Introduction

1Party cohesion is a crucial feature of daily life of parliaments. The existence of unified blocs of legislators constitutes an important requisite in many normative accounts of liberal democracy. Cohesiveness directly impacts government’s survival (Saafeld 2009), coalition behaviour and bargaining power of the party in public office (Bäck 2008, Giannetti and Benoit 2009, Pedersen 2010, Tsebelis 2002). It shapes the electoral strategies, influences the electoral success or failure of the party in the electorate (Boucek 2012, McAllister 1991) and may have important consequences at the intra-party level, ranging from altering candidate selection procedures (Cox and Rosenbluth 1996, Kohno 1992) to shaping the party’s positions (Harmel and Tan 2003). Yet cohesiveness varies across time, across parliaments and across parties (Depauw 2003, Malloy 2003, Patzelt 2003). Scholarly effort has been dedicated to put forth some explanations to this variance –mostly, institutional explanations (Carey 2007, Coman 2015, Depauw and Martin 2009, Sieberer 2006). This article aims at improving our knowledge of the mechanisms that may explain the different degrees of party cohesion found in political parties. It does so by clarifying the concepts of agreement and homogeneity of preferences and by bringing party ideology to the forefront of the explanation.

2According to the literature, party agreement refers to a situation of cohesion “stemming from homogeneous policy preferences among legislators of the same party” (Van Vonno et al. 2014: 112). Whereas theoretically intra-party agreement results from the homogeneity of preferences among MPs, empirically speaking, scholars have often equated both concepts, by measuring the extent to which an MP disagrees with her/his party as the distance between the MP’s position and the party’s placement (Andeweg and Thomassen 2011, Kam 2001, Kam 2009). However, these two concepts do not cover identical realities. At the individual level, whereas agreement refers to a situation in which an MP evaluates whether (s)he agrees with the concrete policy proposal s/he has to vote for, s/he may, at a more abstract level, have preferences that are more or less distant with the party’s average placement. At the aggregate-level, MPs’ preferences within a party might be heterogeneous, but this does not systematically result in a high level of disagreement within the party. When legislators feel that their views on important issues are not aligned with those of the party, they are more likely to express higher levels of disagreement. This paper puts that hypothesis to the test, by developing distinct measures of intra-party disagreement and ideological distance. Delineating between the two concepts is fundamental for a better understanding and a better conceptualization of party cohesion.

3If party cohesion has sometimes been used as a synonym of unity in parliamentary votes (Özbudun 1970), the present study delineates between the two concepts. It considers cohesion as the voluntary pathway to reach voting unity (Andeweg and Thomassen 2011, Bowler et al. 1999, Hazan 2003), by contrast to discipline that constitutes the compulsory one. Party cohesion would be made of ideological homogeneity or agreement, as well as of loyalty, deriving from legislators’ internalization of the norms of party unity and solidarity. Discipline rather refers to the tools and institutional mechanisms –e.g. use of carrots and sticks, whipping [1], division of labour [2], candidate selection procedures [3] etc.- available to the party leader to enforce unity when neither agreement nor loyalty is effective (Andeweg and Thomassen 2011, Van Vonno et al. 2014). This research focuses on the agreement dimension of cohesion, and thus leaves aside both loyalty and the disciplinary mechanisms affecting legislators’ final decision to toe or not the party line.

4In a second step, this article argues that the relationship between heterogeneous preferences and intra-party disagreement is dependent on two interrelated factors: party ideology –in terms of party family– and the type of issues on which preferences are heterogeneous. Looking at party family is quite innovative. Indeed, most studies including party ideology as a factor of cohesion have grasped the concept only in terms of left versus right placement, and have therefore missed an important part of the story. This article contends that ideology matters in the sense that party families have emerged around specific cleavages and have been structured around core issues. We argue that if parliamentarians’ preferences are heterogeneous around these core issues, it will translate into disagreement more visibly. The analysis partly supports that disagreement increases when MPs’ preferences are heterogeneous around issues that are at the core of the party’s ideology. In that way, the findings suggest that party ideology works as a mediator between the heterogeneity of preferences and intra-party disagreement

5The demonstration is based on the analysis of attitudinal data collected among 840 parliamentarians from 15 European national assemblies, in the frame of the PartiRep Comparative MP Survey [4]. The empirical contribution of the paper resides in aggregating at the party level both subjective and objective assessments of ideology and preferences. The advantage of focusing on the aggregate level is that it helps us to “uncover relationships that do not appear at the individual level” (Curtice 2007: 906). In this case, the aggregate-level perspective provides interesting insights into the way ideology influences the relationship between heterogeneous preferences and disagreement within a party.

6The article proceeds as follows. First, the main hypotheses and expectations are presented regarding the impact of heterogeneity and ideology on disagreement. Second, the data and method used in the analyses are described. Third, the article proceeds with the analyses and results. A final section summarizes and discusses the research’s main findings.

Heterogeneity, ideology and disagreement: hypotheses

7Introduced by Krehbiel (1993), the “preference-driven” approach claims that individual MPs’ policy preferences “are crucial to any understanding of legislative behaviour” (Willumsen and Öhberg 2013: 2). Legislators’ voting cohesion would primarily result from their shared preferences over the legislation voted in parliament: MPs within the same party have similar policy preferences, and they simply vote because of ideological convictions (Hazan 2003: 3). Cohesion is reached voluntarily rather than coercively.

8The literature suggests that the heterogeneity of preferences within a parliamentary party leads to voting disunity, because it increases the MPs’ propensity to disagree with their party. If, theoretically, disagreement results from the heterogeneity of preferences among MPs from the same party, the chain of causation has rarely been demonstrated. Indeed, empirically, disagreement has been equated to heterogeneous preferences. At the individual level, disagreement has been measured as the ideological distance between an MP’s position and her/his party’s placement (Andeweg and Thomassen 2011, Kam 2001, Kam 2009, Willumsen 2011), either on left-right or policy scales. At the aggregate level, general indexes of party agreement have been based on indicators such as the standard deviation of a party’s MPs’ self-positioning on left-right or policy scale, or on measures of dispersion – e.g. Van der Eijk’s (2001) agreement index [5].

9We argue that the two concepts do not cover identical realities, and should therefore be measured distinctly. Whereas (dis)agreement refers to a situation in which an MP evaluates whether (s)he agrees with a concrete policy proposal s/he has to vote for, on a broader and more abstract level his/her preferences may be more or less distant with the party’s average position. At the aggregate level, one might expect that if MPs within the same parliamentary party share very disperse ideological opinions and positions, it will be more difficult to find a high level of agreement with the party line when it comes to actual voting in parliament. Yet there is a lack of empirical studies on this relationship at the aggregate level. Therefore, the relationship between heterogeneity of preferences and intra-party disagreement needs to be put to the test, and this constitutes the first aim of this article. In order to achieve this aim, we develop distinct measures for intra-party disagreement and MP-party ideological distance, and we test the hypothesis that, ceteris paribus, i.e. assuming that all other factors (discipline and other institutional mechanisms) remain constant:

10H.1. The more dispersed ideological positions within a party are, the higher the level of disagreement.

11If the ideological heterogeneity within a party matters in determining its level of cohesion, the very nature of a party’s ideology itself should also be taken into account. Actually, even if ideology constitutes a fundamental element for political parties, few works have investigated its impact on party cohesion (Owens 2003: 26). The few studies investigating the relationship between a party’s ideology and its level of cohesion have focused on the party’s placement in the left-right spectrum: some studies argue that left-wing parties have a higher degree of cohesion or discipline (Bailer et al. 2011: 5-6, Damgaard 1995: 322), while others make this statement for right-wing or centrist parties (Owens 2003: 26). These studies have mostly focused on the linear relationship between the party’s positioning in the left-right spectrum and its level of voting cohesion, thus neglecting other dimensions of ideology. This article departs from these studies, by considering the effect of the party’s family instead of the effect of the party’s placement in the left-right spectrum.

12We argue that the relationship between ideological heterogeneity and the probability of disagreement is mediated by the ideological nature of the issues at stake.. This paper contends that MPs’ preferences should be more homogeneous on what we called the party’s core issues; but also that, given their centrality and visibility in the party’s identity and day-to-day legislative work, disagreement is more likely to occur when MPs’ preferences are heterogeneous on these core issues. This concept stems from the discussion on issue ownership (Budge and Farlie 1983, Petrocik 1996) defined by Walgrave et al. (2012) as “the fact that specific political parties are, in voters’ minds, identified with specific policy issues and considered best able to deal with them” (2012:771). These authors point out the fact that this definition mixes a “competence component” –which party is perceived by citizens to deal best with an issue– and an “associative component” –the link of specific parties with specific issues. In this sense, our approach would rather be focused on those issues to which parties pay more attention; that is, to the associative component. We argue that voters are most able to link parties with certain issues when parties pay more attention to them. Typically this attention translates in holding more extreme positions on these topics. As such parties’ positioning on these core issues allows differentiating between parties in the minds of voters. This approach therefore also relates to van der Brug’s (2004) view of issue ownership as a question of party “priorities” (Bellucci 2006, Bélanger and Meguid 2008).

13These core issues are fundamental from both an exogenous and an endogenous point of view. Firstly, from an exogenous point of view, the party’s core issues constitute the basis of the framework of reference among voters (Pennings et al. 2006: 206) ; or, in other words, the party’s “brand” identity –how the party wants to be perceived by the voters and what makes the party or party family different from other competitors in the electoral arena. Parties “‘own’ specific issues in the sense that they are the natural bearers of this ideology” (Pennings et al. 2006: 206).

14The link between party ideology and these core issues stems from the fact that party families in the European political space have emerged and have been structured around specific cleavages or divisions (Lipset and Rokkan 1967, Rokkan 1970): historically, the Church/state; the rural/urban; and the owner/worker or ‘class-cleavage’. These divisions have varied in duration and strength across Europe (Bartolini and Mair 1990). As Kriesi argues, today these four classic cleavages “essentially came to boil down to two: a cultural (religion) and a social-economic one (class)” (Kriesi et al. 2006: 923). In addition, a new structuring conflict seems to have developed since the late 1960s and is now “embedded into existing two-dimensional national political spaces” (Kriesi et al. 2006: 921). Various labels apply to this new cleavage: the “new value” (Inglehart 1977, Inglehart 1990), the “new politics” (Franklin 1992) or the “postmaterialism/materialism” cleavage (Ignazi 2003: 201). Hooghe et al. (2002) also talk about an opposition between a “GAL” pole (Green-Alternative-Libertarian) and a “TAN” pole (Traditional-Authoritarian-Nationalism).

15We assume that the policy positions of parties can be located in those conflicts that structure the actual cleavage system. Parties’ preferences “may be described in terms of a position in these conflict [or issue] dimensions” (Narud and Skare 1999: 50). After all, “a cleavage delineates the social base on which parties build their support” (Narud and Skare 1999: 50). Issue dimensions can be associated with the traditional left-right socio-economic policy dimension (class cleavage), but can also pertain to other dimensions: morality issues (derived from the old Church/state cleavage), environmental issues (embedded in the new value cleavage) etc. Parties differ in the weight or the relative importance they attach to these dimensions and in terms of the position they adopt on these conflict dimensions (Narud and Skare 1999: 50-51).

16Secondly, these core policy areas are important from an endogenous perspective: the recruitment of political personnel is linked to these areas. For instance, in the case of a Green MP, it is possible that s/he has had previous working experience in the field of NGO of environmental protection and therefore may pay more attention to the party’s positions in this area than, say, in the field of justice reforms. In the case of Social democratic parties, part of their political personnel is likely to be recruited in workers’ associations and unions and among employees; while in the case of Liberal and Conservative parties, candidates are likely to be recruited among managers and self-employed. MPs from these parties might therefore pay special attention to their party’s placement on socio-economic issues.

17Parliamentarians’ preferences are expected to be very homogenous around the party’s core issues. Homogeneity is expected because of a double process of (self)selection: “parties select candidates whose preferences are most in line with that of the party, and candidates self-select into parties […], adopting the party label that most closely suits their own policy preferences” (Van Vonno et al. 2014: 114). However, heterogeneous preferences on the party’s core issues may occur. Given the saliency of these issues in the party’s ideology, such heterogeneity is more likely to be translated into disagreement between the MP and her/his party. The impact of heterogeneous preferences on disagreement is thus mediated by both the party’s ideology and the type of issues on which preferences are heterogeneous.

18H2: The more preferences are heterogeneous on the core policy issues of the party ideology, the greater the level of disagreement.

19Next section describes the data on which this study in based, then presents the conceptualization, expectations and operationalization of each ideological family.

Data and method

Ideology and ideological distance: operationalizing the independent variables

26The dataset includes parties from diverse ideological horizons (see table 2 in Appendix). Seven party families are identified, from the more leftist to the more rightist: the Radical Left, the Social Democrats, the Greens, the Liberals, the Christian Democrats, the Conservatives and the Radical right. Some ideological families (Radical left, Green and Radical Right) include very few parties (three, two and three respectively) compared to the Social democratic (14 parties), Conservative (12), Liberal (9) and Christian democratic (7) families. Since the Green family only includes two parties, our second hypothesis can hardly be tested for this group of parties, but we include it for informative and comparative purposes. In addition, since the Radical Left and Radical Right families only include 3 parties each, the results should be considered carefully.

27As regard the other independent variable –heterogeneity of preferences–, our analyses are based on two distinct measures. First, a ‘subjective’ measure [12], for which we compute the left-right ideological distance between each MP and her/his party, using a question asking MPs to position themselves and their national party on a 10 point left-right scale [13]. Second, we focus on positions within each party around specific policy issues. Five types of policy issues are considered, which are derived from traditional and more recent cleavages: two socio-economic issues (class cleavage), a morality issue (Church/state or cultural cleavage), a security issue and an immigration issue (both related to the new politics cleavage).

28These are based on questions that ask MPs to position themselves on several policy issues –socio-economic, morality, security, and immigration issues [14]. The party’s position on each policy has been computed by equating it to the position of its average MP. Heterogeneity of preferences is measured as the standard deviation of each party’s average position. In spite of the limitations [15] that this type of scales pose, using different measures contributes to improve conventional measures of intraparty heterogeneity of preferences and provides a better understanding of the mechanisms underlying party cohesion.

29Table 1 shows the policy dimension issues on which we base our analysis as provided by the PartiRep dataset, the cleavages to which they are linked and the ideological families with more extreme (positive or negative) positions. We depart from the assumption that paying more attention to certain policy dimensions with the aim of being associated with them usually entails that those parties express more extreme positive or negative positions on these policy dimensions. According to our second hypothesis, we expect that a higher heterogeneity of preferences on core issues is related to a higher disagreement; therefore, that the relationship between heterogeneity on specific issues and disagreement varies across party families.

Table 1. Ideological families’ core issues

Policy item	Cleavages	Ideological families with more extreme positions*
Larger income differences are needed as incentives for individual effort.	class cleavage	Liberals (3.23) Radical Left (1.52) Social Democrats (1.89)
Government should play a smaller role in the management of the economy	class cleavage	Liberals (3.44) Radical left (1.4) Social Democrats (2.03)
People who break the law should be given stiffer sentences	new politics cleavage	Radical Right (4.04) Green (2.44) Radical Left (2.17)
Immigrants should be required to adapt to the customs of our country	new politics cleavage	Radical Right (4.38) Green (2.69) Radical Left (2.92)
Government should make sure that films and magazines uphold moral standards	Church/state or cultural cleavage	Conservatives (3.04) Radical Left (2.05) Liberal (2.08)

Table 1. Ideological families’ core issues

* Average position across parties in the Partirep dataset in parentheses.

30The socio-economic cleavage “has dominated the political conflict for most of this century” (Narud and Skare 1999: 50). Accordingly, we could have similar expectations for all party families. But we expect that heterogeneity should strongly translate into disagreement in those groups with a clearer (more extreme) position on the socio-economic cleavage: Liberals, Radical Left and Social Democrats. These are the ones for which a more positive (Liberals) or a more negative (Radical Left and Social Democrats) position is found on the items asking about the need for income differences to incentivize individual effort and about the need for a smaller role of government in the management of the economy.

31Two issues have been selected that are representative of the new politics cleavage: security, and most notably, immigration. Arguably, these issues are at the core of the ideology of the party families that have crystallized around the new politics cleavage: the Radical Right and the Radical Left. On one side, the Radical Right has been “fuelled [by] the need for self-defence and self-reassurance” (Ignazi 2003: 201), and has promoted the value of law and order (Hagtvet 1994). The relationship between the heterogeneity of preferences on the immigration and security issues and intra-party disagreement should be manifest in the Radical Right family, as these issues are central in these parties’ program –especially, the immigration issue (Delwit 2007: 11, Mayer 1997: 17-8, Mudde 1999). On the other side, Radical Left parties have aggregated postmaterialist values of self-affirmation or self-fulfilment, they have been “combating all kinds of diehard domination” (Escalona and Vieira 2013: 8) and have been at the forefront of the fight for immigrants’ rights. The Green family [16] can also be said to be rooted on this cleavage based on postmaterialist values (Bennulf and Holmberg 1990, Müller-Rommel 1990). Issues related to the new cleavage politics are at the core of their ideology: we therefore expect a positive relationship between the heterogeneity of MPs’ preferences and intra-party disagreement.

32As regard the Church/state or cultural cleavage we analyse a cultural or moral issue that measures MPs’ agreement with the fact that governments should guarantee moral standards in the media. The traditional cultural or religious opposition has mostly concerned the Liberals and, in many cases, the Radical Left, on the secular and progressive side, and the Conservatives, on the traditional side of the conflict. As such the expectation is that heterogeneous preferences –regardless of their direction– on that issue should result in higher disagreement in the case of these three families. This expectation should also be applicable to the Christian Democratic family, but its average position on this item is not as extreme as in the three families.

Heterogeneity and disagreement: the mediating role of ideology

33While at the individual level it might seem straightforward to expect more disagreement from those MPs who feel that their position is different from their party´s, this does not hold at the party level. Figure 1 displays the average subjective ideological distance for each party and the percentage of MPs that frequently disagree with their party. It shows that there is no correlation between both variables (P=-0.122, sig=0.398, n=50) even when other variables such as country or party size are controlled for. Hypothesis 1 is therefore not confirmed. The claim often made in the literature that heterogeneity of preferences can be equated to disagreement should accordingly be revised.

34

Figure 1. Subjective ideological heterogeneity and level of disagreement.

Figure 1. Subjective ideological heterogeneity and level of disagreement.

35Figure 2 introduces party ideology as a potential factor of disagreement. Figure 2 shows the average left-right positioning of each party family as expressed by the parties’ MPs and the level of disagreement. At first sight, one could conclude that there is a greater disagreement in those families that are situated more to the left of the left-right continuum. Yet, a closer look at the graph instead reveals that traditional families –Social Democrats, Conservative and Christian Democrats– have a quite similar level of disagreement. Great differences appear between the two radical families, with the Radical Left showing high levels of intra-party disagreement, and the Radical Right showing very low levels. Variances are relatively important inside the party families. Yet, an analysis of variance reveals significant differences across groups: Radical right parties have an average proportion of disagreeing MPs significantly lower than Liberal and Radical Left parties (LSD test, p value<0.05). These findings suggest that the role of party ideologies should not be understood simply in terms of left-right effect, but well in terms of families.

36

Figure 2. Left-right position and level of disagreement.

Figure 2. Left-right position and level of disagreement.

37Introducing party ideology as a mediating factor between ideological heterogeneity and disagreement provides further interesting perspectives. Figure 3 shows that the ideological heterogeneity measured through the subjective assessment of ideological distance becomes much more informative when party families are considered. It can be observed that in some party families both variables –heterogeneity and disagreement– share a common, though not perfect, pattern. Families that are located more to the right tend to have lower ideological distances and, consequently, less disagreement. Within the Social Democratic family, both variables seem to be congruent. However, in some families, the relationship is not as expected. This is the case, for instance, of the Radical Left family, and to a lesser extent of the Liberal family: they show an ideological distance that is below what should be expected given they high levels of disagreement. Another interesting distinction is between Christian democrats and Conservatives: while both have a relatively similar degree of disagreement (around 35% of MPs frequently disagree), the ideological heterogeneity seems to be higher in the former than in the latter. A reason for this relative lack of fit might be found on Hypothesis 2, which relates the heterogeneity of preferences around the party ideology core issues to the level of disagreement.

38

Figure 3. Subjective distance and level of disagreement

Figure 3. Subjective distance and level of disagreement

39In the theoretical section, we argued that greater disagreement is expected to stem from more dispersed positions, especially when this dispersion takes place around the party ideology core issues. In this section we look into this relationship through two different perspectives. On the one hand, we analyse absolute levels of dispersion across party families in each policy issue (Table 2) and, on the other, we examine whether the relationship holds within each party family (Figure 5). Erreur ! Source du renvoi introuvable. shows the dispersion of policy preferences of each party family in each of the policy items considered: distribution of income in society, size of government, the need for stiffer sentences, the adaptation of immigrants to the country´s customs and the position on imposing moral standards in media. Figures 5 to 9 show the extent to which there is a relationship between the heterogeneity of preferences and disagreement within each party family for each policy item.

40Obviously, while these figures do help to compare two general dimensions –the dispersion of MPs’ preferences on issues and their self-reported frequency of disagreement–, they are plagued with serious empirical limitations. On the one hand, the issues are very general topics that may not be much grounded in the reality of parliamentary debates in the legislatures under study. This is certainly the case for instance of the moral issue, that appears a bit outdated. On the other hand, what should be examined, but is not available in the dataset, is the self-reported frequency of disagreement on these particular issues, and not MPs’ very general feeling. Such data would have allowed to better assess the sequential relationship between ideological distance and disagreement depending on the issue at stake. Unfortunately, we do not possess such empirical material, but we encourage future researches to test this hypothesis with more accurate data. Nevertheless, we think that the findings presented below are already telling regarding the potential mediating role of ideology in this sequential mechanism. Indeed, these figures permit the exploration of the dynamics that occur inside party families. In the following paragraphs we address the interpretation of these figures for each party family.

Radical Right

49The Radical Right family stands out as the ideological family that displays the lowest level of disagreement and, at the same time, the most cohesive preferences. This ideological family exceeds the level of homogeneity of the rest of ideologies in almost all the items. The only item for which the Radical Right does not follow the same pattern is on the position on government size. For this item, this family has the highest score of heterogeneity, which might also be related to the differences regarding this topic across parties of this family.

50Radical Right parties appear to have a more extreme position –core issues– in two of the items analysed here: the one on the need for stiffer sentences and the one that points at the adaptation of immigrants to the country´s customs (New Politics cleavage). Although in the former item the relationship is not confirmed (the slope being nearly flat), in the latter we can see a clear positive slope.

Table 2: Heterogeneity across ideological families and disagreement

Table 2: Heterogeneity across ideological families and disagreement

51

Figure 4. Socio-economic issue (larger income differences are needed as incentives for individual effort)

Figure 4. Socio-economic issue (larger income differences are needed as incentives for individual effort)

Figure 5. Socio-economic issue (position on size of government)

Figure 5. Socio-economic issue (position on size of government)

52

Figure 6. Moral issue

Figure 6. Moral issue

Figure 7. Security issue

Figure 7. Security issue

53

Figure 8. Immigration issue

Figure 8. Immigration issue

Discussion

54In sum, several interesting patterns emerge from this analysis. First of all, equating heterogeneity of preferences to disagreement is misleading. This research has shown that this relationship does not hold at the party level [17].

55Second, associating the degree of disagreement with the party family’s core issues reveals interesting patterns. Parties tend to have on average more cohesive preferences around topics that are at the core of their ideology. Those ideologies that are more heavily centred on issues along the socio-economic cleavage (Radical Left, Social Democrats or Liberals) display more homogeneous preferences than the other ideologies on these issues. Similarly, the families that pay more attention to issues related to the New Politics cleavage (Radical Right, Radical Left, Conservatives) also share less dispersed views on these core issues.

56Third, the positive relationship between heterogeneity of preferences on core issues and disagreement has been confirmed in most of the cases, although, as suggested, our measure of disagreement is too general to grasp the accuracy of the sequential relationship. The only instances in which such positive relationship is not found are the Radical Left (only for the item on size of government), the Conservatives (on moral issues), and the Radical Right (no relationship is found on the item of judicial sentences). However, we also find unexpected positive relationships, which do not support the idea that disagreement only stems from heterogeneity on core issues. Positive relationships are found for most issues within the Christian Democratic family, although we had no specific expectation given the ideological centrality –or uncertainty– of this family. Positive relationships are found within the Radical Right family on the socio-economic issue, and within Liberals and Conservative on the New politics issues. Interestingly, while these issues were not defined as core issues for these families, they might well constitute important topics on which these families are actually realigning –a process that certainly result in a high frequency of disagreement. Our second hypothesis should be revised accordingly. On the basis of these results, we can nevertheless argue that the relationship between heterogeneity in policy positioning and the frequency of disagreement is mediated by the party ideology.

57Fourth, party ideology matters in this relationship, and not only in terms of left-right positioning. In some party families, heterogeneity (either measured subjectively or on policy issues) follows the same pattern as disagreement: this is the case of the Conservatives and the Radical Right. But in others, the relationship between ideological heterogeneity and disagreement contradicts what is usually assumed. The Radical Left and Liberal families show at the same time relatively high levels of intra-party disagreement and homogeneous preferences in terms of policy issues, but also in terms of subjective distance on the left-right scale. By contrast, the Social Democratic family displays rather high levels of ideological heterogeneity, with relatively low levels of disagreement.

58The more or less important mismatch observed between the degree of homogeneity of preferences and MPs’ reported frequency of disagreement in fact constitutes one of the most important contribution of this research. Explaining these variations require digging more into the role of the political culture and identity of the party families, but also into their common organizational structure (see Close 2015). Party ideology, intended as a set of values and principles, can directly affect legislators’ perception of role and integration of behavioural norms. Liberal or libertarian values embodied in Liberal, Green and certain Radical left parties might bring about a greater emphasis on self-affirmation and self-fulfilment and thus a more individualistic style of representation. By contrast, right-wing values of order and discipline might translate into a lesser tendency for MPs to state their disagreement. At the organizational level, some families tend to foster internal democracy and MPs’ freedom of choice more than others –e.g. Green parties– , which rather promote a greater centralization of intra-party decision around a charismatic leader – e.g. the Conservatives and the Radical right. MPs in Social-democratic parties might be more prompt to appear as a unified bloc, as these parties usually promote a collective style of representation and are organized accordingly. Such organizational devices often reflect the party families’ foundational ideological values (Gauja 2013), and certainly impact on the importance MPs attribute to party unity or freedom to dissent.

59Obviously, the mismatch between homogeneity of preferences and level of disagreement informs on the values embodied in the different ideologies, but also gives some insights into the relative capacities and willingness of the parties to socialize their members to the norm of party unity or to appear as democratic organizations. The impact of both the ideological and organizational components of party families on MPs’ socialisation to the norm of party unity and on the psychological dynamics of cohesion (Raymong and Overby 2014, Russell 2014) certainly deserve greater attention.

Conclusion

60This article provides new insights regarding the roots of parliamentary party cohesion. At the empirical level this paper brings evidence of its claims at the party level by aggregating survey responses from 840 parliamentarians from 50 parties across 15 national parliaments. While the relatively small number of parties in our sample makes it difficult to run satisfactorily more complex multivariate statistical analyses, our results shed light over the relationship between party cohesion and ideology.

61Several contributions need to be highlighted. The first is conceptual. Our results show that homogeneity of preferences and agreement with the party line are different phenomena. Preferences within parties might be heterogeneous and yet the party might enjoy high levels of agreement. This goes counter a common practice in the field: the one consisting on equating agreement and homogeneity of preferences. The second contribution of this article is theoretical. Most studies of party cohesion have focused on institutional and individual-level factors. The effect of ideology has been reduced to the placement in the left-right scale and has been so far underexplored. In this article we contend that there is a need to include in party cohesion studies a more complex and rich concept of ideology. In so doing we examine the dynamics of party cohesion across political parties according to their political family. Ideologies are hence considered to be primarily frameworks of shared ideas. We base our argument on the policy dimensions to which each political family attributes more importance: their core dimensions. We show how the frequency of disagreement within parties is related to heterogeneous differences regarding these core values.

62The results we present in this article show that there is a need to delve more deeply into the factors that lead to party cohesion. More in-depth analyses of the process of agreement and preference formation within party families are therefore needed to unveil the precise mechanism that would lead to greater disagreement. While here the focus has been put on the party, there is a need to explore more precisely how the possible underlying mechanisms work across the party and the individual levels, and how individuals and parties influence each other’s. Party ideology and ideological families should therefore remain on the party cohesion research agenda.

Acknowledgement

63Our thanks go to Émilie van Haute, Pascal Delwit, Jean-Benoit Pilet, Kris Deschouwer, Rudy Andeweg and David Farrell for their encouragements and helpful comments.

Funding

64This work was supported by a F.R.S.-FNRS grant and the PARTIREP research project, an Interuniversity Attraction Pole (IAP) funded by the Belgian Science Policy.