«Values and the European Parliament Vote 2009: Voting Left-right or Europe? A Multi-level Analisys» André Freire - Professor at ISCTE-IUL (Lisbon University Institute), and Senior Researcher at CIES-IUL (Centre for Research and Sociological Studies) Paper to be presented at the IPSA XXII World Congress of Political Science (7-12 Jully 2012), Madrid, RC 17 (Comparative Public Opinion), Panel Comparative studies of social structure, value orientations and party choice. Focus on values, 10 July 15:00 16:45 in School of Journalism, room C007, Chair Oddbjørn Knutsen (oddbjorn.knutsen@stv.uio.no), Discussant Mauro Barisione (mauro.barisione@unimi.it), University of Millan. Draft version: please do not quote without author s permission.
Abstract Research about the impact of values upon the vote in EP elections is rare. True, there are several papers showing an impact of left-right selfplacement on both eurosceptic attitudes and behaviour: an inverted U curve has been said to best describe the relationship between left-right ideology and both attitudes and behaviour toward european integration. But almost no research has been done about the relation between substantive issue positions/value orientations and the vote in EP elections. By including several indicators of value orientations/issue positions the 2009 EES (European Election Study PIREDEU) allow us to test the impact of value orientations upon the vote in EP elections as measured either by the vote in parties ordered in a pro-anti (further) european integration scale and, due to the second-order nature of EP elections, alternatively also in a left-right scale. By using the Hierachical Linear Modelling (HLM), we will try to explain variation both across individuals and countries (27).
Introduction Research about the impact of values upon the vote in EP elections is rare. True, there are several papers showing an impact of left-right selfplacement (and EU attiudes) on both eurosceptic attitudes and behaviour: an inverted U curve has been said to best describe the relationship between leftright ideology and both attitudes and behaviour toward european integration (van der Brug et al, 2007; van der Brug et al, 2008; van der Eijk and Franklin, 2004 and 2007; van der Eijk et al, 1996; De Vries et al, 2011). But almost no research has been done about the relation between substantive issue positions/value orientations and the vote in EP elections (except in what concerns attiudes towards immigration and attitudes in terms of pro vs. anti further EU integration). By including several indicators of value orientations/issue positions the 2009 EES (European Election Study PIREDEU) allow us to test the impact of value orientations upon the vote in EP elections as measured either by the vote in parties ordered in a pro-anti (further) European Union (EU) integration scale and, due to the second-order nature of EP elections, alternatively also in a leftright scale. By using the Hierachical Linear Modelling (HLM), we will try to explain variation both across individuals and countries (27). Thus, our major purpose in the present paper is to know not only what is the impact of values in the vote for EP elections but also to understand what are the (cross country) contextual conditions that may explain the increase or the decrease in the impact of values upon the vote in EP elections. Additionally, in terms of the impact of values on the EP vote, we also want to know what is the more relevant
dimension of competition, competition along the left-right divide or competition in terms of pro vs. anti (further) EU integration. And why is it relevant to study the impact of values/issue positions on the EP vote? Here we must bear in mind that elections are about meaningful choices concerning the type of society we want to live in. Additionally, elections are about choices concerning the major policy options to reach the diferent desirable types of society we want ot live in. Thus, first of all, it is important to study this topic (namely the contextual conditions that may explain the increase or the decrease in the impact of values upon the vote in EP elections ) because it may help us understand what are the factors that might increase the quality of political representation are at the EU level: the more the vote is anchored in policy preferences the higher the quality of political representation (Pierce 1999: 9). Therefore, considering these values can be said to be the underlying constructs behind issue packages, the level of value voting can be said to be a necessary precondition for the quality of political representation at the EU level. Some authors do use left-right seflplacement as a proxy measure of policy preferences: policy distances between the respondent on the one hand, and the respective party in the stacked data matrix were computed for two policy scales: left-right and the issue of Europena integration (van der Brug et al, 2007: p. 175). However, LR sefl-placement if far from adequate as a measure of substantive policy preferences, for several reasons. LR is dependent not only on issue positions but also and often more strongly on partisan orientations and social factors (Inglehart and Klingemann, 1976; Knutsen, 1995 and 1997; Freire, 2006b and 2008a). Moreover, in some countries (Austria, Estonia, Latvia, Ireland, Portugal,
etc.), namely the less party polarized ones and also in new democratic regimes, LR sefl-placement is barely correlated with issue positions and underlynig value orientations (Freire, 2008a; Freire and Kivistik, 2012). Thus, to proper analyze the topic we really need to check this using direct indicators of substantive policy preferences. Second, acordihg to Hooghe and Marks at least until the 1980s some schollars of public opinion and elections in the EU polity have not taken public opinion seriously (non-attitudes, non-structured public opinion concerning EU issues, voting in the EU as a mere reflection of domestic politics, etc.) (see Hooghe and Marks, 2008, for a review), but if we find that voting in EP elections is indeed anchored in policy preferences (as measured by the underlying value orientations) than we can conclude that the conflict over the EU is anchored in major value conflicts and thus there is a structured public opinion concerning EU issues. Third, several authors have argued the impact of values on political attitudes and behaviour is increasing in post-modern societies (Inglehart, 1997; Knutsen and Scarbourgh, 1995; Thomassen, 2005); therefore, it is important to understand in which contextual conditions this is more likely to happen, especially at the EU level. Fourth, this is an understudied topic both at the national (Knutsen and Kumlin, 2005; Freire, 2009) and, above all, at the EU level (especially in what concern the impact of values on the vote at the mass level) (see the next section for a literature review, and corresponding references). Thus, the present study can contribute to fill this gap in the literature.
To reach our objectives in the present paper, we will proceed in the following way. In the next section, we will review the literature about values / issue positions and both EU support and voting to the EP, and we will also especify our micro and macro level hypotheses. In the third section, we will present the data and methods used in the paper. In the fourth section, we will test both our micro and macro hypotheses about the impact of values on the EP vote when the dimension of competition considered is the LR divide. In the fifth section, we will test both our micro and macro hypotheses about the impact of values on the EP vote when the dimension of competition considered is the divide over pro vs. anti (further) EU integration/unification. The paper ends with some concluding remarks. Theory and hypotheses Since we want to investigate the impact of values on the EP vote, the first thing we need to agree upon is on a definition of values. Here we follow van Deth s and Scarbrough s definition of values as not directly observable conceptions of the desirable world (Deth and Scarbrough 1995: 21 47). Values are usually said to be long-term predispositions with both direct and indirect influence on the vote. In terms of the funnel of causality metaphor, they are located near the base of the funnel, while (position) issues are usually considered to be short-term factors. However, as Thomassen points out when referring to position issues, they often are a reflection of the same domains of conflict as the value dimensions just discussed (Thomassen 2005: 17). Bearing in
mind all this, we will use Principal Component Analysis (PCA) to reduce the 10 position issue items (about immigrantion, libertarian-authoritarian topics, and socioeconomic topics) present in EES 2009 mass survey into their underlying value factors measuring orientations related either with libertarian-authoritarian values (Flanagan, 1987; Flanagan and Lee, 2003), that overlap at partly with post materialism values (Flanagan, 1987), or with socioeconomic (LR) values, sometimes all called left-right materialism (Knutsen, 1995 and 1997; Freire, 2006b, 2008a). To measure the impact of values on the EP vote 2009 we will control also for indicators of social positions (social class subjective social class, and education -, religion - church attendance -, and habitat urban vs. rural residence) which are location before values in the base of the funnel of causality. The reader should also bear in mind that we don t want to test the full model of the EU vote, we just want to see test the impact of values on the vote in European elections and see if they are more (or less) important when the dimension of party competition if the LR divide or when it is the pro vs. anti (further) EU integration, and what are the contextual factors that explain variantions across countries/political systems. As we said before, research about the impact of values / issue atitudes on EU support and voting for the European Parliament is rare, especially in the latter case. For exemple, Inglehart (1970), and Inglehart and Reif (1991), found that cognitive mobilization and post materialist values are positively correlated with support for European integration. However, these results were rather mixed across time, namely because it was found that the positive impact of cognitive mobilization and post materialist values on EU support was declining with the passage of time. Moreover, questioning these findings, Janssen (1991) argued
that the relation between post materialism and EU support is a spurious relationship, i.e., it vanishes when the relevant controls are introduced. Additionally, it has been argued that GAL (Green, Alternative, Libertarian) - TAN (Traditional, Authoritarian, Nationalist) party divide, which is clearly associated with the authoritarian-libertarian value conflict, has a relevant impact on EU support, but the results focused only on the party level: TAN parties tend to be more euroceptic; GAL parties tend to be more supportive of the EU (several papers in Marks and Steenbergen, 2004; see also Hooghe and Marks, 2008, for a review). The TAN stance is due mainly to the cultural (and also economic) threat represented by non-nationals in each country, which are allegedly favoured by the EU trend of openness of frontiers, as well as to challenge to the nation-state authority/soverignty represented by the EU unification. Moreover, the EU is said to be friendly to the rights of minorities (women, immigrants, gays, etc.), and thus is fought by those opposing the liberalization of life styles. On the contrarty, GAL parties are said be more prone to support (further) EU integration because they are more cosmopoliton in outlook, are more prone to support minorities rights, and are also less supportive of traditional forms of authority (associated with the Nation state or otherwise). At the mass level, however, studies about the impact of these types of values on EU support are rather rare and, first, do not concern the EP vote but attitudes towards the EU (i.e., EU support), and, second, concern only EU 15 (Ray, 2001; see also Leconte, 2008, for a review, and in 2010, pp. 161-191, for some specific findings). Moreover, it was found that attitudes towards immigrants do have a major impact on both attitudes towards the EU and EU voting, due to the cultural threat vs. interests divide (McLaren, 2002-2007).
Religion and moral issues have also been shown to be relevant to mass attitudes towards the EU, at least indirectly through the authoritarian- libertarian value orientations, or through attiudes towards immigration (Boomgarden and Freire, 2009). However, overall the effects of religion on EU support tend to cancel each other out: on the one hand, religion is positively associated with authoritarian values and thus to less EU support mediated by the authoritarianlibertarian divide; on the other hand, religion is positively associated with more tolerance vis-à-vis immigrants and thus to show more supportive EU attitudes mediated through positive atitudes towards immigrants (Boomgarden and Freire, 2009). There is also some controversy over what is the major dimension of value/ideological conflict across the EU: either neoliberalism (market deregulation) vs. regulated capitalism, dividing right and left, or GAL-TAN. Again, existant studies have focused mainly on the party/system level and studies at the mass level are pretty rare or non-existant (see several papers in Marks and Steenbergen, 2004; see also Hooghe and Marks, 2008, for a review). Additionally, a debate is also going on about the degree of overlap of the LR dimension of competion and the pro vs. anti (further) EU integration dimension (Hix, Noury, and Roland, 2007: 161-181), even if both studies at the mass level (van der Eik and Franklin, 2004 and 2008), and the level of MEP s voting in the EP Parliament (Hix, Noury, and Roland, 2007: 161-181), especially the former, do show that the overlap is only partial and highly variable across countries. Moreover, studies about the impact of the traditional (LR) socioeconomic values (i.e., those concerning inequalities, atitudes towards the market and the state, state intervention in the economy, social protection, the
role of the unions, etc.) on both mass EU support and citizens voting to the EP are even more scarce (at least partly, because the indicators measuring socieconomic values have not been included in EES studies until the EES 2009: one of its major innovations) and have been concentrated exclusively at the party system/meps levels (see Hooghe and Marks, 2008, and Leconte, 2010, for reviews; Hix, Noury, and Roland, 2007, pp. 161-181). Of course, as we said before Ideological left-right self-placement is often taken as proxy for socioeconomic values (Ray, 2001; van der Brug et al, 2007; van der Brug et al, 2008; ) but, as we said before, left-right is a multi-dimensional construct (Inglehart and Klingemann, 1976; Knutsen, 1995-1997; Freire, 2006b-2008a) and, more importantly, in many countries (like Austria, Estonia, Latvia, Ireland, Portugal, etc.) it is barely correlated with vaues/issue positions (Knutsen, 1995-1997; Freire, 2006b-2008a; Freire and Kivistik, 2012). Thus, from our standpoint, LR self-placement is not in many countries a good proxy for substantive policy preferences, and that is a major reason why we really need direct mesures of substantive issue preferences and their underlying value orientations. This can be done by using the EES 2009 because, as we said before, it contains ten items on substantive issue preference that uncover major value conflicts in modern industrialized societies, namely value divides associated with both old (LR socioeconomic issues) and new politics (issues related with the authoritarianlibertarian divide). Thus, this EES 2009 data allow us to overcome the gap in the literature concerning the impact of values/substantive issue preferences on the EP vote. Concerning the relationship between leftwing socioeconomic values (i.e., those sets of attiudes favouring strong measures to fight inequalities,
sckeptical towards market liberalization and that favour more state intervention in the economy, more social protection, and stronger unions, etc.) and support for furhter EU integration, it has been argued that «in the 1960s and 1970s, when European integration primarly related to the liberalisation of trade between member states, parties on the left tended to be anti-european where as parties on the right tended to be pro-european. Then, the relationship reversed in the 1990s. Once the single market was created, the left become more supportive of furhter political centralization than the right, in order to enable to state to regain control of the market (Hix, Noury, and Roland, 2007, p. 163).» However, we can also conceive the reverse situation: since the Maastricht Treaty and the strong neoliberal tendancies that the EU has known, we can also expect that individuals (and parties) which share more pro state/leftwing socioeconomic values are increasingly more euro skeptical about the state of EU affairs, and fear that further unification can bring more liberalization not less. Additionally, it has been shown that the patterns of the relationships of between socioeconomic values and EU support/suppport for further EU integration might be signficantly variable across countries (see Ray, 2004; Hooghe, Marks and Wilson, 2004). Namely, data for the EU 15 showed that the left tend to be more pro European in countries with less developed welfare states (in the liberal UK and Ireland, and in Southern Europe), expecting that EU integration might bring some upgrade on their respective low levels of social protection; on the contrary, in countries with more developed welfare states (maxime in Scandinavia) the left tend to be more euro sckeptical because fears some dowgrading of their respective high levels of social protection if at least some social unformization at the EU level is to take place (Ray, 2004).
Considering alll these debates about the impact of values on the EP vote, either when the dimension of competition in the LR divide or when it the pro vs. anti EU integration divide, it is now time to present our hypotheses. I - Micro-level Hypotheses: H1.1 - Individuals that are more tolerant vis-à-vis immigrants and/or that are more prone to show more libertarian atitudes (Factor 1: Libertarian and Group Tolerant Values) are also more prone to vote in lefttwing parties (Knutsen, 1995 and 1997; Huber, 1989; Flanagan, 1987; Flanagan and Lee, 2003; Freire, 2006b and 2008a). H1.2 Individuals that are more tolerant vis-à-vis immigrants and/or that are more prone to show more libertarian atitudes (Factor 1: Libertarian and Group Tolerant Values) are also more prone to vote in parties that favour further European unification (Ray, 2001; Mclaren, 2002-2007; GAL-TAN theories: Hooghe and Marks, 2008; Marks, 2004). H2.1 - Individuals that share libertarian values in terms of life styles/are more liberal in terms of life styles like gay marriage, abortion liberalization, etc., (Factor 2: Libertarian Life-Styles-, are also more prone to vote in lefttwing parties (Flanagan, 1987; Flangan and Lee, 2003; Knutsen, 1995 and 1997; Knutsen and Scarbrough, 1995) H2.2 - Individuals that share libertarian values in terms of life styles/are more liberal in terms of life styles like gay marriage, abortion liberalization, etc., (Factor 2: Libertarian Life-Styles), are also more prone to vote in favour further of
parties that favour further European unification (Ray, 2001; Leconte, 2008 and 2010; GAL-TAN theories: Hooghe and Marks, 2008; Marks, 2004). H3.1 - Individuals that are more oriented towards fighting inequalities and favouring state intervention/regulation in the economy and society (Factor 3: Socioeconomic Values) are also more prone to vote in lefttwing parties at the EU level (Huber, 1989; Knutsen, 1995-1997; Knutsen and Scarbrough, 1995; Freire, 2006b and 2008a). H3.2 Some authors argue that, since the 1990s, Individuals that are more oriented towards fighting inequalities and favouring state intervention and/or stronger regulation in the economy and society (Factor 3: Socioeconomic Values) are also more prone to vote in parties that favour further European unification (Hix, Noury, and Roland, 2007). However, as we said before, the reverse situation is also conceivable: strong neoliberal tendancies at the EU level are in place at least since the Maastricht Treaty, and that might lead to a loss of EU support from tindividuals (and parties) which share more pro state/leftwing socioeconomic values. Additonally, some authors (Ray, 2004; Hooghe, Marks and Wilson, 2004) argue that the relathionship between socioeconomic values and EU support / support for further EU unification can be highly variable across countries, in fact working in different directions in different types of contexts. So, we have no clear and one-size-fitsall predictions for all the countries in this respect, expecting instead mixed results across countries.
II - Macro-level Hypotheses: H4 (Clarity of Policy Alternatives) The higher the clarity of policy alternatives (measured by the polarization index see below) presented by parties to the voters, the stronger the impact of values on the vote in EP elections structured either across the the Left-Right dimension of competition or across the European Integration dimension of competition. The rationale behind this hypothesis is the following: the greater the ideological polarization between the parties, the greater the clarity of (LR or EU Integration) policy alternatives they present to the voters, and, thus, the more easily and frequently the citizens will use their value orientations to guide them in their voting choices (van der Eijk et al, 2005; Knutsen and Kumlin, 2005; Freire, 2008b; Berglund et al, 2005; Wessels and Schmitt, 2008; Lachat, 2008; Schmitt, 2009; Freire, Lobo, and Magalhães, 2009; Schmitt and Freire, 2012). H5 (level fo party System Institutionalization/years of democracy) The higher the level of party system institutionalization (as measured by the number of years of the demcratic regime since 1920s) the stronger the impact of values on the vote in EP elections structured in terms of the LR dimension of competition. The rationale behind this hypothesis is the following: we expect that in the newer democratic regimes the impact of values on party choice will tend to be weaker than in the more consolidated democratic regimes because new democratic regimes are less likely to exhibit high levels of party system institutionalization, and where these elements are absent, party LR ideological placements and electoral choices tend to exhibit high levels of
instability and fluidity (see Schmitt and Freire, 2012; Freire, 2009; Tworzecki, 2002; Rose and Munro, 2003) H6 (Net contributor to the EU) - The higher the level of net contribution of a country to the EU budget the weaker the impact of values on the vote in EP elections, when the dimension of competiton is the pro vs. anti (further) European Integration, because economic interests (pay less for the EU) will weight more than values to explain the vote, thus depressing the overall level of EU suppport across the board (Maclaren, 2002-2007). H7 (Social Expenditure) - The higher the level of a country s social expenses the less individuals with pro state intervention values (Factor 3: Socioeconomic Values) will be tend to vote in parties favouring (further) European unification. The rationale behing this hyphothesis is the one already presented above: the left will tend to be more pro European in countries with less developed welfare states, expecting that EU integration might bring some upgrade on their respective low levels of social protection; on the contrary, in countries with more developed welfare states (socioeconomic) left voters will tend to be more euro sckeptical because they fear some dowgrading of their respective high levels of social protection if at least some social unformization at the EU level is to take place (Ray, 2004; Brinegar et al, 2004). H8 - (Immigration) The higher a country s percentage of immigrants in the resident population, the stronger the impact of attitudes vis-à-vis immigrants (Factor 1: Libertarian and Group Tolerant Values) in the vote in EP elections (structured either across the European Integration dimension of competition or across the Left-Right dimension of competition). The rationale behind this hypothesis is that a strong presence of immigrants (especially if coupled with
high levels of unemployment) can boost xenphobic attitudes and, thus, increase their impact on the vote (Mclaren, 2001-2007). H9 - (Secularization) The higher the level of secularization in a country, as measured by the percentage of persons that declared never go to church, the stronger the impact of authoritarian vs libertarian values (Factor 2: Libertarian Life-Styles) upon the vote in EP elections (structured either across the European Integration dimension of competition or across the Left-Right dimension of competition). The rationale behind this hypothesis is the following: authoritarian vs libertarian values/ the new politics divide is typical of advanced post industrial societies, also usually more secularized (Flanagan, 1987; Flangan and Lee, 2003). Data and Methods To estimate the impact of values on the 2009 EP vote, we use data from the 2009 European Election Study (PIREDEU). Data refers to representative samples of adult populations in 27 EU countries.
Table 1.1: List of dependent and independent variables at the individual level Variables Categories Dependent variable Party choice_eu: party voted for in the last 2009 European election ordered in a antipro European unification scale (q25receu) Ordered from 0 (European unification has already gone too far) to 10 (European unification should be pushed further) according to the average position of each party following citizens perceptions (EES 2009 mass survey): only citizens with secondary education or more were Party choice_lr: party voted for in the the last 2009 European election ordered in a Left-Right scale (q25reverse) Control variables (independent) Social Class 4Q114r Social class (subjective) Q200r Education (cross country comp.) Religion Q118 attendance religious services Habitat Q115 rural-urban area Independent variables: Values Factor 1: Libertarian & Tolerant (Group) Factor 2: Libertarian Life-style Orientations Factor 3: Socioeconomic values: pro state (vs. pro market) considered Ordered from 0 (Right) to 10 (Left) according to average position of each party following citizens perceptions (EES 2009 mass survey): only citizens with secondary education or more were considered 1 Upper class; 5 Working class 1 Level 6; 17 Level 0 1 Several times a week; 6 Never 1 Rural area or Village; 4 Large town or City See Table 1.2 below See Table 1.2 below See Table 1.2 below
Table 1.2: List of issues and value orientations extracted through PCA Value orientations Issue items Factor 1: Libertarian & Tolerant (Group) Factor 2: Libertarian Life-style Orientations Immigration issues Q56 - Immigrants required to adapt to customs of [country] Q67 - Immigration to [country] should be decreased significantly Socio-economic issues Q57 - Private enterprise best to solve [country's] economic problems Q59r - public services and industries should be in state ownership Q63r - Income and wealth should be redistributed towards ordinary people 0.689 0.647 Factor 3: Socioeconomic values: pro state (vs. pro market) 0.536 0.727 0.622 Libertarian and authoritarian issues Q58 - Same-sex marriages 0.614 should be prohibited by law Q60r - Women should be free to decide on matters of 0.743 abortion Q62 - People who break law 0.697 should get much harsher sentences than now Q64 - Schools must teach children to obey authority 0.630 Q66 - A woman should cut 0.636 down on paid work for her family Variance explained 21.968 14.733 13.091 Cumulative var. explained 21.968 36.701 49.793 Notes: 1) The answers were all originally coded in an ordinal scale: from strongly agree (1) to strongly disagree (5); when there is the mentioned r is present after the variable name, it means that the coding was reversed. Thus, in each variable lower values indicates either intolerant attitudes towards immigration (and to alternative life-styles) or pro market attitudes or authoritarian attitudes. For higher values it is the other way around: either tolerant attitudes towards immigration (and alternative life styles) or pro state attitudes or libertarian attitudes 2) KMO = 0.757; Significance = 0.00; Varimax rotation; Scores >+ 0.4 and sorted by size.
Table 1.3: List of independent variables at the macro-level Independent and control variables Categories For Party Choice LR Left-Right polarization index (PI_Left-Right) Minimum: 0,146; Maximum: 0,528. YearsDemocracy Years of democratic regime, 1920-2009, in each country (polity >= 7) Anti-pro European unification polarization index (PI_European) Minimum: 19; Maximum: 90. For Party Choice EU Minimum: 0,037; Maximum: 0,352. Netcontributor_ReceiverEU - Net contributor (or receiver) to EU Budget 2009 (in per head: negative indicates ctry net receiver) Minimum: -2364,00; Maximum: 211,00. SocialExpenditure - Social Expenditure in each country - Euro per inhabitant Eurostat 2007 Percent_Immigrants - Percentage nonnationals in resident Population by country, Eurostat 2009 (BE & UK: 2008) Minimum: 568,30; Maximum: 15081,30. For Party Choice LR & EU Minimum: 0,1; Maximum: 43,5. Secularization - Percentage of respondents (EES 2009) that declared (q118) that "never" go to church Minimum: 4,70; Maximum: 61,10. By using the PCA (Principal Component Analysis) technique, first we reduce the 10 issue items about immigrantion, libertarian-authoritarian topics, and socioeconomic themes into three factors measuring the following underlying value orientations: Libertarian Values & Tolerance Towards Immigrants (only the positive polarity is titled), Libertarian Values Towards Alternative Life-Styles, and Socioeconomic Values (see Table 1.2). The first two sets can be said to be related / to measure the underlying Libertarian- Authoritarian Value Dimension (Flanagan, 1987, Flanagan and Lee, 2003), and the third one can be said to measure substantive preferences related with «Left-Right Socioeconomic Values», sometimes also called «LR materialism»
(Freire, 2006b, 2008a, 2008b; Knutsen, 1995 and 1997; Huber, 1989; Knutsen and Scarbrough, 1995). In Level 1 (individual level) of the Multilevel Analysis, the three Value Factors are then used as predictors to explain Party Choice in the EP Elections across the 27 EU countries hypotheses about the expected relationships were already presented in the previous section. Besides, indicators of social class, religion, and habitat are used as control variables (for Level 1 in HLM) (see Tables 1.1 and 1.2). Note that Party Choice, the dependent variable, is ordered either in a Left-Right or in a Pro-Anti (further) EU integration scale we used information about voters perceptions of parties locations in each scale to transform vote choice in that sense (the relevant questions were asked in the EES 2009 mass survey to the citizens in each of the 27 EU countries, i.e., the questions about where the voters locate the parties in a (reversed) LR scale, 0 (Right) -10 (Left)), and in Pro (10) - Anti (0) (further) EU integration scale) (see Table 1.3). Although, on the one hand, this procedure is used in the comparative political science literature (see for example Norris, 2004; Freire, 2009), namely because it allow us to use an elegant and easy to use way to compare the vote across different countries, on the other hand some people might argue that we are imposing two specific dimensions of competition to the EP vote (LR and the pro vs. anti further - EU Integration) but the vote is a nominal discrete variable in each country and diferent dimensions can exist in the diferent countries, besides the two above mentioned, even in same country. Although we agree that is argument is somehow pertinent, we nevertheless believe that our option is adequate to our research for several reasons. First, because we want to test and compare the impact of values on the EP vote when the
dimension of competition is LR and when the dimension of competition is the level of EU integration to be pursued, namely to see see which one of the two dimensions is more structured by substantive value preferences. Second, because the literature about dimensions of conflict at the EU level has shown that these are the two more important dimensions of political conflict either in terms of mass EU attitudes and voting (Ray, 2001; van der Eijik and Franklin, 2004 and 2008; van der Brug et al, 2008; De Vries et al, 2011), in terms parties ideological locations, and in termos MEPs voting behaviour and atitudes (Hix, Noury and Roland, 2007; Hooghe, Marks, and Wilson, 2004; Marks, 2004). Third, because alternative ways of treating the dependent variable, namely the PTV (Probability To Vote) variables are also far from adequate considering our objectives and data: 1) since in the stacked data matrix, using the PTVs, multiple observations refer to the same respondent, these observations are not independent of each other (van der Brug et al, 2007, p. 184) ; 2) because when no appropriate measure of distance or similarity can be straightforwardly computed, for instance because relevant party characteristic are unmeasured (as is the case here because we don t have strict and comparable measures of parties policy orientations), then the independent variables have to be constructed by means of specific transformations of the original independent variables and this solutiin prevents differences among parties in the average level of support from being incoporated in the newly created independent variable (van der Brug et al, 2007, pp. 175 and 184) ; 3) because using conditional logit regression techniques (which treat the dependent variable as a nominal discrete one) creates not only problems of comparability across countries (in pooled datasets) but also computational difficulties (De Vries et al, 2011, p. 20).
In Level 2 of the Multilevel Analysis (Macro Level), we want to explain cross country variantion in the impact of values on the EP vote. Several macro level independent variables are used to that purpose: see Table 1.3. Namely, when considering the EP vote ordered in a Left-Right dimension, we will use macro level explanatory factors such as the level of LR party system polarization 1, the age of the democratic regime, as well as the percentage of immigrants in the country population (only to explain the impact of F1), and the 1 The left-right polarization index, which was originally designed by Hermann Schmitt (see Schmitt and Freire, 2012, forthcoming), is as follows (for the anti vs. pro (further) European unification the values for LR are replaced by the respective values for the anti vs. pro EU): IP= ( LRm-LRpx *EPpx ) / IPmax Where: IP=ideological polarization LRm=the numerical centre of left-right scale; LRpx=the (average) position of party x on the left-right scale; and EPpx=the proportion of party x of the valid vote in the last election. IPmax=maximum ideological polarization defined by two equally strong parties located at opposite poles of the ideological spectrum. Consider Spain in 2004, for example, where the LRM is 10.5 (on a 0 20 scale); now consider the LRpx for the Spanish socialist party (PSOE Partido Socialista Obrero Español) (8.2). Calculate the difference between LRM and this latter value, which gives 2.3. Then calculate this value and multiply it by the proportion of the vote for each party ( LRm-LRpx *EPpx): 2.3*0.426. Add the result of this calculation to the values calculated for the other parties with parliamentary representation then divide it by IPmax (9.5) to obtain IP: 0.414.
level of secularization in each country (only to explain the impact of F2) (see Table 1.3). Additonally, when considering the EP vote ordered in a Anti-Pro EU integration dimension, we will use macro level explanatory factors such as the level of EU party system polarization, the country net countribution to the EU budget, and the level of social expenditure in a country, as well as the percentage of immigrants in the country population (only to explain the impact of F1), and the level of secularization in each country (only to explain the impact of F1) (see Table 1.3).. Although we will be using the HLM 6.2 software to estimate the Multi-level impact of values on the EP vote, due to the fact the Pseudo R 2 calculated from the HLM data are very crude measures of the variance explained (Snijders and Bosker, 1994), only for that purpose, i.e., for the calcultation of the variance explained at both level 1 and level 2, we rely instead on the two step hierarchical regression (Achen, 2005; Jusko and Shively, 2005). The latter is a especific multi-level estimation technique which is both equally efficient and precise vis-à-vis HLM, and more flexible. In level 1, OLS regressions country by country (and in a pooled data) are estimated. The R 2 for the pooled data set is the variance explained for level 1. In level 2, or step two of the hierarchical regressions, the unstandardized regression coefficients (for values: F1, F2, F3) are now considered as the dependent variables to be explained by the macrolevel explanatory factors. And the variance explained in each case (for F1, F2, and F3), and also averaged for the three sets of value indicators, is the R 2 for level 2. Values and voting to the European Parliament considering the Left-Right dimension of political competition
In the present section we will test the impact of values on the EP vote when the latter is structured in terms of the LR dimension of political competition. From the individual level analysis (OLS regressions country by country, using the two step hierarchical regression), we confirm that with two exceptions (Latvia, for F2 and F3: negative coefficients, and Poland, F3: negative coefficient) our individual level hypotheses (H1.1, H2.1, and H3.1) are confirmed in all the countries where the regressions coefficients are statistically significant (data not shown due to spatial limitations, but that can be furnished upon request). I.e., in each and every country and whenever the relationships are singificant, individuals with more libertarian values (either in terms of group tolerance or in termos of social issues/life-styles) and/or with more pro-state values are also more prone to vote in leftwing parties. The two exceptions are Latvia, where libertarian individuals (in terms of life-styles) and pro-state individuals (socieeconomic values) are more prone to vote for the right, and Poland, where individuals with pro-state attitudes tend to vote more for the right. Overall, in terms of variance explained at level 1, the value for the Adjusted R 2 is 0,136, but there is considerable and significant variance to be explained across countries in terms of the value of the regression coefficients. So now we proceed to level 2. Here, the average Adjusted R 2 calculated through the two step hierarchical regresson is 0,210 (for each value factor the Adjusted R 2 is F1: 0,164; F2: 0,224; F3: 0,242). Thus, country level variation seems to be better explained by our macro level indicators thant individual level variations for all countries taken together in the pooled data set.
The final formal HLM 6.2 model for the EP vote 2009 ordered in a LR dimension of competition reads as follows: Level-1 Model V25REVER ij = β 0j + β 1j *(Q115 ij ) + β 2j *(Q118 ij ) + β 3j *(Q114R ij ) + β 4j *(V200R ij ) + β 5j *(FAC1LIBE ij ) + β 6j *(FAC2LIBE ij ) + β 7j *(FAC3SEV ij ) + r ij Level-2 Model β 0j = γ 00 + γ 01 *(PI_LEFT j ) + γ 02 *(YEARSDEM j ) + γ 03 *(PERCENT j ) + γ 04 *(SECULARI j ) + u 0j β 1j = γ 10 + u 1j β 2j = γ 20 + u 2j β 3j = γ 30 + u 3j β 4j = γ 40 + u 4j β 5j = γ 50 + γ 51 *(PI_LEFT j ) + γ 52 *(YEARSDEM j ) + γ 53 *(PERCENT j ) + u 5j β 6j = γ 60 + γ 61 *(PI_LEFT j ) + γ 62 *(YEARSDEM j ) + γ 63 *(SECULARI j ) + u 6j β 7j = γ 70 + γ 71 *(PI_LEFT j ) + γ 72 *(YEARSDEM j ) + u 7j Table 2 - Final HLM (Hierarchical Linear Modelling) 6.2 estimation of fixed effects (with robust standard errors) with the EP Vote 2009 ordered in a LR dimension of competition Fixed Effect Coefficient Standard Approx. t-ratio error d.f. p-value For INTRCPT1, β0 INTRCPT2, γ00 46.974102 9.033251 5.200 22 <0.001 PI_LEFT, γ01-5.634582 23.891467-0.236 22 0.816 YEARSDEM, γ02 0.023481 0.064740 0.363 22 0.720 PERCENT, γ03-0.085191 0.176301-0.483 22 0.634 SECULARI, γ04-0.169030 0.122687-1.378 22 0.182 For Q115 slope, β1 INTRCPT2, γ10 0.741707 0.666312 1.113 26 0.276 For Q118 slope, β2 INTRCPT2, γ20 4.129281 0.745815 5.537 26 <0.001 For Q114R slope, β3 INTRCPT2, γ30 4.713310 0.907711 5.193 26 <0.001 For V200R slope, β4 INTRCPT2, γ40 0.278432 0.158260 1.759 26 0.090 For FAC1LIBE slope, β5 INTRCPT2, γ50-7.540228 2.675889-2.818 23 0.010 PI_LEFT, γ51 32.634966 9.369310 3.483 23 0.002 YEARSDEM, γ52 0.103754 0.027932 3.714 23 0.001 PERCENT, γ53-0.015793 0.048687-0.324 23 0.749 For FAC2LIBE slope, β6 INTRCPT2, γ60-6.509181 2.479624-2.625 23 0.015 PI_LEFT, γ61 34.049564 6.531613 5.213 23 <0.001 YEARSDEM, γ62 0.073433 0.029207 2.514 23 0.019 SECULARI, γ63-0.109275 0.043871-2.491 23 0.020 For FAC3SEV slope, β7 INTRCPT2, γ70-11.182678 4.788079-2.336 24 0.028 PI_LEFT, γ71 53.673669 14.262579 3.763 24 <0.001 YEARSDEM, γ72 0.093361 0.035649 2.619 24 0.015 Source: data elaborated by the author form the European Eelection Study 2009 / PIREDEU, for the individual level data and also some macro data.
Figure 1: Libertarian & Group Tolerant on LR Vote, cond. LR Polarization 91.23 PI_LEFT = 0.240 PI_LEFT = 0.397 EP Vote 2009 (LR order) 84.18 77.14 70.10 63.05-1.27-0.60 0.07 0.74 1.41 Libertarian and Group Tolerant Values Figure 2: Libertarian & Group Tolerant on LR Vote, cond. Years Democracy 94.09 YEARSDEM = 20 YEARSDEM = 85 EP Vote 2009 (LR order) 86.30 78.51 70.72 62.93-1.27-0.60 0.07 0.74 1.41 Libertarian and Group Tolerant Values
Figure 3: Libertarian Life-Style on LR Vote, cond. LR Polarization 86.61 PI_LEFT = 0.240 PI_LEFT = 0.397 EP Vote 2009 (LR order) 81.66 76.71 71.76 66.81-1.27-0.60 0.07 0.73 Libertarian Life-Style Values Figure 4: Libertarian Life-Style on LR Vote, cond. Years Democracy 87.87 YEARSDEM = 20 YEARSDEM = 85 EP Vote 2009 (LR order) 82.95 78.03 73.11 68.19-1.27-0.60 0.07 0.73 Libertarian Life-Style Values
Figure 5: Libertarian Life-Style on LR Vote, cond. Secularization 87.76 SECULARI = 17.300 SECULARI = 42.600 EP Vote 2009 (LR order) 83.45 79.15 74.84 70.54-1.27-0.60 0.07 0.73 Libertarian Life-Style Values Figure 6: Socioeconomic Values on LR Vote, cond. Polarization 95.14 PI_LEFT = 0.240 PI_LEFT = 0.397 EP Vote 2009 (LR order) 85.59 76.03 66.48 56.92-1.27-0.62 0.02 0.67 Socioeconomic Values (High: pro state)
Figure 7: Socioeconomic Values on LR Vote, cond. Years Democracy 95.41 YEARSDEM = 20 YEARSDEM = 85 EP Vote 2009 (LR order) 86.40 77.39 68.38 59.37-1.27-0.62 0.02 0.67 Socioeconomic Values (High: pro state) Looking now at Table 2 to see which macro-level variables do explain variation across countries in the EP vote when the dimension of competition is Left versus Right, and considering for graphical representation only the singificant macro-level factors, we can conclude the following. In all the situations (Factor 1, Factor 2, and Factor 3), both the clarity of policy alternatives (confirming H1) and the level of party system insttitutionalization (confirming H2) do have a positive impact on the strenght of the relationship between values and the EP vote, as expected: in more polarized polities and/or in polities with higher levels of party system institutionalization, individuals with libertarian attitudes (either in terms of group tolerance, or in terms tolerance visà-vis alternative life-styles) and individuals with more pro-state values tend to vote more for leftwing parties. The percentage of immigrants in the resident population has no effect on the relationships between values and the EP vote, thus not confirmining H8, and secularization (H9) has an effect contrary to the one expected, i.e., more impact of libertarian values (in terms tolerance vis-à-
vis alternative life-styles) in less secularized societies. But these are basically auxiliary hypotheses so we won t loose much time with them. Values and voting to the European Parliament considering the pro vs. anti further EU unification dimension of political competition Passing now to the impact of values on the EP vote when the latter is structured in terms of the pro vs. anti (further) EU Integration dimension of political competition, we can conclude from the individual level analysis (OLS regressions country by country, using the two step hierarchical regression, that it is not only for H3.2 (impact of socioeconomic values on the EP vote ordered in a EU scale of political competition) that we find mixed results. Here, considering only the siginificant coefficients, we can see that, out of 27 countries, in 10 polities that impact is negative, i.e., individuals that share more pro-state intervention attitudes, that want a stronger fight to inequalities, tend to vote more for parties that are closer to the anti further EU unification pole of this dimension of competition. Only in Italy, Poland, and Slovenia it is the other way around. And in the remaining 14 countries the coefficients are not signficant. Of course, this clearly contradicts Hix s, Noury s, and Roland s (2007: p. 163) thesis that, since the 1990s, Individuals that are more oriented towards fighting inequalities and favouring stronger state regulation in the economy are also more prone to vote in parties that favour further European unification. Although the results are somehow mixed, they nevetheless point overwhelmingly to the other direction: in most of the countries, individuals with substantive policy preferences close to socioeconomic left are becoming increasingly alienated from the process of EU integration.
But the results are also mixed, although in a lower extent, both for H1.2 (impact of F1: Libertarian attitudes and tolerance vis-á-vis immigrants) and for H2.2 (impact of F2: Libertarian attitudes and tolerance vis-á-vis alternative lifestyles). In the first case, H1.2, nine coefficients are positive as expected, namely by the GAL-TAN theories (Hooghe and Marks, 2008) and the theory of cultural threat vs. EU support (Mclaren, 2001-2007): more libertarian and group tolerant individuals tend to vote more for parties favouring furhter EU unification. However, there are five countries (Estonia, Greece, Lithuania, Sweden, and Romania) where it is precisely the other way around. In the remaining 14 cases the coefficients are not significant. Again, also for H2.2 (impact of F2: Libertarian attitudes and tolerance vis-á-vis alternative life-styles) the results do point in the expected direction: in eigth countries the coefficients are positive, so libertarian individuals tend to vote more for parties favouring furhter EU unification. But in four countries (Deutschland, Cyprus, Poland, and Portugal) it is precisely the other way around. So, here too (H1.2 and H2.2) we have mixed results, even if they are overwhelmingly in the expected direction. Overall, those mixed results (especially for H3.2, but also for H1.2 and H2.2) are translated in much lower levels of variance explained (vis-à-vis what we found when LR was the assumed to be the relevant dimension of competition): for all the countries take together in level 1, the value for the Adjusted R 2 (calculated through the two step hierarchical regresson procedure) when the 2009 EP is ordered in a EU scale is 0,034 (recall that it was 0,136 for the 2009 EP vote ordered in a LR scale). Thus, even at the individual level (for the 27 countries) the EP vote is much more structured by substantive policy preferences when the dimension of competition is LR than when the dimension of competition is EU unification. And this is in large part due to the mixed results
across countries in the latter case, which question the possibility of a unified electorate (across the 27 EU countries) in this respect; the opposite can be said about LR competition. And this findings also translate for level 2 (macro level): the average Adjusted R 2 (calculated through the two step hierarchical regresson) is 0,090 (F1: 0,119; F2: 0,108; F3: 0,042), and they are especially poor for the case of the impact of socioeconomic values on the vote ordered in a EU scale of policy competition (recall that equivalent results were 0,210, overall, and F1: 0,164; F2: 0,224; F3: 0,242, when the EP vote was ordered in a LR scale). Thus, the 2009 EP vote is much more structured at the across the EU countries when the dimension of competition is LR than when it is EU unification. Let us pass now to the final formal HLM 6.2 model for the EP vote 2009 ordered in a pro vs. anti (further) EU Unification dimension of competition, which reads as follows: Level-1 Model Y (Vote25RecEU) = B0 + B1*(Q115) + B2*(Q118) + B3*(Q114R) + B4*(V200R) + B5*(FAC1LIBE) + B6*(FAC2LIBE) + B7*(FAC3SEV) + R Level-2 Model B0 = G00 + G01*(PI_EUROP) + G02*(NETCONTR) + G03*(SOCIALEX) + G04*(PERCENT) + G05*(SECULARI) + U0 B1 = G10 + U1 B2 = G20 + U2 B3 = G30 + U3 B4 = G40 + U4 B5 = G50 + G51*(PI_EUROP) + G52*(NETCONTR) + G53*(PERCENT) + U5 B6 = G60 + G61*(PI_EUROP) + G62*(NETCONTR) + G63*(SECULARI) + U6 B7 = G70 + G71*(PI_EUROP) + G72*(NETCONTR) + G73*(SOCIALEX) + U7
Table 3 - Final HLM (Hierarchical Linear Modelling) 6.2 estimation of fixed effects (with robust standard errors) with the EP Vote 2009 ordered in a pro v.s anti (further) EU Unification dimension of competition ---------------------------------------------------------------------------- Standard Approx. Fixed Effect Coefficient Error T-ratio d.f. P-value ---------------------------------------------------------------------------- For INTRCPT1, B0 INTRCPT2, G00 5.541123 0.285850 19.385 21 0.000 PI_EUROP, G01 1.946956 1.369157 1.422 21 0.170 NETCONTR, G02-0.000355 0.000250-1.418 21 0.171 SOCIALEX, G03 0.000034 0.000025 1.371 21 0.185 PERCENT, G04-0.000800 0.017818-0.045 21 0.965 SECULARI, G05-0.004474 0.004270-1.048 21 0.307 For Q115 slope, B1 INTRCPT2, G10 0.005792 0.007604 0.762 26 0.453 For Q118 slope, B2 INTRCPT2, G20-0.064962 0.016144-4.024 26 0.001 For Q114R slope, B3 INTRCPT2, G30-0.091065 0.013873-6.564 26 0.000 For V200R slope, B4 INTRCPT2, G40-0.001605 0.002760-0.582 26 0.565 For FAC1LIBE slope, B5 INTRCPT2, G50-0.053212 0.057963-0.918 23 0.368 PI_EUROP, G51 0.513935 0.409593 1.255 23 0.222 NETCONTR, G52 0.000209 0.000069 3.034 23 0.006 PERCENT, G53 0.004422 0.002744 1.612 23 0.120 For FAC2LIBE slope, B6 INTRCPT2, G60 0.022972 0.036910 0.622 23 0.540 PI_EUROP, G61-0.263185 0.285175-0.923 23 0.366 NETCONTR, G62-0.000005 0.000024-0.214 23 0.832 SECULARI, G63 0.001446 0.000847 1.707 23 0.101 For FAC3SEV slope, B7 INTRCPT2, G70-0.180963 0.059979-3.017 23 0.007 PI_EUROP, G71 0.796752 0.356159 2.237 23 0.035 NETCONTR, G72 0.000025 0.000027 0.936 23 0.359 SOCIALEX, G73-0.000006 0.000005-1.210 23 0.239 ---------------------------------------------------------------------------- Source: data elaborated by the author form the European Eelection Study 2009 / PIREDEU, for the individual level data and also some macro data.