather, in an ideal world, a prospective project should collect data on variables that capture the theoretical concept of subnational political opportunity structures. This might include a content analysis of the local and regional media so as to capture its tenor (see Boomgaarden & Vliegenthart 2007 for a recent application to the national media in the Netherlands), an assessment of the organizational strength of local parties (see Pedahzur & Brichta 2002 on the institutionalization of the FN), and in-depth interviews with local political elites to probe their stances on radical right issues.





In their article, Kestilä and Söderlund highlight an important point, which although sometimes discussed in theoretical terms (e.g. Eatwell 2003), has largely been overlooked in empirical studies of the success of the radical right in Western Europe: local and regional contexts should not be ignored. Unfortunately, however, the importance of this message is somewhat obscured by the actual analysis that Kestilä and Söderlund carry out. For the reasons outlined above, we believe that there are difficulties with both Kestilä and Söderlund’s conceptualization of subnational political opportunity structures and their empirical findings.


A large data collection exercise focussing on factors that capture the concept of subnational political opportunity structures could potentially resolve many of the problems that Kestilä and Söderlund encountered in their study. Moreover, if time and resources were invested in any such future project, it would be all the more useful to analyse the relevant variables in a cross-national perspective. After all, authors such as Lubbers and Scheepers (2001, 2002) and Dülmer and Klein (2005) have already applied standard models of radical right voting to subnational units in individual countries.


That said, we fully concede that constructing cross-national models of radical right voting that contain rich information on very small subnational units (smaller even than French départements) would be a substantial accomplishment. Although collecting suitable data on one country is possible – as the British Election Study demonstrated more than ten years ago – gathering appropriate, comparable data across many countries would be a formidable feat. What is more, a cross-national study of subnational political opportunity structures would have to grapple with difficulties that are inherent to comparative analyses of this kind. That is, it would have to deal with the trade-off that exists between being able to draw conclusions that may be generalized beyond the cases in question and being able to gain an understanding of the intricacies of the particular contexts being examined. Indeed, some of the difficulties that Kestilä and Söderlund faced in their study reflect this very point: on the one hand their model is very sensitive to the selection of cases and hence does not allow for generalizable inferences to be made beyond the context of the 2004 French regional elections, but yet, on the other, it does contain rich information on the characteristics of the French regions and départements. This trade-off between generalizability and richness of data might raise questions over the very utility of any cross-national study of subnational political opportunity structures. Yet, if an appropriate balance can somehow be struck between these two concerns, we might learn a great deal more about the impact of local and regional contexts on the vote for radical right parties.





Achen, C.H. & Shively, W.P. (1995). Cross-level inference. Chicago: University of Chicago Press.


Alker, H.R. (1969). A typology of ecological fallacies. In M. Dogan & S. Rokkan (eds), Quantitative ecological analysis in the social sciences. Cambridge, MA/London: MIT Press.


Arzheimer, K. & Carter, E. (2006). Political opportunity structures and right-wing extremist party success. European Journal of Political Research 45(3): 419–443.


Auberger, A. (2008). The National Front vote and turnout in the French Presidential elections. French Politics 6(1): 94–100.


Berk, R.A. & Freedman, D.A. (2003). Statistical assumptions as empirical commitments. In T.G. Blomberg & S. Cohen (eds), Punishment and social control: Essays in honor of Sheldon L. Messinger. 2 ed. New York: Aldine de Gruyter.


Boomgaarden, H.G. & Vliegenthart, R. (2007). Explaining the rise of anti-immigrant parties. The role of news media content. Electoral Studies 26(2): 404–417.


Bréchon, P. & Mitra, S.K. (1992). The National Front in France. The emergence of an extreme right protest movement. Comparative Politics 25(1): 63–82.


Dülmer, H. & Klein, M. (2005). Extreme right-wing voting in Germany in a multilevel perspective. A rejoinder to Lubbers and Scheepers. European Journal of Political Research 44(2): 243–263.


Eatwell, R. (2003). Ten theories of the extreme right. In P.H. Merkl & L. Weinberg (eds), Right-wing extremism in the twenty-first century. London: Frank Cass.


Fox, J. (1997). Applied regression analysis, linear models, and related methods. Thousand Oaks, CA: Sage.


Golder, M. (2003). Explaining variation in the success of extreme right parties in Western Europe. Comparative Political Studies 36(4): 432–466.


Jackman, R.W. & Volpert, K. (1996). Conditions favouring parties of the extreme right in Western Europe. British Journal of Political Science 26(4): 501–521.


Kestilä, E. & Söderlund, P. (2007). Subnational political opportunity structures and the success of the radical right. Evidence from the March 2004 regional elections in France. European Journal of Political Research 46(6): 773–796.


Knigge, P. (1998). The ecological correlates of right-wing extremism in Western Europe. European Journal of Political Research 34(2): 249–279.


Kuhn, R. (2005). The French regional and European elections, 2004. Representation 41(2): 96–105.


Lijphart, A. (1994). Electoral systems and party systems. A study of twenty-seven democracies, 1945-1990. New York: Oxford University Press.


Lipset, S.M. & Rokkan, S. (1967). Cleavage structures, party systems, and voter alignments: An introduction. In S.M. Lipset & S. Rokkan (eds), Party systems and voter alignments: Cross-national perspectives. London: Collier-Macmillan.


Lubbers, M. & Scheepers, P. (2001). Explaining the trend in extreme right-wing voting. Germany 1989-1998. European Sociological Review 17(4): 431–449.


Lubbers, M. & Scheepers, P. (2002). French Front National voting: A micro and macro perspective. Ethnic and Racial Studies 25(1): 120–149.


Lubbers, M., Gijsberts, M. & Scheepers, P. (2002). Extreme right-wing voting in Western Europe. European Journal of Political Research 41(3): 345–378.


Pedahzur, A. & Brichta A. (2002). The institutionalization of extreme right-wing charismatic parties: a paradox? Party Politics 8(1): 31–49.


Reif, K., Schmitt, H. & Norris, P. (1997). Second-order elections. European Journal of Political Research 31(1-2): 109-124.


Robinson, W.S. (1950). Ecological correlation and the behavior of individuals. American Sociological Review 15(3): 351–357.


Swank, D. & Betz, H-G. (2003). Globalization, the welfare state and right-wing populism in Western Europe. Socio-Economic Review 1(2): 215–245.


Taagepera, R. & Shugart, M.S. (1989). Seats and votes. The effects and determinants of electoral systems. New Haven, CT: Yale University Press.


van der Eijk, C., Franklin, M. & Marsh, M. (1996). What voters teach us about Europe-wide elections. What Europe-wide elections teach us about voters. Electoral Studies 15(2): 149-166.



Figure 1: The relationship between the effective number of party lists and the vote share of the FN in four French départements in the 1998 regional elections



Table 1: Alternative models of FN support in the French 2004 regional elections







Kestilä & Söderlund’s model

Kestilä & Söderlund’s model plus Le Pen vote

Le Pen vote only

District magnitude 1998 (ln)








Effective number of lists 1998








Turnout 2004 (per cent)








Immigrants born outside EU (per cent)








Unemployment (per cent)








MNR running








Moderate right lists: 2








Moderate right lists: 3+








Vote for Le Pen 2002
















Adj. R2




Root MSE




















Standard errors in parentheses

* p < 0.05, **p < 0.01, ***p < 0.001


Table 2: The effect of the effective number of party lists in 1998 and ideological competition in 2004 on FN support in the French 2004 regional elections








Effective number of party lists

Effective number of party lists plus Le Pen vote

Ideological Competition

Ideological Competition plus Le Pen vote

Effective number of lists 1998










Vote for Le Pen 2002










MNR running










Moderate right lists: 2










Moderate right lists: 3+




















Adj. R2





Root MSE

























Standard errors in parentheses

* p < 0.05, **p < 0.01, ***p < 0.001



[1] The choice between these two terms seems to be largely a matter of taste. To avoid unnecessary confusion, we follow Kestilä and Söderlund who chiefly use the adjective ‘radical’.

[2] Since the findings for the index of electoral success are largely comparable, our discussion will focus on the more straightforward measure of electoral support (i.e. vote share).

[3] In other words, in this simulation we assumed that voters would move between the FN list and mainstream right lists (i.e. RPR, UDF, and ‘divers droite’ lists) but not between left and right blocs. We further assumed that movements within the right bloc would not trigger movements between other parties. If we drop these assumptions and instead suppose that support for the FN comes from and goes to all other parties the results are almost identical. In both simulations, the upper threshold is the FN’s empirical vote share plus ten percentage points. The lower threshold is either the empirical vote share minus 20 percentage points or zero.

[4] The Pearson correlation for these four curves, which picks up the linear component, varies between .73 (Vaucluse) and .96 (Haute-Vienne and Indre).

[5] The share of the vote a party must win in order to gain parliamentary representation is determined either by the district magnitude or by the existence of a legal threshold if that legal threshold overrides the impact of district magnitude. To ascertain whether it is the district magnitude or the legal threshold which determines the vote a party needs for representation, or indeed to compare electoral systems with and without legal thresholds, we can make use of either Taagepera and Shugart’s ‘effective magnitude’ (1989: 135–141), or Lijphart’s ‘effective threshold’ (1994:182-183, note 29). Using the latter, the formula for which is 75/(M+1), we can see that if there had been no legal threshold in place in the regional elections of 1998, parties would have needed to win 18.75 per cent of the vote to gain representation in the district with the smallest magnitude (the Lozère which had a district magnitude of 3), whereas they would have needed only 1.03 per cent of the vote to win representation in the départment with the largest magnitude (i.e. the Nord which, as the most populous département, had a district magnitude of 72). In the Nord the legal threshold clearly overrides the effects of district magnitude. Indeed, the point at which the legal threshold starts overriding the effect of district magnitude is 14, since a district magnitude of 14 implies an effective threshold of 5 per cent.

[6] The electoral system used in 2004 included a number of legal thresholds. The law stipulated that in order for a party list to proceed from the first round of the election to the second it had to win at least ten per cent of the valid votes in the region. Lists that won five per cent in the region could also proceed if they fused with a list that had won ten pent of the valid votes in the region. The party list which won an absolute majority at the first round (if this occurred) or a plurality at the second round was given an automatic 25 per cent of the seats. The remaining seats were distributed proportionally among all party lists that had won at least five per cent of the votes in the region (Kuhn 2005; As it turned out, no party list won an absolute majority in the first round of the 2004 elections in any of the 22 regions of metropolitan France so all contests went to a second round. Had there been no legal thresholds in place in 2004 parties would have been able to win seats with very small percentages of the vote: district magnitudes of between 43 and 209 would infer effective thresholds of between 1.7 per cent and 0.36 per cent (see note 5). As such, the legal thresholds overrode the effects of district magnitude in all cases.

[7] Figures that pertain to the number of immigrants born outside of the EU do not capture the racial, ethnic, and/or religious characteristics of immigrants, something which is less than ideal in this instance given the FN’s appeals centre on notions of race, ethnicity and religion.  That said, data on the racial, ethnic and religious attributes of immigrants in France have not been collected.

[8] When both micro- and macro-data are available, separating these effects (by way of a multi-level model) is relatively straightforward. When only macro-level data is at hand (as in this instance), things are much more difficult, however. Indeed, the interpretation of pure macro-data leads almost inevitably to cross-level inferences which are highly problematic unless very specific assumptions hold (Achen & Shively 1995; Alker 1969; Robinson 1950). These assumptions include the need for extreme distributions (i.e. départements with almost no unemployment or immigration and départements with almost full or no unemployment or immigration) which would then enable the calculation of a range of individual correlations that are compatible with the observed aggregate correlations. Even then, however, one would need to be very cautious. What is more, things are even more complicated in this instance because, since Kestilä and Söderlund are interested in contextual effects, their study is not a straightforward ecological analysis of individual behaviour. Rather, their interest in contextual effects means that they imply a two-level model: conditions vary at the level of the département; these then affect whether party lists are presented at all, whether parties chose to present individual or joint lists, and just how much parties try to mobilize voters; and these two factors then affect the behaviour of individual voters.

[9] This does not include people born in the départements d’outre-mer and the territoires d’outre-mer (DOM-TOM), which are considered part of France for census purposes.

[10] See note 8.

[11] For instance, immigrants have a well above-average propensity of being unemployed, but individual unemployment status will in all likelihood have a different effect on the probability of an FN vote for immigrants and non-immigrants.

[12] The latter problem could be rectified by weighting the départements according to their population. In the event the coefficients do not actually change that much if départements are weighted by population, though the coefficient for immigration is effectively reduced to zero and the adjusted R2 decreases.

[13] ln(22)-ln(10)×3.45

[14] What is more, these calculations ignore the fact that the effects of district magnitude are effectively cancelled out in districts with a magnitude of 14 or more because of the existence of a five per cent legal threshold – see note 5. Had the effects of this legal threshold been taken into consideration, the effect of district magnitude would have been even smaller.

[15] Bompard was re-elected as mayor of Orange in 2001. Then, in September 2005 he resigned from the FN and joined the Mouvement pour la France (MPF) three months later. He was again re-elected as mayor in March 2008.

[16] The augmented model also allows much better predictions than the original one: the adjusted R2 almost doubles, while the mean squared error of the prediction is reduced by roughly 50 per cent. Given that just one additional parameter is estimated, the drop in the log-likelihood is massive. Accordingly, the Bayesian Information Criterion (BIC), which relates the improved fit of a more complex model to the ‘costs’ of adding parameters, drops substantially, indicating that the augmented model is indeed preferable to the original one.

[17] This simple model fits the data almost exactly as well as the augmented model, resulting in an even lower BIC.

[18] Mégret announced his retirement from politics in May 2008 and the following month it was decided at the MNR’s National Council that the party would henceforth be led by the 7-member Executive Bureau.