% & '!%!(%) *+,-,./01,.112

Transcription

1 !"#" $" % & '!%!(%) *+,-,./01,.112

2 Introduction The role of political parties in legitimizing regimes and integrating citizens into the democratic political order is often praised but rarely demonstrated. If parties have indeed such a role, they must exercise it, among other ways, through the emotional links that exist among parties and citizens. But identification with parties is declining in most established democracies, and seems to be relatively weak in new democracies. It becomes therefore particularly important to identify the conditions under which partisan attachments develop and to assess their attitudinal consequences. The present paper tries to find out what are the correlates of party identification (PID) in both New and Old Europe. While we have also analyzed sociological explanations, attitudes and institutions proved to be more consequential. We find bright and civic feelings behind PID as efficacy and positive evaluation of how democracy works, but also, and more importantly, darker sentiments, like ideological radicalism, highly differentiated feelings towards parties (that is, both love and hate) and the perception of parties as standing far apart. Next to the length of democratic experience, polarized ideological space and institutions that foster adversarial politics also contribute to the level of partisan attachments Institutional regimes that encourage clear governmental responsibility and concentrate power proved to be beneficial as far as PID is concerned. Interpretative frameworks. General hypotheses Party identification has been discussed in various frameworks in political science. The Michigan school (Campbell et al. 1960, etc.) was mainly interested in the sources of PID and worked within the framework of socialization. The individual was seen as being provided with an identity that was ready-made by parents, churches, neighbourhoods, trade unions and various social groups. PID appeared as a perceptual screen that filters and distorts information and thereby solves problems of cognitive dissonance. Rationalist perspectives have been developed by Downs, Fiorina, Budge and Fairlie, Robertson and others, who detected values and interests behind PID and emphasized the information-cost-saving-device aspect of party identification. This school experimented more with predictive models of party competition, and focused more on the consequences and less on the sources of PID. Party identification typically appeared as just one instance of predispositions, like social group membership or pure value orientation. Its significance was defined not only through its ability to structure vote, but also to confine party competition. Scholars working in the tradition of V. O. Key (e.g. Beck 1974, Burnham 1991, etc.) approached party identification from the angle of party system development. Party systems, according to this body of literature, develop through the phases of alignments, dealignments and realignments. The alignment period polarizes the public into distinct camps and creates strong partisan attachments. In this framework the focus is not on the stability, but on the periodic changes of the aggregate levels of party identification. The life cycle of the party systems, and not of the individuals, structures the narrative. The fourth major framework views PID from the perspective of modernization. This approach, relying on the definition of PID as a cost-reducing device, predicts the decline of PID, as cognitive mobilization makes partisan cues superfluous. People develop skills that enable them to make sense of politics without relying on identification with large scale bureaucratic organizations (Dalton 2000).

3 In the last decades one could witness a gradual strengthening of a fifth approach, the one that takes political context seriously. The characteristics of the political context have been repeatedly found to shape both the level of party identification and its relationship to other phenomena. Formal political institutions, like parliamentarism, the party-centered nature of the electoral system, the frequency of the votes, the number of elected offices, the complexity of voting, or the possibility of ticket splitting (Butler and Stokes 1969: 43, Norris 2005, etc.) and less formal ones, like ideological polarization and political culture (Percheron and Jennings 1981, Westholm and Niemi 1992, Schmitt and Holmberg 1995, Berglund at al 2005, Richardson 1991) have been discovered as major sources of crossnational variation in the level of PID. All of the above discussed frameworks assume established democracies. Those scholars who study party identification in new democracies discuss PID in the context of consolidation. In this framework the main question is whether people develop emotional attachment to parties at all (Huntington 1968, Rose and Mishler 1998, Morlino and Montero 1995). This issue is regarded as a paramount question for the assessment of the level of democracy, since parties are central players in democratic regimes. Or, if they are not, they should be, according to most scholars. Widespread alienation from party competition, expressed in the lack of party identification, is seen as a worrying sign of the fragility of representative democracy. 1 This literature regards therefore party identification not as an irrational perceptual screen, nor as a neutral social phenomenon, but as one of the best indicators of the deep-rootedness of the party system, and henceforth, of liberal democracy (Mainwaring and Scully 1994). Most empirical studies, whichever framework formed their basis, come up with similar profile of identifiers (e.g., Campbell 1960, Miller and Shanks 1966, Norris 2005, Schmitt and Holmberg 1995, etc). Men, older people, who are better integrated into families, urban dwellers, leftists, and those who are ideologically more committed are more likely to have PID. Norris (2005) found the religious to be also characterized by stronger partisanship, but we suspect that this is related to the presence or absence of religious parties. On other aspects there is less consensus. Or example, Berglund at al. (2005) found a negative relationship between education and PID 2, while Norris (2005) found a positive one. If one perceives party identification as a special form of partisan mobilization, then the dividing line is between those who place the emphasis on partisan and those who accentuate mobilization. If PID is just one form of mobilization we should find the most politicized segments to have the strongest PID: the educated, the rich, the sophisticated, and those who feel more efficacious. If, however, one regards parties as rather old fashioned institutions, appealing less to the elite than social movements and single issue groups, then one should expect rather the more traditional and less educated segments to be attached to parties, in line with Dalton s and Inglehart s cognitive mobilization thesis. This above difference emerges also concerning the clashing perspectives on the social utility of partisan attachments. One the one hand, following the functional theory of PID, it is obvious that the poor and the uneducated are the ones who are most in need of party cues. They are the ones who may have problems orienting themselves via an individualised consumption of political news. On the other hand, these people can very well choose (although choice is an obvious misnomer in this case) to ignore politics entirely. Alienation is typical of segments having lower status, and lack of PID can be conceives as a specific form of alienation from the regime. 1 Converse (1969: 142) also treated strong party identification as a precondition to democratic stability. 2 Note that the authors found this relationship weakening and among the youngest cohorts disappearing.

4 There is also some uncertainty concerning the attitudinal profile of the identifiers. From a civic culture perspective, those should be attached to parties who are integrated into the (partisan) liberal democratic regimes: who are satisfied and who have a positive attitude towards elections. But engagement in politics is often rooted in dissatisfaction. A polarized view of politics is also likely to characterize the identifiers. Party identification is increasingly a trait of a minority group, for which politics is particularly important. Actually, not so much politics in general, but rather the differences between parties. Only people in whose minds and hearts parties occupy very different positions are likely to have strong partisan attachments. Therefore primarily those should be identifiers who see large ideological differences between parties and who give very discriminating emotional responses to them. At the aggregate level, we expect countries dominated by older parties to have more identifiers. Political systems that present fewer constraints in front of party government are also expected to produce higher levels of identification. In these regimes responsibility can be clearly assigned to particular parties, and the election results have a definite impact on governmental policies. Therefore unicameral parliamentary regimes, without direct election of the president should be more hospitable to the phenomenon of PID. Ideologically polarized party systems should have the same effect. 3 Finally, although Norris (2005) found no relationship between party system fragmentation and PID, we entered this variable as well into our explanatory models. It was unclear, however, what relationship to expect. On the one hand if there are few parties less depends on coalition bargaining and responsibility is more clearly assigned. On the other hand, if there are more parties, people can choose from a larger menu and therefore it is more likely that they will find a party they can identify with. Those variables that had an impact on PID at individual level are also expected to have an effect at the aggregate level: the political contexts dominated by the aggregated individual-level characteristics are expected to create a certain climate that encourages or discourages partisan attachments. West versus East. Region-specific hypotheses Post-communist Eastern Europe is one of those areas of the world where the question of party system consolidation has an obvious relevance. Early assessments concerning the anchorage of parties in the society have been rather bleak, and Peter Mair even raised the possibility that the emotional involvement of citizens in party politics may remain at a significantly lower level than observed in the West (Mair 1997). Party system institutionalization has indeed encountered considerable obstacles all over the postcommunist world. Scholars most often point to the following hostile factors: weakness and instability of socio-political differentiation (meaning, in its moderate version, the lack of cleavages, or, in its more radical formulation, the complete atomization of the society), widespread alienation from the political system, elite-driven political transition, the particularly large influence of electronic media, inherited anti-party sentiments, weak civil society, international constrains on government activities, and the shortness of democratic experience (Enyedi 2006, Evans and Whitefield 1993, Katz 1996, Mair 1997, Hanley 2001, etc.). These background factors lead to the low popularity of parties (e.g. Rose 1995, Wyman et al., 1995), high electoral volatility, relatively low turnouts, small party 3 Cf. Schmitt and Holmberg Polarization has also been found to decrease volatility in Eastern Europe (Tavits 2005),

5 membership, weak grounding of parties in civil society, and the low level of organizational loyalty among politicians. Logically feeble party identification should be also part of the package. Whether this is indeed so is unclear, the literature is somewhat contradictory in this regard. Empirical studies typically report weak partisan identification (e.g., Rose 1995, Plasser et al. 1998, etc.), but Norris ( ) found no difference between Western and Eastern Europe. The expectation for the weakness of PID is somewhat paradoxical in the light of the information-cost-reducing function of PID (Butler and Stokes 1969, Goldberg 1969, Shively 1979, etc.). People would need the guidance provided by PID exactly in the midst of large scale social, political and economic changes. On the other hand, the central role of mass media in Eastern European politics may make PID less necessary than in the West. According to Shively (1979: 1049) PID does not form if there are other, more effective guides. While the guides he had in mind discussing Europe, strong social groups and mass organizations, are absent in Eastern Europe, mass media may serve as their functional equivalent. There are further reasons to expect low levels of PID in Eastern Europe. The high degree of volatility at the elite level and the frequent changes of party labels are likely to impede the development of stabile party identification, whatever the psychological needs of the individual voters are. As Budge and Fairlie (1976) put it, PID assumes that parties exist to some extent independently of the leaders, and also survive the mistakes of leaders. As the disappearance of many parties attests, in Eastern Europe this has often not been the case. The peculiarities of post-communist democracies make some of the important questions of the PID literature largely irrelevant. It is obvious, for example, that direct parental transfer of partisanship can occur only in relatively few cases. Parties that competed prior to the Communist takeover did resurrect in a number of countries, but they managed to capture only a minority of the vote. This does not mean that party preferences today are unrelated to the preferences that existed in the past in one s family, neigborhoud or social group. Wittenberg (2006) has recently demonstrated a surprisingly high level of continuity on voting preferences between 1947 and 1990 in Hungary, and was even able to trace back the transmission of political orientation to such social actors as local priests. But the lack of continuity of democracy, of party labels, and of politicized mass organizations, factors listed as paramount by Converse (1969), reduce the relevance of classic socialization mechanisms. The above listed contextual differences between the standard geographical area of the PID literature and Eastern Europe do not render the analysis of Eastern European PID meaningless. It is still perfectly sensible to detect the degree of closeness of citizens to parties, and to explain the causes and consequences of their attachments. Based on our knowledge of the two regions, we expect some differences between East and West, although the fundamental structure of PID should be the same. Given that classic mechanisms of socialization must play a smaller role in Eastern Europe than in the West, we expect that the role of ideological and emotional polarization will have a larger effect on PID in the East. We also expect that partisans should have a higher social status in the East, because post-communist politics is more elitist than Western politics (Mishler and Rose 1998). 4 As Norris, we use CSES data, but we have more countries and use slightly different indicators.

6 In case of education and age we expect opposite relationships in the two regions. If the cognitive mobilization theory of new politics is right, education should be negatively related to PID in the postindustrial West, and positively in the industrial East. An even more drastic difference between East and West is expected as far as PID s relationship with age is concerned. PID has been found to be related positively to age in almost all of the studies done in established democracies (see Abramson 1979 for an exception). This relationship with life cycle has been at the very core of the Michigan theory (Campbell et al. 1960: 161-9; Butler and Stokes 1969: 55-9); party identification is supposed to crystallize with age and with its repeated confirmations (i.e. vote). In new democracies, however, a negative correlation can be expected given that older people were socialised in authoritarian regimes. Young people are repeatedly found as the main reservoir of democratic party politics, being more open to liberal-democratic principles (Rose and Carnaghan 1995, Miller et al. 1998). Therefore they are more likely to shape their identity around political parties. Data Our data come from the Comparative Study of Electoral Systems (CSES). Although the CSES data-sets contain information from around the world, for the sake of comparability 5 we analyze only European countries, grouped into two categories: West (Finland, Belgium, Netherlands, France, Britain, Norway, Sweden, Switzerland, Germany, Iceland, Portugal, Denmark, Ireland and Spain), and East (Bulgaria, Hungary, Belarus, Lithuania, Czech Republic, Ukraine, Poland, Russia, Slovenia and Romania). That is, we have fourteen Western and ten Eastern countries. But we have altogether thirty-nine samples, because there are two samples (from two different time points), from Switzerland, Czech Republic, Spain, Hungary, the Netherlands, Germany, Sweden, Denmark, Britain, Norway, Portugal, Poland, Russia and Spain: altogether respondents. The data set was weighted in order to give equal weight to each sample. When country level analyses were conducted we manipulated the data so that the weight of countries have been equalized. Next to the CSS data, we also relied on the Comparative Political Data Set (Klaus Armingeon, Philipp Leimgruber, Michelle Beyeler, Sarah Menegale, Institute of Political Science, University of Berne) and on the Comparative Data Set for 28 Post- Communist Countries (Klaus Armingeon and Romana Careja, Institute of Political Science, University of Berne). Party identification is operationalized with the help of the CSES question: Do you usually think of yourself as close to any particular political party? (variable B3028 in the datafile). Next to treating the yes answers to this question as the indicator of PID, we have also created an eight value PID scale. For the construction of this, more fine grained, measurement tool, we used some further questions from the CSES survey in the following way: Respondents received eight points if they answered yes to the question above and also chose very close on the question: Do you feel very close to this party, somewhat close, or not very close?. If they answered with somewhat close or don t know to the latter question, they received seven and if they answered not very close received six. Missing value for the second question was treated as don t know. In case they chose don t know to the first question, but yes, to the further (third) question on identification Do you feel yourself a little closer to one of the political parties than the others? they received five. In case the answer was negative to the first question, but affirmative to the 5 By including Belarus we already stretch the limits

7 question inquiring about little closeness, they received four. Three was given to those who had don t know for the first question and no to the third question, while two was given to those respondents who rejected the first question and gave a don t know to the third. Finally, we gave one to those who said no to both first and third question. Ideological extremism was defined as a squared distance between respondents left-right self-placement and theoretical neutral point (score 5 on the 11-point CSES scale B3045). This variable ranges from 0 indicating exact centrist position, to 25 indicating the most extreme position (regardless whether left or right). To capture whether a political system creates obstacles in front of the (party-based) governments we relied on a scale entitled lack of constraints, consisting of two components: whether the president is directly elected and whether there is an effective second chamber. 6 If a country lacked an effective second chamber it received one point. If it also lacked a directly elected president, it received a second point. The information on the direct election of the president comes from the CSES survey. Strength of bicameralism has been calculated from the Armingeon data sets (with some modifications): Bulgaria, Finland, France, Britain, Hungary, Lithuania, Norway, Sweden, Iceland, Portugal, Spain and Ukraine were classified as being unicameral or quasi unicameral, that is with a very weak second chamber, wile Belgium, Belarus, Czech Republic, the Netherlands, Poland, Russia, Slovenia, Switzerland, Germany and Romania were classified as having a relatively effective second chamber. Party system fragmentation was measured by the Rae indices found in the Armingeon data while polarization is calculated from the CSES expert judgments of the left-right position of individual parties, simply calculating the distance between the two most-extreme parties (B5018). Affective polarization was measured by the standard deviation of the values assigned to the parties on the like-dislike items (B3037: I'd like to know what you think about each of our political parties. After I read the name of a political party, please rate it on a scale from 0 to 10, where 0 means you strongly dislike that party and 10 means that you strongly like that party). Higher score indicates larger differences in liking (and disliking) different parties. The perception of ideological polarization was measured by the standard deviation of the values assigned to the parties on the left-right dimension (B3038: In politics people sometimes talk of left and right. Where would you place party x. on a scale from 0 to 10 where 0 means the left and 10 means the right? Higher score indicates perception of larger ideological differences between the parties. For the operationalization of the other variables see the Appendix. Results Table 1 reports the percentage of those who answered with yes to the question Do you feel close to any particular party. Note that the surveys were usually within a few weeks after an actual elections, and therefore the results may somewhat overestimate the degree of partisanship. The table shows that post-communist systems do not form a separate group in terms of the level of party identification. Obviously, minor differences in question wording may have led to different results (cf. Kaase 1976, Abramson and Ostrom 1991). While we feel 6 Needless to say, these two features allow only a very rough assessment of the underlying phenomenon.

8 confident that the CSES question used here conveys rather well what is meant traditionally by party identification, some of the figures reported in Table 1 raise some concern. Particularly the very high (59.3%) identification in Ukraine is striking, given the country s fragile party system and the central role of individual politicians in that country. But in the same year another study has been run in Ukraine, with a different question format, and that produced the same percentage (Miller and Klobucar 2000). We have no reason to question that those who answered with yes feel as strongly for their party in Eastern Europe as anywhere else. But given the very high electoral volatility it is justified to question the stability of these feeling. To sum up, we are dealing here with a phenomenon that has different sources (socialization in the West, something else in postcommunist Europe), different level of stability, but, in all likelihood, the same degree of emotional significance. Examining those countries where more than one time point exists one can see that the temporal tendencies do not show clear regional differences either. An increase could be witnessed in the case of Switzerland, Czech Republic, Spain, Hungary, the Netherlands, and Sweden, while there was a decline in Denmark, Britain, Norway, Portugal, Poland, and Russia. Table 1 The number of identifiers as percentage of all respondents Election study % PID Election study % PID Election study % PID BLR_ ESP_ RUS_ SVN_ CHE_ SWE_ IRL_ ESP_ POL_ NLD_ BGR_ DNK_ LTU_ PRT_ ISL_ BEL_ ROU_ HUN_ HUN_ CZE_ FRA_ DEU_ GBR NOR_ CHE_ FIN_ ISL_ DEU GBR_ CZE_ POL_ SWE_ RUS_ NLD_ DNK_ ESP_ NOR_ PRT_ UKR_ Notes: Percentage calculated comparing the number of YES answers to B3028 against all other responses. Weighted by CSES sample weight (B1010_1). The difference between East European countries taken together (41.4%) and West (43.6%) is small, but statistically significant (weighted for the equal representation of each sample). The difference is, however, influenced by the relatively large proportion of don t know answers in Eastern Europe (7.9%). If PID is calculated just as the ratio between YES and NO answers, East (45.5) and West (44.9) are virtually equal. The differences are substantially larger across countries (eta=.26, missing included, PID as dependent) than across the two regions (eta=.13) in terms of our 2 value PID measure. In terms of our 8 value PID measure, Eastern Europe has somewhat lower average of PID (4.27, compared to 4.58, p<.001). In the light of the literature reviewed above, the level of PID in Eastern Europe appears as surprisingly high. And if we relate these figures to turnout, the weight of PID appears even

9 larger. Turnout in Eastern Europe is about ten percent lower than in the West. Therefore, the percentage of the active electorate that lacks party identification is larger in Western Europe than in the East: it is about 29% in the West, and only 20% in the East. Bivariate relations at individual level Those who are or have been married have, as expected, higher level of identification than the singles both in East and in the West. Here instead of correlations we simply report the percentages who answered yes to the main PID survey question: In the East 43% of the (once) married vs. 35% of the singles, in the West 45% of the (once) married vs. 35% of the singles have a party they feel close to. A similar degree of relationship was found with gender: 45% of men, vs. 39% of women have PID in the East, and 46% vs. 41 in the West. Contrary to our expectations, older people have stronger identification in Eastern Europe as well, although in some samples the relationship is not significant (Hungary 2002, Romania 1996, Belarus 2001, also Spain 1996) and in general the correlation coefficients 7 are small, and in Eastern Europe somewhat smaller (on average.11, p<.001) than in Western Europe (.13, p<.001). Education is related weakly and positively to PID both in the East and in the West (.08 each). This correlation goes up to.10 (East) and.13 (West) if one controls for age. That is, as opposed to Berglund at al (2005), we find no trace of education-based demobilisation in either of the regions. The strongest association is observed in Switzerland 1999 (r=.20, p<.001), Poland 1997 (.21), and Romania (.24). The relationship is insignificant in a number of cases, but no significant negative association was obtained. Income (household income divided by the number of family members) was also positively, but even more weakly related to PID (.07 in both regions). Stronger association was observed in Germany and Poland (.16 and.17 respectively), and again we found no significant negative association (though Denmark, Russia and Ukraine approached significance in that direction). The socio-economic status of the respondents (whether one is white-collar, worker, farmer or self-employed) has somewhat stronger impact in the East (contingency coefficient is.18) than in the West (C=.12). Eastern white-collars are somewhat more likely to be identifiers than their Western counterparts, while Western workers have somewhat more PID than the Eastern workers. The difference is large among the farmers the Westerner having more PID, but these findings are based on small number of cases. The degree of urbanization is not associated with PID in either of the regions. Positive (as expected) association materialized in Poland 1997 (.13) and Romania (.14), but negative coefficients appeared in Russia 2000 (-.12), and Denmark 1998 (-.13). In sum, as opposed to what we expected, social status is only slightly more related to PID in the East, and in general the relationships are rather weak. But it is also clear that the less well-off, less integrated segments of the society are the least likely to have party identification. Table 2 Correlation between PID and socio-economic status indicators 7 Below we report Pearson correlation coefficients between various background variables and our 8 point PID scale.

10 Age Education Income SES (B2012) Urbanization Church att. East white-collar more PID West workers and farmers more PID Note: Entries are Pearson correlation coefficients except in SES column contingency coefficients. All coefficients are significant at p<.01. Church attendance has again weak association with PID, but in this case the relationships have the opposite directions in the two regions: it is negative (-.04) in the East and positive (.04) in the West. The negative relationship in the East is due to the mobilization of the non-attending. The same difference, though even weaker, appears on the religious selfdefinition item (-.01 in East and.04 in West). One of the countries that contributes to this negative sign in East is, somewhat surprisingly, Poland. The Polish example goes against our expectation that the presence of religious parties will create a positive relationship 8. In this country a number of parties existed that used Christian labels, but they did not seem to be able to emotionally mobilize their supporters. Moreover, the religious core seems to have been demobilised between 1997 and Those who have PID are more satisfied with how democracy works in the West (.11), but in the East this relationship is very weak (.03), and in a number of countries, like Romania, the more dissatisfied are actually slightly more likely to be attached to parties. Those who think it matters who is in power are more likely to have PID in both East (.17) and West (.22), and the two regions are equal in the association between PID and finding vote meaningful (.20). Table 3 Correlation between PID and attitudes Satisfaction with democracy Political efficacy average Perception of ideological polarization Affective polarization Ideological extremism Left-Right selfplacement East West Note: Political efficacy average constructed by averaging the summarized responses to B3013 and B3014 (in the common direction). All coefficient significant (p<.01) except in the last column. In agreement with our expectations, emotional involvement and the perception of ideological (left-right) polarization among parties are among the strongest correlates of PID. The perception of polarization measured by the standard deviation of the party leftright scores assigned by respondents is related moderately to PID in the West (.17) and in the East (.16). More robust relations were found when standard deviations were calculated for the liking scores given to parties. The correlation of PID with this affective polarization score reached.34 in the West and.38 in the East. 9 Partisan attachments are also getting stronger as one moves from the center towards either of the extremes on the left-right self-placement scale. The shape of the relationship is 8 Some other examples, like the Dutch one, with r=.22, are more supportive. 9 Note that East has somewhat more affective polarization (2.9 vs. 2.6) and also somewhat more subjective ideological polarization (2.8 vs.3.1) than the West.

11 fundamentally similar in the East and in the West (see Figure 1), although the correlation between the two dimensions is higher in the East (.29) than in the West (.21) PID and Ideological extremism East West PID Centrism High extremism Figure 1 PID Ideological extremism in two regions Contradicting our expectation, left and right are equally identification-based. The extreme left has a slight edge over the extreme right in Eastern Europe 11 (Figure 2) PID and Left-Rght self-identification PID East West LEFT RIGHT Figure 2 PID and left-right self-identification in two regions Party family membership is not a particularly good predictor of the level of identification. As opposed to Norris (2005), we found that in Eastern Europe the Ecology parties have the most identity based electorates, followed by the Nationalists, Socialists, Right Wing 10 The correlation was particularly high in Bulgaria (.30), Switzerland 2003 (.29), Czech Republic 1996 (.36), Iceland 1999 (.30), Poland 1997 and 2001 (both.30), and Russia 2000 (.35). 11 Right wingers have stronger party identification in Denmark 2001, Finland, Ireland, Iceland, Lithuania, Portugal 2002, and Romania, while the left is more identity based in Switzerland 2003, Spain 1996, Russia and Sweden. The direction of the relationship changed from positive (1997) to negative (2001) in Poland.

12 Liberals and Communists. At the end of the list one finds the Agrarians, Social Democrats, Liberals and Conservatives. In Western Europe the Nationalists, Left Liberals, Agrarians, Conservatives and Socialists base their appeal most on identity. At the other extreme one finds the Nationalist Left, the Extreme Right, the Greens, and the Extreme Left. That is, in the West one can detect some relationship between old and new politics (the families belonging to the latter have weaker identification levels), while in the East identifiers are somewhat more likely to vote for radical parties. The East-West differences are particularly sharp for the Greens, the Social Democrats, the Agrarian and Conservatives. Regression analyses with individual PID as dependent variable Regressing the PID scale on social background variables (age, education, income, gender, urbanization, marital status, religious service attendance, and socio-economic status) one can explain six percent in the East (Adjusted R 2 =.060) and three percent in the West (.031). Age, education and gender retain their significance in both analyses. The Western model is dominated by age, while in the East education is the most important. Table 4 Socio-demographic predictors of PID B Std. Error Beta t Sig. Eastern (Constant) Age Gender Education Socio economic status Religious services attendance Rural or urban residence Single Family income Western (Constant) Age Gender Education Socio economic status Religious services attendance Rural or urban residence Single Family income Attitudes (affective polarization, perception of ideological polarization, left-right position, ideological extremism, satisfaction with democracy, and efficacy do a better job, producing Adjusted R 2 of.178 (East) and.140 (West). Left-right position has no impact in the Western analysis. The strongest influence is that of affective polarization, followed by political efficacy and ideological extremism in both regions. Table 5 Attitudinal predictors of PID

13 B Std. Error Beta t Sig. Eastern (Constant) Ideological extremism Political efficacy a Affective polarization Polarization of respondent's party L-R estimates Satisfaction with democratic process Left-Right Self Western (Constant) Ideological extremism Political efficacy a Affective polarization Polarization of respondent's party L-R estimates Satisfaction with democratic process Left-Right Self Notes: a Average of reversed B3013 and B3014. Attitudes and social background together explain more than one fifth of the variance on PID in the East (Adjusted R 2 =.204) and one sixth in the West (.149). Marital status and income have no effect, joined by socio-economic status and education in the East, and religiosity and urbanization in the West. After the controls, affective polarization remains the strongest predictor, followed by efficacy and in Eastern Europe by ideological extremism. Table 6 Combined attitudinal and socio-demographic predictors of PID B Std. Error Beta T Sig. Eastern (constant) Ideological extremism Political efficacy Affective polarization Polarization of respondent's party L-R estimates Satisfaction with democratic process Left-right self Age Gender Education Socio economic status Religious services attendance Rural or urban residence Single Family income Western (constant) Ideological extremism Political efficacy

14 Affective polarization Polarization of respondent's party L-R estimates Satisfaction with democratic process Left-right self Age Gender Education Socio economic status Religious services attendance Rural or urban residence Single Family Income While the picture does not differ radically across the regions, there are certain interesting divergences. Religiosity and urbanization in the West and education in the East seem to have an impact on PID only through certain attitudes: their direct impact is insignificant. Western identifiers reproduce better the traits described in the literature, for example satisfaction with the status quo and age matters more in their case. In the East the darker side of democratic politics dominates: identifiers come primarily from those who are ideologically committed (mainly to the left), the non-religious, and those who have a strongly polarized perception of politics. Bivariate relations at country level We managed to explain only a relatively small percentage of the variance of PID at the individual level. The aggregate level PID can be thought of being the function of the quantity of those variables that mattered at the individual level (context), with the addition of institutional specificities of the particular countries. Among the aggregated individuallevel variables (that is, the country averages) only the general level of affective polarization is significantly related to PID in bivariate (Pearson) correlations. Given the small sample size (39) it is not surprising that the variables are in general not related significantly to PID, but the direction of the relationships is as expected: ideologically more radical, more leftist, more satisfied, more efficacious and subjectively more polarized societies seem to have more party identifiers. Table 7 Correlation between average PID and attitudes Average affective polarization Leftright Ideological extremism Average satisfaction with democracy Average efficacy Average perception of polarization.372(*) Institutional level variables have a larger impact. With the exception of fragmentation, the relationships are all statistically significant. Older and more polarized party systems, lacking alternative centers of power (no directly elected president, no second chamber) have higher number of identifiers. The larger number of parties seems to depress the level of identification, though in this case the relation is statistically insignificant. But this piece

15 of information also fits well to the rest of the findings, since the smaller number of parties implies more power concentration and more clear responsibilities. Table 8 Correlation between average PID and institutional characteristics Lack of constraints Polarization Fragmentation Average age of party system.479(**).381(*) (*) Regression with country level PID as dependent variable Regressing average country PID scores on the aggregate level variables discussed above, the explained variance reaches 48%, a much more respectable figure than at the individual level. Among the examined variables only polarization and, with weaker effect, efficacy retains a significant impact. (If one uses the benchmark of.1 for establishing statistical significance, then also the age of the party system). Table 9 Macro-level model of PID B Std. Error Beta t Sig. (Constant) Polarization of respondent's party L-R estimates Affective polarization Left-right Ideological extremism Satisfaction with democracy Political Efficacy Party system polarization Fractionalization (Rae) No constraint for parties Average age of party system Given the small sample size one cannot run the regression separately on East and West. But entering region as a dummy variable, the explained variance goes up to 60%, and the impact of affective polarization becomes significant. That is, in the West, other things being equal, people are more likely to have party identification, but this difference between the regions is counterbalanced by the fact that affective polarization is somewhat higher in the East. Table 10 Macro-level model of PID including the region variable B Std. Error Beta t Sig. (Constant) Polarization of respondent's party L-R estimates Affective polarization Left-right Ideological extremism Satisfaction with democracy

16 Political Efficacy Party system polarization Fractionalization (Rae) No constraint for parties Average age of party system East-West Conclusions According to our results, whether one is attached to parties depends only to a small extent on one s social background. But as far as relationships exist, it is the more resourceful people (men, older, educated, married, higher income) who are likely to be attached to parties. That is, we haven t arrived yet to a postmodern era where the elite could leave parties behind. PID goes together with more positive views on democracy and more efficacy, on the one hand, but more ideological extremism and more polarized emotional and cognitive map of politics on the other hand. These results may sound trivial and paradoxical at the same time. They are paradoxical in so far as democracy is said to stabilize when the mentality of the people is moderate and inclusive, when party competition is not seen as a life-or-death struggle but as a routine practice that evaluates the incumbents and sets the course for new government policies. Citizens with this mindset should see most of the parties as potentially acceptable agents and should support the best one according to their interests and values. But those with PID rather see the struggle of parties as a fight between good and evil, taking the side of the good. The very fact that they take sides in the democratic struggles seems to help them to integrate into the system and is probably partly responsible for their pro-establishment feelings. Note however, that we could explain individual level PID to a small extent. Obviously, important factors are still missing from our model. We haven t included ideological distance from the preferred party, the liking of the party leader, or vote, because we felt that they are too close to the studied concept. Country level PID was explainable to a much larger degree, although we did not have the best data for measuring the political context and institutions. Older party systems have more identifiers. The lack of division of power (indirect election of the head of state and unicameralism) has a positive effect on PID. The same influence is shown for ideological polarization. In systems with few parties PID is slightly higher, maybe because citizens rarely have the experience of shifting from one party to another. These relations point to the same direction as individual level analysis: the key to party identification, next to the established and internalized democratic practice, is adversarial politics. East-West differences proved to be smaller than expected, obviously partly due to the fact that both regions, are very heterogeneous. The general model of the identifiers applies to post-communist Eastern Europe as well. The differences expected on social status, age, and education materialized only to a very limited extent. In Eastern Europe the leftists, the urban dwellers and the non-religious are somewhat more attached to parties, and age matters less. More importantly, PID goes together in the East with a lower level of satisfaction with democracy and with a higher level of ideological extremism (expressed by left-right identification and by the vote for radical parties). In general attitudes seem to have a larger role in Eastern Europe, and the adversarial aspect of PID is more pronounced.

17 Notwithstanding these differences, it must be emphasized that PID has a positive relationship with pro-democratic, civic attitudes even in the East. Rose and Mishler wrote that Whereas party identification in established democracies is regarded as showing civic virtue, in post-communist countries it is more likely to indicate that a person was formerly a member of the Communist party. (1998: 221). Our analysis in fact shows that partisanship is related to civic virtues also in Eastern Europe, though to a slightly smaller extent than in the West. Rose and Mishler also characterized Eastern Europe as a region where negative partisanship dominates. Trying to give a more optimistic interpretation to their data, they referred to Wattenberg (1994), who noted that negative feelings can eventually turn into positive identification. Our analysis points in the same direction with somewhat more confidence. Dislike of certain parties, the perception of the political world as a dangerous jungle, and the readiness to join the fight has a positive potential, and PID may be the mechanism that transforms these darker feelings into lighter ones. 12 In the West positive attitudes have a larger role, perhaps exactly because PID had a longer time to do the trick. But our data show that conflict, the perception of the conflict and the commitment to one of the sides is the fundamental source of PID in the West as well. Conflict and polarization can be supported by the psyche of the individuals and by the macro institutions of the polity. Our analysis indicates that both dimensions work, and they work in a similar direction, pointing to the role of adversarial politics as the foundation of strong partisanship. Institutions and individual perceptions are linked by party strategies. As shown elsewhere (Enyedi 2005), elite discourse that teaches citizens to see politics as a struggle between mutually exclusive camps, if coupled with adequate organizational strategies, can create strong political identities even under the conditions of the 21 st century media politics. And since old style parental socialization is loosing relevance across the whole Europe, we can expect that polarization-based explanations will gain more relevance. Future research should therefore rely more on theoretical frameworks emphasizing context, party competition and consolidation, and less on those that are centered on socialization and modernization. A major task is to disentangle the causal order between PID and attitudes. In American context it has been demonstrated repeatedly that PID is more rightly perceived as cause, but given the lower stability in Europe, and especially Eastern Europe, the opposite causal sequence equally likely. Appendix- Operationalization Education (B2003) was measured with an 8 value scale (1. none, 2. incomplete primary, 3. primary completed, 4. incomplete secondary, 5. secondary completed, 6. post-secondary trade/vocational school, 7. university undergraduate degree incomplete, 8. university undergraduate degree completed). For Czech 2002 study, respondents with special code 12 If one places in a regression model PID between the darker feelings (cause) and satisfaction with democracy and efficacy (effect), the direct impact of polarization-related feelings (subjective polarization, affective polarization, ideological extremism) becomes negative on the civic virtue feelings, while the indirect impact, the one that goes through PID, stays positive. We don t discuss these regressions because we cannot be certain concerning the actual causal order.