POLITICAL IDEOLOGY AND ETHNIC PERCEPTIONS: THE TWO EUROPES

Transcription

1 POLITICAL IDEOLOGY AND ETHNIC PERCEPTIONS: THE TWO EUROPES Immigration, Minorities and Multiculturalism In Democracies Conference Ethnicity and Democratic Governance MCRI project October 25-27, 2007 Montreal, QC, Canada Work in progress. Please no not cite. Sergi Pardos-Prado PhD student at the European University Institute, Florence

2 1 ABSTRACT: While economic and socio-demographic features of countries and individuals have been widely considered so far as determinants of the hostility towards migrants, the role of political ideology, values and inherited axiological structures has been surprisingly unattended in Europe. This paper aims to fill this gap and shows that political ideology shapes and frames ethnic perceptions to a remarkable and robust extent, and that categories like left and right emerge as important psychological tools to reduce complexity and address people s opinion when the competitive and threatening dynamics of the interrelation with migrants is less evident. Overall, the findings presented here draw a binary portrait of Europe according to two overlapping criteria: the post-communist vs. the non post-communist Europe (where ideology matters more); and the less educated and located in the right-authoritarian ideological axis vs. the more educated and located in the left-libertarian pole (who do not perceive migrants as dangerous competitors and rely more upon ideological categorisations). - INTRODUCTION Understanding the way in which the resident population of a given country constructs the image of a foreigner has been a matter of interest for migration studies since more than four decades now. Among the many topics regarding the analysis of immigration in political science, sociology and social psychology, the public attitudes towards migrants still touch one of the core essential aspects of people movements across countries. i.e., why does xenophobia emerge and how to integrate culturally different people in a unified but diverse demos? In spite of the relevance of the topic and the broad theoretical coverage that has deserved, however, the impact of political ideology in framing ethnic perceptions on a comparative European basis has been practically unattended. The previous theories mainly stress the socio-structural and economic aspects of it. The social position of citizens is understood as the main explanatory factor of ethnic perceptions, whereas the role of values and ideology as psychological tools, shortcuts and schemes to reduce complexity and address people s opinions about immigration is relatively unknown in Europe. Since the general framework established by Adorno and his colleagues nearly three decades ago, the categories left and right have been considered as a psychological tool framing tolerant versus authoritarian opinions towards different sorts of people and ways of life (Adorno 1969). This paper relies upon this school of thought and not only shows that political ideology indeed contributes to frame ethnic perceptions, but that it also has a stronger effect as a psychological tool among those citizens less likely to see migrants as competitors. Ethnic and realistic group conflict theory is one of the main set of principles shedding light on the causes of xenophobia and predicts that the higher the direct

3 2 competition for scarce resources with immigrants, the more likely the emergence of hostile attitudes towards them (Blumer 1958; Sherif and Sherif 1969; Bobo and Hutchings 1996; Bergesen and Herman 1998; Oliver and Wong 2003; Kauffman 2003b; Coenders 2004). The findings presented here suggest that when this perception of competition is more evident, people tend to articulate their ethnic perception in a more straightforward manner (without the mediation of previously inherited and predefined values). When the competition is less evident, however, people tend more to reduce complexity, make sense of immigration and interpret its social impact through the usage of ideological labels. This paper is divided into seven sections. The first conceptualises political ideology and justifies the use of the left-right axis as its main operative measure. The second and the third present the hypothesis to be tested and explain the individual and contextual elements which interact with the framing effect of left-right in ethnic perceptions. The fourth section justifies the necessity of disentangling different dimensions of attitudes towards migrants and the use of three different dependent variables. The fifth section presents the data and the method, the sixth reports the results, and the seventh concludes. - POLITICAL IDEOLOGY: A CONCEPTUALISATION In this section I attempt to define the concept of political ideology through three components (stable, rational and symbolic), and to justify the left-right axis as its first basic operasionalisation. The idea that issues need to be framed in the normal ideological structure of a society to become institutionalised in the political dynamics is not new. In one of their more classic works, Nie, Verba and Petrocik have argued that as time goes by and as issues become more settled in the public agenda, public opinions towards different topics tend to correlate more with each other. These authors have theorised about this phenomenon through the notions of issue constraint (conservative people in economic issues are more likely to also become conservative in other issues, and viceversa) and issue consistency (all these coherent issue positions can be framed and located in an ideological axis, which correspond to the liberal/conservative schema in the American context) (Nie, Verba et al : ).

4 3 The relationship between an ideological axis and issue opinions is an indicator of stability. Ideological values are understood as products of the socialisation processes and are thus relatively unchangeable over time. Even if previous researchers have acknowledged a certain progressive variation within societies or individuals (because of the adaptation to new social contexts or because of radical revolutionary changes), one of the necessary conditions of ideological orientations is its relative predictability over time (Eckstein 1988; Inglehart 1988). Indeed, this stability of the link between ideology and public perceptions was particularly notorious precisely in the opinions about race in the USA during the last century, when this issue was first politicised (Converse 1964; Nie, Verba et al :104). More recent research in cognitive theories also supports the idea that ideology generates stability in people s political worldview. According to these experiments, even experts with a high level of knowledge and political sophistication use ideology to articulate predictions about the future and base their expectations on past events (Tetlock 1999). In this case, ideology becomes a connective link between past events and future expectations generating an evident cognitive coherence over time. Moreover, ideology not only affects citizens political and social opinions but it also influences other political actors movements such as parties. The stability implied in ideological orientations has been shown in the supply side and not only in the demand side of the electoral competition dynamics. Adams, Clark, Ezrow and Glasgow find evidence that, surprisingly, parties do not adjust their ideologies in response to past election results. They only react to shifts in public opinion, and these effects are only significant in situations where public opinion is clearly shifting away from the party s policy positions (Adams, Clark et al. 2004). Ideology is thus a clear indicator of stability, both on the public opinion and on the parties side. Apart from its inherent stability, the notion of ideology can be characterised by two more components: a rational/cognitive and a symbolic/affective. Regarding the rational/cognitive component, ideology has indeed been considered as a tool to frame and organise opinions towards rather disparate issues difficult to know in all their complexity (Nie, Verba et al :117, ; Klingemann 1979). As Sniderman reminds us, the notion of issue framing has not been given a rigorous interpretation even if it is central to understand how people make sense of attitudes towards public policies (1993:222). The question about how far people s ideas of politics are shaped by ideological conceptions was born in the USA, the work of Converse (1964) and Kinder (1983) being two of the main classic inflection points. The rational component of ideology relies on the expression of preferences and on the maximisation of utility. The constant calculation of costs, benefits and utilities inherent

5 4 to any political action requires a tool like the labels left/right or liberal/conservative to organise the information and offer a comprehensible path (usually in terms of a spatial distribution of preferences) to let individuals locate themselves and to minimise the distance from their preferred possible outcomes (Downs 1957). It has been argued, however, that consecrating a rational dynamics like this one as an exclusive feature of ideology would lead to a high variability of attitudes and political acts depending on the context. Since this theoretical expectation has no correlation with the empirical life and self-placements in the left-right continuum or voting preferences seem to be quite stable over time, some authors have defended the consideration of a certain affective attachment and symbolic component in the nature of political ideology (Conover and Feldman 1981; Sniderman 1986; Sniderman, Brody et al. 1991). In this sense, a citizen s political ideology can also be understood as a result of the social codes, world visions, values and cultural background that s/he has inherited through the socialisation processes. In terms of indicators, the self-placement along the left-right continuum has become a virtually unavoidable tool in attitudinal research since the times of the French Revolution. 1 Either from a sociological (Lipset and Rokkan 1967; Bartolini and Mair 1990), psychosociological (see Medina Lindo 2004) or from a rational perspective (Downs 1957), the leftright axis has been considered a reliable and comparable indicator of ideology. Taking into account the previous contributions, the notion of political ideology can be understood as a category within the broader concept of ideology and may be defined as follows: political ideology is a stable set of psychological orientations which serve to frame individuals opinions towards political issues (like immigration) through a cognitive component (which decodes and makes sense of reality and which lets people maximise their utility in a given circumstance of choice) and a symbolic component (which implies an affective attachment to values inherited via socialisation processes). The main comparable indicator of people s political ideology across Europe is their self-placement along the leftright scale. 1 The categories left and right were used for the first time in the first National Assembly in the era of the French Revolution.

6 5 The left-right axis, however, could be considered a too broad ideological dimension to disentangle and grasp different specific values contained in it. For conceptual but also analytical purposes, it is important to take into account to some extent the multidimensional meaning of the left-right schema and to be able to grasp its axiological shades. As I will argue below, at least the differential impact of left-right along its economic (less vs. more economic individualism) and its cultural meaning (less vs. more social intolerance) should be tested. - THE IMPACT OF LEFT-RIGHT AND ITS INDIVIDUAL INTERACTIONS There is a certain consensus about the role of ideology in shaping attitudes towards substantive political issues like economy, immigration, environmental protection or foreign policy. The conceptions of issue opinions as derivations of broader values stress this direction of causality (Van der Eijk 1999; Dalton 2006 :100). Even if ideology has already been conceptualised as a predictor of ethnic perceptions (Citrin, Sears et al. 2001), however, the correct assessment of the direction of causality in attitudinal and behavioural research is always slippery. Therefore, the analyses proposed below do not necessarily assume a relation of causality between ideology and attitudes towards immigration. Rather, the aim here is to observe the strength of the connection between these two elements and then be able to interpret when, where and why immigration has to be embedded in the value system of a society to let the people construct their attitude towards migrants. There are reasons to think that immigration has a significant relation with the main ideological axis structuring a society s distribution of political preferences. In the American context, for instance, Sniderman and Carmines consider that immigration is not a separate issue and that it is framed in the liberal and conservative categories (Sniderman and Carmines 1997). On the other hand, when analysing the nature of extreme right-wing parties in Europe, Kitschelt argues that they do not only respond to a single issue concern regarding immigration. To be electorally successful, they have to be embedded in the particular configuration of citizens ideological preferences in postindustrial societies, which correspond to an axis ranging from left-libertarian to right-authoritarian attitudes 2 2 Kitschelt s argument is that people s preferences are now concentrated along these two poles. The leftlibertarian extreme refers to opinions in favour of government intervention in the economy and pro liberal moral and cultural attitudes, whereas the right-authoritarian pole attracts opinions in favour of economic individualism and against certain moral and cultural ways of living.

7 6 (Kitschelt 1995 :15,35). According to him, if a party expresses anti-immigrant tendencies but forgets to locate its own discourse in this broader right-authoritarian ideological agenda it will fail to attract voters. Immigration must thus be incorporated in the specific ideological framework of contemporary societies to get saliency. Kitschelt s fusion of economic and cultural ideological dispositions is quite similar to the distinction and subsequent integration of the so-called Old Politics and New Politics issues respectively (Dalton 2006 :140). In spite of the previous contributions, however, political ideology is usually considered a simple control variable or even forgotten in the models explaining ethnic perceptions (Coenders 2004 :101; Pantoja 2006 :524). Even if there is a certain basis to assert a prominent impact of ideologies, values and general belief systems in constructing positive and negative attitudes towards migrants beyond economic conditions (Citrin, Green et al. 1997; Burns and Gimpel 2000; Alba, Schmidt et al. 2003), the explanations more usually considered in Europe to explain hostile attitudes and behaviour against migrants are mainly economic and socio-structural. Even if it may sound obvious, it has to be stressed that the relation I expect to find between political ideology and ethnic perceptions cannot be endogenous. In other terms, left-right self-placements and attitudes towards migrants cannot overlap excessively or even be the same indicator from a statistical point of view. Previous research shows indeed that racial resentment and ideology do not have an endogenous relationship (Feldman and Huddy 2005 : ,180). Moreover, since the seminal work of Adorno and his colleagues in The Authoritarian Personality (Adorno 1969), right-wing tendencies have been associated with a more rigid way of thinking, characterised by a Manichean tendency to split the world up between friends and enemies, and to dislike people with an unfamiliar appearance or background. Therefore, I expect to find a positive and not endogenous relationship between ideology and ethnic perceptions. All these arguments lead to the principal hypothesis to be formulated in this section: H1- Political ideology on the basis of the self-placement in the left-right continuum has a positive, significant but not endogenous relationship with the different dimensions of attitudes towards migrants beyond other theories about ethnic perceptions.

8 7 On the other hand, the overall complexity of the link between these two elements might not be completely grasped through the validation of this first hypothesis. The growing complexity and heterogeneity in preferences, interests and ways of life in contemporary societies lead to think of a non common and unidimensional effect of left-right. The notion of causal heterogeneity has become practically unavoidable in the research about public attitudes. It is widely acknowledged that people no longer tend to make up their minds in the same way and that explaining the whole behaviour of a country through a single additive regression model is no longer valid. Citizens are not equally informed and equally interested in issues, and they do not necessarily respond equally to the same stimulus in shaping their opinions to them. Therefore, in words of Sniderman, one of the key questions here is under what conditions is ideology related to policy preference? (1993:221, ). The necessity of taking into account this heterogeneous attribution has also been underlined in the context of the research about ethnic perceptions (Gomez and Wilson 2006) and other issues like the economy (Van der Brug, Van der Eijk et al :52). As regards the possible interactions shaping the impact of left-right in attitudes towards migrants, it is necessary to refer to the different substantive meanings that people might give to this continuum. The main contribution in this sense in the previous literature has been to identify two main substantive connotations of the left-right axis: the economic and the cultural. These two components are already easily distinguishable as independent dimensions in classic definitions of left and right (Lipset, Lazarsfeld et al :1135; Derks 2004). The existence of an economic and a cultural dimension in the meaning of leftright which vary independently from each other has received strong empirical support in analyses about party competition (Kitschelt 1995 :44-45; Dalton 2006 :142). The first meaning deals mainly with questions of economic equality and government intervention. Acceptance of equality is, indeed, one of the main properties distinguishing the historic philosophical meaning of the labels left and right (Bobbio 2000). This can be measured with individualist values, since individualism has been defined as the belief that individuals can and should get ahead on their own merit without government assistance (Pantoja 2006). The inclusion of a certain measure of individualist values has appeared to be very important in studies of ethnic perceptions (Federico 2006 :600; Gomez and Wilson 2006 :618,620; Pantoja 2006). There is a second meaning, however, where authoritarians contrast with more tolerant or tender-minded people in cultural and moral terms (Eysenck 1954; Kitschelt 1994; Enyedi

9 8 1999). This connects with another of the main historic semantic differential linkages of left and right, which is the support towards the ideas of tradition and emancipation (Bobbio 2000). In this case, the cultural and moral dimension of left-right has a lot in common with the notion of cosmopolitanism, a relevant concept when explaining attitudes towards migrants. The higher the support for cosmopolitan values, the lower the tendency to generate hostile attitudes forwards foreigners (Haubert 2006). These findings lead to hypothesise in this paper that the impact of left-right in the hostility to migrants will depend on people s position towards both economic issues such as government intervention and cultural issues such as social tolerance of different ways of life. But which should be the expected direction of the interaction between the left-right axis and its two more prominent meanings? If we rely on the notion of ideology as a relatively permanent cognitive tool which is used to reduce complexity and frame issue opinions through a rational and a symbolic component, we might expect a higher impact among those citizens who do not necessarily have a defined opinion about migrants due to their social, cultural and economic background, and therefore need to frame it according to a consistent and predefined set of values. In other terms, political ideology can be hypothesised as a tool to help people to construct their opinions towards migrants, but especially among those individuals who do not have to have a specific ethnic attitude according to other theories about ethnic perceptions. If we rely on group-conflict theory as one of the main theoretical and strongly empirically based approaches to the study of attitudes towards migrants, we have to expect a higher likelihood to find negative opinions among those individuals who see migrants as competitors in a zero-sum game for both economic and cultural resources. Then, it is reasonable to expect a higher impact of ideology in those individuals who are not in a position of direct competition with migrants, who therefore are not likely to construct a strong opinion about them, and who need to interpret this new phenomenon and to seek issue consistency through other methods. If we apply the principle of political ideology as a tool to frame and make sense of the perceived impact of immigration in a given society which has a stronger impact among those who do not necessarily see migrants as direct competitors for scarce resources, we have to expect a weaker influence of the left-right axis in shaping ethnic perceptions among those economically individualist and culturally intolerant. This is so because those who have economic individualism as a prominent value in their axiological structure base their own success and progression in life in an individualistic effort where the help of the state

10 9 intervention is not particularly appreciated. It is in this mental schema, where everybody depends on his/her own performance, that migrants can be seen as competitors according to this individualistic sense of social progression. On the other hand, those economically leftist who trust social aids and state intervention to reduce inequalities might be less prone to see migrants as competitors because they do not see themselves alone competing with them: the state will take care of the fact that, in spite of the emergence of competition, nobody will necessarily loose completely and there will always be a certain social protection. Something similar might happen when analysing the cultural meaning of the left-right axis. Those morally conservative and less keen to change the cultural status quo existing in a society are those more likely to see migrants as competitors and as a clearer threat because of their different religious and cultural background. The cultural challenge to the status quo code, however, is perceived as less problematic among those people without a position of cultural dominance or among those with a higher tendency to accept a wider pluralist range of cultural ways of life. It is among this latter profile of citizen where ideology might play a stronger role in shaping ethnic perceptions, because the theoretical principles of competition theory work worse and because people need other cognitive tools to make sense of reality and generate opinion consistency in a new issue such as immigration. Still in terms of individual interactions to be included in the models, the role of ideology in framing ethnic perceptions may also depend on the individuals level of education. As stated above, the left-right continuum lets people organise rationally their opinions about very different issues and avoid the extremely high cost of knowing everything from every issue. This framing function of ideology has been related to the research about the coherence of belief systems. Converse (1964) and Nie, Verba and Petrocik (1976) considered that this internal consistence of people s values and opinions depends on the level of individuals political sophistication. The authors of The New American Voter used the term ideologues to describe the citizens that tend to use ideology to frame issues, and near-ideologues to describe those who only use ideology without issues. In their classic contribution, these authors saw an increase in the number of the former and a decrease in the latter due to higher levels of political sophistication (Nie, Verba et al :12-113,117; Sniderman 1993: ). This link between a stronger use of ideology to frame issues and a higher level of political sophistication, however, is not accepted by everyone. More recent works in the context of ethnic perceptions studies acknowledge that individuals with differing levels of political sophistication seek congruence selectively. In contrast to

11 10 Converse s model of the political belief system, they suggest that it is the least sophisticated individuals, not the most, who attempt to maximise proximal consistency. On the contrary, highly sophisticated individuals seek distal consistency in their political belief systems (Gomez and Wilson 2006). Following the same logic which served to articulate the hypothesised effect of the two former interactions, ideology can have a stronger effect among those who do not necessarily see migrants as competitors. When the competition for scarce resources with migrants is less evident and when therefore the perception of vulnerability cannot work as a cognitive tool to interpret the impact of immigration, previous ideological values contained in the axiological background of the individual can reach a stronger role in organising and articulating ethnic perceptions. In the more specific case of education, group-conflict theory expects a higher tendency to have negative attitudes towards migrants among those citizens with lower levels of cognitive resources (because they usually imply a more vulnerable social position). Thus, I expect a stronger conditional impact of ideology among those with higher levels of education who are less likely to frame the impact of immigration in competitive terms. H2- The effect of the left-right self-placement in shaping attitudes towards immigration is stronger among those in favour of the government intervention in the economy (H2a), among those more socially tolerant (H2b), and among those with a higher level of cognitive resources (H2c). - POLITICAL IDEOLOGY AND ETHNIC PERCEPTIONS ACROSS CONTEXTS Apart from the causal heterogeneity due to interactions with individual level variables, the effect of values is also very likely to depend on the context (Sniderman 1993 :233). In terms of ethnic perceptions, it has already been found that the impact of political ideology in shaping these attitudes is differential across culturally distinct geographic areas (Clark and Legge 1997 :910,912 for a comparison between Eastern and Western Germany), different economic contexts and different legal types of citizenship regimes for immigrants (Weldon 2006 : ).

12 11 Since, as argued above, the two most relevant dimensions of the left-right continuum refer to its economic and cultural meaning, it is reasonable to think that its impact may also depend on the economic and cultural features of each country. As for the economic contextual characteristics, some of the features that have proved to be relevant in the field of ethnic perceptions are the country s level of wealth, social expenditure and unemployment. As claimed by group conflict theory, the scarcity of resources in a country increases the likelihood of generating hostile attitudes against migrants (Coenders 2004). Therefore, it could be expected that in those contexts where the scarcity of resources is more evident, there is a higher probability that people will be more negative to migrants no matter their ideological constructs. In better off countries, however, where the eventual competition with immigrants for scarce resources is less evident, citizens might have a higher tendency to use ideological labels to make sense of the impact of immigration. Among the factors shaping contextual economic conditions I will consider the level of unemployment, which is one of the most straightforward indicators directly measuring the scarcity of resources in such an important source of them like the job market, and a good proxy for the global economic health of a country. As regards the cultural features of the context, it could be argued that the impact of citizens self-placement in the left-right axis depends on the differential cultural and historic usage of these categories in each place. In Europe, this theoretical expectation refers first to the distinction between post-communist and non post-communist countries. There are three specific arguments which foster the intuition about a differential impact of the left-right axis in the post-communist area: the historic differential use of such an ideological schema, the presence of left-wing people with comparatively more negative attitudes towards immigrants, and relatively higher overall levels of xenophobia. The different cultural heritage and the enormous social and economic shifts underway across the former communist world have already made it necessary to distinguish between East and West Europe in mass attitudes research (Tucker, Pacek et al :569). The separation between left and right has been historically denaturalised in the post-communist area and thus might have a weaker impact in framing issue opinions 3 (Tismaneanu 2002 :91,93). 3 The different historic cultural usage of the categories left and right in Central and Eastern Europe could be seen for some as an argument of the incomparability of the labels with the West. Nevertheless, as Miller, White and Heywood explain (1998: ), there should be nothing extremely foreign about the political meanings of the words left and right. They derive from the seating arrangements in the French

13 12 It might also be useful to focus specifically on post-communist countries, because the historic status quo corresponded to the left during the old communist regimes, and to the right in the Western European countries. According to Tismaneanu, Eastern European societies have inherited a particularly noticeable anti-modern cultural background and nationalist and xenophobic tendencies. This heritage, together with the uncertainty regarding the consequences of the transition and the loss of status and identity (Tismaneanu 2002 :83,87,90,92) might have had a greater effect on the left, which held a position of dominance before the shift to a market society in post-communist countries. As a consequence, left-wing citizens could have lost more in terms of cultural and even economic resources and thus could be more prone to showing hostile attitudes towards immigrants. The existence of winners and losers in post-communist regimes has been acknowledged in recent academic contributions, but the definition of what is a winner and what is a loser is much less consensual. The distinctions proposed so far have been based on socio-demographic and economic variables (Tucker, Pacek et al :559), but I propose to test here an ideological criterion. Indeed, left-wing ideology has been more related to authoritarian and intolerant practices in the post-communist than in the non post-communist world. The ideological interpretation of the changes after the end of the communist regimes and the eventual image of immigrants in these new free market societies could also have been different in post-communist systems (Enyedi 1999). Therefore, it can be hypothesised that right-wing citizens might be less intolerant towards migrants in post-communist than in non post-communist societies. Finally, the comparatively higher levels of nationalist and xenophobic tendencies detected in post-communist societies in previous research and mentioned above might also inhibit the effect of the left-right ideology. This is so because if the overall level of xenophobia in the country is higher, the individuals can be socialised in this particular code to perceive foreigners and have a stronger tendency to adopt the collective vision. A high global level of hostility towards migrants can also work as an indirect clue for individuals with no particular opinion to adopt the dominant vision about the issue. On the other hand, in those revolutionary assemblies, and all the countries they study (Russia, Ukraine, Czech Republic, Slovakia and Hungary) were invaded by French revolutionary armies. Indeed, the French Revolution and its new paradigm of political values caused a deep impression on the symbolic structures of the Tsarist empire. Thus, the existence of a common organisational political schema makes possible the comparison of the left-right continuum between post-communist and non post-communist systems. Moreover, the findings of Miller, White and Heywood reflect that the left is significantly more committed to socialist values, consistently with the Western meaning (1998:310). On the other hand, Markowski (1997) finds that the correlation between left-right self-placements and economic and cultural opinions in East Central Europe reflects a very similar pattern as what could have been expected in a Western country.

14 13 more moderate countries where the levels of xenophobia are less pronounced, the particular saliency of the collective attitude cannot act as a clue anymore and people might use predefined ideological schemas such as left-right to frame and even generate a coherent attitude towards migrants. H3- The effect of the left-right self-placement in shaping attitudes towards immigration is different across countries. This variance is accounted for by interactions between this ideological continuum and both economic and cultural features of the country. More specifically, left-right matters more in countries where competition for scarce resources in the job market is less evident (H3a), and matters less in post-communist countries where these ideological categories have been historically denaturalised, where left-wing people are comparatively more negative towards immigrants and where the overall levels of xenophobia have been considered comparatively higher (H3b). - ON THE DEPENDENT VARIABLE AND ITS MULTIDIMENSIONALITY In the previous sections I have defined the concept of political ideology and I have hypothesised its role in shaping and framing ethnic perceptions. My objective now is to discuss and assess empirically the proper specification of the dependent variable, that is, ethnic perceptions. The necessity of disentangling different dimensions in the notion of ethnic perceptions, the appropriacy of not assuming ex ante classifications of types of attitudes but observing how people do actually construct their opinion towards migrants, and the opportunity to shed light on contradictions found in previous literature about ethnic attitudes probably due to a bad specification of the dependent variable, justify the existence of this section. The intuition about the existence of a multidimensional semantic structure in the public construction of immigrants comes in part from some contradictions detected in previous analyses of ethnic perceptions (Lahav 2004). For instance, it has been found that in opinion polls thousands of European citizens express a commitment to helping immigrants in an abstract sense, whereas at the same time they tend to have a negative view of the impact of immigrants on jobs, crime and society. This is a similar pattern to that found in the USA, where citizen support for the principle of equality coexists with the remnants of segregation.

15 14 The existence of distinct and relatively independent dimensions in the public perception of immigration can also be inferred through their variation over time. In the USA the racial issues which were intensely contentious in the 1960s (school desegregation, open housing and public accommodations) register more liberal responses today. On the contrary, however, new racial issues (quotas and affirmative action programs) now seem to divide sharply the American public (Dalton 2006 :109,119). As the research in social psychology progresses, there is increasing evidence that issues are not constructed as single unidimensional entities in people s minds but that they come together with each other (Duckitt 1992; Sniderman 1993). When analysing the emergence and success of radical right-wing parties, for instance, Kitschelt argues that this process cannot respond exclusively to a single issue concern regarding immigration. According to him, if this were so, when the saliency of immigration decreases the electoral success of these parties should also decrease. This correlation, however, does not exist in reality and therefore the success of these parties must be linked to a broader issue agenda (Kitschelt 1995 :1). The attempt to identify different dimensions in racial policy preferences has already been made by Kinder and Sanders (1996) and by Sniderman and Piazza (1993). These two latter authors argue that there is no single race issue, but that rather it can be split into three broad categories: the race conscious, social welfare and equal treatment agendas. These categories have been used in subsequent analyses and have been operationalised through indicators of support for positive discrimination towards blacks (race conscious), for public expenditure to improve their social status (welfare state) and for equality in treatment of blacks at the workplace or at school (equal treatment) (Hetherington and Globetti 2002 :271). Nevertheless, I will not accept a priori the validity of this classification. As explained below, I will instead try to assess empirically the actual multidimensional structure of the public perception of immigration in the European public. Moreover, I will not directly adopt this three-fold classification for four reasons: 1- it relies exclusively on the notion of race, whereas immigrants crossing European borders do not necessarily belong to a different race from the native population 4 ; 2- the focus of the classification is based on the particular North American history of inter-ethnic relations 4 It is important to bear in mind at this point the distinction between xenophobia and racism. The former refers to the hostile prejudices or attitudes towards foreigners on the basis of their different identity, whereas the latter is an ideological conception which presumes and justifies a condition of superiority (physical, cultural, economic, etc.) in a given racial or ethnic group. Moreover, racism implies the consideration of race as the ultimate explanatory variable determining people s intelligence and morality, and xenophobia does not. The type of attitudes analysed in the present research corresponds more to the definition of xenophobia than of racism. For the distinction of both concepts see (29/08/07).

16 15 (segregation, value of individualism and social expenditure in perceiving blacks, etc.), which might not be directly applicable to the European context; 3- it only focuses on the sympathy towards policies and not on other perceived impacts of immigration (such as crime, culture, economy or identity); 4- the data available now and that I will use include far more questions about ethnic perceptions which could lead to new attitudinal dimensions not observed before. The need to disentangle distinct dimensions of ethnic perceptions is also justified by works framed in more theoretical paradigms of migration studies. According to some scholars linked to this kind of perspectives, the definitions of identity and ethnicity are constructed in the public sphere through overlapping polarised categories such as universalism/relativism or homogeneity/heterogeneity. These categories, in addition, vary across space and time (Yuval-Davis 2001 :59-63). Furthermore, the contemporary idea of ethnicity emerges from a complex set of social and cultural interactions between people who are tied to different constructs of collectivity and citizenship (Bauböck 2001). It is indeed not rare to see in the literature about ethnic perceptions some apparently unexplained contradictions about the effect of some explanatory variables. The most prominent examples are the effect of age, gender and even perceptions regarding the economy in restrictive attitudes against foreigners (Coenders 2004 :111 for age) (Burns and Gimpel 2000 :210 for age and gender) (Hoskin 1991 for economy; Citrin, Green et al. 1997; Hayes and Dowds 2006 :473 ). One of the most plausible explanations for these apparent contradictions in the previous literature is the specification of the dependent variable. Some authors, for instance, may explain an attitude regarding the presence of migrants, whereas others may explain an attitude regarding the convenience of granting legal rights to foreigners and others the perceptions of threat against the economy. The use of one or another dependent variable is not always theoretically and empirically justified, and may lead to these apparent contradictions simply because the semantic construction of immigration is different in every attitude. Therefore, different social profiles or attitudes may articulate a different tendency towards being anti-immigration, depending on the type of attitude analysed. Opinions regarding the presence of migrants (Citrin, Reingold et al. 1990; Kinder 2003) and regarding whether they should be granted civic rights (Gijsberts 2004) are two of the most common types of attitudes analysed. The so-called perceived ethnic threat is another set of

17 16 relevant attitudes, which are usually understood as a mediating variable between social conditions and other types of perceptions of migrants. In the American context there has also been a debate about the proper conceptualisation and specification of ethnic perceptions. More specifically, the controversy has divided the scholars between those who used measures of overt prejudice (agreement with racial stereotype questions that portray blacks as inherently inferior to whites) and those who work with the concept of new racism (also called symbolic racism, modern racism or racial resentment, which canalize racism through core values like individualism. i.e. the belief that blacks are undeserving of any form of special government assistance) (Feldman and Huddy 2005 :169 Hetherington, 2002 #199 : ). In sum, there is no clear consensus about whether there is a correct classification with a strong empirical basis of the types of attitudes towards immigrants which really do exist in public opinion. Some of the typologies, for instance, mix rather haphazardly attitudes (behavioural intentions) with actual behaviour such as far right wing voting (Coenders 2004), when the distinction should be clearer (Oliver and Mendelberg 2000 :574; Hjerm 2001 :43; Verberk, Scheepers et al :206). As it will be explained in more detail in the following methodological subsection and in the appendix, I have combined dimensional analyses both based on correlation logic among items in the dataset I use and based on an ordinal logic inherent to the construction of Guttman scales. The final result leads to the specification of three scales assessing different aspects on how people construct attitudes towards migrants. These scales will be the three dependent variables to be taken into account in the analyses, and are the result of summing up different items which according to the dimensional analyses appear as constituting single and distinct indexes. The scales refer respectively to the perceived impact of migrants in the resources of a given society (in economic, cultural and legal terms), the adaptability of migrants in the host society from an utilitarian point of view, and the volume or presence of migrants that the respondent cap accept in his/her country. It is important to stress the relative novelty of the second type of attitude in comparison with the dependent variables used in previous research. Apart from perceiving migrants on the basis of their impact in the competition for scarce resources and their volume in the host society, people seem to also articulate their ethnic attitude according to the easiness of migrants to be adapted (religious background, language) and the utility that can be obtained by letting them cross the national borders (working skills, level of education, wealth).

18 17 Table 1- Three scales of ethnic perceptions with their respective items RESOURCES SCALE ADAPTABILITY SCALE PRESENCE SCALE 1-How much migrants put in or take out in terms of taxes and services 2-Migrants take away or create new jobs 3-Migrants make the country a better or worse place to live 4-Migrants are good or bad for country s economy 5-Good or bad thing to promote law against ethnic hatred 6-Good or bad thing to promote law against discrimination in workplace 7-Migrants enrich or undermine cultural life 1- Importance of good working skills of migrants to be accepted in country 2- Importance of talking official language for migrants to be accepted 3- Importance of having a good level of education for migrants to be accepted 4- Importance of being Christian for migrants to be accepted 5- Importance of being wealthy for migrants to be accepted 1- Allow few or many immigrants from poorer countries in Europe 2- Allow few or many immigrants from poorer countries outside Europe 3- Allow few or many immigrants from different ethnic group as majority The sum of the items in each respective dimension generates a 70-point scale in the case of the resources dimension, a 50-point scale in the case of the adaptability attitude, and a 10- point scale in the presence dimension. The three indexes range from a positive to a negative perception of immigrants. - DATA AND METHOD The main database to be used in the empirical analyses is the first wave of the European Social Survey 5, conducted during 2002 and 2003 in 22 different countries (N=42,359). To my knowledge, this database is the most appropriate to carry out the research I propose here and one of the only ones containing all the indicators I need. The different waves of the ESS are conducted every two years. In each wave there is an extended module of questions 5 (29/08/07). The link to the technical details of the dataset such as the weights needed are in this site.

19 18 specific to one particular issue. This module is not repeated in the successive waves, as its objective is to obtain in-depth answers on one particular issue, and thus permitting a large scale comparison of high quality indicators. These indicators, then, assess a broader range of opinions regarding the issue and reveal more complex aspects of its construction across the European public opinion. In the first wave, this issue was immigration. The more than 40 questions existing about it represent an invaluable opportunity to observe more precisely what attitudes towards migrants exist. This broad set of items regarding immigration is much bigger than the majority of other surveys about the topic and is exclusively focused on the European context. At some point I will also use the database of the following wave of the ESS, which was conducted in The amount of items regarding attitudes towards immigration is unfortunately much more reduced there. Instead of testing again the three scales of attitudes towards immigration that I will consider as dependent variables in the analyses, I will only be able to replicate the findings in two of them. The advantage of combining the analysis with the second wave of the ESS, however, is that the number of post-communist countries included is higher, and therefore much more appropriate to test the cross-level interaction hypothesised in H3b. The methodological strategy has followed a two-step procedure. First, I have worked with the more of 40 items assessing people s opinion towards migrants and contained in the ESS 2002 to disentangle different conceptual dimensions underlying these indicators and to be able to properly specify the dependent variables of the models. Second, I have used the three resulting scales as dependent variables in hierarchical linear models to assess the validity of the hypotheses sketched out above. As for the first step, I have combined the correlation logic inherent to exploratory factor and the ordinal logic of Mokken scale analyses (Van Schuur 2003) to obtain the three final dependent variables which have been summarised in the previous section 6. Regarding the second step, once these different clusters of attitudes have been identified and validated, they have been used as dependent continuous variables in regression models. The type of regression used is a multilevel linear model. There are three reasons for using a hierarchical approach instead of a standard OLS model. First, as has already been acknowledged in previous works, the analyses of attitudes towards immigration has been mainly analysed 6 See appendix for the detailed steps followed in the dimensional analysis.

20 19 either from an individual level or from an aggregate point of view. It might instead be more useful to integrate both levels of analyses and thus to deal better with the problems derived from the ecological (Robinson 1950; Seligson 2002) and the individual fallacies (Lijphart 1980 :45). Second, the validation of the third hypothesis suggested above implies the use of cross-level interactions (random-coefficient models which let the slope of an individual variable vary across different geographical contexts), and this is only possible in such models. Third, when analysing attitudes across different contextual units a multilevel regression is a better choice for technical reasons such as avoiding the truncation of the variance 7 and the calculation of biased standard errors and inflated Type I errors (Snijders and Bosker 1999; Steenbergen and Jones 2002). The downside of the use of cross-level interactions, however, is that they may generate instability in the model due to multicollinearity between the independent variables (Aiken and West 1991; Van der Brug, Van der Eijk et al :79). The problem of multicollinearity when introducing interaction terms is controversial from a statistical point of view. Brambor, Clark and Golder argue that the interaction and its constitutive terms must always be included in the model no matter the apparent subsequent high correlation between them. They argue that if this is not done, the calculation of the intercepts in every regression line assessing the impact of left-right in every conditional circumstance will be dramatically biased (Brambor, Clark et al :67,73). My strategy will be to definitely include the constitutive terms together with the interaction in the same model, but to test the individual level interactions (in random-intercept models) and the cross-level interactions (in random-slope models) in separate regressions to avoid increasing unnecessarily the level of multicollinearity among independent variables. As for the indicators to be added in the models as independent variables, the most relevant one is the individual self-placement on an 11-point scale from the most leftist (value 0) to the most rightist position (value 10) 8. The inclusion of this variable will permit us to test H1. On the other hand, H2 will be tested through the inclusion of the following interactions together with their constitutive terms: left-right*opinion towards government intervention in economy (H2a), left-right*opinion towards homosexual couples (H2b) and leftright*education (H2c). Finally, H3 will be tested through the inclusion of two more 7 That is, to apparently find contextual effects that in reality are accounted for individual differences of the populations studied in distinct countries. 8 See appendix for the descriptive statistics of all the variables included in the models.