How politics can matter: Transformation and Fading of the religious cleavage in Italy 1972-2008 Paolo Segatti Cristiano Vezzoni University of Milan University of Trento Draft 20.03.2015 Do not quote without permission Abstract This paper analyses the changing links between religion, politics and voting behaviour in the last thirty years from an Italian perspective. After a brief description of the changes that occurred in the relations between religiosity and the vote in Italy in the last decades, we contrast the explanations of these changes by two paradigm of electoral change. The starting point of both paradigm is that the relation between social identities, namely religious identities, and party choice has weakened. However, the first paradigm, called here Partisan Dealignment theory (Dalton et al. 1984, Norris and Inglehart 2004), proposes an exclusively sociological explanation of these changes. The second one, called here Politics Matters theory (Thomassen 2005; Evans & Dirk de Graaf, 2013), underlines the fact that the political and institutional context can interact with social change in defining and constraining the political behaviour of citizens. The test of the different hypotheses derived from the two theories is based on the empirical evidence from five Italian electoral surveys (1972, 1985, 2001, 2006, 2008). Results suggest that in Italy the relation between religiosity, ideology and party choice, summarized in the concept of church and state cleavage, was definitely disrupted by changes in the politicalinstitutional sphere (namely, the almost complete change of the political alternatives -parties and coalitions - available and the reform of the electoral law at the beginning of the 90s). The strong process of secularization of Italian society only warrants that, after its disappearance, the religious cleavage will not come back to life in the same form it had before, even if political entrepreneurs or parties will try to mobilize voters using religious appeals. It remains open if religion, through mechanism different from the previous ones, can still shape the party choice. 1
1. Introduction The debate about the changing relations between social cleavages, politics and electoral behaviour has seen a number of substantial contributions in the last thirty years. Its premises are in Lipset and Rokkan s standard explanation of the European party system in terms of four structural cleavages (Lipset and Rokkan 1967;Rokkan, 1999). At the time they were writing, the effects of socio-structural cleavages on partisanship were still fully in place and the party systems of many European countries reflected electoral options frozen at the beginning of the 20 th century, when universal male suffrage with proportional electoral system was adopted. Ironically just a few years later, scholars began to argue that voting behaviour was beginning to lose its social anchors in the Rokkanian cleavages. In this perspective, Partisan Dealignment theory or Individualization theory, as it is often called, emerged as the mainstream account of the ongoing electoral change (Dalton et al. 1984). Its basic tenets are that changes in political attitudes and behaviour should be considered as an outcome of the social and cultural change, and voters choices are now affected more by short-term considerations as preferences on issues and leaders appeal, all of them being campaign driven. Recently, however, there is a growing awareness of the fact that this paradigm neglects or, at best, underestimates a crucial element intervening between social and cultural change and electoral behaviour, which is the political-institutional context of voting behaviour (Thomassen 2005, Elff 2007; Evans & Dirk de Graaf, 2013). We will refer to these contributions as the Politics Matters theory. This paper analyses how the relations between religion, politics and voting behaviour changed during the last thirty years in Italy. First, we briefly describe how the relation between religion and vote changed over the last decades. Secondly, we contrast what the two theories of political change, Partisan Dealignment theory and Politics Matters theory, say about religion and politics in countries like Italy. This analysis suggests empirical hypotheses that we test comparing the effects of religion on political orientation and voting behaviour in Italian elections between 1972 and 2008 1. 2. Religion and voting behaviour: The Italian case In Italy, the debate on the relation between religion and politics is of particular interest as the religious cleavage represented the most important social cleavage shaping the political supply structure as well 1 This paper builds on a previous paper adding the election of 2008 to the model and modifying in part the argument. We did not consider the very last Italian election of 2013 for the radical change in the political offer, which would have need an adjustment of the analytical model. We plan to include the 2013 election in the next version. 2
as the electoral options of millions of Italians. The canonical interpretation of Italian electoral behaviour looked at the religious cleavage from a particular perspective, in which the underlying concept was that of territorial subcultures (Galli 1968). Although the notion of territorial subculture was formulated independently from the Lipset-Rokkan theory of structural cleavages and came from a different intellectual history (Sani 1980), it shares with the Lipset-Rokkan theory the idea that electoral behaviour is the outcome of affiliation to a social group rather than the result of an individual calculus driven by psychological motivations or economic preferences. In Italy, party choice was often seen as the expression of social identities mobilized by political actors (Biorcio 2003). Parisi- Pasquino (1977) rightly defined this vote as voto di appartenenza (vote based on group membership). For many Catholic voters that meant, specifically, their (religious) social identity, built outside the political realm, was mobilized as such, and not through their active assent to the appeal of the denominational party (Christian Democratic Party, DC). This implied that Catholic voters in most cases were loyal habitual voters of the DC despite the lack of a high degree of subjective attachment to the party. In other words, for these voters the relation with their party was not direct, but mediated by membership in a social group. The Catholic subculture, also called White subculture was geographically concentrated in the North-East of the country (Galli 1968). Since the sixties, when the subcultures theory was originally formulated, many changes have occurred in Italian society. From the 1970 s onwards, the process of secularisation reduced the pool of practicing Catholics; it almost halved in less than 40 years (Biolcati, Segatti and Vezzoni 2008). 2 Still in the 1960s, they were approximately 50% of the population. At the beginning of the 21 th century, they are estimated to be between one quarter and one third of the population. 3 The change in the religious composition of Italian society contributed to breaking up the shell of the white subculture (Mannheimer and Sani 1987; Segatti 1999, Bellucci & Segatti, 2010)). But in 1992 election the Christian Democratic party still succeeded in getting the majority of the Catholic voters. Giacomo Sani, writing in the aftermath of the 1992 election, wrote that during the 1980s, students of Italian politics noted various signals of impending change in the structure of the nation s electoral market. [ ] (But) until the end of the last decade [ ] electoral returns still exhibited massive continuities with the past. Between the potential for change inferred by scholars and the degree of change actually recorded at the polls, there remained quite a gap. (Sani 1993: 119) Sani then predicted: a restructuring of the party system seems the likely outcome [ ] and perhaps the most decisive factor 2 Throughout the article, practicing Catholics are defined as Catholics attending the Mass on a weekly base. 3 The estimates of the number of practicing Catholics in Italy are controversial and vary quite substantially among surveys using different questions and methods (Pisati 2000, Castegnaro and Dalla Zuanna 2006). However, all agree on the fact that in the last decades there was a sharp drop in the number of practicing Catholics, and this is the relevant information in our discussion. 3
will be the adoption of a new and different electoral system (Sani 1993: 120). He was right. Although the massive social changes in the previous decades diminished the number of practicing Catholics and slightly affected the fortunes of the DC, it was only after the enactment of the new 1993 electoral law (in connection with the discovery of a vast web of corruption) that a real change took place. The reform promoted incentives to new political alternatives. Old parties disappeared and new actors entered the political scene. The voting behaviour of millions of Italians changed in the following elections. And a new electoral story began. Habitual voters changed their habits. In particular Catholic voters, because the DC disappeared, were to some extent forced to scrutinize the electoral market to find a new convenient home (Cartocci 2002, Diamanti 1996). These changes also had an impact on the relation between religiosity and vote. The religious cleavage seemed to disappear in the new majoritarian elections as the political heirs of the DC split between the two main coalitions (centre-left and centre-right) and in the 1996 and 2001 elections religiosity was no longer a strong predictor of the vote (Sani and Segatti 2002, Corbetta 2006 and Corbetta and Segatti 2002). The Italian elections of 2006 left us with a puzzle. Practicing Catholics seemed to have voted for the centre-right coalition more than in the 2001 election (Segatti 2006a; Segatti 2006b; Ceccarini 2007). The same happened in the election of 2008 (Ceccarini, 2008). The two election differ from the previous in many regards. First, in 2005 the center-right government change the electoral law, introducing a proportional system cum a majority premium to the coalition which received the largest shares of vote. Second, in 2005 and 2006 there was a tense political debate on so called moral issue. In 2005 several pro-free choice movements promoted referendum to repeal the 2004 that drastically limited the access to medically assisted fertilization. The referendum campaign was tense. The President of Italian bishop conference invited Catholics to abstain. At the end, the participation was abysmal low (25.9%) with 78% of yes among who voted 4. Yet, in 2008 the situation was again different. There was not debates on moral issues. The electoral campaign was centered about political issues. The fall of the Prodi center-left government for internal infighting among coalition partners and the new strategy of the two big parties, both brand new. On the centerleft the Pd was born in 2007 from the merge of two parties which were allied over the previous elections, DS and Margherita. On the center-right the PdL was born in 2007 from Forza Italia and Alleanza Nazionale, both allied in the same coalitions over the previous elections. Both parties 4 For the Italian Constitution referendum can be only repealing and they need a turnout more 50% to be considered valid. 4
decided to run in coalition with only one minor partner. This was intended to be a step toward a bipartitism instead of bipolarism. The consequence was frequent appeals to a strategic vote by the two parties aimed to drain the vote of the small parties outside the two coalitions. One of these was the Udc, the only party that until then openly claimed to be the heir of Dc and in previous elections has been always allied with the center-right parties. In the past Udc was a party with the largest share of church attendants Catholics within its electorate. In 2006 and 2008, then practising Catholics were more likely to vote for the center right parties than they were in the previous post1992 elections, but for different reasons 5. In 2006 the public debate on moral issues seemed conducive to re-opening a religious division. But this was not the case for 2008, where political issues were at the center of debate, and a party heir of the DC was more visible. More in general, does the 2006 and 2008 new mean a revival of the religion influence on vote choice carried through the same mechanisms in place in the pre1992 in the pre1992 party system? If institutional and political factors were the key variables in de-activating the effects of religion on the vote, would they also be able to revert the process, independently from the enduring effects of social change? Questions like these bring us to the core of two theories of political change: Partisan Dealignment and Politics Matters. 3. Two theories The basic claim of the Partisan Dealignment theory is that voters political orientation in many European democracies is less anchored to their social identities than it was before the 60 s, during the two decades after WWII (Gunther and Montero 2001, Franklin, Mackie, & Valen, 2009; Bellucci & Segatti, 2010). Voters seem to make a choice now, whereas in the past their votes appeared to be mainly driven by their group identities defined in terms of Rokkan s structural cleavages. Although this trend concerns all structural factors indicated by Lipset and Rokkan, influential scholars like Norris and Inglehart assert that religion seems to maintain more influence on voting behaviour than the other factors (Norris and Inglehart 2004) 6. In their analysis of religion and politics, they acknowledge that the link between religiosity and ideology weakened in many countries over the last twenty years. However, they hold that, even in post-industrial societies, religious voters tend to vote for right-wing parties and locate themselves on the right segment of the left-right scale more 5 However in 2008 there was also a No abortion list organized by a leading journalist former communist and selfprofessed devout atheist which got about 135000 votes (0.4%) 6 Franklin et alii (2009) share this view. 5
consistently than non-religious voters. 7 They claim therefore that religion remains more strongly and more consistently related to voting choice than any of the various indicators of socio-economic status (p.201), concluding their analyses with the strong statement: (At an individual level) in postindustrial societies, religious values continue to predict a sense of affiliation with the political right. This religious gap remains significant even after employing a standard battery of societal and individual controls. This gap is also consistently found in many diverse societies, suggesting that there is fairly universal pattern at work in people s ideological orientations (p.211). However, their explanation of this phenomenon is partial. It suggests that the salience of the religious cleavage in a country depends upon the level of secularisation in that country. For Norris and Inglehart, religious effects vary between countries (and possibly also across elections) and only depend on the pace of change of societal secularisation. This approach assumes therefore a single cause, linear explanation. This may be true considering that social factors, like secularisation, are able to diminish the proportion of religious voters and negatively affect the electoral fortunes of religious parties. But it is wrong when it does not consider that political factors can alter the way in which individual voters are linked to a party. This distinction between social (compositional) effects and political (relational) effects is not present in their accounts. Due to these shortcomings, their explanatory model does not provide any interpretation for variation of effects of religion on voting behaviour across different elections, when the level of secularisation remains almost constant. However, such variations have often been observed. In Italy, the strength of the relationship between religion and voting dramatically dropped between 1992 and 1996, although it is implausible that this drop was caused by a contemporaneous increase of secularisation. If we follow Norris and Inglehart, we should attribute the greater alignment between practicing Catholics and the right-right coalition of 2006 and 2008 Italian parliamentary election to a significant reversal of the secularisation process. Which is a quite unlikely hypothesis. The most reasonable explanation looks therefore in the direction of changes in the political and institutional arena. In other words, the explanation of Norris and Inglehart does not consider how choice may be constrained not only by social and cultural factors but also by institutions and the alternatives politics supplies. Those are precisely the main tenets of a set of theories that we label as Politics Matters theory. Over the years, the supply side factors have achieved a new status in electoral studies. There is a growing attention to the role that institutions and politics play in shaping electoral 7 Religious voters are defined as weekly attending church and non-religious voters are those who report they never go to church. This definition is independent from the party voted. A right self location on the left-right scale is defined as positions from 6 to 10. 6
choice. For instance, in comparative analysis of turnout, electoral rules and procedures are credited with a larger impact than social or individual dispositions of voters and there is an increasing awareness that institutions can strongly affect the cost of participation (Franklin 2002; Franklin 2004; Norris 2002). A volume edited by Thomassen (2005), where the predictions of a sociological theory of electoral change are tested against those of a theory focusing on the effects of political context in shaping electoral choice, provides new fuel against an exclusively sociological theory of partisan dealignment. 8 Looking at the individual relationship between voters social background and party choice, the studies edited by Thomassen find that the trend is not as linear as the sociological theory of partisanship would have predicted and furthermore a lot of country variation exists clearly depending on differences in political context. As far as the religious cleavage is concerned, the results are even clearer. The concluding strong message of the book is that politics matters more than social positions for individuals in defining their political choices ( p. 265) and that the politicalinstitutional context rather than any development in society is responsible for the variations in the explanatory power of a particular variable (p.263). 9 According to Thomassen, the role of the political context is crucial: not only can it re-compose an old cleavage, it can also foster its come-back. His final remark is revealing of this idea. He states that, due to the effect of political context, the political relevance of a particular social cleavage can decline or slumber for a while and then can come back to life (p.265, emphasis by us). This claim implies that the return of social factor effects on party choice, made possible by politics, could resurrect the social cleavages in the same form known in the years of frozen social identity politics. Accordingly, there would not be any long-term electoral change, but only a trendless waning and waxing of the social and individual factors impact on party choice. In other words, the disappearing of a cleavage is only one point in the cycle of life, death and resurrection. More recently, Evans and Dirk de Graaf (2013) restated the argument that political choice matters specifying which mechanisms are involved. They argue that distinctiveness of party supply strengthens the association between religiosity or class, not shaping the voters preferences, but presenting voters with more or less restricted choice with respect to ideology or values (p. 6-7). In other words, voters preferences are exogenous to elite cues or party signals, but they are reactive to them, (e.g. Note 4 in p. 7) 10. 8 The volume analyses the electoral change occurred since the 60 s in six European countries: Britain, Denmark, West Germany, The Netherlands, Norway, and Sweden. 9 Their results showed that in all countries polarization between parties determines the strength of some non political factors. In particular they showed that the effect of social class or religion co-varies with the degree of polarization (Thomassen 2005: 265). The argument that politics and institutions matter in party choice has become trendy in the past year, as the Cses series documents (Klingemann, 2009; Dalton & Anderson, 2011;Dalton, Farrell, & McAllister, 2011) 10 In the Evans and Dirk de Graaf volume, Heath and Bellucci (2013) present an analysis of the impact of religion on party choice in Italy, slightly different from ours, but the broad picture is similar. 7
In the light of this brief discussion of Partisan Dealignment theory and Politics Matters theory, we summarize our views as follows: in agreement with the sociological account of partisan de-alignment, we maintain that social identities may influence party choice less than they did before, but also they may have been changing. Following Bartolini and Mair we think of a cleavage not merely as a line of social division, but a three dimensional concept in which social location, organizational membership and subjective orientation have to be considered. When the church and state cleavage in Italy was at the apex of its electoral effectiveness, the party choice was influenced by being part of a religious group (social locations), exposed to organizational cues through weekly church attendance and through other organization membership, and self-identification with the Catholic World, standard expression to indicate the extended web of social community, organizations, and social practices. It is true that the decline of the effects of religious identities has been slower than that of other social factors and religiosity retains some predictive power for voting behaviour, but the key point we want to stress is the following. The religious cleavage broke down, so not only because less people went to the church, but also because regularly church goers less and less followed the cues by their religious leaders and feel part of an integrated world. Thus, one should take into account that over time the meanings attributed to the church attendance changed while remains in the shadow which aspects, if any of a religious attitude may be electorally activated. As a consequence, we think that an exclusively sociological account of voting behaviour is partial and misleading, since social change is seen as the only determinant of de-alignment, and we need to understand how political choice may trigger which aspect of a former integrated religious identity.. An account including political-institutional elements in the explanation of voting behaviour is then convincing and necessary. We actually hold that politics matter by affecting the links between social positions of the individuals and their party choice. On the other hand, we doubt that politics can first bury what was a cleavage and then bring it back to life without introducing any change in the way in which preferences related to social identity are connected to party choice. As far as the religious cleavage is concerned, such an approach neglects the fact that the electorate changes across time, the number of practicing Catholics diminishes and, therefore, politics might mobilize voters on religious issues through new motivations related to a particular aspect of their religious or social identity which became distinct and salient when in past they were subsumed by their social position (e.g. actively belonging to a church) generically activated by traditional cleavage politics. In short, we are not entirely convinced by the Evans and Dirk de Graaf argument that preferences anchored in social positions or in religious identity are exogenous to the political choice. We think that the Italian case might illustrate a thesis that political choice matters, but on the other hand that a social or religious identity may host a plurality of preferences, that can be diversely activated by party supply.as an 8
example, there are robust evidence that within the segment of weekly churchgoers opinion on moral issues may diverge (Segatti and Brunelli,2011, Biolcati, Segatti, and Vezzoni 2008). However, we have not evidence of which aspect of the religious identity is activated by variations of party distinctiveness on moral issues. We can only assess which mechanisms connect a generic index of religious identity, as church attendance is, to party choice, controlling for ideology.. From these observations, we derive the following testable hypotheses (see also fig. 1). H1. Decline of religious cleavage - In the course of the last half century, in Italy the role of religion in defining vote (H1a) and political self-placement on the left-right scale (H1b) decreased since the secularization process altered the religious composition of the Italian electorate; - but the dealigning key factor was the political earthquake in 1993. H2. Politics matters and increased autonomy of the political sphere - In the course of the last half century, ideology becomes increasingly autonomous with respect to socio-structural variables, with a consequent increase of the effect of ideology (left-right self-placement) on the vote (H2). H3. No revival of religious cleavage - The activation of the debate on moral issues does not imply a revival of the traditional religious cleavage indexed as church attendance, as the direct effect of religion on voting remains null and the major indirect effect passes through political ideology and not through orientation on moral issues. 9
Figure 1. Graphical representation of the complete structural model of the relations between religiosity, opinion on issues, ideology and voting The explanation suggested by hypothesis (H1) has two aspects: there is a compositional effect (diminishing number of Catholics) and, secondly, a decline of the relationship between social position and party choice. The first part comes from a mainstream interpretation of de-alignment theory (Inglehart 1977). But does the strength of the relationship between religion and politics at the individual level evolve in the fashion Norris and Inglehart expect? If so, the changes should parallel the development of the process of secularization, they should be slow and smooth, and the selfplacement of religious voters should remain oriented towards the right throughout time, although becoming slightly more heterogeneous. However this perspective does not take into account the impact of changes in the political-institutional context. The contribution of De Graaf et al. 2001 about the evolution of religious voting in the Netherlands actually shows that decline of religiously based voting was not only an effect of the shrinking number of religious persons but also the consequence of the changes in the political alternatives which became available. We hold that this is also the case 10
in Italy. The second part of the Hypothesis (H1) then looks at the role of the political supply side factors. The second hypothesis underlines that while the connection between religion and voting fades away, the ties between ideology (left-right orientation) and vote grew stronger over time and the individual political attitudes and behaviour became more independent from the social position of individuals (Shively 1979, Thomassen 2005:12, van der Eijk et al. 2005). This goes against the view of Partisan Dealignment theory (Dalton et al. 1984) which expects a decrease of the impact of ideology on the vote and a rise of issue politics and more individualistic motivations in party choice. Finally, the third hypothesis focuses on the shape religious cleavage assumes after its decline in the last half of the 20 th century. Contrary to the thesis of Thomassen, we argue that if the religious division in Italy were to become consolidated again, it would take an entirely new form, different from what we call a rokkanian cleavage. 11 Our expectation is that, given the lack of the social conditions (diminishing number of practicing Catholics), there is no room for a revival of the traditional religious cleavage. If religiosity will increase its influence on vote, it will not be through an activation of membership to a traditional religious institution like a church, but by means of a mobilization on religious issues taking place in the political sphere. However, this cannot be defined as traditional cleavage politics as Thomassen himself recognizes: voting for a party out of shared values without being a member of the associated social group does not constitute cleavage politics. (Thomassen 2005: 13, underlined by us. 4. Data, measurements and methods The data comes from four Italian national surveys run respectively in 1972, 1985, 2001, 2006 and 2008. 12 When weighted, the data is representative of the population; however, the analyses reported here did not use the weights, because the main interested lies in tracking the strengths of the relations at stake. 11 In this we agree with Montero et al. 2006 when they claim that in Spain the nature of the strong alignment of Catholics with the PP is different from the classic rokkanian religious cleavage. 12 Barnes and Sani 1972, post-election survey; Four nations 1985; Itanes 2001, 2006 and 2008, post-election surveys. Only for Figure 3 we used also data coming from Barnes 1968 post-election survey and from Itanes 1992, 1994, and 1996 post-election survey. All the data are available at the Itanes site: www.itans.org. 11
The dependent variable is electoral choice, defined as a dichotomous variable: either vote for the party or the coalition which collects the majority of the votes of active Catholics (coded 1), or vote for another party or coalition (coded 0). In 1972 and 1985, Democrazia Cristiana (DC), the main denominational party, was coded 1 while 0 represents all the other parties located to the left of DC. 13 From 2001, the same coding scheme could not be used because Democrazia Cristiana disappeared in 1993 in the wake of the bribing scandals at the beginning of the 90 s. The vote for the centre-right coalition was codes for these two elections as 1, since this coalition receives the votes of the majority of active Catholics, although this proportion is not as high as the proportion of Catholics who voted DC (in the 70s and 80s), while the vote for the centre-left coalition was coded 0. A vote for other parties (a very small percentage, less than 5% of the respondents) was not included in the analysis. 14 The intermediate variables pertaining to religiosity, political orientation and opinions on issues were operationalized as follow: Religiosity: single item measurement, attending Mass, going from never (1) to every week (4). Ideology (Left-right self-placement): ten point scale, with 1 corresponding to extreme left and 10 corresponding to extreme right. Opinions on moral issues were available for 1972, 2001, 2006 and 2008, and differed in these 4 surveys. In 1972 a question on support for legislation facilitating divorce was asked. From 2001, two questions were relevant to our purposes, one about abortion and one about homosexual rights. All these questions were recoded in such a way as to present a 4 point scale going from 1 (progressive moral position) to 4 (conservative moral position). 15 As structural controlling variables, we considered gender, cohort (divided in intervals of ten years), education (four possible outcomes: elementary school, intermediate, secondary school, university), and occupational class (three possible outcomes, compacting the Italian version of the Goldthorpe classification into: upper class, including entrepreneurs, service class and professionals; middle class, 13 In this choice, we follow Montero et al. 2006 in their analysis of the religious vote in Spain and Portugal. As a matter of fact, in 1972 and 1985, we excluded from our samples the individuals who declared a vote for the PLI (Liberal Party) or MSI (the former fascist party). That means that in the 0 coding we include, beside the Communist Party (PCI) voters, also the voters of the Socialist Party (PSI), Social Democratic Party (PSDI), Republican Party (PRI), and other extreme left small parties. 14 For sake of consistency with previous model, in 2008 we included Udc in the center-right coalition. 15 This choice gives to this scale the same substantive orientation of the left-right scale and of the measure of religiosity. In fact, we expect increasing religiosity and more right-wing location to bring to a more conservative moral position. 12
including urban and rural petit bourgeoisie and white collar; lower class, including rural and urban working class). Data were analysed by means of structural equation models (Bollen 1989). Since these models assume the use of continuous variables and many of our variables were ordinal, we repeated the analyses using polychoric and polyserial correlations instead of the conventional Pearson correlations (Bentler 2006: 7, 144-149). This method assures more reliable estimates when the model includes categorical and ordinal variables. Despite some minor changes (usually the parameters slightly increase their values), the conclusions from the models were the same. For this reason only the results obtained by structural equation models using Pearson correlation matrices and maximum likelihood estimation, are presented. 16 All cases presenting missing values were excluded from the analyses. The number of valid cases for each survey was: 802 in 1972, 1097 in 1985, 1846 in 2001, 975 in 2006, 1265 in 2008. 5. Testing hypotheses The first and second hypotheses can be tested in a simplified model presented in Figure 2, without including opinions on issues and can therefore be estimated for all five surveys (1972, 1985, 2001, 2006 and 2008). 16 The program used for the analysis was EQS 6. Cfr. Bentler 2006. 13
Figure 2. Graphical representation of the simplified structural model of the relations between religiosity, ideology and voting In table 1, the relevant coefficients of the model referring to the first hypothesis are reported. These results were computed controlling for the demographic variables (gender, cohort, education, social class) on religiosity and ideology. Table 1. Standardized coefficients of the relation between religiosity, left-right self-placement and vote from the models in 1972, 1985, 2001, 2006 Relation Survey 1972 1985 2001 2006 2008 H1a: Religiosity Vote,34,33,00,00,09 H1b: Religiosity Left-Right,37,37,17,21,21 H2: Left-Right Vote,44,39,76,81,73 (coefficients not significantly different from zero are reported as.00) From the results, it is clear that the direct relation between religion and voting dropped dramatically, whereas the relation between religiosity and ideology lost approximately half of its strength. Only in 2008 the direct effect of religiosity on vote gain back some significant strength, but it is far from the figures observed until mid 1980s. This is in line with the expectation derived from dealignment theory 14
which predicts a decline in the relation between religiosity, political orientation and the vote. However, following this paradigm, predictions should include a) a progressive, linear, slow decline of the relation; b) an increase of the dispersion of religious voters on the left-right self-placement scale (indicating an increasing ideological heterogeneity among religious voters ); and c) also, a persistently more right-wing position of religious voters, due to their adherence to more traditional value orientations. The estimates show that the decline is anything but progressive and linear. Between 1972 and 1985 (13 years) the relations remain almost unchanged. Between 1985 and 2001 (16 years), the strength of the relation between religion and voting moved from 0.33 (standardized coefficients) to 0 (zero). One could suggest that the decline of the relation began around the mid-1980 s and has been linear since then. But our knowledge about political-electoral events in the 90 s does not support this interpretation. In 1992, there was a drop in the Catholic vote for the DC. But really dramatic change occurred after 1993, when the Christian Democratic party disappeared and Catholics almost equally split between the two new coalitions (1996 and later), after a transitional election in 1994, when a centre coalition was still present. The effect of restructuring the political supply in 1994 on voters choice election is nicely illustrated by Figure 3. The connected line in Figure 3 shows the probability change between a church practicing voter and non church practising voter to vote for a party from 1968 to 2008. The key point is the sudden transformation of the landscape from 1992 to 1994.. In 1992 difference between church goers voters and non church goers was still large. In 1994 shrunk and the following elections became even smaller marginal. Only when in the political arena disappeared any remnants of the former Dc party the political relationship between church attendance and vote disrupted. On the other hand the historical difference between parties even when became allied has remained significative. Note, as an example, the difference between former communist Ds and former Dc Margherita /Ppi in 1996 and 2001. The two parties in 2007 join to form the PD. In 2008 the gap between Pd and other parties became a bit larger. 15
Fig. 3 Weekly Church attendance and party choice from 1968 to 2008 Pred. Probability change -.4 -.2 0.2.4 Dc Dc Old Party System Psi,Psdi,Pri Psi,Psdi,Pri Pci Pci Dc Dc Psi,Psdi,Pri Psi,Psdi,Pri Pds,Rc Pci Ppi Ppi Ccd Fi Fi An An Pds Pds New Party System Fi Ppi Udc An Ds Udc Fi An Ulivo Pdl Udc Pd 1968 1972 1983 1992 1996 2001 1994 Electoral years Pred. Prob. of Weekly Church Attendants minus Pred. Prob of Non Weekly Church Attendants across Parties 2008 2006 Note: Probability change is estimated with a multinomial regression in party choice are regressed on gender, age, weekly church attendance, education, class and macroarea of residence. The importance of the earthquake in 1993-1994 supports the idea of the crucial role which politicalinstitutional factors play in re-defining the relations at the individual level between religiosity and vote. This confirms the expectation of Thomassen and is in line with the findings of De Graaf et al. 2001 in the Netherlands, showing that the possible effects of the political-institutional context are heavily underestimated by Partisan Dealignment theory. This view can be supported further by considering the relation between religion and ideology measured by left-right self-placement. In the Dalton-Norris-Inglehart framework, the decline of the connection between the two variables should cause an increase in the heterogeneity of the positions of practicing Catholics (put differently, an increase of the dispersion on the left-right scale of religious voters however measured and here measured as an increase of the standard deviation), even when religious voters maintain more right-wing positions than the rest of the electorate. 16
Table 2 reports the average position on the left-right scale for practicing Catholics versus the others, and it seems that the expectations of dealignment theory are fulfilled. The average self-placement of Catholics seems not to have changed dramatically in the last 30 years (from 5.2 to 5.8) and their position remains more to the right than the other voters (remember that 1 is extreme left and 10 is extreme right); however, the dispersion of the distribution increases (s.d. from 1,8 to 2.6) and moreover the distance between religious people and the others voters decreases through time. Table 2. Mean and standard deviation on the left-right self-placement scale for Catholics and others across time YEAR Mean Practicing Catholics Std Deviation Others N Mean Std Deviation 1972 5,2 1,8 342 3,8 2,0 460 1985 5,4 2,0 295 3,8 2,2 802 2001 5,7 2,3 420 5,3 2,6 1426 2006 5,6 2,5 193 4,8 2,8 782 2008 5,8 2,6 348 5,2 2,7 917 N Nonetheless, against this first sight impression, the situation turns out to be completely different dividing the practicing Catholics between those who declared to have voted for the right (DC before 1993) and those who declared to have voted for the centre-left, as presented in table 3. In the two groups of Catholics the dispersion did not increase at all, although the movements on the scale were in opposite directions. Centre-right Catholics moved towards the right, while centre-left Catholics moved towards the left. The explanation is rather trivial: after the disappearance of the DC, in the 1994 election, the centre coalition (the natural successor of DC) was heavily punished by the majoritarian electoral system and did not overcome the shock, disappearing soon afterwards. Since 1996, the heritage of DC was split almost equally between the centre-left and centre-right coalitions. In the meantime, the long-standing taboo against publicly declaring right-wing sympathy fell, especially thanks to the re-definition of Alleanza Nazionale, the successor of the previous post-fascist party MSI, and respondents (finally) felt free to declare themselves as right-wing supporters. As a consequence the left-right self-placement became much more polarized compared to the DC era, and the Catholics did not escape this logic: those voting for the centre-right coalition moved towards the right. This is a well-known story, but it is essential here to underline that the relation between religiosity, ideology and voting did not evolve gradually in the expected Inglehart fashion but instead was much more influenced by abrupt institutional changes and a dramatic redefinition of the political supply side. 17
Table 3. Mean and standard deviation on the left-right self-placement scale for practising Catholics voting centreleft and centre-right coalitions across time YEAR Catholics voting Center-Left Mean Std Deviation Catholics voting Center-Right N Mean Std Deviation 1972 3,8 1,8 75 5,6 1,6 267 1985 4,3 2,2 89 5,9 1,8 206 2001 3,8 1,7 176 7,1 1,6 244 2006 3,6 1,5 89 7,4 1,7 104 2008 3,5 1,7 142 7,3 1,8 206 N Coming to the second hypothesis, we can see that, differently from the relation between religiosity and vote, the relation between left-right self-placement and vote doubled its strength, as indicated in table 1: the value of the coefficient moved from 0.44 in 1972 to 0.81 in 2006, going back to 0.73 in 2008, indicating a strong identification of the two political dimensions (self-placement and voting behaviour). This information, together with the previous demonstration of the role of political institutional change played in shaping relations at the individual level, confirms our second hypothesis of the increasing autonomy of the political dimension from the other dimensions connected to the social position of individuals. As far as the third hypothesis is concerned, we analyse the data coming from 1972, 2001 2006 and 2008 surveys where indicators for opinions on moral issues are available (divorce in 1972, abortion and rights of homosexual couples 2001 2006 and 2008). To test these hypotheses, we concentrated on a decomposition of the total effect of religiosity on vote and on the evaluation of the direct effect of opinion on moral issues on voting, as shown in table 4. 18
Table 4. Decomposition of the effects of religiosity on vote 1972, 2001, 2006 Survey Effects of religiosity on vote 1972 2001 2006 2008 H3: Total effect 0,51 0,13 0,19,25 Direct effect 0,31 0,00 0,00,05 H3b(derived): Indirect effect via issue 0,05 0,00 0,00,04 H3a: Indirect effect via left-right 0,15 0,13 0,17,15 H3b: Direct effect of issue on vote 1972 2001 2006 Divorce 0,17 - - - Abortion - 0,06 0,05,11 Homosexual rights - 0,00 0,05,04 (coefficients not significantly different from zero are reported as 0.00) First, there was a dramatic drop of the total effect of religion on voting from 1972 to 2001, which was predictable from the previous analysis. Then there appears a slight increase in the total effect of religiosity on the vote, but the whole effect is mediated through the political dimension of left-right self-placement, and the increase of the effect is more a result of the identification between selfplacement and voting than of a genuine revival of the importance of religion in politics. In fact, the indirect effect of religiosity on vote through orientation on moral issues in non-existent or very small (2008). This is mainly due to the fact that the effect of orientation on moral issues and voting is weakening and practically non existent when issues are not activated through political mobilization (an example is the question of homosexual rights which in 2001 was not really discussed during the campaign). These results are in line with our hypothesis, that the attempt to mobilize Catholics through the political activation of the debate on moral issues does not produce a resurrection of the previously known religious cleavage, since the fraction of Catholics in Italian society is diminishing and among Catholics the opinions are quite heterogeneous. The only significant effect is through political orientation. But this is rather different from the situation known until the beginning of the 90 s, when it was still possible to talk of voto di appartenenza (vote on the basis of group membership). 19
Conclusions The main aim of this paper was to test the predictions of two theories of socio-political change at the individual level as far as the relation between religion and politics is concerned. The test was based on the empirical evidence derived from the Italian case, comparing data coming from four surveys (1972, 1985, 2001, 2006 and 2008). The starting point of the analysis rests on the observation that in the last decades the relation between social identities and party choice has weakened. However, the two theories considered offer different accounts of this change. The first theory, called here Partisan Dealignment theory, proposes an exclusively sociological explanation of the changes between religiosity, ideology and party choice. The second theory, called here Politics Matters theory, underlines the fact that the political and institutional context can interact with social change in defining and constraining the political behaviour of citizens. In line with the first theory, we found that social identities influence party choice less nowadays than in the past. From 1972 to 2008, the direct effects of religiosity on party choice disappeared (but in 2008 increased a bit), while the indirect effect through ideology (left-right self-placement) remained but halved its strength. At the same time, the effect of ideology on party choice increased. We read this as an increasing autonomy of politics from other social spheres, a phenomenon in line with the expectation of the Politics Matters theory. This increasing autonomy of politics has two aspects: on one side, it shows how ideology becomes independent from socio-structural variables that represented important predictors of individual voting behaviour in the time of cleavage politics. On the other side, it shows how political and institutional changes have played an important role in the production of these changes. In Italy, where the more important cleavage was the religious one, the collapse of the Christian Democratic party at the beginning of the 90s represented a crucial event in the development of the relation between religiosity and vote. The disappearance of the religious party and the consequent impossibility to express a clear religious vote ended for Catholics the era of voto di appartenenza and forced the supporter of Dc to scrutiny other voting options. Eventually, these voters (many of them practicing Catholics) split almost evenly between the left and right coalitions. As a consequence, the old religious divide disappeared. In our opinion, this is the clear sign that in this process politics mattered a great deal. There is, however, a last puzzle. Several students of Italian politics have seen in the last election of 2006 and 2008 signs of a re-alignment of Catholic voters with the centre-right coalition. Is this event in line with the claim of Thomassen concerning a trendless cycle of disappearing and resumption of 20
the impact of social and individual factors on party choice? From the data this does not appear to be the case: even if an increasing number of religious people gave their vote to the centre-right coalition, this did not produce a revival in the relation between religiosity and party choice. The only influence of religion on the vote that increased is the indirect one passing through ideology, and, as shown, this effect is produced by a movement towards the right on the ideology scale of religious voters voting for the Centre-right. Surely, this phenomenon is again a product of the changes in the political alternatives which are available. But it does not support the view of Thomassen of a possible revival of the traditional religious cleavage because even if the results of the 2006 elections indicate a potentially increasing religious division, this division does not have the content it had 30 years ago. At that time, Catholics were habitual Dc voters since socially encapsulated in a web of organizations. The web fragmented over time and broke down in the early 90 earthquake. Nowadays seem to be marginally more likely to vote for center-right party feeling more close these party to their preferences on moral issues. But their positions on moral issues is increasingly detached by their church attendance, which hardly influence directly the vote choice, while its increasingly influence by ideology. Thus, the mechanisms connecting today religion to party choice in the early 90 are different from those in which the Dc was the denomination party of Catholics. On one hand, this is in line with the Evans and De Graaf (2013) thesis. On the other hand, the limited variations across elections seems to show that when there is a radical change in party supply as it was in Italy in early 90, parties seem less effective in reactivating the link between religions preferences and vote simply re-framing their positions on moral issues. It might so be because religion as all becomes less resilient to the party supply heresthetic or the measurement we adopted for religiosity (church attendance) is inadequate to grasp which religious dimensions is more resilient to the party choice. We need to think of it. Reference list Bellucci, P., & Segatti, P. (2010). Votare in Italia Dall'appartenenza alla scelta, 1968-2008. Bologna: Il Mulino. Bentler, PM (2006) EQS 6 Structural Equations Program Manuals. Encino, CA: Multivariate Software, Inc. Biorcio, R (2003) Sociologia Politica. Bologna: Il Mulino. Bollen, KA (1989) Structural Equations with Latent Variables. New York: Wiley. Cartocci, R (2002) Voto, valori e religione. In Caciagli M.e Corbetta Piergiorgio (ed.) Voto, valori e religione. Bologna: Il Mulino. 21