Intra-party leadership elections: What are its effects on party unity?

Intra-Party Leadership Elections:

What are its effects on Party Unity?

Masters’ Thesis for the specialisation “Parties,

Parliaments, and Democracy”

Keywords: leadership elections, party unity, political parties, party image, Attina index

Tim Coté s1703544 Professor Ekaterina Rashkova Leiden University Date: 10-06-2019 Words: 9781

1

Abstract

This thesis examines the relationship between party leadership elections and party unity. This thesis argues that party leadership elections lower party unity in parliament and make it difficult for voters to have a continuing image of the political party. The difference in party unity between parties with and parties without leadership elections is minimal. Party disunity is problematic because party unity is essential for the working of political parties in the contemporary system of democracy and for the choice citizens have to make in election time. The research question is “What are the consequences on party unity of introducing leadership elections within parties?” And the data used to answer this question are the CSES, European roll-call, EES, and Political Party Databases. Parties with party leadership elections are analysed, and the continuity of the party image and the Attina Index is measured. These parties are then compared to parties without leadership elections within the 8th _{European parliament to}

determine parliamentary party unity. A comparison over time is done to research the influence of holding a leadership election on the image of party unity. This research is partly a comparative analysis of several European democracies over time and partly cross sectional, as an interrupted time series design and a comparison at one moment in time between parties with and without leadership elections.

2 In 2015 a leadership election was held in the United Kingdom Labour party, which resulted in Jeremy Corbyn being elected by the members. Jeremy Corbyn is a classical left politician and, while he was chosen by the party members, the party establishment disagreed with his, according to them, old fashioned ideals. Additionally, the establishment feared this ideology would not stroke with the electorate and it would result in electoral losses. As a result, members of parliament left the party and other members of parliament tried to remove Jeremy Corbyn from power. These attempts failed and Corbyn did not do worse than his predecessor in the national elections. In the end, the tensions within the party leadership are still there. This is an example of the damage that the election of the ‘wrong’ candidate can do to a political party. Another possible consequence of party leadership elections is that the candidate that is preferred by the party leadership is elected, but that the voters would have preferred a different candidate. This happened in the Dutch People’s party for Freedom and Democracy in 2006. One candidate won the leadership elections held within the party, the now prime minister Mark Rutte, but the number two on the list, Rita Verdonk, received more preference votes in the national elections. This led to her contesting for the leadership position because, while the party members might prefer Rutte, the voters preferred her. This tension divided the party, which eventually led to Rita Verdonk being forced out and founding her own party. In the Netherlands this happened to the largest and electorally most important political party, In the United Kingdom it happened in one of the two main political parties. As these examples show, the presence of leadership elections could have negative consequences on the unity of political parties. Whereas these changes are implemented by parties in an attempt to democratize. The subject that will be examined in this master thesis ties to the described situations, it concerns the consequences for party unity of leadership elections. The relevance of this question lies in the increasing number of parties that introduce leadership elections and the negative impact it can have on party unity. Party unity is important in the working of party democracy, which is the primary system of democracy in contemporary times. Most studies, as well as the reigning opinion, argue that introduction of leadership elections is a means of democratizing, and democracy is a valuable good. In this thesis, whether the introduction of leadership elections is democratic, is not important. It focusses on whether leadership elections influence one aspect of democracy: party unity. The data used for this thesis are the Political Party database, the European Election Study, and Dataset of European Parliament roll-call votes. This thesis argues that there are small changes in party unity because of the introduction of leadership elections and small differences in party unity between political parties with and without leadership elections. This thesis proceeds as follows, it will start with the relevance of the chosen subject, then the research question will be introduced, backed up by existing literature, after which data from different sources is introduced to be used in the analyses in the next chapter. After the analyses, for both research questions, the results are

3 presented, which leads to the conclusion that the influence of leadership elections on party unity is minimal.

Relevance of party leadership elections

The expansion of intra-party democracy is an ongoing development in Western democracies (eg. Bille, 2001). Also, European wide organisations like the Venice Commission (2009), the European Commission for Democracy through Law, state that their goal is to reinforce political parties’ internal democracy. This goal is based on the idea that democratisation of the leadership selection process is a positive development, but whether this is the case is still debated. What the consequences of the presence of party leadership elections are, is not yet, or only partially, known, because existing studies only focus on one aspect of democracy or are purely theoretical.

This thesis will focus on party leadership elections, which are the possibilities of members of a political party to vote on their next party leader. Party leadership elections are one aspect of intra-party democracy. Intra-party democracy encompasses all measures taken to increase the influence of party members within a political party. Intra-party democracy, regarding leadership elections, is determined both by the influence the party members have in selecting the leader, as well as by who decides on the eligibility of candidates (Rahat & Hazan, 2010, p.31; Aylott & Bolin, 2017, 57). In this thesis, the focus will only be on the selectorate in leadership elections. The selectorate is the body in charge of selecting the party leader. In most cases, leadership elections are introduced by political parties themselves as a reaction to changes in society, but in some countries political parties are encouraged to change their internal democracy. Several changes in society result in an adaptation of leadership elections: a growing distrust in parties (Shomer & Lavi, 2018), the changing relationship between parties and the state (Wolkenstein, 2016), loss of party members or electoral defeat (Pennings & Hazan, 2001, p.269), or personalisation (Cross & Katz, 2013, p.115; Rahat & Shapira, 2017, p.86). Apart from being a reaction, leadership elections can also be implemented deliberately, to legitimize the political party or distinguish the it from others (Cross & Katz, 2013, p.133; Close, Kelbel & Van Haute, 2017, p.658).

A relation between party unity and leadership selection method is expected in this thesis. The question this research tries to answer is whether democracy within a political party contributes to democracy on the national level. Starting point for this reasoning is the idea that ‘democracy on a large scale is not the sum of many little democracies’ (Sartori, 1965). In practice this would mean that political parties need not be democratic to fulfil their democratic function. Their internal working, especially regarding the leadership selection process, is expected to have a negative impact on several democratic aspects (Teorell,1999, p.364). It can be argued that party disunity will make it harder to govern, and that there is a greater risk

4 of party split-offs and a less clear-cut the party image towards voters. This results in voters being less informed by the party label, and less informed about politics in general. Which would harm democracy and efficiency of the democratic process.

Theoretical Framework

Building on existing research and expectations, the research question is formulated as follows: “What are the consequences on party unity of introducing leadership elections within parties?”

In the research question, party unity is the dependent variable and leadership elections within a political party is the independent variable. Leadership election is conceptualised as closed party primaries. Closed primaries are leadership elections that are open to all members of the political party, regardless of the means of voting. In leadership elections, the party leader is chosen by the party members, but this role within the party can have a different name than “party leader”. In all cases in this research, the term party leader will be used. Furthermore, in this research will be looked at the acquittal of leadership elections, not the formal change of the rules regarding party leadership elections.

Most of the literature on party leadership elections is normative or theoretical. Theoretical literature debates whether the process or the outcome of leadership selection should be democratic, because these different aspects negatively influence one another (Cross & Katz, 2013, p.149; Teorell, 2018, p.363). Leadership elections do strengthen the democratic process within parties, but whether the outcome of leadership elections contributes to democracy is undetermined (Pennings & Hazan, 2001,p.270; Cross & Katz, 2013, p.10). Empirical research on leadership elections focusses on its consequences (eg. Rahat & Shapira, 2017, p.85; Wolkenstein, 2018). The consequences on party membership are researched (Pennings & Hazan, 2001,p.268; Cain et al., 2003; Rahat & Hazan, 2010, p.99). As well as the opinions of voters at large regarding leadership elections (Close, Kelbel & Van Haute, 2017, p.659). Other scholars looked into the relationship between political trust and representation, and party leadership elections (Aylott & Bolin, 2017, p.56; Close, Kelbel & Van Haute, 2017, 662; Kenig 2009, p.240/246; Rahat & Hazan, 2010, p.115; Shomer & Lavi, 2018). The introduction of leadership elections also negatively impacts responsiveness and causes legitimacy loss (Lehrer, 2012, p.1315; Rahat & Hazan, 2010, p.149/159). The electoral fortunes of parties with leadership elections are bleak as well (Bäck, 2008, p.84; Rahat & Hazan, 2010, p.143). In the end the introduction of leadership elections can result in the collapse of party democracy (Rahat & Shapira, 2017, p.86).

Party unity will be conceptualised using definitions of democracy, because party unity is an important aspect of contemporary party democracy. Parties are the gatekeepers in

5 democratic societies today and are the main players in the political arena, (Sartori, 1965, p.120; Schattschneider, 1942). To fulfil the democratic functions, political parties need party unity. Party unity can be divided into unity in parliamentary voting, unity in the party image towards voters, and unity in the party organisation. This division follows the three faces of political parties: in public office, on the ground, and the party central office respectively (Katz & Mair, 1993, p.595). The aspect of voting unity in parliament, is formal unity. The aspect of unity in the party image towards voters can be described as a continuity in party image. The last aspect, the unity in the party organisation, relates to the centralisation of power and to what organ makes the important decisions within the party. This face is hidden from voters and is no part of this thesis. The party in public office needs to act in unity because then laws can be passed and voters can trust the party will do as it promised in its election manifesto, and unity in the party image is needed because it simplifies the choices for voters to make in elections and relates to trust in political parties, because they are continuous and keep promises.

There are two contradicting findings regarding party unity and leadership elections. On the one hand the introduction of leadership elections would increase, or not change, party unity. Leadership elections result in more incumbents being chosen (Aylott & Bolin, 2017, p.56; Kenig 2009, p.240/246; Shomer & Lavi, 2018). Incumbents are more prone to follow the party line, because they have done so in the past. Party unity would also increase, because the influence of the party leader decreases after the introduction of leadership elections. This decrease in power stems from a continuing accountability to voters and an increased influence of opinion polls (Loxbo, 2013, p.550). These developments limit the party leader’s freedom of action and result in fewer possible policy directions, so there will be less conflict on the policy direction. Also, leadership elections influence parliamentarians. They are more prone to follow the party line and contribute to party unity (Rombi & Seddone, 2017, p.584).

On the other hand, there is research arguing that the introduction of leadership elections decreases party unity. Leadership elections result in the candidate party leaders assuming more extreme policy stances, whereas leadership appointments result in policy converging with the position of the median voter (Teorel, 1999, p.365; Rahat & Hazan, 2010, p.116). The explanation for the more extreme policy stances is the need for a candidate to distinguish themselves from other candidates (Kenig 2009, p.243). More moderate views result in broad support, whereas extreme views will erode support within and outside of the party, which will result in friction. Also, a leader with a personal mandate has more policy freedom and has to deal with party factions less (Ceron, 2012, p.700). This can result in the policy of a party changing whenever the party leader changes, and in party factions feeling left out. Apart from unity within the party leadership, the introduction of leadership elections within parties also influences the party members. Introduction of leadership elections invokes greater demands for other internal changes (Lisi & Razzuoli, 2018, p.406). Additionally, voters for the

6 losing candidate are likely to leave the party or withdraw from party activities (Cross & Pruysers, 2017, p.8). So, introduction of leadership elections can distort party discipline and party unity, and weaken the quality of representative democracy (Pennings & Hazan, 2001, p.268; Rahat & Hazan, 2010, p.157; Cain et al., 2003).

Building on previous research, a theory is drawn up about leadership elections undermining party unity (Rahat & Shapira, 2017). This theory is built on the idea that change in the internal structure of political parties will influence their ability to fulfil their democratic function. Leadership elections influence the image voters have of one party and influence trust in parties in general (eg. Rahat & Hazan, 2010, p.9; Shomer & Lavy, 2018). This can have either a positive or a negative result. One of the negative consequences of leadership elections is that democratizing internal mechanisms of the political party can distort the functioning of the party (Pennings & Hazan, 2001). Parties will be unable to function properly because factions will no longer be kept satisfied by the party leadership and might rebel (Ceron, 2012). Leadership selection is a bargaining between factions and results in a compromise, whereas leadership elections are a winner-takes-all system of leadership election. The theory drawn up argues that party unity will suffer due to the introduction of party leadership elections, because the leadership election system of compromise is replaced. This theory will be examined using the following sub-questions central in this thesis:

1. What are the consequences of leadership elections within parties for party unity in parliament?

2. What are the consequences of introducing leadership elections within parties for clarity of the party image to the electorate?

Unity in parliament is conceptualised as voting with the party. Expanding the theory, it is expected that leadership elections within parties negatively impact parliamentary party unity. This expectation stems from party factions not being heard, and new party leaders being unable to control party factions (Cross & Pruysers, 2017; Hazan & Rahat, 2010, p.2), or parliamentarians profiling on their personal agenda (Indriðason Kristinsson, 2015, p.569). It can also be argued that party unity improves, because the choice of voters for a leader’s policy is respected. This leads to the formulation of the neutral hypothesis:

H1: There will be differences in party unity between parties with leadership elections and parties without leadership elections.

The clarity of the party image is conceptualised as a continuity in the image voters have of the political party. It is expected that introducing leadership elections within political parties negatively impact the party image voters have. This expectation stems from the fact that a clear image of a party, that is stable over time, helps voters make an informed decision during

7 elections. When the image of a party changes, voters have to invest to stay up-to-date. When they do not, this may lead to them making wrong choices. This change in party image is caused by the need of leadership candidates to profile themselves on issues in leadership elections (Ceron, 2012, p.689; Loxbo, 2013, p.550). More extreme candidates are chosen and party leaders have more freedom in determining the party policy due to profiling of leadership candidates (Kenig 2009, p.243; Schmitter & Karl 1991 p.82; Teorel, 1999, p.365). When different candidate party leaders of the same party present different policy, voters will have a hard time determining the party image. Consequently, there will be changes in the party image visible in surveys. Taking the Israeli case as an example, it is expected that party image of voters not only changes after the introduction of leadership elections, but that a direction can be determined. In Israel, parties are observed moving to the right of the political spectrum after the introduction of leadership elections (Vick, 2013). This leads to the formulation of several hypotheses:

H2: There will be a change in the image voters have of a political party when leadership elections are introduced.

H3: Parties that introduce leadership elections will move to the right in the image of voters.

H4: The introduction of leadership elections within parties will result in them being more extreme in the eyes of voters.

Data & Method

Party unity is measured using the Dataset of European Roll-call votes (Cherepnalkoski et al., 2016), the Comparative Analysis of Electoral Systems (CSES, 2018), and the European Election study (Schmitt et al., 2009, 2016; Egmond et al, 2013). The Dataset of European Roll-call votes is used for the first sub-question and contains data on parliamentary voting in the European Parliament. Several political parties are compared in the same period. The CSES and the European Election study are used to measure the unity in party image of several political parties, by comparing two moments in time. The independent variable in the research questions is the presence of leadership elections. The total amount of European political parties with leadership elections is 21 over 7 countries. For the first sub-question, twelve political parties, spread over five countries are selected, for the second sub-question eleven parties over five countries. The chosen parties are presented in table 1 and table 2.

General choices

The presence of leadership elections is determined by using the question “C25PLMBRTXT: Role of Individual Members” in “PartyLeader_Selection_Vote_AIPD.” from the Political Party Database (Poguntke et al., 2017). This is an open question about an experts opinion on the

8 possibility leadership elections. The choice is made to exclude unopposed party leadership election candidates and procedural changes that have not yet been put into practice. There may be differences in the leadership selection process within the selected group, but the research questions are about the presence or absence of party leadership elections. European countries are studied in this research, because primaries are spreading in Europe (Bille, 2001, p.365). Furthermore, democracy is established in most European countries, so the influence of a single person should be less compared to the formal rules of the game. It is expected that if there is a change in party unity in Europe as a result of the introduction of leadership elections, it will be universal, using a least-likely-case approach. Also, European countries have a similar history, similar cleavage structure, are joined in the European Union, and have parliamentary systems, which makes comparison of parties from different countries and rejecting alternative explanations for a change in party unity possible (McKay et al., 2014, pp.xxxvii-xlviii). Last, European countries have strong political parties compared to other countries, so party unity is deemed important in the working of the system. Some European countries have a formal necessity of the presence of political parties in their constitution, or electoral law, and there are no politically relevant independents (O’Neil et al., 2015).

The European Parliament is chosen to examine the cases in a similar setting, because all political parties chosen are represented there. This choice results in national parliamentary norms or rules having no impact on the results. European elections are seen as second-order elections (Boomgaarden et al., 2016), and Members of the European Parliament are national legislators, not European (Scully et al., 2012, p.678). As a result, national party leadership elections influence party unity in the European Parliament, although it can be argued that the choice for the European Parliamentary parties gives a different image of party unity than analysing the national parliament would, due to European Parliamentarians having to answer to the national party, the party fraction, and the European party group. Previous research into party unity in the European Parliament argues that party unity is caused by consensus, and is not due to party discipline (Mühlböck, 2013). Additionally, party unity in the European Parliament is rarely unanimous, and enforced by neither the European party group nor the national party. Also, the choice to analyse the European Parliament to research the consequences of national leadership elections can be defended by the fact that those elected in the European Parliament fulfil a role in name of the national party. For newer and smaller parties, the European Parliament is a training centre for their politicians. For anti-European Union parties and older established parties, European Parliamentarians are national veterans to draw votes or focus on certain policy. In the end, it is the national party that determines the career path of their representatives on the European level (Aldrich, 2018, p.1299). In this light it may be expected that national party changes influence behaviour in the European

9 Parliament. Although caution should be taken translating the results of this thesis to the national level.

Parliamentary unity

To examine parliamentary unity, parties with three or more seats in the European parliament are analysed. The timeframe chosen for the analysis is half a year: ranging from after Labour has held their first leadership election (12 September 2015), till march 2016. The cases will be compared in a cross-sectional research design. The political parties with leadership elections will be compared to their ideologically closest national competitor in the European Parliament without leadership elections, or a similar party in a similar country.

This choice for a minimum amount of seats has been made for several reasons, first a one-man party has perfect unity. Second, because of absence of parliamentarians, in practice always one parliamentarian is absent, so a two-man party would show perfect unity as well, and third this choice is made because most large national parties only have three sets in the European Parliament. Parties with less seats are more influenced by a single parliamentarian voting out of line, but the chance of it happening is greater in larger parties. That is why party size is a variable to take into account in the analysis. .

The timeframe is chosen so as many political parties as possible could be included in the analysis and because till this time the data is freely available. The Labour party in the United Kingdom was the last party to change its leadership election procedure, and the first leader chosen under the new rules was installed in 12 September and the first voting instance in the European Parliament after that was the 16th _{of September. There are 1004 instances of}

voting that can be analysed in the selected time period. The choice for a longer time period is made so the influences of elections is minimal, as is the influence of single events. The introduction of leadership elections within parties varies, both within countries as well as between them. Time could theoretically influence the relation this thesis examines and could blur or strengthen the effects found, so the choice is made to compare the selected cases to other parties within the European parliament at the same point in time. Regarding introduction of leadership elections, there are some extreme outliers like Labour in the United Kingdom that introduced leadership elections in 1994 and expanded them in 2015, but most parties in the data introduced leadership elections in the first decade of the millennium, so differences between long term and short term effects should be minimal. In this set-up party unity will be examined at the same moment, for the same laws, and controversial happenings and global disasters will influence all cases equally.

The parties will be compared as a group, so a mix of different ideologies and an equal amount of parties with and without leadership elections is present in the analysis. The selected parties and their characteristics are presented in table 1. For the Netherlands and Spain, a

10 comparison is made with national parties without leadership elections that are represented in the European Parliament and have a similar size and ideology. For Portugal, a national party, and, because of absence of large enough parties, a Spanish party, are used as comparison. Spain and Portugal share a similar history, have a similar economic position in the European Union, have a similar political system and are geographically close. The Partido Popular is chosen because, as the Social Democratic Party, it is the largest party in the country on the ideological right. For the parties in Belgium and the United Kingdom, there are no national parties without leadership elections that are represented in the European Parliament to compare. So, the United Kingdom parties will be compared to German parties, because of the similar size of the countries, the use of electoral districts, a mainly protestant history, both systems being parliamentary and having a two-and-a-half party system (O’Neil et al., 2015). Belgian political parties will be compared to German and Spanish political parties, because these countries share a federal structure, a Catholic history, and the ideology of the political parties is similar.

Table 1. Parties sub-question 1

Party Country Seats EP Compare to Seats EP Conservative Party United Kingdom 19 CDU Germany 29 Labour Party Netherlands 3 D66 Netherlands 4 Labour Party United Kingdom 20 SPD Germany 27 New Flemish Alliance Belgium 4 IU (Spain) 4 Open Flemish Liberals and

Democrats

Belgium 3 FDP (Germany) 3 People's Party for Freedom

and Democracy (VVD)

Netherlands 3 CDA (Netherlands) 5 Reform Movement Belgium 3 Ciudadanos

(Spain)

3 Social-Democratic Party Portugal 6 Partido Popular

(Spain)

16 Socialist Party Belgium 3 Union (Spain) 4 Socialist Party Portugal 8 Communist Party

(Portugal)

3 Socialist Party (PSOE) Spain 14 Podemos (Spain) 5 UK Independence Party United Kingdom 24 PVV (Netherlands) 4

Source: Author’s collection

A quantitative analysis will be done using data for the selected political parties from European Parliament roll-call data (Cherepnalkoski, 2016). Parliamentary unity will be calculated using the Attina Index (Attina, 2006). This is a measurement for party unity in parliament using roll-call voting data. The Attina Index is slightly different than the most used measurement, the Rice Index, because it also takes into account the neutral voting option that is possible in the European parliament. The Attina Index is calculated as follows:

𝐴𝑡𝑡𝑖𝑛𝑎 = ∑|𝑌𝑒𝑠 − 𝑁𝑜 − 𝑁𝑒𝑢𝑡𝑟𝑎𝑙| 𝑌𝑒𝑠 + 𝑁𝑜 + 𝑁𝑒𝑢𝑡𝑟𝑎𝑙

11 Where ‘Yes’, ‘No’ and ‘Neutral’ are the for, against and neutral votes respectively. It is assumed that the political parties votes ‘for’ a law in the formula above. When this is not the case, the option that is chosen most is switched with ‘Yes’ in the formula and the order of the formula changes. The amount of for, against and neutral votes on a certain law that was voted on by roll-call vote need to be filled in and then the Attina Index will vary between 0 (no unity) and 1 (perfect unity). In this thesis, aggregate data will be entered, not the separate voting instances, so that an average Attina Index for a certain time is calculated. The average Attina Index will then be compared between the groups of parties with and without leadership elections following the first hypothesis. For this an independent sample t-test will be used. This test is used when the same dependent variable, namely Attina Index, will be looked at for two unrelated groups, in this case parties with and without leadership elections. There are certain assumptions for this test: the dependent variable is continuous and the independent variable consists of two or more categories, the observations should be independent, there are no outliers, the data is distributed normally, and there is a homogeneity of variances. These assumptions are tested under the heading ‘results’.

Several cases in the data are excluded because parliamentarians of a party did not vote, or the votes were secret. The choice was made to exclude all missing votes, but to include votes by parliamentarians that are absent a lot, whenever they did partake in the vote. In these instances, when the parliamentarian was present, an important decision might have been made, so presence might have been required. This was not taken into account in the data analysis. In short, all voting instances are counted equally.

Party image continuity

To examine party image continuity the question "In politics people sometimes talk of left and right. Where would you place [PARTY A] on a scale from 0 to 10 where 0 means the left and 10 means the right?" (Schmitt et al., 2009, 2016; Egmond et al, 2013; CSES, 2018) will be used. The answer to this question is analysed previous to the introduction of leadership elections within a party and again after the introduction of leadership elections in that party. The selected parties and their properties are presented in table 2.

For all data before 2004, CSES data are used, later data is taken from the EES. The question used the same wording in the CSES as in the EES, so the results of the different datasets can be compared. The wording of the question is general, and does not refer specifically to European parties or European elections, and the question asked refers to national parties. It is expected that information regarding unity in party image for the national parties is examined by using this research design. Although the EES is done during the electoral campaign for the European Parliament.

12 the image voters have of a party is influenced by the presence of party leadership elections. An interrupted time series design is used, to compare the situation before and after a critical juncture, in this case the first leadership election held within a political party. One instance in time before and after the introduction of leadership elections is analysed. In this set-up, four years are analysed, if the choice was made to analyse two elections before and two elections after the introduction of leadership elections, twelve years would be analysed. In twelve years population characteristics might have changed, new parties have risen, and new issues are important, which leads to questions about the measurement validity.

The survey questions are posed in election time. At this time people are involved with politics, so they have a clear image of the party and its leader. If there is an influence of the party leader on the party policy, it should best be noticed by voters during election time. The European election data are: Halfway June 2004, start of June 2009, and the end of May 2014. The CSES used for the Conservative party and PvdA is conducted during national election time, 1997 and 2002 respectively.

Not all political parties that are eligible for the research question, are included in the analysis, either because the parties are too small to be represented in the European Parliament, because the question used in this analysis was not posed in the chosen election year for the selected political party, because political parties had leadership elections since their founding, or because data was unavailable for the period before or after the introduction of leadership elections. The year of introduction of the leadership election and the first leadership election are taken from the Political Party database, complemented by online sources.

Table 2. Parties sub-question 2

Party Country Seats

EP Primary introduced Christian-Democrat and Flemish Belgium 2 May 2008 Conservatives* UK 19 September 2001

Green Belgium 1 November 2014

Open Flemish Liberals and Democrats

Belgium 3 December 2009 Labour Party (PvdA)* Netherlands 3 November 2002 People’s Party Portugal 1 July 2013 People's Party for Freedom

and Democracy (VVD)

Netherlands 3 May 2006

Plaid Cymru UK 1 March 2012

Socialist Party Portugal 8 July 2011 Socialist People's Party Denmark 1 April 2005 UK Independence Party UK 24 September 2006

13 The means and confidence interval (95%) of the left-right placement per political party are calculated before and after the introduction of leadership elections. Then the difference between the means before and after the introduction of leadership elections per party will be calculated. This will be done in two waves, one with the data origination in the EES, the second one with all political parties. These parties are marked with an asterisk in table 2.

Then a t-test will be done using the difference in means as a ratio variable. A t-test is used to establish whether the differences in the means are a result of chance, or whether they could be caused by the change in leadership selection method. Per hypothesis, a t-test is done to analyse the difference in means. Per t-test, the sign of the means differs; for the first hypothesis the absolute difference is used, for the second hypothesis the real difference is used, and for the last hypothesis a decline when the party scored below five, or an increase if the party scored greater than five before the introduction of leadership elections is seen as positive, and a move in the other direction is seen as negative.

The one sample t-test can be used when certain assumptions are met: the dependent variable is interval/ratio, the data are independent, there are no significant outliers, and the data is normal distributed, which can be tested by using Shapiro-Wilk’s test.

Parliamentary unity: the Attina Index

The results for the second sub-question are presented in the table below. In the table the party name is presented in the first column. Then the amount of excluded voting instances, because it was secret, because nobody voted or because only one person voted. This number has then been translated into a percentage. The column contains the amount of voting instances where there was disunity per party, followed by a translation into a percentage of the total amount of votes. Then the amount of valid votes and the amount of different votes are presented, as is the percentage of the amount of different votes compared to the total amount of votes.

In the last column, the Attina Index of party unity is presented, calculated according to the formula in the previous heading. The high value can result in a plafondeffect, which is that high values will not easily get even higher, when they are near the top of the scale. The high score on the Attina Index stems from the fact that for most cases less than ten percent of the votes cast differed within the party. Although this finding varies across parties: there is more disunity in the larger parties and less in the smaller ones. For Ciudadanos, FDP and PvdA, a lot of votes were excluded in the analysis, twenty to forty percent of the votes was excluded.

14

Table 3. Results sub-question 1

Party Excluded Excluded percent Amount non-unity Non unity percent Total valid Total different votes Votes with difference percentage Attina Index CDA 6 0,597609562 36 0,996016 4677 38 3,585657371 0,983750267 CDU 6 0,597609562 186 3,585657 25026 444 18,32669323 0,964516902 Ciudadanos 403 40,13944223 29 18,5259 1202 29 2,888446215 0,951747088 Conservatives 6 0,597609562 218 2,888446 16223 422 21,61354582 0,947975097 D66 7 0,697211155 13 21,71315 3716 15 1,294820717 0,991926803 FDP 205 20,41832669 46 1,294821 2169 48 4,581673307 0,955739972 MR 11 1,09561753 85 4,581673 2742 86 8,466135458 0,937272064 NVA 6 0,597609562 16 8,466135 3788 18 1,593625498 0,990496304 OVLD 8 0,796812749 45 1,593625 2607 46 4,482071713 0,964710395 Podemos 5 0,498007968 24 4,482072 4675 26 2,390438247 0,988877005 Communist 29 2,888446215 21 2,390438 4769 21 2,091633466 0,991193122 SDP Portugal 6 0,597609562 90 2,091633 5560 135 9,462151394 0,951438849 Socialist 6 0,597609562 111 8,964143 7475 155 11,05577689 0,958528428 PP 5 0,498007968 62 11,05578 12956 127 6,175298805 0,980395184 PS 15 1,494023904 22 6,175299 2769 22 2,19123506 0,984109787 PSOE 6 0,597609562 142 2,191235 11795 282 14,14342629 0,952183128 PvdA 212 21,11553785 16 14,14343 2109 16 0,099601594 0,984826932 PVV 9 0,896414343 7 1,593625 3633 7 0,697211155 0,996146435 IU 74 7,370517928 32 0,697211 2957 32 3,187250996 0,978356442 Union Pyd 18 1,792828685 93 3,187251 2665 94 9,262948207 0,92945591 SPD 6 0,597609562 150 9,262948 23329 460 14,94023904 0,960564105 Labour 5 0,498007968 105 14,94024 17931 199 10,45816733 0,977803803 UKIP 7 0,697211155 192 10,45817 17736 427 19,22310757 0,951849346 VVD 21 2,091633466 10 19,12351 2660 11 1,09561753 0,991729323

Source: Author’s calculations

A visual overview of the parties is presented below. It can be noticed that IU, a Spanish party without leadership elections, is an outlier. Apart from a slight concentration of party leadership elections parties on the lower end of the Attina Index, and political parties without primaries on the higher end, there is no clear visual pattern.

15

Image 1. Attina Index per party

The cases are compared as a group, to determine whether there are differences in parliamentary party unity between parties with and without party leadership elections. The data used for comparison can be found in the table below. Comparison shows that the average parliamentary unity is lower for parties with party leadership elections (Mean=0,966 SD=0,019) than parties without leadership elections (Mean=973 SD=0,020). But this difference is not statistical significant (t(22)=-0,831, p=0,415). The median of parties with no leadership elections is 0,979 and that of parties with leadership elections is 0,962. So again, a difference between the two groups can be seen.

Table 4. Comparing the groups

N Mean Standard Deviation Confidence Interval

Primaries 12 0,966076954667 0,0188163285221 0,954121629897 0,978032279436 No Primaries 12 0,972722436250 0,0203321762557 0,959803987780 0,985640884720

Source: Author’s calculations

The different groups are compared using an independent sample t-test, for which the assumptions were tested. There is a continuous dependent variable, the independent variable is a category, the observations of the different parties are independent, there is a normal distribution, and there is homogeinity of variances (Shapiro-Wilks Y:0,201; N:0,251; Levene’s test for variances is 0,337). There may be an outlier in the ‘No primaries’ data, so the same analysis is repeated with the exclusion of one case (Shapiro-Wilks Y:0,201, N:0,220). The differences in the means and confidence interval between the two groups with and without the outlying case are presented in the images below.

16

Image 2. means and confidence interval

Image 3. means and confidence interval without outlier

After exclusion of the outlier, the effect is still the same: political parties with leadership elections (Mean=0,966 SD=0,019) do not have statistically lower party unity than political parties without leadership elections (Mean=977 SD=0,016), (t(21)=-1,452, p=0,161). Except for the fact that three Dutch parties have the highest score, country of origin does not seem to be a factor for the Attina index, neither does left-right position of a political party. A change from government to opposition does not stand out either. Party size on the other hand is a determinant of party unity. Examining image 4, it is found that smaller parties have a more extreme Attina Index, both the parties with the most and the least party unity are smaller parties. This finding is remarkable, because according to literature, larger parties are found to have more party unity than smaller ones (Sieberer, 2006, p.159). This finding can be explained by referring to the increased influence of a single defector in smaller parties. The finding that smaller parties also have the most party unity, can be explained be small parties, below a certain threshold of seats, being more cohesive through personal contact (Desposato, 2005, p.733).

17

Image 4. Party size and party unity Party image continuity: comparing means

For the second sub-question, there are three hypotheses that are tested, namely that there is a change in voters’ perception of the left-right position of political parties after the introduction of leadership elections, that voters perceive parties more to the right after the introduction of leadership elections, and that political parties will be perceived more extreme after the introduction of leadership elections. The assumptions of the method of data analysis were tested. The assumptions are the use of a ratio variable, use of independent data, there are no outliers and normality is tested by Shapiro-Wilks because there is a small amount of cases (p=0,177). And Shapiro-Wilks is tested again for the inclusion of all parties (H2:p=0,164; H3:p=0,399; H4:p=0,433).

The data used is presented in table 5; from left to right the party name, amount of respondents, means of left-right placement before and after the introduction of leadership elections, with their respective confidence intervals and standard deviation can be found. What stands out in the data is that Plaid Cymru has less observations in the ‘before’ survey than other political parties. Also, from all respondents in the dataset, around a tenth did not answer the question used for the analysis. Other data in the table show that the smallest difference between means of left-right placement of parties before and after the introduction of leadership elections is 0,26 and the largest difference is 1,46. Table 5 also shows the confidence intervals for the before and after situations, in eight of the nine cases these do not overlap and a difference can be seen. Only in the case of the Belgian party ‘Green’ the differences could be caused by chance.

18

Table 5. properties of variables

Party name N Mean CI low CI high Standard Deviation CDV Before 856 5,02 4,86 5,1 2,355 After 986 5,46 5,31 5,6 2,353 Conservative Before 2324 7,17 7,07 7,28 2,626 After 1266 6,39 6,25 6,52 2,409 Green Before 870 4,13 3,98 4,29 2,298 After 881 3,93 3,79 4,08 2,202 OVLD Before 854 5,73 5,56 5,91 2,597 After 631 7,19 7,08 7,29 1,304 People’s Party Before 860 7,65 7,49 7,81 2,386 After 886 8,81 8,69 8,94 1,955 Plaid Cymru Before 597 3,35 3,16 3,54 2,382 After 212 4,68 4,36 5,01 2,420 PvdA Before 1530 3,43 3,25 3,51 1,593 After 1450 3,79 3,69 3,89 1,925 Socialist Party Before 875 4,74 4,61 4,88 2,015 After 912 5,22 5,11 5,34 1,769 Socialist People’s Party Before 1118 2,95 2,86 3,05 1,650 After 934 3,21 3,08 3,34 2,045 UKIP Before 1020 5,94 5,77 6,12 2,898 After 804 5,43 5,23 5,63 2,952 VVD Before 1438 7,48 7,37 7,59 2,095 After 929 6,77 6,63 6,90 2,093

Source: Author’s Calculations

To test the hypotheses, the difference between the means is presented differently for every hypothesis. It is presented absolute, then contains a minus sign if the party moved to the left, then it contains a minus sign if the party moved to the middle.

To examine the first hypothesis a one sample t-test was run to examine whether there is a difference in left-right image of a party between the instances it did have and did not have leadership elections. The mean difference between before and after for all parties (Mean=0,7278; SD=0,15701) was higher than zero, on a scale of 1 to 10 and is statistical significant (p=0,002; t(8)=4,635). So the null hypothesis of there being no effect of the introduction of leadership elections can be rejected.

To examine the second hypothesis a one sample t-test was run to examine whether parties move to the right on the left-right scale after the introduction of leadership elections in the eyes of voters. The mean difference between before and after for all parties (Mean=0,4122; SD=0,26385) was higher than zero, on a scale of 1 to 10, but is not statistical significant (p=0,157; t(8)=1,562). So the null hypothesis of there being no positive or negative effect of the introduction of leadership elections still stands.

19 To examine the last hypothesis a one sample t-test was run to examine whether the left-right image of a party becomes more extreme after the introduction of leadership elections. The mean difference between before and after for all parties (Mean=-0,0033; SD=0,90428) was not zero/0, on a scale of 1 to 10, but is not statistical significant (p=0,991; t(8)= -0,011). So the alternative hypothesis of polarization because of the introduction of leadership elections can be rejected.

Table 6. Differences between means

Party Mean before Mean After H2 difference H3 difference H4 difference CDV 5,02 5,46 0,44 0,44 0,44 Conservative 7,17 6,39 0,78 -0,78 -0,78 Green 3,79 3,93 0,20 -0,20 0,20 OVLD 5,73 7,19 1,46 1,46 1,46 Peoplesparty 7,65 8,81 1,16 1,16 1,16 PlaidCymru 3,35 4,68 1,33 1,33 -1,33 PvdA 3,43 3,79 0,36 0,36 -0,36 SocialistParty 4,74 4,74 0,48 0,48 -0,48 SPeoplesP 2,95 3,21 0,26 0,26 -0,26 UKIP 5,94 5,43 0,51 -0,51 -0,51 VVD 7,48 6,77 0,71 -0,71 -0,71

Source: Author’s calculations

Running the tests again, using all the parties. The first hypothesis, whether there is a difference in left-right image of a party between the instances it did have and did not have leadership elections. The mean difference between before and after for all parties (Mean=0,6991; SD=0,43634) was higher than zero, on a scale of 1 to 10 and is statistical significant (p<0,001; t(10)=5,314). So the null hypothesis of there being no effect of the introduction of leadership elections can be rejected.

To examine the second hypothesis a one sample t-test was run to examine whether parties move to the right on the left-right scale after the introduction of leadership elections in the eyes of voters. The mean difference between before and after for all parties (Mean=0,2991; SD=0,79347) was higher than zero, on a scale of 1 to 10, but is not statistical significant (p=0,240; t(10)=1,250). So the null hypothesis of there being no positive or negative effect of the introduction of leadership elections still stands.

To examine the last hypothesis a one sample t-test was run to examine whether the left-right image of a party becomes more extreme after the introduction of leadership elections. The mean difference between before and after for all parties (Mean=-0,1064; SD=0,84590) was smaller than zero, on a scale of 1 to 10, but is not statistical significant (p=0,685; t(10)= -0,417). So the alternative hypothesis of polarization because of the introduction of leadership elections can be rejected.

20

Conclusion

This thesis argues that there is less unity in voting in parliament and less continuity in party image towards their voters in political parties with party leadership elections compared to political parties without leadership elections, regardless of country. The differences that are found are not statistical significant, which may have to do with the small sample size, and the effect that are found are rather small. An explanation for this can be the presence of a plafond effect regarding parliamentary party unity, because the Attina Index for all parties is larger than 0,93 and the maximum is 1. Additionally, it is found that party image changes after the party image introduction of leadership elections, and that most parties move to the right after the introduction of leadership elections, but that this is not a large or a significant effect. The hypothesis expecting polarization is rejected, and the slight effect that is visible is contrary to the hypothesis. This leads to the answer to the research question “What are the consequences of introducing leadership elections within parties for party unity?” being there is indeed a slight negative impact of the introduction of leadership elections on party unity, both in parliament and regarding the continuity of the.

The findings of this thesis can be seen in the light of existing literature. The results of this thesis were expected regarding the literature on the negative consequences of the introduction of leadership elections on party unity. Also, small effects are found because of the high party unity in existence, which is typical for the European systems. The expectations build on the literature regarding more extreme policy stances, following the introduction of leadership elections, is not supported.

While interpreting these findings several things need to be highlighted. Party unity can be culturally determined or ideologically determined, which has been tried to take into account by comparing parties nationally and ideologically, but it may still result in a European bias. This bias would influence the external validity of the research. Just as the overrepresentation of leftist parties may skew the results. The ideological difference between left and right parties regarding party leadership elections may detriment the measurement validity.

Furthermore, roll-call votes may be an overestimation of party unity in parties that know strong party discipline or party whipping (Ceron, 2015, p.253). On top of that, the use of secret ballot influences party unity in parliament, but what exactly is its effect is unclear (Raymond, 2014, p.386). It should also be noted that the fact that a political party image is not continuous, does not give any information on voting, and does not contain a positive or negative assessment by voters. Also, in the researched period, three parties changed from government to opposition or vice versa. A political party being in government or opposition can influence the image voters have of it and the need within a political party for party unity (Carey, 2007; Sieberer, 2006). For the other cases, a mix of both government and opposition parties is made, so the results should be applicable to both kinds of parties.

21 Then, in the process, several choices have been made that may influence the results and conclusions reached. Unopposed party leadership election candidates are excluded, just as formal changes regarding the rules concerning leadership elections. This results in not knowing whether a change of the rules influences on party unity. This may influence the results, because members get the formal chance to vote in both excluded situations. Another influence on the results may be the inclusion of cases (eg. VVD Netherlands) where the chosen leader has been in power for quite some time at the time taken in this research. This may result in them having control of the party and in the effect of the leadership election having faded. Also, only political parties that are represented in the European Parliament and have at least three seats are analysed for the research question. Only relatively large parties can be examined because of the amount of seats on the European level compared to the national level. Another bias in the data can be the case selection, which is based upon a national political scientist’s judgement.

Furthermore, to answer the first sub-question a comparison between parties is made. This results in differences over time not being included, for example not all cases analysed introduced leadership elections at the same time, or had party leaders just installed, or for a longer period of time at the moment examined. If there are effects of the introduction of leadership elections that are time dependent, they will not be found by using this analysis. Theoretically, It may also be the case that the time period examined, is exceptional. To answer the second sub-question a comparison over time within the same political party is chosen. It could be that events influence the image voters have of a political party. In this case, no significant global political events happened, but smaller national or even regional changes can still have influenced the image voters have of parties. Another choice made is analysing only one election before and after the introduction of leadership elections, which may result in overestimating differences between the two instances. Had a longer period in time been analysed, natural differences in party image between elections could be identified, and be taken into account. Last, the use of the election studies narrows the available time of researching the implementation of leadership elections to before or after elections. Finally, there are some limits to the data used. The data contain secret votes and absent parliamentarians, which are not taken into account. Absence is a relatively large part of the data, so depending on the motivations of the absent parliamentarian, the results could be different. Strategic abstention cannot be taken into account. Also, in some parties the same parliamentarians is the one breaking the unity. Furthermore, still the most instances are unitary voting instances.

22 Aldrich, A.S. (2018). National Political Parties and Career Paths to the European Parliament,

JCMS: Journal of Common Market Studies, 56(6), 1283-1304.

Attina, F. (2006). The Voting Behaviour of the European Parliament Members and the Problem of Europarties, European Journal of Political Research 18(5), 557-579. Aylott, N., Bolin, N. & Polk, J. (2017). Managed intra-party democracy: Precursory

delegation and party leader selection, Party Politics, 23(1), pp.55-65.

Bäck, H. (2008) ‘Intra-Party Politics and Coalition Formation: Evidence from Swedish Local Government’, Party Politics, 14, 71–89.

Bille, L. (2001). Democratizing a Democratic Procedure: Myth or Reality?: Candidate Selection in Western European Parties, 1960-1990, Party Politics, 7(3), 363-380. Boomgaarden, H.G., Johann, D. & Kritzinger, S. (2016). Voting at National versus European

Elections: An Individual Level Test of the Second Order Paradigm for the 2014 EuropeanParliament Elections. Politics and Governance, 4(1), 130-145. Bryman, A. (2016). Social Research Methods. Oxford: Oxford University Press. Cain, Bruce E., Russell J. Dalton, and Susan E. Scarrow (eds.) (2003), Democracy

Transformed? Expanding Political Opportunities in Advanced Industrial Democracies. Oxford: Oxford University Press.

Carey, J.M. (2007). Competing Principals, Political Institutions, and Party Unity in Legislative Voting, American Journal of Political Science,51, (1), 92-107. Published by: Midwest Political Science Association

Ceron, A. (2012). Bounded oligarchy: How and when factions constrain leaders in party position-taking, Electoral Studies, Vol.31(4), 689-701.

Ceron, A. (2015). Brave rebels stay home: Assessing the effect of intra-party ideological heterogeneity and party whip on roll-call votes, Party Politics, 21(2), 246-258. Cherepnalkoski, D., Karpf, A., Mozetič, I. & Grčar, M. (2016). Dataset of European

Parliament roll-call votes and Twitter activities MEP 1.0. Accessed on 12-03-2019, https://www.clarin.si/repository/xmlui/handle/11356/1071

Close, C., Kelbel, C. & Haute, E. van (2017). What Citizens Want in Terms of Intra-Party Democracy: Popular Attitudes towards Alternative Candidate Selection Procedures.

Political Studies, 65(3), 646-664.

Cross, W.P. & Katz, R.S. (2013). The Challenges of intra-party democracy. Oxford: Oxford University Press.

Cross, W. & Pruysers, S. (2017). Sore losers? The costs of intra-party democracy, Party

Politics, 23(4), 1-12.

23 Desposato, S.W. (2005). Correcting for Small Group Inflation of Roll-Call Cohesion Scores

British Journal of Political Science, 35(4), 731-744.

Egmond, M. van, Brug, W. van der, Hobolt, S., Franklin, M. & Sapir, E.V. (2013): European Parliament Election Study 2009, Voter Study. GESIS Data Archive, Cologne. Field, A. (2014). Discovering Statistics using IBM SPSS Statistics. London: Sage

Publications.

Hazan, R.Y. & Rahat, G. (2010). Democracy within Parties: Candidate Selection Methods

and their Political Consequences. Oxford: Oxford University Press.

Hix, S., Noury, A. & Roland, G. Roll-call Votes in the European Parliament. Received on 12 03-2019 from: http://personal.lse.ac.uk/HIX/HixNouryRolandEPdata.HTM.

Indriðason, I.H. & Kristinsson, G.H. (2015). Primary consequences: The effects of candidate selection through party primaries in Iceland, Party Politics, 21(4), 565-576.

Katz, R.S. & Mair, P. (1993). The Evolution of Party Organizations in Europe: The Three Faces of Party Organization, The American Review of Politics, 14(4), 593-617. Kenig, O. (2009) ‘Democratization of Party Leadership Selection: Do Wider Selectorates

Produce More Competitive Contests?’, Electoral Studies, 28, 240–247. Louwerse, T., Otjes, S. & van Vonno, C. (2018). Acta Politica, 53(1), 149-166.

Lehrer, R. (2012). Intra-Party Democracy and Party Responsiveness, Routledge West

European Politics, 35(6), 1295-1319.

Lisi, M. & Razzuoli, I. (2018). Intra-party democracy from members’ viewpoint: the case of left-wing parties in Portugal, Acta Politica, 53(3), 391-408.

Loxbo, K. (2013). The fate of intra-party democracy: Leadership autonomy and activist influence in the mass party and the cartel party, Party Politics, 19(4), 537-554. McKay, J.P., Hill, B.D., Buckler, J., Crowston, C.H., Wiesner-Hanks, M.E. & Perry, J. (2014).

A History Of Western Society. Basingstoke: Palgrave MacMillan.

Mühlböck, M. (2013) Linking Council and European Parliament? Voting unity of national parties in bicameral EU decision-making, Journal of European Public Policy, 20(4), 571-588.

O’Neil, P.H., Fields, K. & Share, D. (2015). Cases in Comparative Politics. London: Norton & Company Inc.

Pennings, P. & Hazan, R.Y. (2001). Democratizing Candidate Selection: Causes and Consequences, Party Politics, 7(3), 267-275.

Poguntke, T., Scarrow, S. & Webb, P. (2017), Political Party Database Version 1a.3. Rahat, G. & Shapira, A. (2017). An Intra-Party Democracy Index: Theory, Design and A

Demonstration, Parliamentary Affairs, 70(1), 84-110.

Raymond, C.D. (2014). The Effect of the Secret Ballot on Party System Fragmentation: A Test of Three Competing Arguments, Politics, 34(4), 378-390.

24 Rombi, S. & Seddone, A. (2017). Rebel Rebel. Do Primary Elections Affect Legislators’

Behaviour? Insights from Italy, Parliamentary Affairs, 70(3), 569-588. Sartori, G. (1965). Democratic Theory. New York: Frederick A. Praeger, Inc. Schattschneider, E. E. (1942) Party Government, New York, Farrar and Rinehart.

Schmitt, H., Bartolini, S., Brug, W. van der, Eijk, C. van der, Franklin, M., Fuchs, D., Toka, G., Marsh, M. & Thomassen, J. (2009): European Election Study 2004 (2nd edition). GESIS Data Archive, Cologne.

Schmitt, H., Hobolt, S. B., Popa, S. A., Teperoglou, E., & European Parliament, Directorate-General for Communication, Public Monitoring Unit (2016): European Parliament Election Study 2014, Voter Study, First Post-Election Survey. GESIS Data Archive, Cologne.

Schmitter, P. C. and Karl, T. L. (1991) ‘What is Democracy . . . and is Not’, Journal of Democracy, 2, 75–88.

Scully, R., Simon, H. & Farrell, D.M. (2012). National or European Parliamentarians? Evidence from a New Survey of the Members of the European Parliament* JCMS:

Journal of Common Market Studies, 50(4), 670-683.

Shomer, Y. Lavy, E.G. (2018). Does intra-party democracy affect levels of trust in parties? The cases of Belgium and Israel Shomer, Acta Politica, 53(2), 167-183.

Sieberer, U, (2006). Party unity in parliamentary democracies: A comparative analysis. The

Journal of Legislative Studies, 12(2), 150–178.

Teorell, J. (1999). A Deliberative Defence of Intra-Party Democracy, Party Politics, 5(3), 363-382.

The Comparative Study of Electoral Systems (www.cses.org). CSES INTEGRATED

MODULE DATASET (IMD) [dataset and documentation]. December 4, 2018 version Venice Commission (2009) ‘Code of Good Practice in the Field of Political Parties’, adopted

by the Venice Commission at its 77th plenary session, Venice, 12–13 December 2008.

Vick, K. (2013). Israel’s right turn, Time, 181(3), 1.

Wolkenstein, F. (2016). A Deliberative Model of Intra‐Party Democracy, Journal of Political

Philosophy, 24(3), 297-320.

Wolkenstein, F. (2018). Membership ballots and the value of intra-party democracy,

Routledge Critical Review of International Social and Political Philosophy, 21(4), 433