Unpacking Party Unity: The Combined Effects of Electoral Systems and Candidate Selection Methods on Legislative Attitudes and Behavioural Norms

Abstract

This article analyses the effect of electoral systems, candidate selection methods and the interplay between them on individual legislative attitudes and behavioural norms, specifically on two facets of party unity: party agreement and party loyalty. Our main argument is that one must take into account the effect of inter- as well as intra-party competition, and the interaction between the two, in order to explain individual legislative attitudes and norms. Using data from 34 European parties across 10 countries, we show that under exclusive candidate selection methods, there are large differences between proportional representation and single-member district electoral systems in their effect on party agreement and party loyalty. Under inclusive candidate selection methods, however, such differences are much less apparent. In other words, the candidate selection method conditions the influence of the electoral system on legislative attitudes and behavioural norms.

Keywords

electoral systems candidate selection party unity

Introduction

This article empirically tests the combined effect of electoral systems (inter-party competition) and candidate selection methods (intra-party competition) on legislative attitudes and behavioural norms. We focus on two facets of party unity: party agreement – defined as the extent to which co-partisans share similar policy preferences with one another when voting on legislation – and party loyalty – defined as the extent to which legislators subscribe to the norm of party unity in the absence of such shared preferences. The academic literature has yet to address this two-dimensional approach empirically, moving beyond explanations of legislative attitudes and behaviour that only rest on the influence of electoral systems. We argue that the candidate selection method conditions the effect of the electoral system on legislative attitudes and behavioural norms. Under exclusive candidate selection methods, we will see substantial differences in party agreement and loyalty between legislators elected by proportional representation (PR) electoral systems and those elected by plurality/majority single-member district (SMD) electoral systems. On the contrary, under inclusive candidate selection methods, such differences will shrink and legislators elected by both systems will resemble each other much more regarding party agreement and party loyalty.

Using attitudinal data from the PARTIREP¹ project on elected representatives in 15 countries, combined with data collected on candidate selection methods in 34 European parties in 10 countries (for a full list of the countries and parties included in this research, see Appendix 1), we are able to show that the candidate selection method and its characteristics condition the effect of the electoral system on party agreement and loyalty. More specifically, when the party leadership controls legislators’ chances for reselection, we find substantial differences in the effect of the electoral system on party agreement and loyalty: legislators elected by party-focused PR electoral systems demonstrate higher levels of party agreement and loyalty compared to legislators elected by candidate-based SMD systems. However, when the party leaders do not control the nomination process, the type of electoral formula has a much smaller influence on legislative attitudes and behavioural norms.

Party unity is conceptualised in this article as an end result of the legislators’ sequential decision-making process. That is, we do not examine the final outcome – legislators’ voting behaviour – but rather focus on legislators’ pre-floor attitudes: party agreement and party loyalty. By doing so, we follow an increasing number of scholars and studies who adopt the attitudinal approach to the study of party unity (Andeweg and Thomassen, 2011; Kam, 2009; Krehbiel, 1993). The reason for this is first and foremost theoretical: party unity can be the result of many different mechanisms, which cannot be bound together in a uniform way (which is what we do when we only look at the final outcome). Unity can be the result of legislators’ shared preferences (party agreement), of a voluntary subscription to the norm that legislators should toe their party line even when they personally disagree with it (party loyalty) or of the use of involuntary disciplinary instruments by the party leadership (party discipline). While the end result in all of these cases is the same voting unity, the intra-party dynamics leading to it can vary dramatically. Clearly, the dynamics in a party where there is high party agreement is different from the dynamics in a party whose legislators lack party agreement but possess high levels of party loyalty; and these dynamics are very different from those in a party whose legislators lack both party agreement and party loyalty, thus requiring discipline. This changes the atmosphere within the party, the dynamics between co-partisans and, most importantly, the nature of the relationship between the party leadership and its members of parliament (MPs). Therefore, it is important to flesh out these processes and analyse the factors affecting them. Following that, we cannot claim that institutions, such as the ones studied in this article, affect party unity directly. On the contrary, we claim that institutions do not affect voting behaviour directly; rather, they influence the legislators’ decision-making process in determining whether to cast their vote with their party or against it.

But there is also an empirical reason for our decision to use attitudinal data on party unity. While the legislators’ voting behaviour is heavily constrained – West European parliamentary systems provide very little variance in party unity (Carey, 2009; Depauw and Martin, 2009; Heidar, 2006; Sieberer, 2006) – their attitudes and norms of behaviour regarding party unity are far less constrained and present much more variance. In order to overcome the no-variance problem we face, we can investigate party unity – both as it is perceived by the legislators and how it exhibits itself in the legislature – by conceptualising it as a sequential decision-making process. Hence, while this research does not address legislators’ actual voting behaviour, it provides an understanding of the inner processes legislators and parties undergo before a legislator’s vote is cast in the plenum.

Background

Almost any prospective or incumbent politician in any country must pass two barriers, not one, in order to become an elected representative – they must first be chosen by a party selectorate, be it a party elite or the party members, and only subsequently be elected by the voters. Each of these two barriers can vary in relevance and in nature – each can be either more personal or more under party control – but both are present, both are significant and both must be taken into account if we want to understand legislative behaviour. The second barrier, the electoral system, has received a plethora of scholarly attention, producing numerous studies of its diverse effects on politics in general and on the legislative branch in particular.² However, the first barrier, the candidate selection method – and especially its influence on legislative behaviour – remains largely disregarded by the academic literature and has only recently begun to attract scholarly attention (Alvarez and Sinclair, 2012; Hazan and Rahat, 2010, 2006; Katz and Cross, 2013; Narud et al., 2002; Rahat and Hazan, 2001; Tavits, 2008, 2005). In other words, most scholars attribute legislative behaviour to the way legislators are elected by the general public, and not to the way they are selected by their parties.

Since electoral systems shape the conditions under which legislators compete for re-election, they are expected to affect both party agreement and party loyalty, depending on the candidate-centred versus party-based nature of the system (Van Vonno et al., 2014). While the majority of the literature finds the ballot structure to be the key characteristic in shaping the candidate-centred versus party-based nature of the electoral system, there are studies that emphasise the importance of the electoral formula and the district magnitude in shaping it. For example, a recent study on how electoral systems affect legislative behaviour in France, Portugal and Belgium found that of the three main characteristics of the electoral system, the most influential on the constituency versus party focus of representation among legislators is the electoral formula. ‘When “representing district/region” versus “party” are compared, majoritarian systems with SMD, as opposed to PR systems with multi-member districts (MMD), seem to be more important in explaining the focus on district/region than either district magnitude … or ballot structure’ (Pilet et al., 2012: 366).

The study by Pilet et al. (2012) is the most recent to address the importance of the electoral formula in shaping legislative behaviour, but it is by no means the first. Authors like Mayhew (1974), Fenno (1978), Bowler and Farrell (1993) and Stratmann and Baur (2002) claimed that plurality/majority SMD systems create stronger ties between legislators and their constituency and provide incentives for legislators to engage in constituency service or even pork-barrel politics. Thus, Farrell and Scully (2007) write:

Research suggests that representatives working under SMD systems are much more likely than those under closed-list systems to feel the need to nurse a personal vote; equally, voters under SMD systems are more likely to pay attention to the work of individual MPs (Farrell and Scully, 2007: 52).

Indeed, in a study on Australia, Canada, Ireland, New Zealand and the United Kingdom, Heitshusen et al. (2005) found that MPs’ constituency attention is greater under SMDs compared to MMDs. Bradbury and Mitchell (2007) found that in Wales and Scotland, district members tend to hold many more surgeries, compared to list members. Lundberg (2007) found that in Germany and the United Kingdom, district members tend to contact with constituents more frequently than list members. Stratmann and Baur (2002) found that in Germany committee membership is divided asymmetrically, with district members holding more positions in the so-called pork committees than list members. Finally, André and Depauw (2013) found that MPs elected in SMD electoral systems spend more working hours in their district and consider themselves as ‘local promoters’ even more so than MPs elected in open-list, multi-member, PR electoral systems.

In this article, we chose to use the electoral formula as the defining characteristic of the electoral system since, as exemplified by the abovementioned studies, it allows us to distinguish between a candidate-centred SMD electoral system associated with a constituency-based focus of representation and a party-centred PR electoral system associated with a more generalist party-based focus and to test their effect on party unity. However, while these studies examined the effect of the electoral system on legislators’ behaviour, we examine how it affects legislators’ attitudes and norms of behaviour. We clearly acknowledge that attitudes are not behaviours, but we believe it is important to study them, especially with regard to party unity. In other words, we believe we can be more accurate in our understanding of party unity if we conceptualise it as an end result of a multi-stage process and attempt to explain how different institutional arrangements affect the different mechanisms involved in it. Those mechanisms, we claim, are based on legislators’ attitudes and norms of behaviour and that is why it is important to detect the causal mechanisms that affect them.

We expect the relationship between the electoral system and party agreement/loyalty to be as follows: legislators elected under SMD electoral systems will be highly committed to their constituency, a fact which may harm party agreement insofar as the aggregate preferences of the constituencies legislators represent differ from the national party policy (Kam, 2009). Alternatively, legislators elected through PR multi-member electoral systems lack a strong connection to their constituencies and can therefore share their party’s policy preferences more easily. This is true even if such legislators are elected by a preference vote, as those often tend to be quite random, and do not represent an organised, well-defined interest such as that of the constituency. Similarly, the strong ties between legislators elected by an SMD electoral system and their constituents may result in the creation of ‘competing principals’ (Carey, 2009) for legislators, since legislators who feel they owe their election to their constituents, and not to their co-partisans, are more likely to adopt a norm according to which legislators may sometimes vote against the party line when the latter is at odds with their personal bases of support.

When it comes to candidate selection methods, they can be classified according to several criteria, but the central criterion is the inclusiveness of the selectorate – who may take part in selecting the party candidates (Hazan and Rahat, 2010). At the inclusive pole are parties that allow all voters to select the party candidates, with no restrictions whatsoever. At the exclusive pole are parties that allow a party leader to single-handedly select the party candidates. In between are parties that allow the party elite, party delegates or party members to select the party candidates. The common convention in the literature is that the more inclusive the selectorate, the less control the party leadership has over the process of candidate selection and, as a result, over the behaviour of the candidates selected (Bowler et al., 1999; Cross, 2008; Katz and Cross, 2013). The inclusiveness of the selectorate thus affects party unity in at least two ways. First, it affects party agreement. Selectorates that are more inclusive hamper the ability of the party leadership to create a united, ideologically coherent, homogeneous list. As a result, the heterogeneity of preferences among legislators who belong to parties that use inclusive selectorates is higher than among legislators who belong to parties that select their candidates by methods that are more exclusive. Second, inclusive selectorates affect party loyalty since they create a dual source of legitimacy for legislators. Legislators are thus exposed to various external pressures which could be different from, or even contradictory to, that of the party programme and to which they must be responsive (Atmor et al., 2011; Hazan and Rahat, 2010; Rahat, 2008).³

The literature thus leads us to expect that electoral systems and candidate selection methods both have an effect on party agreement and party loyalty. While studies of candidate selection have increased, there is as yet no systemic analysis of the connection between the two dimensions – electoral systems and candidate selection methods. A few authors mention both dimensions; however, this is usually done only in passing and without empirically examining the connection between the two.⁴ The two-dimensional approach has strong theoretical foundations, but it has never been tested. This study tests the two-dimensional approach empirically, for the first time.

Argument

This article argues that the influence of the electoral system on legislative behavioural norms is contingent on the nature of the candidate selection method. That is, we hypothesise that the effect of the first institution depends on the type of the second institution. We base our hypothesis on the fact that selection mechanisms precede electoral ones: before a candidate can run in a general election, he or she has to survive the selection stage. This enables the candidate selection method to set the ground rules and to shape the arena in which legislators operate. The electoral system does not have the ability to completely counter the general context set by the candidate selection method, but it certainly has the ability to make significant changes both within and to the given framework. Our hypothesis is best explained using Figure 1.

Figure 1.

The Hypothesised Effect of Electoral Systems, Candidate Selection Methods and Their Interaction on Party Agreement and Party Loyalty.

Figure 1 shows the independent effect which the electoral system and the candidate selection method have on party agreement and loyalty, as well as the interaction between the two. While the figure presents a simplified illustration of our hypothesis, there are a few assumptions underlying its creation.⁵ First, the independent effect of the electoral system is captured by the fact that PR systems are expected to produce higher levels of agreement and loyalty compared to SMD systems, regardless of the nature of the candidate selection method (hence the difference in the intercepts). The independent effect of the candidate selection method is captured by the fact that exclusive selectorates are expected to produce higher levels of party agreement and party loyalty compared to inclusive selectorates, regardless of the nature of the electoral system. Moving from an exclusive selectorate to an inclusive one, the slope of both lines, representing PR and SMD electoral systems, is negative. However, the slope is clearly sharper in PR systems than in SMD ones because we expect the decline in party agreement and loyalty to be more dramatic in party-based electoral systems – which start off with rather high levels of agreement and loyalty – than in candidate-based ones – in which there are lower levels of agreement and loyalty to begin with.⁶

The difference in slopes between PR and SMD systems – which creates the change in distance between the two lines – represents the interaction effect between the electoral system and the candidate selection method in their effect on party agreement and loyalty. As candidate selection methods become more inclusive, the gap between the two lines grows smaller. Hence, the largest difference between PR and SMD electoral systems exists when the selectorate is most exclusive. As the selectorate becomes more inclusive, the differences between the two electoral systems shrink, demonstrating the limited ability of the electoral system to affect party agreement and party loyalty under inclusive selectorates.

When the party uses an exclusive selectorate to choose its candidates – be that the party leader or a small party caucus – it can select candidates whose ideological preferences are as close as possible to the party’s agenda and as a result ensure high levels of party agreement. Additionally, since the party controls the most important resource for legislators – their ability to be reselected, and as a result re-elected – an exclusive candidate selection method provides clear, party-centred incentives to legislators. Legislators know that if they wish to be reselected by their party, they must behave as team players, demonstrating high levels of party loyalty.

However, as mentioned above, legislators not only need to be reselected by their party, they also need to be re-elected by the voters. This is where the electoral system comes into play, affecting the incentives offered to legislators and, as a result, their attitudes and norms of behaviour. The electoral system can either reinforce the incentives the candidate selection method offers legislators or it can contradict them.

When an exclusive party-based selectorate is coupled with a PR party-based electoral system, the two reinforce each other, resulting in high levels of party agreement and loyalty. Unsurprisingly, legislators who are both selected and elected under party-based rules are expected to share the ideological preferences of their party and to demonstrate substantial loyalty towards it. However, when the party-based incentives of the candidate selection method are coupled with the personally focused incentives of an SMD electoral system, legislators are faced with conflicting incentives, which may pull in opposite directions. Introducing a candidate-based electoral system to an exclusive selectorate may thus decrease the level of shared preferences legislators share with their party, as well as create competing principals legislators must be accountable to. This harms party agreement and loyalty and creates the gap seen in Figure 1 between legislators elected by a PR electoral system and exclusive selectorates to those elected by an SMD system.

The dynamics portrayed above – of reinforcing and conflicting incentives given the nature of the candidate selection method and the electoral system – are relevant, we claim, mostly for exclusive selectorates. When the party uses inclusive selectorates, the nature of the electoral system matters much less. Once a party democratises its candidate selection method, making it more inclusive, it loses its control over the process. Inclusive selectorates offer legislators personally focused incentives and limit, to some extent, the ability of the party to control the identity of the candidates selected. While it is true that here too we can portray a reinforcing incentives scenario – when an inclusive selectorate is coupled with an SMD electoral system – and a conflicting incentives scenario – when such a selectorate is coupled with a PR electoral system – we claim that once the selectorate becomes inclusive, this dynamic becomes largely irrelevant. In other words, inclusive selectorates shape party agreement and loyalty to such an extent that the electoral system cannot do much about it, whether it is a party-focused PR electoral system or a personally based SMD one. Once legislators understand that they owe their reselection – which is a prerequisite for their re-election – to someone other than the party and its leadership, there is no going back. Given the fact that the dynamics of democratising candidate selection are so influential (and destructive) to party agreement and loyalty, we argue that the electoral system does not have much of a say on these issues, if any. That is, PR electoral systems – countering the strong personally based incentives offered by an inclusive candidate selection method – will not be able to strengthen party agreement and loyalty, and SMD electoral systems – reinforcing those strong personally based incentives – will not further weaken party agreement or loyalty any more than it already is under an inclusive candidate selection method.⁷

Method

In order to account for the effects of electoral systems and candidate selection methods on legislative attitudes and behavioural norms, this article relies on the attitudinal data collected for the international PARTIREP project. These data are integrated with data on candidate selection methods gathered from various sources such as party regulations, expert surveys and published articles. The PARTIREP project involved a quantitative cross-national survey carried out among members of 15 national European parliaments during the period 2009–2011. We analyse cases from 10 of the 15 countries, encompassing 34 political parties. We chose these cases because they are the ones for which we also have reliable data on their candidate selection methods.

In terms of the representativeness of the data, the average response rate in the dataset created is nearly 25%, which is considered rather satisfactory in this type of survey. Response rates fluctuate from country to country, ranging from 7.2% (France) to 34.6% (Belgium) (see Appendix 1). Despite the difficulty in obtaining the data and its limitations, it is relatively representative: it presents an adequate division between PR and SMD systems, as well as encompassing parties from the entire political spectrum, varying in size and in coalition participation. The Duncan index of dissimilarity, which measures the percentage differences between the distribution in the population and the sample, shows that the selection closely resembles the population in most respects (for an elaboration of the representativeness of the PARTIREP survey, see Deschouwer et al., 2014).

The distribution of the selectorate variable in the dataset is also quite representative, with a majority of cases categorised as inclusive. This resembles the global trend of democratising candidate selection methods (see histograms in Appendix 2). Due to the unequal number of observations we have from different parties (see Appendix 1), we use weights in all analyses that correct for these differences.

As mentioned earlier, we conceptualise party unity as the end result of the legislators’ three-staged sequential decision-making process (see Figure 2). In the first stage of the process – the party agreement stage – legislators consider whether or not they agree with their party’s position. If they agree, then party unity is the result of party agreement, hence shared preferences. If, however, legislators do not agree with the position of their party, they then proceed to the second stage – the party loyalty stage – where they decide whether or not to subscribe to the norm of party loyalty and vote according to the party line, despite their disagreement. If legislators decide not to subscribe to the norm of party loyalty, the third and last stage of the process – the party discipline stage – includes the use of disciplinary measures by the party leadership in order to achieve party unity.

Figure 2.

A Sequential Multi-Stage Model of Party Unity: Agreement, Loyalty and Discipline.

In order to measure party agreement, legislators were asked, ‘How often, in the last year, would you say you have found yourself in the position that your party had one opinion on a vote in parliament, and you personally had a different opinion?’⁸ In order to measure party loyalty, legislators were asked, ‘How should a Member of Parliament vote if his or her own opinion on a vote in parliament does not correspond with the opinion of his/her party?’⁹ It is important to stress that MPs’ responses to these questions cannot be seen as predictors of how they would actually vote in cases of disagreement because if they were, then parliamentary dissent would be much greater than it is. Here, we must resort to our sequential model, claiming that those MPs who believe an MP should not vote according to the party line when faced with a disagreement will move on to the third stage of the model, where they will be disciplined by their party.

Table 1 presents country descriptive statistics for our two dependent variables. While 60% (352/588) of legislators indicated that they are frequently in agreement with their party – they disagree with their party only about once a year to almost never – 40% frequently disagreed with their party positions. Country differences are apparent: the percentage of legislators who frequently agree with their party and therefore demonstrate party agreement is highest in Ireland (80%), Spain (76%) and the Netherlands (67%), and lowest in Hungary (51%) and Israel (30%).

Table 1.

Party Agreement and Loyalty – Descriptive Statistics.

Country	Party agreement	Party loyalty
Belgium	0.61	0.71
France	0.57	0.36
Germany	0.54	0.44
Hungary	0.51	0.61
Ireland	0.80	0.79
Israel	0.30	0.46
The Netherlands	0.66	0.87
Norway	0.61	0.89
Spain	0.76	0.80
United Kingdom	0.53	0.46
Total	0.60	0.62

The table presents the mean values of party agreement and loyalty per country. Both variables are binary.

As for party loyalty, 62% (345/553) of legislators responded that legislators should vote according to the party line when it conflicts with their own opinion, whereas 38% think that a legislator should stick to his or her own opinion and break party unity. In terms of country differences, the percentage of legislators who subscribe to the norm of party loyalty is highest in Norway (89%), the Netherlands (87%) and Spain (80%), and lowest in Germany (44%) and France (36%).

As explained above, of the three main characteristics of the electoral system – electoral formula, district magnitude and ballot structure¹⁰ – we chose to focus on the electoral formula as the defining variable of the electoral system, differentiating between plurality/majority SMD electoral systems and PR MMD electoral systems.^11,12

As for the candidate selection method, we chose to focus on what the recent literature points to as the most relevant factor when examining legislative behaviour – the inclusiveness of the selectorate (Best and Cotta, 2000; Hazan and Rahat, 2010; Norris et al., 1990). The inclusiveness of the selectorate is a scale variable borrowed from Hazan and Rahat’s (2010) 24-point inclusiveness index (see Hazan and Rahat, 2010: 48–49). In this scale, every 6 points relate to one selectorate or category: single leader (0), party elite (6), party delegates (12), party members (18) and, finally, the general electorate (24).¹³ The distance of 6 points between each category is designed to systematically account for more complex selectorates which involve more than a single selectorate in the candidate selection process (for the selectorate score given to each party in the dataset, see Appendix 1; for elaboration on the operationalisation of the selectorate variable, see Appendix 2).

When analysing the data, and in order to leverage the sequential logic presented earlier, we use a fixed-effects multinomial logistic regression model. That is, our dependent variable is a categorical index of party unity formed by the four possible combinations between the party agreement and party loyalty variables. This way, instead of modelling the two dummy dependent variables independently, we are able to model them together and truly account for the sequential logic. The ‘unity index’ is created by a 2-by-2 matrix presenting the four possible combinations between our two dummy dependent variables: from the most unified category to the most non-unified one (see Table 2).

Table 2.

Party Unity Index.

	Frequently agree with party	Frequently disagree with party
Vote with party in case of disagreement	Agree and loyal	Disagree but loyal
Vote own opinion in case of disagreement	Agree but disloyal	Disagree and disloyal

The highest degree of party unity is demonstrated when legislators frequently agree with their party and also adopt the norm of party loyalty (agree and loyal). In this case, legislators toe the party line simply because they agree with it and unity is the result of shared preferences (party loyalty hardly comes into play in these cases). The second option is less straightforward since it includes legislators who claim to frequently agree with their party yet who believe that in a case of disagreement a legislator should vote against the party line (agree but disloyal). We consider this a small and interim category of legislators, who in the vast majority of cases share their parties’ preferences and therefore will vote with their parties; however, should they disagree with their parties, they will not necessarily comply with the norm of party loyalty. While these legislators still demonstrate relatively high values of party unity, it is not as high as the ones in the previous category. The third option is the classic party loyalty category, which includes those legislators who frequently disagree with their party but believe that despite such dis-agreement an MP should vote according to the party line, as a result of internalised norms (disagree but loyal). Finally, the fourth option includes those legislators who both frequently disagree with their party and did not internalise the norm of party loyalty (disagree and disloyal). These legislators are those who will potentially dissent, unless they will be effectively disciplined by their party.

Our model also includes a number of control variables: parliamentary seniority, government participation, gender, MP’s age and party size. Those were included as they are often addressed in the literature as variables that may affect our dependent variables (see, for example, André and Depauw, 2014; André et al., 2014; Shomer, 2014; Van Vonno et al., 2014). For example, we can postulate that long-serving MPs may represent their party’s views better or, alternatively, have greater freedom because of their seniority.

The Combined Effect of Electoral and Selectoral Mechanisms on Party Unity

Table 3 presents the results of our multinomial logistic model, with the base category being the first option discussed above, agree and loyal.¹⁴ That is, the results should be analysed with reference to the highest degree of party unity in our index: legislators who frequently agree with their party and also believe that in a case of disagreement a legislator should still vote according to the party line.

Table 3.

The Effect of the Electoral Formula and the Inclusiveness of the Selectorate on Party Unity.

	Agree but disloyal (2)		Disagree but loyal (3)		Disagree and disloyal (4)
	B	SE	B	SE	B	SE
Electoral formula (SMD)	3.56***	1.29	1.79	1.33	2.22*	1.20
Selectorate inclusiveness	0.13**	0.06	0.13***	0.05	0.08*	0.04
Electoral formula × Selectorate inclusiveness	−0.29***	0.09	−0.15	0.09	−0.18**	0.08
Government participation	.042	0.38	0.43	0.35	0.59	0.38
MP’s age	−0.01	0.01	0.01	0.01	−0.03**	0.01
Party size	−4.5**	1.93	−0.88	2.13	−6.48***	1.85
Female	−0.33	0.29	−0.56**	0.28	−0.58*	0.32
Seniority	0.02	0.02	−0.00	0.01	0.02	0.02
Belgium	−1.69*	0.93	0.44	0.96	−3.23***	0.95
France	−0.32	0.74	−0.29	0.97	0.10	0.71
Germany	−0.91	0.58	0.15	0.65	−0.92	0.56
Hungary	−0.98	0.66	0.57	0.67	−0.48	0.61
Ireland	−2.15**	0.84	−1.05	0.88	−2.86***	0.97
Israel	−1.77*	1.06	1.02	1.02	−0.77	0.93
The Netherlands	−2.88***	0.91	−0.26	0.85	−4.18***	1.20
Norway	−2.78***	1.06	0.44	0.84	−3.27***	1.02
Spain	−0.98	0.80	0.11	0.80	−1.97**	0.86
United Kingdom	Omitted		Omitted		Omitted
Intercept	−12.38	38.6	−8.48	36.3	22.1	40.4
Log likelihood	−620.81
Pseudo R²	0.11
N	527

SE: standard error; SMD: single-member district; MP: member of parliament.

Results are presented as log of the odds.

The reference category is that of the highest level of party unity – agree and loyal.

p < 0.1; **p < 0.05; ***p < 0.01.

Looking at Table 3, we can see that the electoral system coefficient has a positive value in all three response categories.¹⁵ That is, when changing the electoral formula from PR to SMD, the relative risk of a legislator demonstrating a lower level of party unity, compared to the highest category of unity in our index (agree and loyal), increases. The same is true for the selectorate variable. In all response categories, the selectorate has a positive, statistically significant effect, meaning that as the selectorate becomes more inclusive, the relative risk of a legislator demonstrating disunity increases. As for the interaction term, it was found to be statistically significant only in response categories 2 and 4 of our dependent variable.

Notwithstanding the above analysis, since this is an interaction model, we cannot interpret these coefficients at face value. Each one of the lower order terms (the coefficients for the electoral formula and the inclusiveness of the selectorate) only has a substantive meaning at a specific value of the other (Brambor et al., 2006). For example, the value of the coefficient for the electoral system refers only to the magnitude of the change from a PR to an SMD electoral system when the inclusiveness of the selectorate is equal to zero, that is, most exclusive. For example, a change from a PR to an SMD electoral system increases the relative risk of possible dissent (the category of disagree and disloyal) by a factor of 9.2, compared to the highest category of party unity (agree and loyal). However, we do not know what happens to this effect once the selectorate is different from zero. The same is true for the selectorate coefficient. Therefore, in order to understand the interaction dynamics, we plot the predicted probabilities of the highest level of unity in our index (the category of agree and loyal) in PR and SMD systems given different levels of the inclusiveness of the selectorate.¹⁶

Figure 3 supports our hypothesis in three ways. First, it shows that unity is higher in PR electoral systems compared to SMD systems, regardless of the nature of the selectorate. From the most exclusive selectorates to the most inclusive ones, the predicted values of unity are higher for legislators elected in PR electoral systems than for those elected in SMD ones. Second, it shows that unity is higher in exclusive selectorates compared to inclusive ones, regardless of the nature of the electoral system. Democratising the candidate selection method decreases the predicted probability of unity for legislators elected in both PR and SMD electoral systems. Third, it demonstrates the interaction between electoral and selectoral mechanisms, showing the capability of the candidate selection method to condition the effect of the electoral system on party unity. The narrowing gap between PR and SMD electoral systems shows how inclusive selectorates limit the ability of the electoral system to affect party agreement and loyalty, whereas under exclusive selectorates it is much more influential. True, the gap does not narrow as dramatically as we hypothesised, yet there is still a noteworthy difference in the gap between PR and SMD electoral systems under exclusive selectorates (a 0.52 difference in the predicted probability of party unity) compared to the same gap under inclusive selectorates (a 0.38 difference in the predicted probability of party unity).

Figure 3.

Predicted Probabilities of Party Unity for PR and SMD Electoral Systems at Different Levels of Selectorate Inclusiveness (with 95% CIs).

It is here that we should also consider the effect of institutional rules on the behaviour of the selectorates themselves or the rationality of these actors. This is based on the assumption that different selectorates have different motivations when choosing candidates: exclusive selectorates are oriented towards office-seeking behaviour and will thus adapt their choice to ensure their candidates get re-elected, whereas inclusive selectorates might be more policy-seeking or have a motivation to select candidates who are more oriented towards the constituency-based style of representation. According to these distinct motivations, different selectorates might be more or less responsive to the electoral system.

This brings us to present yet another way to interpret our findings, according to which exclusive selectorates are more responsive to the electoral rules than inclusive ones: they strategically select more loyal and more ‘agreeing’ legislators in party-based systems, while in candidate-based systems, they will allow a lower integration of the norm of party loyalty and a higher frequency of disagreement because they believe it will be more rewarding electorally. Alternatively, when inclusive selectorates choose candidates, they tend to choose individuals they believe will best defend their interest and preferences, regardless of the electoral system.

This interpretation suggests that it is also possible for the electoral system to condition the effect of the candidate selection method. While we claim that it is the other way around and back this claim with strong theoretical reasoning, we acknowledge that mutual effects may be at work here as demonstrated above and that the conditioning effect may work in both directions, to some extent. Still we believe that the main conditioning effect is reserved to the candidate selection method, for reasons detailed earlier in the article.¹⁷

Neither one of the control variables was very consistent in their effect on our dependent variable. MP’s age proved to negatively affect the possibility of a legislator being found in the lowest category of our unity index (disagree and disloyal), compared to the highest category of the index, but did not have a significant effect in the other two response categories. Party size and gender had a similar effect to that of MP’s age in only two response categories: 2 and 4, and 3 and 4, respectively. The same is true for the country fixed-effects, which proved to have a rather sporadic influence on party unity.

The interaction between the electoral system and the candidate selection rules shows that the key to understanding internal party dynamics lies first and foremost in the nature of the candidate selection method. In other words, the identity of those who control the nomination process has a substantial influence on legislative attitudes and behavioural norms, not only directly but also indirectly by conditioning the ability of the electoral system to affect those attributes.

This can help us evaluate the need for, and the use of, party discipline under different institutional scenarios. It is important because party discipline is a last resort, one that party leaders wish to avoid. Frequent out-in-the-open use of party discipline can result in bad publicity for the party and its leaders, who are perceived as incapable of controlling their representatives. Here, we reemphasise that candidate selection is in itself a disciplinary measure that leaders can mobilise in case of dissenting legislators. Our findings show that when leaders can make use of such a tool, they are able to control, to some extent, the level of potential dissent by selecting legislators who agree more often with the party and are more loyal to it. Given the fact that parties prefer to avoid the use of party discipline when possible, or to use it in ways that are hidden from the public eye, this enriches our understanding of the sequential logic (Heidar and Koole, 2000; Olson, 2003).

Our findings with regard to the sequential model can thus explain how electoral institutions, candidate selection rules and the interaction between them affect the frequency of shared preferences and the probability of legislators voluntarily subscribing to the norm of party loyalty in the absence of shared preferences. As a result, we can assess when party leaders will need to use discipline in order to achieve the high unity levels we are so accustomed to seeing and tap into the inner dynamics of parties operating under different electoral and candidate selection rules.

Conclusion

This article analysed the effects of electoral systems, candidate selection methods and the interplay between them on legislative attitudes and behavioural norms, and specifically focused on two facets of party unity: party agreement and party loyalty. Our main argument was that one must take into account the effects of both inter- and intra-party competition and the interaction between them, in order to explain legislators’ attitudes and norms of behaviour better.

Using attitudinal data from the PARTIREP project covering 15 countries, combined with data collected on candidate selection methods in 34 European parties in 10 countries, we argued that the candidate selection method – the first barrier legislators must cross in order to fulfil their goal of re-election – has the ability to condition the effect of the electoral system on legislative attitudes and behavioural norms. More specifically, we claimed that the less control the party holds over the nomination process, the more limited is the ability of the electoral system to affect party agreement and party loyalty, given the damage already done to party agreement and party loyalty in the nomination stage.

Our elected representatives function in a highly complex environment, and their behaviour is influenced by a variety of factors: cultural, sociological, institutional, electoral and more. The literature has shown that electoral systems matter, and we posit that so do candidate selection methods. But what matters the most is the combination between them and the way they interact with one another. We believe that this research promotes and advances our efforts to understand what influences legislator’s attitudes and behavioural norms with regard to party unity, and in doing so, it moves us from the traditional unidimensional perspective to a two-dimensional inter- and intra-party approach. Yet, we should not be circumspect in our attempt to explain, much less predict, legislative behaviour. Our findings are preliminary, and more research is needed, but they do shed light on an important explanatory variable that has only recently begun to receive scholarly attention. Legislative behaviour is indeed difficult to explain, yet that does not mean that we should be satisfied with partial arguments based largely on a single explanatory variable.

Footnotes

Appendix 1. Countries and Parties Included in the Research.

Country	Ballot structure	Electoral formula	Number of responses	Return rate	Party	Selectorate score	N (party)^a
Belgium	Semi-open list	PR	52	34.6	MR (Reform Movement)	9	8
					CD&V (Christian Democratic and Flemish)	11	12
					Open VLD (Open Flemish Liberals and Democrats)	9	6
					PS (Socialist Party)	11	10
					SPA (Socialist Party Differently)	11	9
					VB (Flemish Interest)	6	7
France	N/A	Plurality/majority	42	7.2	UMP (Union for Popular Movement)	7	21
France	N/A	Plurality/majority	42	7.2	PS (Socialist Party)	17	21
Germany	Closed-list (for those elected through a list)	Mixed	115	18.2	CDU/CSU (Christian Democratic Union of Germany/Christian Social Union of Bavaria)	15	38
					SPD (Social Democratic Party of Germany)	15	33
					Die Linke (The Left)	15	26
					G (Green Party)	15	18
Hungary	Closed-list (for those elected through a list)	Mixed	84	21.7	Fidesz (Hungarian Civic Union)	4	36
Hungary	Closed-list (for those elected through a list)	Mixed	84	21.7	MSZP (Hungarian Socialist Party)	14	48
Ireland	N/A (open ballot with no party lists)	PR (STV)	30	18	Fianna Fáil	11	14
					Fine Gael	17	10
					Labour	17	6
Israel	Closed-list	PR	27	22.5	Likud	18	4
					Kadima	18	11
					Labour	18	5
					Yisrael Beytenu	0	4
					Shas	0	3
The Netherlands	Semi-open list	PR	51	34	CDA (Christian Democratic Appeal)	16	21
					PvdA (Labour Party)	9	19
					VVD (The People’s Party for Freedom and Democracy)	9	11
Norway	Semi-open list	PR	40	23.6	DNA (Labour party)	9	16
					FRP (Progress Party)	7	11
					Høyre (Conservative Party)	9	7
					SV (Socialist Left Party)	9	6
Spain	Closed-list	PR	94	26.8	PP (People’s Party)	8	26
Spain	Closed-list	PR	94	26.8	PSOE (Spanish Socialist Workers’ Party)	9	68
United Kingdom	N/A	Plurality/majority	59	9.1	Conservative	14	22
					Labour	14	30
					Liberal Democrats	16	7

PR: proportional representation; STV: single transferable vote.

To account for differences in the number of respondents from different parties, we use party weights in all analyses.

Appendix 2 Appendix 3. Robustness Checks.

	Model 1 (bivariate effect – electoral system)		Model 2 (bivariate effect – candidate selection)		Model 3 (interaction without controls)
	B	SE	B	SE	B	SE
Frequently agree and vote own opinion (2)
Electoral formula (SMD)	0.75***	0.24			3.19***	0.86
Selectorate inclusiveness			0.08***	0.03	0.13***	0.04
Electoral formula × Selectorate inclusiveness					−0.21***	0.06
Infrequently agree and vote with party (3)
Electoral formula (SMD)	0.11	0.25			0.94	0.99
Selectorate inclusiveness			0.07**	0.02	0.09***	0.03
Electoral formula × Selectorate inclusiveness					−0.08	0.07
Infrequently agree and vote own opinion (4)
Electoral formula (SMD)	1.06***	0.24			3.4***	0.89
Selectorate inclusiveness			0.11***	0.03	0.16***	0.04
Electoral formula × Selectorate inclusiveness					−0.20***	0.06
Constant	−1.19***	0.15	−2.10***	0.39	−3.01***	0.52)
Log likelihood	−718.02		−721.35		−705.07
Pseudo R²	0.02		0.01		0.03
N	553		553		553

SE: standard error; SMD: single-member district.

Results are presented as log of the odds.

The reference category is that of the highest level of party unity – frequently agree and vote with party.

p < 0.1; **p < 0.05; ***p < 0.01.

Acknowledgements

The authors would like to thank Gideon Rahat, Orit Kedar and the three anonymous reviewers for their comments on earlier drafts of this paper; Cynthia Van Vonno for helping elaborate the sequential model; and Nir Atmor for access to his data.

Funding

The authors acknowledge BELSPO for funding the PARTIREP ‘Participation and Representation in Modern Democracies’ international study.

Notes

Author Biographies

Reut Itzkovitch-Malka is an Assistant Professor in the Department of Sociology, Political Science and Communications at the Open University of Israel. Prior to joining the academic faculty of the Open University, she was a Postdoctoral Fellow at Stanford University’s Center on Democracy, Development and the Rule of Law. Her research interests include legislative studies, political representation, electoral systems, candidate selection methods and gender and politics. Her articles have appeared in journals such as Representation. She is co-author of The Representation of Women in Israeli Politics: A Comparative Perspective (The Israel Democracy Institute, 2013) and co-author of a chapter in Representing the People: A Survey among Members of Statewide and Sub-State Parliaments (Oxford University Press, 2014).

Reuven Y Hazan is a Professor in and former Chairperson of the Department of Political Science at the Hebrew University of Jerusalem. His research interests are elections and electoral systems, political parties and party systems, and legislative studies. He is the co-author of Democracy within Parties: Candidate Selection Methods and Their Political Consequences (Oxford University Press, 2010) and co-editor of Understanding Electoral Reform (Routledge, 2012). His articles have appeared in Comparative Political Studies, Electoral Studies, International Political Science Review, Legislative Studies Quarterly, Journal of Legislative Studies, Journal of Theoretical Politics, Party Politics, Political Geography and West European Politics, among others.

References

Alvarez

Sinclair

(2012) Electoral Institutions and Legislative Behavior: The Effects of Primary Elections. Political Research Quarterly 65 (3): 544–557.

Andeweg

Thomassen

(2011) Pathways to Party Unity: Sanctions, Loyalty, Homogeneity, and Division of Labor in the Dutch Parliament. Party Politics 17 (5): 655–672.

André

Depauw

(2013) District Magnitude and Home Styles of Representation in European Democracies. West European Politics 36 (5): 986–1006.

André

Depauw

(2014) District Magnitude and the Personal Vote. Electoral Studies 35: 102–114.

André

Gallagher

Sandri

(2014) Legislators’ Constituency Orientation. In: Deschouwer

Depauw

(eds) Representing the People: A Survey among Members of Statewide and Sub-State Parliaments. Oxford: Oxford University Press, pp.166–187.

Atmor

Hazan

Rahat

(2011) Moving beyond the Electoral System: The Influence of Candidate Selection on Personal and Party Representation. In: Colomer

(ed.) Personal Representation: The Neglected Dimension of Electoral Systems. Colchester: European Consortium for Political Research Press, pp.21–36.

Best

Cotta

(2000) Elite Transformation and Modes of Representation since the Mid-Nineteenth Century: Some Theoretical Considerations. In: Best

Cotta

(eds) Parliamentary Representatives in Europe 1848–2000: Legislative Recruitment and Careers in Eleven European Countries. Oxford: Oxford University Press, pp.1–28.

Bille

(2001) Democratizing a Democratic Procedure: Myth or Reality? Candidate Selection in Western European Parties 1960–90. Party Politics 7 (3): 363–380.

Blais

(1988) The Classification of Electoral Systems. European Journal of Political Research 16 (1): 99–110.

10.

Blais

Carty

(1991) The Psychological Impact of Electoral Laws: Measuring Duverger’s Elusive Factor. British Journal of Political Science 21 (1): 79–93.

11.

Blais

Massicotte

(2002) Electoral Systems. In: LeDuc

Niemi

Norris

(eds) Comparing Democracies 2: New Challenges in the Study of Elections and Voting. London: SAGE, pp.40–69.

12.

Bowler

Farrell

(1993) Legislator Shirking and Voter Monitoring: Impacts of European Parliament Electoral Systems upon Legislator-Voter Relationships. Journal of Common Market Studies 31 (1): 45–70.

13.

Bowler

Farrell

Katz

(1999) Party Cohesion, Party Discipline, and Parliaments. In: Bowler

Farrell

Katz

(eds) Party Discipline and Parliamentary Government. Columbus, OH: Ohio State University Press, pp.3–22.

14.

Bradbury

Mitchell

(2007) The Constituency Work of Members of the Scottish Parliament and National Assembly for Wales: Approaches, Relationships, and Rules. Regional & Federal Studies 17 (1): 117–146.

15.

Brambor

Clark

Golder

(2006) Understanding Interaction Models: Improving Empirical Analyses. Political Analysis 14 (2): 63–82.

16.

Carey

(2009) Legislative Voting and Accountability. Cambridge: Cambridge University Press.

17.

Carey

Shugart

(1995) Incentives to Cultivate a Personal Vote: A Rank Ordering of Electoral Formulas. Electoral Studies 14 (4): 417–439.

18.

Cross

(2008) Democratic Norms and Party Candidate Selection: Taking Contextual Factors into Account. Party Politics 14 (5): 596–619.

19.

Depauw

Martin

(2009) Legislative Party Discipline and Cohesion in Comparative Perspective. In: Giannetti

Benoit

(eds) Intra-Party Politics and Coalition Governments. New York: Taylor & Francis, pp.103–120.

20.

Deschouwer

Depauw

André

(2014) Representing the People in Parliaments. In: Deschouwer

Depauw

(eds) Representing the People: A Survey among Members of Statewide and Sub-State Parliaments. Oxford: Oxford University Press, pp.1–18.

21.

Duverger

(1954) Political Parties: Their Organization and Activity in the Modern State. London: Methuen Books.

22.

Esteban

Guerra

(1985) Electoral Rules and Candidate Selection. In: Penniman

Mujal-Leon

(eds) Spain at the Polls 1977, 1979 and 1982. Washington, DC: American Enterprise Institute, pp.48–72.

23.

Farrell

(2011) Electoral Systems: A Comparative Introduction. Basingstoke: Palgrave Macmillan.

24.

Farrell

Scully

(2007) Representing Europe’s Citizens? Electoral Institutions and the Failure of Parliamentary Representation. Oxford: Oxford University Press.

25.

Fenno

(1978) Home Style: House Members in Their Districts. Boston, MA: Little, Brown.

26.

Gallagher

(1988) Introduction. In: Gallagher

Marsh

(eds) Candidate Selection in Comparative Perspective: The Secret Garden of Politics. London: SAGE, pp.1–19.

27.

Hazan

Rahat

(2006) The Influence of Candidate Selection Methods on Legislatures and Legislators: Theoretical Propositions, Methodological Suggestions and Empirical Evidence. Journal of Legislative Studies 12 (3–4): 366–385.

28.

Hazan

Rahat

(2010) Democracy within Parties: Candidate Selection Methods and Their Political Consequences. Oxford: Oxford University Press.

29.

Hazan

Voerman

(2006) Electoral Systems and Candidate Selection. Acta Politica 41 (2): 146–162.

30.

Heidar

(2006) Parliamentary Party Group Unity: Does the Electoral System Matter? Acta Politica 41 (3): 249–266.

31.

Heidar

Koole

(2000) Parliamentary Party Groups in European Democracies: Political Parties behind Closed Doors. London; New York: Routledge.

32.

Heitshusen

Young

Wood

(2005) Electoral Context and MP Constituency Focus in Australia, Canada, Ireland, New Zealand, and the United Kingdom. American Journal of Political Science 49 (1): 32–45.

33.

Kam

(2009) Party Discipline and Parliamentary Politics. Cambridge: Cambridge University Press.

34.

Katz

Cross

(2013) Problematizing Intra-Party Democracy. In: Cross

Katz

(eds) The Challenges of Intra-Party Democracy. Oxford: Oxford University Press, pp.170–176.

35.

Kingdon

(1977) Models of Legislative Voting. Journal of Politics 39 (3): 563–595.

36.

Krehbiel

(1993) Where’s the Party? British Journal of Political Science 23 (2): 235–266.

37.

Lundberg

(2007) Proportional Representation and the Constituency Role in Britain. Basingstoke: Palgrave Macmillan.

38.

Lundell

(2004) Determinants of Candidate Selection. Party Politics 10 (1): 25–47.

39.

Mayhew

(1974) Congress: The Electoral Connection. New Haven, CT: Yale University Press.

40.

Narud

Pedersen

Valen

(2002) Party Sovereignty and Citizen Control: Selecting Candidates for Parliamentary Elections in Denmark, Finland, Iceland and Norway. Odense: University Press of Southern Denmark.

41.

Norris

Carty

Erickson

. (1990) Party Selectorates in Australia, Britain and Canada: Prolegomena for Research in the 1990s. The Journal of Commonwealth & Comparative Politics 28 (2): 219–245.

42.

Olson

(2003) Cohesion and Discipline Revisited: Contingent Unity in the Parliamentary Party Group. Journal of Legislative Studies 9 (4): 164–178.

43.

Pilet

Freire

Costa

(2012) Ballot Structure, District Magnitude and Constituency-Orientation of MPs in Proportional Representation and Majority Electoral Systems. Representation 48 (4): 359–372.

44.

Rae

(1967) The Political Consequences of Electoral Laws. New Haven, CT: Yale University Press.

45.

Rahat

(2008) Entering through the Back Door: Non-Party Actors in Intra-Party (S)electoral Politics. In: Farrell

Schmitt-Beck

(eds) Non-Party Actors in Electoral Politics: The Role of Interest Groups and Independent Citizens in Contemporary Election Campaigns. Baden-Baden: Nomos Verlagsgesellschaft, pp.25–44.

46.

Rahat

Hazan

(2001) Candidate Selection Methods: An Analytical Framework. Party Politics 7 (3): 297–322.

47.

Rahat

Hazan

Katz

(2008) Democracy and Political Parties: On the Uneasy Relationships between Participation, Competition and Representation. Party Politics 14 (6): 663–683.

48.

Reynolds

Reilly

Ellis

(2005) Electoral System Design: The New International IDEA Handbook. Stockholm: International Institute for Democracy and Electoral Assistance.

49.

Sartori

(1976) Parties and Party Systems: A Framework for Analysis. Cambridge: Cambridge University Press.

50.

Shomer

(2010) Electoral Systems and Intra-Party Candidate Selection Processes: Influences on Legislators’ Behavior. PhD Dissertation, Washington University in Saint Louis, St. Louis, MO.

51.

Shomer

(2014) What Affects Candidate Selection Processes? A Cross-National Examination. Party Politics 20 (4): 533–546.

52.

Sieberer

(2006) Party Unity in Parliamentary Democracies: A Comparative Analysis. Journal of Legislative Studies 12 (2): 150–178.

53.

Stratmann

Baur

(2002) Plurality Rule, Proportional Representation and the German Bundestag: How Incentives to Pork-Barrel Differ across Electoral Systems. American Journal of Political Science 46 (3): 506–514.

54.

Taagepera

Shugart

(1989) Seats and Votes: The Effects and Determinants of Electoral Systems. New Haven, CT: Yale University Press.

55.

Tavits

(2005) The Development of Stable Party Support: Electoral Dynamics in Post-Communist Europe. American Journal of Political Science 49 (2): 283–298.

56.

Tavits

(2008) Party Systems in the Making: The Emergence and Success of New Parties in New Democracies. British Journal of Political Science 38 (1): 113–133.

57.

Van Vonno

Itzkovitch-Malka

Depauw

. (2014) Agreement, Loyalty and Discipline: A Sequential Approach to Party Unity. In: Deschouwer

Depauw

(eds) Representing the People: A Survey among Members of Statewide and Sub-State Parliaments. Oxford: Oxford University Press, pp.110–136.