Party Organisational Change and Leader Effects on Voting Behaviour: Intra-Party Democracy, Leadership Strength and the Electoral Impact of Leaders

Transcription

1 Party Organisational Change and Leader Effects on Voting Behaviour: Intra-Party Democracy, Leadership Strength and the Electoral Impact of Leaders Frederico Ferreira da Silva European University Institute ECPR Joint Sessions April 2017, Nottingham, UK <First draft, please do not cite> Word count: 6882 Abstract Recent studies indicate that changes in party organisation have been leading to more internally democratic parties, where more participatory and deliberative methods are being progressively implemented, particularly with regard to the appointment of party leaders. In parallel, research also highlights that political parties have become more presidentialized, with an increasingly prominent role of the leader within the party. This paper analyses the impact of these recent trends in party organisation intra-party democracy and increasing leadership power on the extent to which leaders have an effect on voting behaviour. In doing so, it analyses the interconnection between the electoral and the partisan consequences of the personalisation of politics thesis. First, it examines the relationship between the candidate selection method and the impact of voters evaluations of party leaders on voting behaviour. Specifically, is focuses on the possible moderator role of the candidate selection procedure on the leader s electoral impact. It explores the hypothesis that in parties with more open candidate selection procedures (e.g., open to all party members; open primaries), where leaders are selected by a broader selectorate, which empowers them and provides them with a broader support base among the electorate, leaders electoral potential is enhanced. In addition, the relationship between the leadership power, as granted by party statutes, and leader effects on voting behaviour is explored. In particular, it is analysed to what extent a stronger formal power within the party organisation corresponds to stronger leadership effects on voting behaviour. The paper is a comparative study on Western democracies, combining individual-level data from the Comparative of Study of Electoral Systems with contextual data recently made available by the Political Party Database Project. 1

2 Introduction The personalisation of politics has set party leaders as crucial actors in contemporary politics. Leaders play a central role in campaigning, are the main focus of media attention and primary channel of political communication. Electorally, voters evaluations of political leaders have been demonstrated to have a growing relevance in electoral outcomes (Garzia, 2013), in a trend of increasing importance of short-term determinants of voting behaviour. This process towards candidate-centered politics (Wattenberg, 1991) puts leaders as key electoral actors in contemporary democracies and has been described as the electoral face of the presidentialization of politics (Poguntke and Webb, 2005). However, leaders role has also become more dominant within political parties. The transformations undergone by Western political parties in the last decades entailed organizational changes that empowered leaders at the cost of intermediate party strata, reinforced their autonomy and emphasized their role as visible representatives of the parties. These changes happened in closed relation to the growing electoral role of leaders. For this reason, besides the electoral face, the same authors have also referred to a party face of the presidentialization of politics, accounting for the increased influence of leaders on party decision-making. Among other aspects, this is characterized by rule changes empowering the leader, growing autonomy in the draft of party programmes, attempts to bypass sub-leader strata and the institutionalization of direct leadership elections. The latter is part of a relatively recent trend towards intra-party democracy as an attempt to reconnect voters with their representatives. Despite a few important works (Blondel et al., 2010; Passarelli, 2015; Samuels and Shugart, 2010), the party face has been the most understudied dimension of the presidentialization of politics. In particular, its interconnection with the remaining dimensions is yet to be researched. This article establishes a link between the party and electoral face of the presidentialization of politics, with the intent to investigate to what extent the presidentialization of political parties contributes to the increased electoral role of party leaders as short-term determinants of voting behaviour. To what extent do presidentialized parties produce leaders with stronger effects on voting behaviour? Do more open selection procedures enhance the electoral impact of leaders? Are more powerful leaders more determinant in the voting calculus? On the one hand, answering these questions would allow for a better comprehension of how the changes affecting Western political parties in the last decades contributed to the personalisation of politics. Given that the theory names it as a fundamental cause for the 2

3 growing importance of leaders, it would be relevant to have a better understanding of the mechanisms through which party change produced more electorally relevant leaders. On the other hand, it is relevant from the point of view the democratization of selection procedures and, more generally, of intra-party democracy. If leaders who are elected by broader selectorates have then stronger effects on voting behaviour, it could be argued that voters may feel more represented and therefore consider them more thoroughly in the voting decisions. The answers to these questions carry implications also regarding parties electoral strategy. If more open selection procedures produce stronger electoral leaders, this particular measure of intra-party democracy may be advantageous for the maximisation of the leader s electoral potential. Moreover, as office-seeking is argued to be the fundamental goal of contemporary political parties (Katz and Mair, 2001), empowering leaders within the party structure may be functional to the party as a whole from an electoralist perspective. In this sense, leadership can be seen as a growing resource on which parties can capitalise. Emphasising the role of the leader and widening the selectorates may then be conceived as attempts to maximise the electoral utility of leaders by engaging citizens at large and thus enriching the support base of the leader and the party. This article attempts to provide answers to these questions and clarify the interrelationship between the electoral and the party face of the presidentialization of politics. It uses individuallevel data on voting behaviour drawn from the Comparative Study of Electoral Systems and party-level contextual data recently made available by the Political Party Database. The article proceeds as follows: in the first section, the theoretical framework on the basis of the hypotheses is set out; this is followed by a presentation of the data sources and considerations on the methods used in the empirical analysis; next, follows a presentation and discussion of the results; finally, the last section elaborates on the conclusions of the study and its potential contributes. Party change, the role of party leadership and leader effects The transformations undergone by Western political parties in the second half of the last century, resulted in the decline of denominational and mass parties and in the emergence of new party types with distinct characteristics and organizational features. The cleavages in the origin of the long-term attachments of voters to denominational and mass parties have largely eroded, depriving parties of much of their traditional electoral base. In such a context, vote choice is no longer a predisposition constrained by the encapsulation of the mass of the 3

4 electorate into one of the subcultural groups that the parties represent (Katz and Mair, 1995: 7) but it is much more subject to individual choice. Other advances such as economic growth and the development of the Welfare State, depoliticized many of the traditional cleavages; individualisation further accentuated the trend towards more volatile, independent electorates, who do not blindly pledge long-term allegiance to the party; and changes in mass communication made it more attractive for parties to rely on their leaders as key agents of political communication. Consequently, new party types have emerged in reaction to these changes and a vast body of research has developed accounting for new typologies of contemporary Western political parties. The transformations undergone by these parties comprised not only ideological mutation but, importantly, also organizational changes altering intra-party balances of power between leaders and members (Webb, Poguntke and Kolodny, 2007: 79). Early works on the organizational transformations provoked by the changes towards contemporary party types highlighted the shift of power towards the leader, who concentrates more resources and enjoys a higher degree of autonomy from middle-rank members in the decision-making process (Kirchheimer, 1966; Panebianco, 1988). Notwithstanding the important distinctions between them, whether we refer to Kirchheimer s (1996) catch-all party, to Panebianco s (1988) electoral-professional, to Katz and Mair s (1995) cartel party, or any of the several types comprised within Gunther and Diamond s (2003) electoralist genus, contemporary types of political parties emerged after the age of the mass or denominational party types are organizationally thin, maintaining a relatively skeletal existence except at election time ( ), use modern campaign techniques stressing television and mass-communication media over mobilization of party members in order to win votes ( ) [and are] characterized also by the large importance that the party leadership acquires for election purposes (Lobo, 2014: 364). These developments in political parties have contributed to, over the closing decades of the last century, turn leaders into the central political figures in contemporary parliamentary democracies (Cross and Blais, 2012: 1): technological innovations in mass communication, particularly with the advent of television, set leaders as the visible representatives of parties and the main vehicles of political communication; voters assessments of party leaders have become an increasingly important determinant of voting behaviour not only in presidential but also in parliamentary Western democracies (Garzia, 2014); party organisational change increased the dominant role of leaders within parties, that are now steered by the leaders political strategy, as intermediate structures have lost power and influence (Pilet and Cross, 2015: 4). 4

5 This process has recently been intensifying and, as noted by Cross and Pilet, (2014: 2), while political parties have long operated at the centre of the democratic politics, the role played by their leaders has expanded in the last years in a trend towards a presidentialization of politics (Poguntke and Webb, 2005a). According to this thesis, presidentialization denominates a process by which regimes are becoming more presidential in their actual practice without, in most cases, changing their formal structure, that is, their regime-type (Poguntke and Webb, 2005b: 1). A key feature of this trend lies in the growing power of political leaders in parliamentary regimes over the last decades, resulting in increasing leadership power resources, autonomy within the party and an extensive personalization of the electoral process. These features result in what the authors have named as the three faces of presidentialization: a) the executive face, corresponding to the growing independence and accumulation of power resources at the disposal of the head of government; b) the party face, involving a structural shift in intra-party power to the benefit of the leader, who enjoys further autonomy from the party s middle-level elites and is freer to pursue a more personalized form of party leadership; c) the electoral face, translating into a higher consideration of party leaders in the voting decision process, of which increasingly candidate-centered campaigns are an example. While the authors provide preliminary evidence of a trend towards a presidentialization of politics in all three realms, the study has been particularly successful in setting the agenda for further research taking place in the next decade. Research steaming from this theory has been especially prolific in providing evidence concerning the electoral face of the presidentialization, with numerous studies confirming the impact of voters assessment of party leaders in their voting decisions (Garzia, 2013, 2014; Mughan, 2015; Lobo and Curtice, 2015). This line of research fits within what is generally described as the personalisation of politics 1, defined as the process in which the political weight of the individual actor in the political process increases over time, while the centrality of the political group (i.e., political party) declines (Rahat and Sheafer, 2007: 65). Notwithstanding the upmost relevant contribution of this stream of research in demonstrating the paramount role leaders nowadays play in the media and the electoral arena, this research has been carried in disconnection with the increasing importance of leaders within the political parties. A meritorious exception is the study of Lobo (2008), which demonstrated that party leaders of electoralist parties have a stronger impact on vote choice, thus providing evidence of the relationship between party type and leader effects. However, the possible interrelationship between the party and the electoral 1 For a distinction between the presidentialization of politics and the personalisation of politics, please see Passarelli (2015). 5

6 faces of the presidentialization of politics remains largely unstudied. Moreover, theory on the personalisation of politics indicates as one of its main causes precisely the abovementioned transformations in Western political parties in the second half of the last century, that are in the origin of the party face of the presidentialization of politics. Therefore, there are strong theoretical reasons to expect an interdependence between the electoral and the party face of the presidentialization of politics. More specifically, it raises the question of whether a stronger role within internal party organization corresponds to a stronger electoral weight of the political leader. If such proposition is to be confirmed, we should observe stronger leader effects where we also find indicators of the party face of presidentialization to be more pronounced. Arguably one of the most relevant indicators of an intra-party power shift towards the leader, according to Poguntke and Webb (2005b: 20), is the institutionalization of direct leadership elections. In fact, an increasingly relevant stream of literature has developed in recent years, accounting for the democratization of leader selection procedures and confirming this trend. Many authors found evidence of a widespread trend towards more inclusive leader selection procedures in Western parties, where the leader is appointed by broader selectorates such as all party members, or even extended to supporters in open primaries (LeDuc, 2001; Sandri, Seddone and Venturino, 2015). Cross and Pilet (2014: 6-7) identify three main factors in the origin of this democratizing trend. First, as parties adapt to changes in their surrounding context, they have placed leaders at the forefront as their visible representatives and broadened their selection procedures in association with the overall personalisation of politics trend. Second, the proliferation of democratic values, as an outcome of increased educational levels, called for more inclusive means of political participation and transparency in politics, putting pressure on parties to be more responsive and accountable to both their members and citizens through the promotion of intra-party democracy. Third, the democratization of political parties is largely a reaction to the trends of partisan detachment and overall disengangement with politics as a consequence of the erosion of the long-term aligning cleavages that once structured partisan bonds. This translated into a decline of party identification and membership to which parties have responded by granting their members a stronger role in the selection of the leader. Therefore, the trend towards more inclusive leader selection procedures can be understood as a reflection social developments, such as liberalization and the personalization of politics, combined with the decline of ideology and party identification (Barnea and Rahat, 2007: 382). In a context of membership decline and mediatisation of politics, emphasising the leader s role was also functional for political parties, as leaders enjoy a capacity to appeal to the electorate 6

7 at large, an electorate made up of voters who were learning to behave more like consumers than active participants as a result of individualization and dealignment (Katz and Mair, 1995: 7). By privileging the personalities of leaders over programmes or ideology, an effective media strategy with telegenic leaders significantly reduces the utility of mass membership as a vehicle for electoral mobilization (Gunther and Diamond, 2003: 168). Facing the decline in party membership across Western contemporary parties (van Biezen, Mair and Poguntke, 2011; Scarrow, 2002), it became a crucial alternative to the role traditionally played by membership. This connects to broader transformations on what party membership currently entails. Although member rights continue to exist, the fact that these parties try to appeal also to a wider range of supporters makes leaders not only accountable to party members but to the whole wider electorate. As a consequence, on the one hand, at times members are empowered by specific measures destined to enhance their role within the party s organisation such as plebiscitary forms of intra-party democracy. On the other hand, because they share their some participation rights in party activities and decisions with non-member supporters, the distinction between them becomes somewhat blurred (Young, 2013). Furthermore, members also lose much of their influence on the elaboration of policy programmes, as these become eminently defined by the leadership, which imposes a top-down authority flow (Katz and Mair, 1995: 7-8, 20-21). From the perspective of internal power distribution in the party s everyday activity, there seems to be a trade-off between enlarged selectorates and concentration of power in the hands of the leadership: while leaders become more accountable to party members or supporters when it comes to the particular moment of appointment or removal from the leadership, they also become more independent to steer the party at their will during their mandates. This is an important consequence linked to another indicator of presidentialization at the party face: the leadership increasingly attempts to by-pass sub-leader or activist strata of the party by communicating directly with the grass roots regarding strategic questions (Poguntke and Webb, 2005b: 20). The same had been noted by Mair (1997: ), who argues that the cartel party leaders formally empower the ordinary party members, or even a broader range of supporters, in order to gain the necessary autonomy. Likewise, Katz (2001: 281) highlights that the increase in the nominal power of the rank-and-file members comes at the expenses of the middle-level party structures, which became deprived of most of its ability to control the leadership, instead of the leadership itself. More recently, Cross and Pilet (2014: 2) subscribed the same idea, arguing that parties controlled by middle-level elites have lost ground. Thus, it seems that paradoxically, a democratization of leader selection procedures, often intended to overcome an intra-party democratic deficit by giving voice to rank-and-file members and reconnect citizens 7

8 with the political process while refraining oligarchic tendencies within the parties (Kenig, 2008: 241), does not necessarily correspond to a more disperse power distribution within the party and can also (un)intentionally promote more concentration of power at the disposal of the leadership. In such a setting, party leaders benefit from a greater pool of resources to exercise their influence and affirm their predominant role within the party. This could have repercussions in voters perceived role of leadership and in the attribution of relevance to leader evaluations in the voting calculus: if leaders become more important in determining the party s fate, they are an increasingly relevant factor to consider in anticipating the party s policy-decisions and overall performance. Also, more powerful leaders have more resources to carry an influential leadership mandate and consequently are more likely to be focused in the media, gaining visibility and a more noticeable profile. They stand out from the party more clearly and thus are more salient to voters. For these reasons, when studying the presidentialization at the party face and its possible effects on voting behaviour, it is also important to take into account the powers at the disposal of the party leader another of the indicators proposed by Poguntke and Webb (2005b: 20). Considering the discussed theoretical framework, the following hypothesis can be formulated regarding the role of the party face of the presidentialization on leader effects on voting behaviour. H1: Leaders appointed by broader selectorates benefit from a wider support basis and legitimacy that boosts their electoral effect H2: Leaders that concentrate a larger set of powers at their disposal, as granted by party statutes, enjoy further autonomy within the party and are perceived by the electorate as more influential, therefore impacting voters behaviour more decisively Data and Methods In order to test the hypotheses, both data on individual-level voting behaviour and on party characteristics is required. The former type of data is provided by the Comparative Study of Electoral Systems (CSES) dataset, which compiles post-electoral studies over a multitude of countries, harmonized according to a common module. Modules 3 ( ) and 4 ( ) of CSES were used for this study because they are the more recent ones. Considering that the discussed party changes, particularly regarding the leader selection procedure, happened essentially in the last two decades, this is the key period of interest for this study and thus older 8

9 modules were not considered. In addition to CSES countries, some countries were added through their respective national election studies, whenever data on the two party-level variables of interest was available. This resulted in a total of 20 countries analysed. 2 The main advantages of the CSES data are its reliability and comparative nature, due to the application of the same set of questions in different contexts. Nevertheless, it has the shortcoming of containing very little information at the party level, which was fundamental to the purpose of this study. Therefore, in the elaboration of the dataset used in this article, the CSES data was complemented with the recently made available data of the Political Party Database (PPDB). This useful resource provides contextual data at the party level for 122 parties in 19 countries, including information on party rules, resources, party membership, intra-party democracy and party organization (Poguntke, Scarrow and Webb, 2016). The PPDB provided the necessary data on the power resources at the disposal of the party leader. This variable was constructed as an index of leadership power, corresponding to the sum of number of powers held by the party leader 3, theoretically ranging from 0 to 9 but in practice ranging from 1 to 7. Another type of data drawn from the PPDB refers to the leader selection procedure. This variable was coded as follows: 1. Single individual; 2. Party elite; 3. Party delegates; 4. Party members; 5. All voters. However, there are no cases in this study s sample where the leader is appointed by a single individual. One limitation of the PPDB is the relatively large amount of missing data on some variables. The variables of interest of this study are no exception data on the leader selection procedure was not always available in the dataset. This issue was tackled by using data on the leader selection procedure from the growing body of literature that has been recently published on the topic (Bucur and McMenamin, 2015; Pilet and Cross, 2014; Sandri, Seddone and Venturino, 2015). Some specificities of the CSES dataset, inherent to its comparative nature, also impose some constraints on how to analyse parties, their leaders and vote choice. Given that parties, leaders and their number vary across time and countries, it would not be possible to analyse their impact 2 Australia (2013), Canada (2011), Israel (2013), Finland (2011), France (2012), Germany (2013), Iceland (2013), Ireland (2011), Denmark (2011), Italy (2013), the Netherlands (2010), United Kingdom (2010), Poland (2011), Spain (2011), Austria (2012), Sweden (2014), Czech Republic (2013), Norway (2013), Portugal (2015) and Hungary (2006). 3 Leader may help select deputy leader; may summon party officials; may summon party congress; may attend party executive; may attend party congress; may appoint at least one member of party executive; must consent to coalition agreements; is designated party s external representative ; is expressly accountable to party congress. Each right is coded as 1 and the absence of the right is coded as 0, except for is expressly accountable to party congress, where the logic is reversed. 9

10 on voting behaviour comparatively without some important transformations on the data matrix. A frequently used approach to deal with this issue is to stack the data matrix according to the number of parties/leaders, so that each line corresponds to a party-respondent combination (see van der Eijk and Franklin 1996). The result is that each respondent appears in the dataset as many times as the total number of parties/leaders running for election in that country in that election year. It follows that the dependent variable vote choice, instead of consisting on a nominal configuration of the several parties running for election in a country in an election-year (what would be incomparable across countries), becomes a binary choice (0: Did not vote for this party; 1: Vote for this party), repeated as many times per respondent as many parties run for that election. In short, the dependent variable is a dummy capturing the combination of party-voters choice. Across that given number of respondent-observations, individual-level constant variables remain constant (i.e., age, etc.) and party-level variables vary (i.e., leader; selection procedure, etc.). At the individual-level, the following variables were included in the models: age is measured in years; gender (0. Female; 1. Male); education (0. None; 1. Primary education; 2. Secondary education; 3. Tertiary education); organizational membership, built from a compound index of the following dichotomous questions (0: No; 1: Yes): Are you a member of a union?, Are you a member of a business or employers association?, Are you a member of a farmers association?, and Are you a member of a professional association? ; party identification (-1. Identifies with another party; 0. No party identification; 1. Identifies with this party); political knowledge, built out of three political knowledge questions (0. No correct answers; 1. One correct answer; 2. Two correct answers; 3. All correct answers); retrospective sociotropic economic evaluations (1. Much better; 2. Better; 3. Stayed the same; 4. Worse; 5. Much worse); and a leader likeability scale (0. Strongly dislikes; 1. Strongly likes). The latter is the key variable of interest, which cross-level interacted with the selection procedure and the leadership power index provides the test to the previously formulated hypotheses. Results Given the hierarchical structure of the data, especially considering its restructuration to a stacked data matrix, a multilevel approach was considered as most adequate. The stacked dataset carries a number of consequences from the data estimation point of view: first, it multiplies the number of actual respondents by the number of parties in each country and, in doing so, substantially increases the N, which is computationally more demanding; second, 10

11 because observations are nested in individuals, the number of second-level units becomes the actual number of unique respondents, which is rather high and demands more from the data estimation procedure. In the model estimation, these constraints had to be taken into account. To reduce the complexity of the models, the choice was to test each hypothesis in an independent model. Besides making the estimation more fluid, it carries the advantage of impeding a reduction of the sample size due to missing cases. As was previously discussed, the PPDB dataset s variables have a considerable amount of missing values. Although this limitation was partially solved by using secondary sources whenever available, the joint estimation of a model including the two PPDB party-level variables would result in the exclusion of cases due to missing values regarding one of the variables that might be valid and pertinent to test the remaining hypothesis. Therefore, two separate models were estimated: the first tests whether leaders selected in more open selection procedures are more determinant on voters electoral decisions (H1); the second tests whether leaders who concentrate more power resources at their disposal have a stronger effect on voting behaviour (H2). A random intercept model was estimated to test each of the hypotheses. After careful consideration, a two-level model was preferred. Although there could be a case for the inclusion of a third level, nesting observations and individuals in countries, an estimation of such model demonstrated that there was virtually no variance to be explained at the country level (Appendix 1 and 2). In light of this finding, it was preferred to reduce the complexity of the model and, as much as possible, keep it parsimonious. A case could also be made for random coefficient models the impact of the party-level variables could vary across countries, for example. However, the several random coefficient models tested revealed extremely low random coefficients, and thus they were also disregarded. It is also worth highlighting that even the random intercept models for all two hypotheses show almost no variance in the second-level intercepts, reinforcing the previous conclusion that, although the data structure is theoretically hierarchical, the hierarchical structural does not receive much empirical sustentation. Therefore, a fixed effects models would perhaps be as adequate. Nevertheless, a more conservative approach prevailed in favour of random intercept models. Finally, given the little differences found in results estimated through logistic and linear probability models, that extreme probabilities are not being modelled and thus are quite unlikely to be greater than 1 or lesser than 0, and that the models are particularly demanding, a linear probability function was preferred over a logistic one. This facilitates both the model estimation and the interpretation of the results, while not compromising the accuracy of the estimates (Hellevik, 2009). 11

12 The first model tests the possible mediating impact of the leader selection procedure on the impact of voters assessments of party leaders in their voting behaviour. In line with the theory, the expectation was that more open selection procedures would produce leaders with a broader basis of support and legitimacy, which would in its turn reflect on their electoral impact. Table 1 - The moderating effect of the selection procedure on leader effects on vote choice: random intercept model COEF. S.E. 95% CONF. INTERVAL AGE.001*** GENDER EDUCATION -.023*** ORG. MEMBERSHIP -.024*** PID.284*** POLITICAL KNOWLEDGE ECONOMY.007*** LEADER EVALUATIONS.035*** SELECTION PROCEDURE PARTY DELEGATES.021** PARTY MEMBERS.026** ALL VOTERS.048*** LEADER*SELECTION PARTY DELEGATES PARTY MEMBERS.011*** ALL VOTERS.010*** CONSTANT.136*** VARIANCE(CONSTANT) VARIANCE(RESIDUAL) AIC BIC N(OBSERVATIONS) N(INDIVIDUALS) * p<.05; **p<.01; ***p<

13 The results, presented in Table 1, confirm a positive, statistically significant interaction effect between the openness of selection procedure and leader evaluations on vote choice. This provides evidence in support of the proposition of H1, confirming that leaders elected by broader selectorates have a stronger impact on voters decisions. However, as can be verified through the analysis of the coefficients, the openness of the leader selection procedure has a positive moderator effect on leader evaluations only when the selection procedure is open to, at least, all party members. Selection procedures open only to party delegates do not have a significant interaction effect with leader evaluations. Thus, it is confirmed that leaders of parties with the two most open modalities of leader selection have a stronger effect on voters electoral decisions. It is also interesting to ascertain if there is any practical difference between selection procedures open to all party members or to all voters in open primaries. Looking at the coefficients, their effect size seems essentially the same, but given the difficulties in interpreting interaction coefficients, the best way to compare their effect is to plot the marginal effects. Figure 1 - Interaction between the leader selection procedure and leader evaluations. Marginal effects on vote choice. The marginal effects confirm that there is virtually no distinction between selection procedures open to all party members and open to all voters with regard to the effects of leader evaluations 13

14 on vote choice. Although open primaries have the strongest effect, the comparison with a selection from all party members reveals very small differences and fall within the confidence intervals. Furthermore, as had become clear before, the results are distinguishable according to two main groups (lower effects, not significant: party elite and party delegates; higher effects, significant: party members and all voters), so the qualitative jump occurs when the selection procedure moves from being restricted to party delegates to being extended to all party members. At the top of the leader evaluations scale, there is an increase of about 15 percentage points on the probability to vote for the party if the leader was selected through one of two more open procedures, compared to the two more closed ones. Therefore, there seems to be an incentive for parties to open the selection procedures to, at least, all party members, in order to maximize the electoral potential of their party leaders. The second hypothesis established a relationship between the leader s power resources and her impact on voting behaviour. The expectation was that leaders who have more powers at their disposal as granted by party statutes, have more leverage to exercise their influence, play a more determinant role within the party, and therefore are perceived as more influential by voters, who will then consider them as more important factors in their voting calculus. Table 2 - The moderating effect of leadership power on leader effects on vote choice: random intercept model COEF. S.E. 95% CONF. INTERVAL AGE.001*** GENDER EDUCATION -.026*** ORG. MEMBERSHIP -.016*** PID.300*** POLITICAL KNOWLEDGE.006*** ECONOMY LEADER EVALUATIONS.017*** LEADERSHIP POWER -.010*** LEADER*L.POWER.004*** CONSTANT.233*** VARIANCE(CONSTANT) VARIANCE(RESIDUAL)

15 AIC BIC N(OBSERVATIONS) N(INDIVIDUALS) * p<.05; **p<.01; ***p<.001 The results presented in Table 2 confirm the expectations formulated in H2. A positive and significant interaction effect was found between the number of powers granted to the leader and leader evaluations. Therefore, more powerful leaders have stronger effects on vote choice. Again, the interaction effects were plotted for a better interpretation (Figure 2). Figure 2 - Interaction between leadership power and leader evaluations. Marginal effects on vote choice. There is an incremental increase on leaders effects on vote choice across the leadership power index which although ranges from 1 to 9, empirically never goes beyond 7. At the top of the leader evaluation scale, there is a difference of about 20 percentage points on the probability to vote for the party between a leader with only one power at her disposal and a leader with seven powers at her disposal. 15

16 Together, the results from both models confirm the two hypotheses set out for this article. This provides further evidence that the transformations in contemporary Western political parties, by emphasising the role of the leader within the party, also enhance leaders electoral potential. Thus, the personalisation of politics at the electoral level seems to be favoured by both more open selection procedures and a concentration of power on the leadership of the party. Conclusions The main goal of this article was to assess to what extent the presidentialization of political parties through the accumulation of power in the leadership and broader leader selection procedures contributes to higher leader effects on voting behaviour. The results from the empirical analysis corroborated the hypotheses. Leaders who have a higher number of powers at their disposal, as granted by party statutes, have stronger effects on voting behaviour. Also, leaders selected from more open selection procedures have a higher impact on vote choice. However, there is no statistically significant difference between leader effects for a leader selected by all party members and a leader selected through open primaries. To be sure, once the selection procedure is opened to all party members it faces a ceiling effect and there is no distinguishable effect regarding a selectorate composed of all voters. Hence, from a purely utilitarian electoral point of view, there is no incentive for parties to implement open primaries in order to maximise the electoral potential of the leader. The conclusions of this study provide evidence of the interdependence between the party and the electoral faces of the presidentialization of politics. As leaders reinforce their dominant position within political parties, they are able to more effectively capitalise on their electoral potential. This corroborates the established theoretical argument that transformations in Western political parties, by emphasising the role of the leader within new party types, contributed to the development of the personalisation of politics, namely through higher leadership effects on voting behaviour. In the present context of the declining role of party attachments, as leader evaluations become an increasingly important determinant of voting behaviour, the means through which leaders can potentiate their electoral gains are a strategic aspect to be considered by office-seeking political parties. Moreover, the fact that voters are sensible to differences in power across leaders is significant of the relevance they attribute to the role of party leadership. Also, the fact that they attribute a stronger electoral weight to leaders appointed by broader selectorates suggests that they recognize higher legitimacy and representational potential to these leaders. 16

17 Yet, this does not necessarily mean that leaders have become undisputed rulers higher intraparty democracy may also increase leaders accountability to their selectorates and facilitate their removal when underperforming. Less supported by the allegiance of middle-level layers, leadership survival is now more contingent on the support from the selectorate and thus it is also more vulnerable to growing contestation both from within and outside the political party. 17

18 Bibliography Barnea, S. and G. Rahat (2007). Reforming Candidate Selection Methods: a three-level approach. Party Politics, 13 (3): van Biezen, I., Mair, P. and T. Poguntke (2011). Going, going, gone? The decline of party membership in contemporary Europe. European Journal of Political Research, 51 (1): Blondel, J., Thiébault, J, Czernicka, K., Inoguchi, T., Pathmanand, U. and F. Venturino (2010). Political Leadership, Parties and Citizens: the personalisation of leadership. New York, NY: Routledge Bucur, C. and I. McMenamin (2015). Poland: The Presidentialization of Parties in a Young Democracy. In G. Passarelli (ed.), The Presidentialization of Political Parties: organizations, institutions and leaders. Houndmills, Basingstoke, Hampshire: Palgrave Macmillan Cross, W. and A. Blais (2012). Politics at the Centre: the selection and removal of party leaders in the Anglo parliamentary democracies. Oxford: Oxford University Press Cross, W. and J. Pilet (2014). The Selection of Party Leaders in Contemporary Parliamentary Democracies. In J. Pilet and W. Cross (eds.), The Selection of Political Party Leaders in Contemporary Parliamentary Democracies: a comparative study. New York, NY: Routdlege Cross, W. and J. Pilet (2016). The Politics of Party Leadership: A Cross-national Perspective. Oxford University Press. van der Eijk, C. and M. Franklin (1996). Choosing Europe. Ann Arbor, MI: University of Michigan Press Garzia, D. (2013). Changing parties, changing partisans. The Personalization of Party Attachments in Western Europe. Political Psychology, 34 (1): Garzia, D. (2014). Personalization of Politics and Electoral Change. London: Palgrave Macmillan. Gunther, R. and L. Diamond (2003). Species of Political Parties: a new typology. Party Politics, 9 (2): Hellevik, O. (2009). Linear versus logistic regression when the dependent variable is a dichotomy. Quality & Quantity, 43 (1): Katz, R. (2001). The Problems of Candidate Selection and Models of Party Democracy. Party Politics, 7 (3):

19 Katz, R. and P. Mair (1995). Changing Models of Party Organization and Party Democracy: the emergence of the cartel party. Party Politics, 1 (1): 5-28 Kenig, O. (2008). Democratization of party leadership selection: Do wider selectorates produce more competitive contests? Electoral Studies, 28 (2): Kirchheimer, O. (1966). The transformation of Western European Party Systems. In J. LaPalombara and M. Weiner (eds.), Political Parties and Political Development. Princeton, NJ: Princeton University Press, LeDuc, L. (2001). Democratizing Party Leadership Selection. Party Politics, 7 (3): Lobo, M. C. (2008). Parties and Leader Effects: Impact of Leaders in the Vote for Different Types of Parties. Party Politics, 14 (3): Lobo, M. C., and J. Curtice (2015). Personality Politics?: The Role of Leader Evaluations in Democratic Elections. Oxford University Press. Mair, P. (1997). Party System Change: approaches and interpretations. Oxford: Oxford University Press Mughan, A. (2015). Parties, Conditionality and Leader Effects in Parliamentary Elections. Party Politics, 21 (1): Panebianco, A. (1988). Political Parties: Organization and Power. Cambridge: Cambridge University Press Passarelli, G. (2015). The Presidentialization of Political Parties: organizations, institutions and leaders. Houndmills, Basingstoke, Hampshire: Palgrave Macmillan Pilet, J. and W. Cross (eds.), (2014). The Selection of Political Party Leaders in Contemporary Parliamentary Democracies: a comparative study. New York, NY: Routdlege Poguntke, T. and P. Webb (2005a). The Presidentialization of Politics: a comparative study of modern democracies. Oxford: Oxford University Press Poguntke, T. and P. Webb (2005b). The Presidentialization of Politics in Democratic Societies: a framework for analysis. In T. Poguntke and P. Webb (eds.), The Presidentialization of Politics: a comparative study of modern democracies. Oxford: Oxford University Press Poguntke, T., Scarrow, S. and P. Webb (2016). Party rules, party resources and the politics of parliamentary democracies: How parties organize in the 21 st century. Party Politics, 22 (6):

20 Rahat, G. and T. Sheafer (2007). The Personalization(s) of Politics: Israel, Political Communication, 24 (1): Samuels, D. J. and M. S. Shugart (2010). Presidents, Parties, and Prime Ministers: How the Separation of Powers Affects Party Organization and Behaviour. Cambridge: Cambridge University Press Sandri, G., Seddonne, A. and F. Venturino (2015). Party primaries in comparative perspective. Farnham, Surrey: Ashgate Scarrow, S. (2002). Parties Without Members?. In R. Dalton and M. P. Wattenberg (eds.), Parties Without Partisans: Political Change in Advanced Industrial Democracies. Oxford: Oxford University Press: Webb, P., Poguntke, T. and R. Kolodny (2012). The Presidentialization of Party Leadership? Evaluating Party Leadership and Party Government in the Democratic World. In L. Helms (ed.), Comparative Political Leadership. Houndmills, Basingstoke, Hampshire: Palgrave Macmillan Wattenberg, M. P. (1991). The Rise of Candidate-Centered Politics. Harvard: Harvard University Press Young, L. (2013). Party Members and Intra-Party Democracy. In W. P. Cross and R. S. Katz (eds.), The Challenges of Intra-Party Democracy. Oxford: Oxford University Press 20

21 Appendix 1. The moderating effect of the selection procedure on leader effects on vote choice: random intercept model (three-level) COEF. S.E. 95% CONF. INTERVAL AGE.001*** GENDER EDUCATION -.005*** ORG. MEMBERSHIP PID.285*** POLITICAL KNOWLEDGE ECONOMY LEADER EVALUATIONS.035*** SELECTION PROCEDURE PARTY DELEGATES.030** PARTY MEMBERS.039** ALL VOTERS -.028* LEADER*SELECTION PARTY DELEGATES PARTY MEMBERS.008*** ALL VOTERS.014*** CONSTANT.088*** COUNTRY:VAR(CONSTANT) INDIVIDUAL:VAR(CONSTANT) VARIANCE(RESIDUAL) AIC BIC N(OBSERVATIONS) N(INDIVIDUALS) N(COUNTRIES) 16 21

22 2. The moderating effect of leadership power on leader effects on vote choice: random intercept model (three-level) COEF. S.E. 95% CONF. INTERVAL AGE.001*** GENDER EDUCATION -.004** ORG. MEMBERSHIP PID.295*** POLITICAL KNOWLEDGE.010*** ECONOMY LEADER EVALUATIONS.020*** LEADERSHIP POWER -.012*** LEADER*L.POWER.003*** CONSTANT.162*** COUNTRY:VAR (CONSTANT) INDIVIDUAL:VAR(CONSTANT) VARIANCE(RESIDUAL) AIC BIC N(OBSERVATIONS) N(INDIVIDUALS) N(COUNTRIES) 12 22