Leader Effects and Dealigned Electorates in Southern Europe

Transcription

1 Leader Effects and Dealigned Electorates in Southern Europe Abstract: According to some studies, party attachment has been declining in consolidated democracies. What does that trend signify for the importance of leader effects? That is the main question which this article tries to answer. We approach the issue by analyzing those without a party identification, the late deciders and those who switch parties from one election to the next and leader effects in three recent elections held in Italy, Portugal and Spain. We have been able to determine with some degree of certainty that leader effects tend to be higher among those with a party identification, and for those who do not switch parties, with evidence on late deciders being more mixed. The evidence signals that, contrary to what is often assumed, liking leaders is a phenomenon that is more prevalent among those who feel closer to the political parties, and therefore also more involved in the political system as such. Attachment to leaders is thus not a sign that individuals are increasingly deciding on the basis of beauty contests rather than on political issues, a promising trend for the future of consolidated democracies. Marina Costa Lobo, Institute of Social Sciences, University of Lisbon Marina.costalobo@ics.ul.pt Paper Prepared for the IPSA-ECPR Congress in São Paulo, Brasil 17 February First draft, please do not cite.

2 Political analysts regularly attribute legislative election victories and defeats to leaders. This view has been corroborated in some literature which has argued that democracies are becoming increasingly personalised, i.e. that irrespective of the formal constitutional position party leaders rather than political parties are now the decisive actors in the political system. (Poguntke and Webb, 2005; McAllister, 2007). And yet, parliamentary elections are still fought on the basis of party labels, and in all cases, voters choose between party labels, not future Prime-Ministers. It is extremely difficult to disentangle the two effects leader and party effects - since the two are highly endogenous, as has been shown by several studies (Curtice and Holmberg, 2005; Brettschneider, 2005, Lobo, 2006). Not only is this an interesting question in itself, but it also leads to another, perhaps more meaningful one, namely what does that finding represent for democracy? This article tries to address that issue by looking at the relationship between different aspects of party attachment and leader effects. According to some studies, party attachment has been declining in consolidated democracies (Schmitt-Beck 2009). Evidence from one of the oldest democracies in the world, the US, shows that in 2000, the percentage of independents reached its highest level in the half-century series of the American National Election Studies at 40% of the electorate (Dalton, 2009). For Western Europe, it has recently been found that partisanship and the vote have become more independent over the past decades (Berglund et al., 2005, p. 123). In a recent article, Anthony Mughan agrees with the idea that partisanship has been transformed in the last years. According to this author, whereas the assumption of partisan exclusivism may have been valid in an era of strong partisan alignments when identifiers were durably loyal it has become less so in a contemporary democratic world where partisan alignments have weakened, parties have converged ideologically and party systems have depolarized to some degree (Mughan, 2009: 414). The emphasis is on changes in parties and party systems themselves to explain the loosening of ties between electors and party identification. On a similar note, Kriesi et. al., 2006, have argued that the growing convergence on economic issues such as wealth redistribution between left and right wing parties, has left leftist identifiers freer to be attracted to right-wing parties who defend conservative stances on certain issues such as EU integration, immigration, and law and order. These issues often cross-cut the socio-economic cleavage line which underpins traditional partisan identification. As the non-economic issues gain salience, so the strength of party identification will tend to decline, if it was sustained on the economic cleavage, as occurred

3 in most West European cases. Thus, a considerable number of scholars agree that partisan identification is becoming looser. What does that trend signify for the importance of leader effects? That is the main question which this article tries to answer. Compared to the literature on the relationship between political sophistication and leader effects ((Kroh 2004 and Gidengil 2000, Rico 2009) there are very few studies which deal directly with the issue of the politically dealigned and leader effects. A recent longitudinal study on the German case found that those without partisan attachments are much more vulnerable to leader effects than those who have them (Brettschneider, 2006: 489:490), and that the relationship holds steady across elections, with only one exception, the election of According to Russell Dalton, the impact of the candidate s image is expected to be stronger on undecided voters who tend to make their choice at the last minute (Dalton, 2006). That is due to the fact that the undecided are characterized by having a relatively unstructured vote, and thus would be more susceptible to leader images during the campaign. Similarly, we would expect that voters who switch from party to party or from abstention to party may be less anchored in party identification and therefore more likely to be impacted by leaders in making their vote choice (Bartels, 1993). But these claims about the relationship between electors and leaders are based on the premise that leaders are not a proxy for political values. Indeed, in hallmark studies the importance of leaders has been interpreted as a signal of a decrease in the political content of electoral choices (Campbell, 1966; Nie, et al., 1976). In this view, using leaders as a cue for voting symbolizes a decline in the accountability between voters and politicians: the former are wooed by appealing media images rather than through party political mobilisation. Leaders are in effect a symbol of the mediatisation of elections, whose political content is progressively thining. On the contrary, if we find that leaders matter because of the political positions they hold or their representation of a given part then we must conclude that leader effects are a sign of a politicized, rational electorate. In this scenario, a personalization of politics would signify a relationship between the electorate and politicians whereby the latter are closely monitored. In this scenario, leader effects are in effect a proxy for party identification, with those who have a party identification also caring more about leaders, than those electors who have no particular liking for any party. If leader liking is a proxy for political positioning, then it is those who are most politically mobilized who will be more sensitive to leaders, since the latter are carriers of ideology and political positioning. Thus, this article s findings

4 are relevant to understand the nature of leader effects, whether they serve as political/party cues for the electorate, or whether they are media objects to attract the demobilized. This is important not only to understand the impact that leader effects have, but also because it give us an idea of whether the importance of leader effects is a worrying trend, signaling the depoliticisation of electoral contests, and their transformation into beauty contests, or whether on the contrary, leader effects are actually party politics conducted in an era where the mediatisation of elections has made the leader as vehicle for communicating the political message, more important. Our focus then is the relationship between a loose relationship with parties and the importance of leader effects, and we will approach the issue by analyzing those without a party identification, the late deciders and those who switch parties from one election to the next and leader effects. The reason for analyzing these three facets of a loose relationship with parties is the following: individuals may manifest their relationship to a party both in attitudes (having a party id or not) and in behavioural terms (deciding late, switching between parties or not). These are not, as we will show, the same electors, but they share in their loose relation to party. To understand the differentiated importance which leaders may have to those with such characteristics, we have opted to analyse these three groups in three recent elections held in Italy, Portugal and Spain. The reason to choose these three counties is due to the fact that they exhibit not only different magnitudes of leader effects, but also aggregate levels of party identification, party switching and late decision for voting thus providing different contexts where we can observe the relationship at work. In Spain and Portugal, existing studies on leader effects have shown that these seem to have an important impact on the vote (Gunther and Montero, 2001, Lobo, 2006, Rico, 2007; Magalhães 2007; Barisione, 2007). Initial studies tried to measure the magnitude of leader effects, although more recently, the conditionality of effects has also been studied (Rico 2007). Lobo has also looked at the substantive meaning of leader effects in Portugal, to show that Prime Ministerial competence (rather than candidate s more superficial traits) is the most explicative candidate trait (Lobo 2009) for the 2005 Portuguese elections. In Italy, leader effects seem to more modest (Barisione, 2004). Nonetheless, the few comparative studies which exist on Southern Europe have not attempted to enquire on the relative difference between dealigned and non-dealigned electors. In the following section we discuss the data which we used as well as the methodology which we employed. Then, we present descriptive data on the electorate with

5 loose attachment to parties in the three countries considered. In the last section we present the analysis and the results which were obtained to test the relationship between partisanship and leader effects. Data and Methods In each case, the most recent databases available which included the relevant variables were selected (Portugal: 2005 n=2801; Italy: 2006, n=2011; and Spain 2008, n=6083). All of these surveys were carried out post-electorally. Some, but not all employ the CSES questionnaire. Our goal is to establish whether there is an interaction effect between those with no party identification, the late deciders and the party switchers, and leader effects. In order to do so, we will run three models to test each interaction. The first is a simple model with only the leader barometer and the interaction as independent variables; then we complexify the model by adding a party barometer. The third model is a fully specified one which, beyond the variables included previously, also includes some socio-demographic models, as well as Left-right self-placement and a short-term variable measuring perceptions on the economy. The goal is to test the robustness of the interactions, both in terms of their sign as well as in terms of significance, across increasingly well-specified models. Our dependent variable dichotomises vote for the incumbent (1) vs. other parties ie all parties which polled more than 3% in the election (0). In Spain, the centre-left PSOE and the right-wing Popular Party together have controlled on average 90% of the seats in the Parliament, for the elections considered here (Appendix 1). The only other national party which is also considered in the Spanish data is its Communist party the IU a really minor party, which polled 3.8% of the votes in In 2005 Portugal also had a majoritarian party system. However, contrary to what occurred in Spain this has not been a constant in Portugal. It was only from the mid-1980s onwards that the NEPP in Portugal reduced substantiatially. Electors switched their voting to the two main parties, the Socialist PS and the centre-right PSD. The smaller parties have seen their percentage of seats in Parliament decline accordingly. Even so, they are larger than the corresponding parties in Spain. The Italian case is a relative outlier in terms of number of effective parliamentary parties. The electoral system change which occurred in 1994, transformed Italy s PR into a mixed system (Bartolini, 2004). Since then, the NEPP in Italy has started to decline. The long-term impact of electoral system change has been to create large new parties on the left and on the right, the Democratici di Sinstra and the Forza Italia, which themselves are at the

6 centre of two broad right and left wing parties/ coalitions. The other two parties analysed are Alleanza Nazionale (AN), and the Rifondazione Comunista (RC). The AN is a populist version of the traditional extreme-right party MSI, lead by the charismatic Fini which in 2006 had governed in a coalition government with Berlusconi s Forza Italia for the previous five years. RC, on its part is the remaining orthodox Communists, remnant of the more left-wing of the Italian Communist Party. Italy had a mixed-member proportional electoral system in place in 2001 and Each voter had two ballots, one to cast a vote on the majoritarian list and another to cast a vote on the proportional list. The dependent variable chosen is the one which asks the respondent to state for which party she voted in the proportional list in the Chamber of Deputies. The reason for this choice is that in the majoritarian part of the election respondents selected mostly among broad coalitions. The leader variable is coded from the response to the following question. Using the scale (1 to 10), I d like to ask you how much you like or dislike some political leaders. Again, if I come to a leader you haven t heard of or you feel you do not know enough about, just say so. The first political leader is [LEADER A]. In each model, two leader variables are included, namely one measuring attitudes towards the leader of the main incumbent party, and to the main opposition party. In Model 2, the independent variable included, apart from the leader variable is a variable which measures proximity to the incumbent party. It is identical to the leader variable, but enquires the respondent on the likeness for each party. This variable was chosen over the party identification one for it all respondents, both those with party identification, and those without responded to that question. Model 3 adds a host of other variables to Model 2. Namely, it includes the following socio-demographic controls: education (a five-point scale variable from minimum to maximum); church attendance (a five point scale variable from never to once a week or more); and union membership (0= no; 1=yes); Due to the fact that there were discrepancies from one database to another we harmonized the codings so that in each case the same scales were used. The respondent s left-right self-placement is also included in Model 3. In some cases (Portugal 2002, 2005) the scale included an eleven point scale (from 0 to 10), whereas in all other cases a ten-point scale (from 1 to 10) was employed. In order to harmonise all Left-right self-placement variables we standardized the Portuguese measurement. Finally, Model 3 also includes a short-term variable, measuring attitudes towards the economy. To measure the former we used the standard sociotropic retrospective economic variable. Although present in Italy and Portugal, in Spain the formulation is slightly different, with the

7 variable tapping the opinion of the electors of the economy now rather than in the previous twelve months. In each Model, apart from the independent variables mentioned, interaction terms are included between the leader barometer and each of the variables which represent different facets of party dealignment. The Dealigned electorate: Lack of Party identification, Late Deciders and Swing Voters In this first section we present a descriptive analysis of the key variables which will subsequently used as interaction terms in the regression models we build. Thus we determine the actual proportion of the electorate who can be characterized as having a weak relationship with parties, being undecided or switching between parties, as well as the differences in these segments of the electorate in the three country/elections analysed here. In Figure 1, we present the electorate of the three countries in terms of their strength of party identification. For constructing the variable we grouped those who professed to having no party identification with the scale which distinguishes between those who are mere sympathizers, to those who either feel quite close, or very close to a party (1= very close to a party to 4=no party id). We can see that Spain stands out in one important respect: over 40 percent of respondents claim that they do not identify with any party. By comparison, Italian voters are much more engaged with parties, with only 21 per cent admitting that they do not feel close to any party. Portugal in this respect holds an intermediate position, with 32 per cent sharing the same view. Considering those who do agree that they identify with a certain party, Portugal stands out for the large proportion of the electorate (44 percent) who declares that they are mere sympathizers. Indeed, if we add the percentages of those who have no party id, or are mere sympathizers, we can see that in Portugal that percentage reaches 76 percent of the electorate, in Spain it is 62 percent and in Italy 45 percent. Thus, in this respect, Spain and Portugal stand out for the weakness of party-citizen links. Despite the successive party (and party system changes) which have occurred in Italy, the attachment to parties in considerably higher in that country. (Figure 1 about here) One of the reasons for the weak anchoring of parties in Iberia is the fact that they developed at a relatively late stage. Diamandorous et alli emphasise that in Spain and

8 Portugal there were leapfrogging processes of change which means that some phases of party consolidation have been completely absent in these democracies: In [.] Portugal and Spain, the establishment of new party organizations following the suspension of party politics occurred in the absence of the institutionalized legacy of older forms of party organization and electoral behaviour. The absence of these institutionalized legacies enabled these countries to pass over developmental stages and organizational forms of earlier eras that had become entrenched features of many parties in the longer-established democracies of Western and Northern Europe. Accordingly, and somewhat paradoxically, the new democracies of Southern Europe, unencumbered by these institutionalized legacies of the past, more quickly and thoroughly took on features of the late twentieth century version of modernity than have many previously established democratic systems (p.4). What this in effect means, as has been shown in the previous graphs is that the link with political parties is considerably weaker than in Italy which followed a more typically West European path in terms of post-war democratization as well as party system institutionalization. Italian political parties established in the 1940s embraced the traditional mass-party models, but the recent upheavals in party system format make it an interesting case, and show that despite this, it still compares favorably to both Spain and Portugal in terms of party identification. The figure below shows us that between a fifth and a quarter of Italians, Portuguese and Spanish electors decide during the campaign for which party to vote. For this indicator, there are not huge differences between the countries. Portugal holds the highest percentage of late deciders (33 percent). (Figure 2 about here) The next figure shows the degree to which electors in Southern Europe are not maintaining the same vote choices from one election to the next. This variable was coded in the following way. For each election a new variable was computed using the recall vote variable from the most recent election and the one before that. Those who had voted for a different party, had abstained in election t-1 and chosen any given party in election t, or were too young to vote in election t-1 and voted in t1 were coded 1. Those who had voted for the same party in both elections were coded 0. The graph shows that Italy stands out for the degree to which its electors switch parties (34 percent), compared to Portugal (25 percent) or Spain (20 percent). However, the party switches in Italy have to be seen in the context of

9 party labels changes which have occurred. In part this data is overstated given the degree of party changes which exist. (Figure 3 about here) Thus overall, the picture presented distinguishes Italy from Portugal and Spain, with the Iberian countries appearing as the most dealigned in terms of party identification. Considering late deciders and party switchers, differences seem smaller between the three countries, although Portugal and Spain are also the ones where decisions on which party to vote for are taken later. To what extent are these electors the same? In order to argue that deciding late, switching votes and being a weak party identifier are components of dealignment, but not exactly the same phenomenon under different names, it is necessary to look at the correlations between the variables just presented. Table 1 indeed shows that correlations between the three variables are positive, and significant. I.e. those who decide late, or who switch parties tend to also be to have none or a weak part id. Moreover, in the three countries, late-deciders are more correlated with strength of party id than swing voters. However, values are not very highly correlated, suggesting that these groups, although related, are not the same. (Table 1 about here) Being a party switcher and deciding late on who to vote is a behavioural sign of a loose relationship with the parties. Thus we can test the relationship between each of these indicators and leader effects. In the following section we present the data and methodology which was used in order to build the models where we test the importance of these variables in interaction with leader effects. Results The first set of results report to the three models that include an interaction with degree of party identification. In the first model tested, the only two variables included are the latter a four-point scale which ranges from no party id (0) to strong party id (4), as well as

10 the leader thermometer. The model also includes the leader barometer interaction with the party identification variable. The simple model is run for each of the three countries. We find that in all three cases, the model fit is substantial, as would be expected. Without further controls, leaders seem quite explicative of voting choices, especially in Italy, and more moderately although quite substantially in Spain and Portugal. In all three cases the interaction is positive, meaning that as party identification is stronger leader effects also matter more. Nonetheless, the coefficients are significant only in Italy and Portugal. In Model 2, we add a party barometer variable to this Model. The party barometer variable is used instead of a classic party identification variable in order to capture the idea of party closeness, which is supposedly more prevalent today than a close ended party identification. We find that in this Model, which includes leaders and parties, plus the interaction term including strength of party id and leaders improves the model fit very substantially, (Italy 83%, Portugal and Spain, circ 60%). Party closeness subsumes a host of socio-demographic variables, and therefore this model is both parsimonious and robust. It shows that the interactions run in each country are positive and significant. This means that in this Model, leader effects increase as party identification is higher. Model 3 is a fully specified one, and therefore includes a few key socio-demographic variables, namely education, union membership, church attendance and left rightselfplacement. Bearing in mind that in Italy and Portugal the incumbents were right-wing parties (Berlusconi and Barroso, respectively), and in Spain the incumbent was left-wing (Zapatero), the coefficients vary in sign as is expected. Both union membership and left right selfplacement have a negative, significant coefficient in Italy and Portugal, and positive in Spain. Church attendance and education have a positive and significative coefficient in Italy and Portugal, whereas they have a negative and significative coefficient in Spain. Left-right selfplacement and the party barometer are the most important variables in this model. As for the variable which we are interested in, the interaction between party identification and leader effects it is always positive, although it only retains its significance in Portugal. This is a very demanding model, with r2 reaching between 70-90%. In all cases, the interaction is positive, and in 6 out of 9 cases it is significant, thus suggesting that leader effects are stronger among those who have a party identification. Next, we evaluated the same three Models, but used a different interaction term, namely one between leader effects and latedeciders, which is coded in the following way (0=decided early, 1=decided late). The model performs well, better in Italy than in Spain and

11 Portugal, although it also has a good fit in these latter cases. More importantly, in all three countries, the coefficients are negative, and positive in two cases (Spain and Portugal). Looking at Model 2, where the barometer on party is added, we see the following: not all coefficients have the same sign. Whereas in Italy and in Spain, the sign is negative, ie those who have decided early when to vote also are more influenced by leader effects, the opposite is the case in Portugal, with late deciders being more affected by them. Moreover, the coefficients are significant for only in Portugal and Spain. Looking at the fully specified Model, we can see that in all cases the sociodemographic variables behave as expected, considering the ideological position of the incumbents in the three countries considered. As with the previous Model 3 presented for the interaction with pary identification, both the party barometers and the left right selfpositioning emerge as the variables with highest significance and the model fit is similar. Finally, in what concerns the variable we are interested in, namely the interaction between late deciders and leader effects, it is now negative in Portugal and Spain, with a positive coefficient in Italy. It has nevertheless remained significant only in Spain. Turning now to the relationship between party switchers and leader effects, we begin with the simplest Model, which includes only as variables the leader barometer, the dichotomous variables which distinguishes between loyal voters (0) and party switchers (1), as well as the interaction term between the two. We verify that in all three countries considered the coefficient is negative, meaning that leader effects are stronger for those who have not switched parties and all coefficients are significant. Looking next at Model 2 where a variable which measures degree of liking of the incumbent party, the following occurs: Model fit improves substantially, reaching between 58% in Portugal and 81% in Italy of variance explained. In all cases the interaction. Again, in all cases the variables are negative, but remain significant in only two cases, namely in Italy and in Spain. Moving to the fully specified model we can see that the inclusion of the additional sociodemographic variables as well as one on left-right selfplacement and liking for party has the consequence that the interaction term does not reach significance, even though the coefficients in Italy, Spain and Portugal remain negative. Thus, we have created several interaction terms to measure the impact of different aspects of party identification on the strength of leader effects. We have been able to determine with some degree of certainty that leader effects tend to be higher among those with a party identification, and for party loyalties, with evidence on late deciders being more

12 mixed. What does this say about the relationship between leaders and parties in these three countries where the level and attachment to partisanship is quite differentiated? It seems to be the case that leaders are more important to those who are more attached to parties, which is a first signal that liking leaders is a phenomenon that is more prevalent among those who feel closer to the political parties, and therefore also more involved in the political system as such. On the contrary, those who have no party identification, and who switch parties are less influenced generally by leaders. In order to really be able to conclude that leader and party effects tend to go together and signify a political choice, it is important to show that it is also the more sophisticated which tend to have a greater tendency to have a party identification, and to be loyal to one party. In the next section, which will be the final section of the paper [results to be presented in São Paulo] we will test an interaction term between political sophistication and strength of party identification, party loyalty and late deciders. The relationship between political sophistication, leader effects and party attachment will show the degree to which democracy in Southern Europe is attracting those who are interested and understand politics through parties and leaders. Conclusions: Even though individual characteristics have often been considered one of the crucial research questions in the leader effects literature, seldom has the dealigned electorate, and its relationship with leader effects been examined. We do so here, looking at non-party identifiers, late deciders and party switchers in three democracies, Portugal, Spain and Italy. We started by showing that in terms of party identification, there are substantial differences between these polities. Whereas Portugal and Spain share similar amounts of electors without a party identification, in Italy most still profess to have a party identification. This is as expected for despite recent shifts, most studies on Italian political behavior emphasise the continuities in political links between electors and parties which may be new, but represent old and very important cleavages in political society. Portugal, in this respect stands out due to the high number of weak identifiers which exist in the country. Thus, it seems that the Portuguese have the weakest relationship to parties of the three countries analysed. Concerning the models and the results which were run, the objective was to show the robustness of findings concerning an interaction term between one facet of party attachment

13 (party identification, vote decision, and party switching) and leader effects. The most robust finding concerns the positive relationship between degree of party identification and the importance of leader effects. In the majority of cases for the countries concerned, these interaction terms remained significant, in each of the Models which were tested. The Models run with the interaction between party switching and leader effects came next in terms of robustness. Across the Models presented, coefficients for this interaction term were always negative, signifying that those who do not switch parties are more prone to leader effects, even though not all coefficients were significant. Notably, in the fully specified model none of the coefficients are significant. The relationship between late deciders and leader effects seems the least stable with coefficients in Portugal exhibiting different signs from Model to Model, and often not significant in the three Models presented. It seems then that leaders are more important for those individuals whose degree of party attachment is greater, as shown by their stated party identification or loyalty in voting. We still have to investigate the relationship between partisans and political sophistication in order to be able to claim that attachment to leaders is not a sign that individuals are increasingly deciding on the basis of beauty contests rather than on political issues. But the evidence which we have gathered here seems strongly suggestive of that already, and is therefore quite promising for the future of consolidated democracies.

14 FIGURES AND TABLES Figure 1: The Strength of Party Identification in Southern Europe Figure 2: Late Deciders in Southern Europe Figure 3: Party Switchers in Southern Europe

15 Table 1: Correlations between Strength of Party Id, Swing Voters and Late-Deciders. Italy 2006 Portugal 2005 Swing Voters and Strength of Party Id.217** 162** (1014) (1567) Late-Deciders and Strength of Party Id.355**.172** (1337) (2048) Swing Voters and Late-Deciders.250**.284** (1244) (1630) Spain ** (3991).248** (4922).279** (4036) Table 2: Model I - Strength of Party Identification Independent Italy Variables 2006 Portugal 2005 Spain 2008 Incumbent Leader.68*** (.15) Strength Pid -.50 (.32) Inc. Leader *.13** Strength Pid.26*** (.07) -.52** (.19).13*** (.03).96 (.10) -.82 (.27).11 (.04) Intercept -3.79*** (.81) -2.10*** (.40) (.60) Pseudo- R N

16 Table 3: Model II Strength of Party Identification Independent Variables Italy 2006 Portugal 2005 Spain 2008 Incumbent Leader.21 (.16) Inc. Party Like.84*** Strength Pid -.39 (.33) Inc. Leader *.11* Strength Pid.42*** (.03) (.21).13*** (.04) -.70***.92*** -.66** (.29).08* Intercept -5.81** (.89) -3.33*** (.46) -3.26*** (.70) Pseudo- R N Table 4: Model II Strength of Party Identification Independent Variables Italy 2006 Portugal 2005 Spain 2008 Education.03 (.23) Church.24* (.14) Union.11 (.44) LRSP 1.44*** (.16) Inc. Party Like.48** (.15) Economy.23 (.24) Incumbent Leader.17 (.25) Strength Pid -.31 (.56) Inc. Leader *.16 Strength Pid.10.11* -.84** (.28).53***.47*** (.04).32** (.10).01 (.10) -.41 (.25).10* -.10* (.17) -.72*** (.04) -.67*** (.09).78*** -.54* (.31).04 Intercept -13.5*** (1.97) -7.0*** (.81) 1.63* (.84) Pseudo- R N

17 Table 5: Model I - Late Deciders Independent Italy Variables 2006 Incumbent Leader 1.01*** Late Decision -.01 (.41) Inc. Leader * Late -.01 Deciders (.07) Portugal ***.32** (.12) -.06** (.02) Spain *** 3.11*** (.63) -.46*** (.10) Intercept -4.97*** (.60) -3.82*** (.67) -7.50*** (.31) Pseudo- R N Table 6: Model II Late Deciders Independent Italy Variables 2006 Incumbent Leader.86*** Inc. Party Like.40** (.13) Late Decision -.20 (.45) Inc. Leader * Late.04 Deciders Portugal ***.21** (.07).03.02* (.01) Spain *** -.66*** 2.98*** (.67) -.43*** Intercept (.70) -4.22*** (.34) -5.18*** (.42) Pseudo- R N

18 Table 7: Model III Late Deciders Independent Italy Variables 2006 Education.09 (.18) Church.04 Union -.18 (.34) LRSP 1.21*** Inc. Party Like.50*** (.12) Economy.15 (.21) Incumbent Leader.25 (.18) Late Decision -1.20** (.59) Inc. Leader * Late.15 Deciders Portugal * -.72** (.27).53***.45*** (.04).25** (.10).30*** (.07).22 (.16) -.05 (.03) Spain ** (.17) -.70*** (.04) -.60*** (.09).90*** 2.06** (.73) -.32** Intercept (1.41) (.69).14 (.59) Pseudo- R N Table 8: Model I - Party Switchers Independent Italy Variables 2006 Incumbent Leader 1.10*** Vote Switchers.83 (.60) Inc. Leader * Vote -.24** Switchers Portugal *** (.03).36 (.35) -.16** Spain *** 2.05** (.62) -.39*** (.10) Intercept -5.47*** (.41) (.19) -7.60*** (.34) Pseudo- R N

19 Table 9: Model II Party Switchers Independent Italy Variables 2006 Incumbent Leader.54*** (.09) Inc. Party Like.92*** (.09) Vote Switchers.78 (.67) Inc. Leader * Vote -.22* Switchers (.12) Portugal ***.21** -.37* (.17).02 (.01) Spain *** 1.21*** 2.08** (.65) -.41*** (.10) Intercept (.55) -4.15*** (.35) -5.28*** (.45) Pseudo- R N Table 10: Model III Party Switchers Independent Italy Variables 2006 Education -.08 (.21) Church -.11 (.12) Union.02 (.40) LRSP 1.22*** (.13) Inc. Party Like.59*** (.14) Economy.15 (.24) Incumbent Leader.43** (.14) Vote Switchers -.20 (.96) Inc. Leader * Vote -.03 Switchers (.17) Portugal * (.07) -.70** (.28).55***.47*** (.04).27** (.10).24*** -.30 (.48) -.05 (.09) Spain (.18) -.70 (.04) -.62 (.09) (.73) -.27 Intercept *** (1.47) -8.23*** (.67).25 (.62) Pseudo- R N

20 Bibliography: Bellucci, P., The parliamentary election in Italy, April 2006, Electoral Studies, Volume 27, Issue 1, March 2008, Pages Berglund, Frode, Holmberg, Soren, Schmitt, Hermann & Thomassen, Jacques (2005) Party identification and party choice, in: Jacques Thomassen (ed.) The European Voter (Oxford: Oxford University Press), pp Booroah, 2002, Logit and Probit, Ordered and Multinomial Models, London, Sage. Brettschneider, Neller, and Anderson, 2006, Candidate Images in the 2005 German National Elections, in German Politics, vol. 15 nr.4, pp Campbell, A., (1966), Elections and the Political Order, NY: Riley. Curtice, J. and Holmberg, S., (2005), Party Leaders and Party Choice, in Thomassen, The European Voter, Oxford: OUP. Dalton, R.J., (2009), Review Symposium: Parties, Partisanship and Democratic Politics, in Journal of Politics, Sept. vol.7, no.3, pp Dalton, R. J., (2002), Citizen Politics, Ney York, Chatham House. Dalton, R. J., Wattenberg, M., McAllister, I., (2002), Democracy and Mass Partisanship in Advanced Industrial Societies Paper presented to the Fulbright Brainstorm Conference, Elections and Democracy, Lisbon, 1-2 February. Field, B., (2009), The general election in Spain, 2008, Electoral Studies, volume 28, issue 1, March 2009, pp Freire, e Lobo, (2006), Election Report on 2005 Legislative Elections, in West European Politics, vol.29, no.3, pp Jalali, e Lobo, (2006), Local and Presidential Elections in Portugal, in South European Society and Politics, vol.11, no.2. King, A., (2002), Leaders personalities and the outcome of democratic elections, Oxford: OUP. Kroh, 2004, Personal Voting: Individual and Contextual Determinants of Personal Voting, mimeo. Lobo, M.C., (2006), Short-term voting determinants in a young democracy: Leader effects in Portugal in the 2002 legislative elections, in Electoral Studies, 25, pp Lobo, M.C:, (2004), O impacto dos líderes partidários: uma escolha entre candidatos pouco populares, in Freire, A., Lobo, M.C., Magalhães, P., Portugal a Votos Lisboa: ICS.

21 McAllister, I. (1996), Leadership, in LeDuc, L., Niemi, R., and Norris, P., Comparing Democracies, London: Sage. Mughan, A., 2009, Partisan Dealignment, Party Attachments and Leader Effects, in JEPOP, Vol. 19, No. 4, Scarrow, (2000), Parties without members?, in Dalton, R., and Wattenberg, M., Parties without Partisans, Oxford, OUP. Schmitt, H., and Holmberg, S., (1995), in Klingemann, H-D. and Fuchs, D., Citizens and the State, Oxford: OUP.