A House Divided against Itself. The Intra ‐ institutional Con ﬂ ict about the Powers of the European Parliament

Many authors suggest that MEPs seek to extend the power of the EP in their daily parliamentary work. A central assumption in this literature is that the EP is a unitary actor that desires more power for itself. In other words, scholars assume that the con ﬂ ict over the power of the EP is inter ‐ institutional and not intra ‐ institutional. In this article we seek to test this idea. Informed by the history of the parliamentarization of Europe ’ s national parliaments, we argue that the con ﬂ ict about the power of the EP is likely to have an intra ‐ institutional component. An empirical analysis of voting in the EP supports this notion. We ﬁ nd that the question of how much power the EP should have is as important as the question of how much power the EU should have in understanding how MEPs vote.


Introduction 1
The literature on the parliamentarization of the EU is extensive. Many authors suggest that, in their daily work, MEPs seek to extend the power of the EP (Crum, 2006;Héritier, 2003, 2007;Hix, 2002;Yordanova, 2011;Rittberger, 2014). The EP is assumed to be a unitary actor that desires more power for itself Høyland, 2013, p. 174 Schoeller andHéritier, 2019, p. 288). In this article, we build upon existing work that examines the political conflict within the EP over the strenghtening of this body itself (Benedetto and Hix, 2007;Kreppel and Tsebelis, 1999;Rittberger, 2003). We investigate the extent to which MEPs are unified when it comes to supporting more power for the Parliament in their daily work. The central question that we seek to answer is therefore: to what extent do differences of opinion among MEPs over the powers of the EP divide them in their voting behaviour?
The EU is in constant constitutional development. Part of this development is a process of parliamentarization, that is, the 'gradual evolution towards a system of government at the European level that would resemble (…) parliamentary democracy in a nation state' (Lehman and Schunz, 2005, p. 10). European states have seen a similar process of parliamentarization. Parliaments have gradually won the time to deliberate, the ability to represent, the power to hold the government accountable and sovereignty over decision-making (Ihalainen et al., 2017, pp. 6-8). The powers these bodies have today is the result of an intra-institutional struggle between those who favoured greater parliamentary power and those who opposed it (Ihalainen et al., 2017, p. 1).
However, most existing studies on the parliamentarization of the EU understand it as the result of an inter-institutional conflict between the Council and the Parliament. In this conflict, the Parliament is often depicted as a unitary actor that prefers to extend its powers (Hix and Høyland, 2013, p. 174). Yet, given evidence that the extension of this power has been contested by MEPs, we propose that parlementarization is also subject to a intra-institutional conflict between parties with different preferences on the Parliament's power.
Our aim is to contribute to two fields of research: on the one hand, we seek to extend the comparative-historical literature on parliamentarization. We will show that intra-institutional conflict about the reach of parliamentary powers is still fought out at the European level (Tiilikainen and Wiesner, 2016). On the other hand, we seek to contribute to the literature on the dimensionality of the EP. We propose to move beyond the dominant Hix-Lord model, which focuses on macro-level policy differences between MEPs, and instead delve into the actual conflicts between MEPs, which are more complex than this two-dimensional model suggests.
We also aim to make methodological contributions to both of these fields. The study of parliamentarization has mainly relied on methods that focus on the historicity and contigency of this process and on the role of specific actors (but see Häge, 2017;Herrmann and Sieberer, 2018). By using a quantative method we are able to show that the process of parliamentarization affects voting in a parliament in general and not just on specific constitutional issues. Specifically, we use a dyadic approach ( Van der Veer, 2018). This is a relatively new method of studying voting in the Parliament and it allows us to assess the strength of different possible explanations of votes cast directly in the Parliament.
This article is structured as follows. The first three sections serve to construct the theoretical framework for the article and provide theoretical support for our hypothesis. We discuss the literature on both the parliamentarization of European states in general and the Parliament specifically. This literature shows that in national parliaments in the early and late modern period, the conflict over parliamentary powers was an integral part of day-to-day politics and possessed an intra-institutional component. While the literature on the EP recognizes that the conflict about its powers is part of daily politics, it remains to be tested whether in the day-to-day work of parliament an intra-institutional division on this issue is visible. This is what we will test in the article. In the third theoretical section of the article, we will derive control variables for our analysis from the literature on voting in the Parliament. Then we discuss the dyadic approach and our results. Finally, we discuss what our conclusions mean for our understanding of the political conflict in the Parliament and for the literature on parliamentarization.
further by arguing that the Parliament has become more powerful than many national parliaments.
In order to understand the parliamentarization of the EU, it is useful to understand the parliamentarization of European states. Ihalainen et al. (2017, pp. 6-8) propose that parliamentarization is the process by which parliaments gain four characteristics: the ability to deliberate between opposed points of view, the role as representive of the citizens, the ability to hold the government responsible, and sovereignty. Parliamentary government based on these pillars is 'a product of a series of political disputes' (Ihalainen et al., 2017, p. 1).
In a number of early modern monarchies, conflicts arose between the factions that supported royal sovereignty and those that defended parliamentary privileges. In that period such conflicts could turn violent (Van Zanden, Buringh, & Bosker, 2012, p. 852). 2 The best example of this is the English Civil War between royalists and parliamentarians. The roots of this conflict can be traced to different views on the powers of the House of Commons and the Crown (Seaward and Ihalainen, 2016, pp. 35-6;Tilly, 1997). After the Restoration, the Whigs and Tories were divided by the question of how much power the House of Commons should have. 3 Similar conflicts existed in the Dutch and French Estates-General, in the Bohemian Estates and in the Spanish Cortes (Van Zanden et al., 2012, p. 852).
During the nineteenth century the conflict between supporters of parliamentary power and the supporters of royal power intensified in many countries. The question of parliamentary power was one of the sources of the conflict between the liberal left and the conservative right. In Sweden, Denmark, Norway and the Netherlands the limits of royal and parliamentary power were the focus of the liberal-conservative conflict in the nineteenth century (Beyen and Te Velde, 2016;Ihalainen, 2015, pp. 72-3;Narud and Strøm, 2011, pp. 200-1;Rossvoll, 1966, p. 66;Sandelius, 1934, p. 365). For instance, the Norwegian liberals rallied around the slogan 'all power in the Halls of the Storting' (Strøm, 1995, p. 53, note 2). The consequence of these conflicts was a series of reforms that increased the powers of parliaments at the cost of those of the monarch.
The issue of parliamentary powers also divided political factions in countries with tumultuous political histories. Revolutionary assemblies like the French Legislative Assembly (1791-1792) and the German Frankfurter Assembly (1848) were divided over the scope of royal and parliamentary power (Garrigues and Anceau, 2016, p. 50;Herrmann and Sieberer, 2018, p. 3). Those questions returned during the Bourbon Restoration in France and the Bismarck period in Germany (Biefang and Schulz, 2016, p. 69;Garrigues and Anceau, 2016, pp. 53-4;Häge, 2017, p. 187).
After the First World War the question of parliamentary power often was no longer an intra-institutional issue. 4 In most countries conflict about parliamentary power is inter-institutional in nature: in conflicts between the executive and the legislative, MPs defend parliament's prerogatives irrespective of their party (Andeweg, 1992, p. 163).
2 There is no linear progress towards greater parliamentary power: parliaments have often been stripped of the powers they previously had (Butterfield, 1931). 3 Notably, these parties were divided about parliamentary power (Seaward and Ihalainen, 2016, p. 44). 4 An exception is Weimar Germany, where both the national socialists and the communists contested the legitimacy of the Reichstag until its demise (Biefang and Schulz, 2016, p. 72). During the formation of the French Fifth Republic the communists contested the decrease of power for parliament (Macridis and Brown, 1976, pp. 211-22). During the UK 2019 Brexit debate the prorogation of Parliament was contested by the opposition parties.
All in all, for most of their history, one of the sources of conflict within Europe's parliaments was the question how much power these bodies should have relative to the King. Some factions favoured greater power for these bodies while other factions opposed it. In other words, conflicts about the scope of parliamentary power were intra-institutional in nature and not just inter-institutional. The central question of this article is whether such a division is also visible in the EP.

II. Parliamentarization of the EU
The EP has seen its power and domain of influence grow over time. In the six decades of its existence it has been involved in a process of continuous constitution-building (Farrell and Héritier, 2003). Its powers have expanded in two ways: treaty changes and daily politics (Farrell and Héritier, 2007). Revisions to treaties have granted the Parliament more power by growing and pruning different legislative decision-making procedures (Hix and Høyland, 2013;Maurer, 2003, pp. 228-31). The Parliament started as a consultative assembly whose advise could be ignored by the Council. The first direct elections in 1979 gave the Parliament its representative function (Tiilikainen and Wiesner, 2016, pp. 294-5). The 1986 Single European Act (SEA) granted it legislative power (Hix, 2002;Rittberger, 2003). The 1992 Maastricht Treaty introduced a legislative procedure where the Parliament had a veto (Hix, 2002, p. 261). Finally, the 2007 Treaty of Lisbon ensured that this legislative procedure became the standard for most policy areas (Hix and Høyland, 2013).
In the period in between treaty revisions the powers of the Parliament grew informally. Institutional rules are incomplete contracts that are subject to renegotiation when they are applied in reality (Farrell and Héritier, 2007;Hix, 2002). Ever since the single european act (SEA), the semi-informal rules about decision-making have proliferated (Farrell and Héritier, 2003;Stacey, 2003, p. 939). The Parliament has a 'voracious appetite' for such informal rules (Stacey, 2003, p. 949). They create cracks in the legislative procedure which the Parliament can exploit. The formal extensions of the Parliament's powers in the treaties mostly codified the informal procedures that it itself created (Farrell and Héritier, 2007;Hix, 2002).
In various policy areas MEPs have extended their policy-making competences by parliamentary activism, by manipulating the institutional parameters and by exploiting legal ambiguities (Bailer and Schneider, 2000;Maurer et al., 2005;Ripoll Servent, 2014;Crum, 2006;Häge and Kaeding, 2007;Roederer-Rynning, 2003). During the reform of the Economic and Monetary Union in the wake of the sovereign debt crisis, the majority in Parliament attempted to increase its power beyond the limits of the treaties (Rittberger, 2014;O'Keeffe et al., 2016;Schoeller and Héritier, 2019). This can be seen directly in the negotiations on the two-pack, the post-crisis budgetary framework (Roger et al., 2017). Originally, the two-pack was an intergovernmental agreement outside EU treaties. The focus of the Parliament in the negotiations with the Council was not on the substance of the agreement but on embedding the agreement in EU law and moving it within the Parliament's area of control.
A similar process can be seen in selecting the Commission, a key aspect of responsible government. The Parliament exploited ambiguities in the treaties during and after the fall of the Santer Commission to expand its powers (Judge and Earnshaw, 2002;Majone, 2002, p. 384;Moury, 2007;Stacey, 2003, p. 945). In 2014 Parliament forced the Council to select the Commission President from the Spitzenkandidaten of the European parties that ran in the EP election (Christiansen, 2016;Fabbrini, 2015). The article in the Treaty of Lisbon about the appointment of the Commission President was ambiguous; it was an incomplete contract ready to be exploited. The majority in Parliament believed that the Spitzenkandidaten procedure was a good way to interpret this rule. The Parliament then forced the Council to accept its interpretation. The selection of Von der Leyen as Commission president over Spitzenkandidaten Timmermans and Weber in 2019 shows that the Parliament is not always on the winning side of this conflict.
Most studies see the expansion of the powers of the Parliament as the result of an inter-institutional conflict between the Council and Parliament and not of an intra-institutional conflict between parties that hold different preferences over the question of the powers of the Parliament. This literature assumes that the Parliament is a unitary actor which desires more power for itself (Hix and Høyland, 2013, p. 174). In contrast, we propose that the Parliament is not a unitary actor but a political arena where MEPs with different views about how much power the institution should have compete and cooperate. There are political parties in the Parliament that do not wish to extend its power (pace Schoeller and Héritier, 2019, p. 288). Historically, the French Gaullists and British conservatives opposed empowering this supranational parliament (Rittberger, 2003, p. 215, pp. 219-20).
There are several studies acknowledging that views about the power of the Parliament can structure parliamentary voting. Kreppel and Tsebelis (1999), for instance, argue that during intense inter-institutional negotiations, the centre-right European People's Party (EPP) and the centre-left Progressive Alliance of Socialists and Democrats (S&D) 5 form a united front to defend their common compromises against the Council. MEPs do not decide only on whether they support the substance of a policy but also whether they want to press for the position that the majority of the Parliament has agreed on, or whether they are willing to settle for the Council's offer. This does not just reflect policy agreement, but the legitimacy of the Parliament.
As we have shown above, many historical parliaments in Europe were subject to intra-institutional conflicts about whether they wanted to gain more powers for the parliament or not. The central question of this article is whether such a division is also visible in the EP. Existing work (Benedetto and Hix, 2007;Rittberger, 2003) suggests that there are parties in the Parliament that oppose extending its power. The work on how the Parliament exploits incomplete contracts suggests that we should not be able to see this division in voting on treaty changes alone but in the daily work of parliament as well. We therefore seek to test the hypothesis that in the everyday voting in the Parliament, there is a division in EP voting between those who favour greater power for the Parliament and those who oppose it.
We formulate this expectation in the form of testable hypothesis. The wording of the hypothesis is somewhat technical and reflects our method (discussed below): H1: The greater the extent to which MEPs agree with each other on the level of power that the EP should have, the more likely they are to vote similarly. (Parliamentary power hypothesis)

III. The Hix-Lord Model of the EP
The debate on the question of which dimensions structure voting in the EP has come to consensus on the Hix-Lord model (Hix and Lord, 1997). This holds that the there are two dimensions that structure voting in the EP: a pro or anti-EU dimension and a leftright dimension. The first dimension concerns the level and speed of integration. The second dimension concerns the EU's role as a market regulator that can choose more or for less regulation. Many studies of the Parliament use this two-dimensional approach (Hix, 2001;Klüver and Spoon, 2015;Otjes and van der Veer, 2016). Given the importance of these ideological dimensions in the literature, we include them in our analyses as control variables. We expect a division in voting in Parliament between left-wing and right-wing MEPs and another between pro-European and eurosceptic MEPs.
MEPs are likely to follow their political group in deciding how to vote. The groups' positions are determined by the groups' policy specialists, and group meetings can be a way in which these groups form a consensus. Therefore, we include sharing a party group in the analysis, with the expectation that those in the same political group are more likely to vote the same. However, MEPs may deviate from the common position of their political group due to pressure to follow their own minister in the Council (Mühlböck, 2013). Therefore, we include government participation as a control variable with the expectation that there is a division in voting in the EP between MEPs from parties that are in the national government and those that are in the opposition. As observed above, MEPs are also pressured by the 'Grand Coalition' of S&D and EPP. The S&D and EPP effectively function as an informal governing coalition (Kreppel and Tsebelis, 1999). We include this variable in the analysis expecting to find a division in voting in the Parliament between EPP and S&D MEPs and those in other groups.

IV. Supporting EU Integration and Supporting the EP
One may ask whether there actually are parties whose views on how EU integration should proceed differ from their views on how the role of the Parliament should develop. To assess this, we use the Chapel Hill Expert Survey (CHES). This is a crucial part of the infrastructure of contemporary European political science. It regularly surveys country experts to position political parties from EU member states on dimensions concerning European and national political issues. The 2014 survey used over 300 experts on party politics in EU member states. We use questions about party position on European integration 6 6 The specific question is: 'What position did the party leadership take over the course of [year] on the following policies? First, take the position of the party leadership in [year] on the powers of the European Parliament (EP),' with eight options: 'strongly opposed', 'opposed', 'somewhat opposed', 'neutral', 'somewhat in favour', 'in favour', 'strongly in favour' and 'do not know'. and on the powers of the EP. 7 These two items are used to measure party positions on European integration and on the powers of the EP. Figure 1 shows the positions of European parties on these two dimensions in five waves of the CHES (Bakker et al., 2015;Polk et al., 2017). It shows that these two dimensions are correlated (Pearson's r = 0.80, significant at the P = 0.01 level). Yet there also are notable differences between parties. There are parties that score significantly lower on the EU integration dimension than on the EP dimension and vice versa. Parties that favour EU integration strongly but are lukewarm about the Parliament are mainly centre-right parties, such as the Finnish conservative National Coalition Party (Kansallinen Kokoomus, KOK). The Spanish social-democratic Spanish Socialist Workers' Party (Partido Socialista Obrero Españo, PSOE) shares this combination of preferences. Parties that score significantly lower on the EU integration dimension than on the Parliament dimension are mainly left-wing opposition parties, such as the Greek socialist Coalition of the Radical Left (Συνασπισμός Ριζ οσπαστικής Αριστεράς, SYRIZA). The radical right-wing populist Flemish Bloc (Vlaams Blok, VB) also shares this combination. 8 All in all, support for EU integration is not identitical to support for the EP. 7 The specific question is: 'How would you describe the general position on European integration that the party leadership took over the course of [year]?' with eight options: 'strongly opposed', 'opposed', 'somewhat opposed', 'neutral', 'somewhat in favour', 'in favour', 'strongly in favour' and 'do not know'. 8 Determining why parties favour a stronger EP but not necessarily a more integrated Europe is outside of the remit of this article. Ideological and strategic calculations may play a role: national opposition parties may be able to influence policy in the Parliament. It may also reflect different views about the EU. Those who favour EU integration but do not favour strengthening the Parliament may prefer a 'Europe of fatherlands', where the governments assembled in the Council dominate decision-making. Those who favour the Parliament but are opposed to further EU integration may believe that EU decision-making is dominated by business interests. They may believe that if citizens exert greater power over policy-making through Parliament, the direction of EU integration may change.

V. Methods
There are different ways to study patterns of voting in the EP: (1) one could focus on specific votes (Roger et al., 2017); (2) one could use a dimension reduction method like NOMINATE (Hix, 2001); or (3) one could use the dyadic approach developed by Van der Veer (2018). While all of these methods have their own drawbacks and advantages, we have opted for the dyadic approach. We discuss our reasoning for this below.

Case-based Approach
Some research into the Parliament uses a case-based approach (Benedetto and Hix, 2007;Earnshaw and Judge, 1993;Roger et al., 2017). This approach focuses on a limited number of cases to determine whether particular considerations matter. One might, for instance, study a limited number of specific votes over the extension of the powers of the Parliament. This allows one to place a specific vote in its historical context and is likely to provide insight into the motivations of the actors involved. Benedetto and Hix (2007) have used this approach to show that in terms of votes on the European constitutional treaty, the division between those who favoured and those who opposed greater power of the Parliament was significant.
One drawback of such a study is that it is a most likely case design. When we study cases in which a particular conflict is likely to appear and find that it does in fact appear, this alone tells us relatively little about whether this division plays a role in the Parliament in general. Yet whether the division plays a role in general is exactly what we seek to test here; namely, the notion that constitution building in the EU is a continuous process (Farrell and Héritier, 2003). Therefore, the division between those who favour greater or lesser power of the Parliament should also be visible in its day-to-day politics. To test this, we additionally need a method that looks at all votes in the Parliament, including those where one would not necessarily expect this division to matter.
By necessity, such a large-N method would be based on correlation. One may wonder whether such a correlational analysis can capture the complexities of voting in the Parliament, and where conflicts about the power of the Parliament are embedded in other conflicts about policy. For instance, when it comes to an issue of consumer safety, parties that prioritize consumer rights may fight for the power of the Parliament more than those who prioritize business interests. Much of the research on parliamentarization does not merely focus on correlations between variables but instead focuses on the historicity, actorship, contigency and context of decision-making. We recognize that a large-N correlation-based approach is unlikely to capture the true complexity of parliamentary decision-making concerning the conflict for the power of the Parliament. This is both a weakness and a strength of a quantitative approach. It is a weakness because we can never capture all the motivations and considerations of MEPs that are linked to the power of the Parliament. We can merely assess whether their voting reflects a unidimensional model of this conflict. It is also a strength, however, because it makes the study more conservative. By using it we are reducing a complex element of parliamentary decision-making into one dimension, which is unlikely to measure all the nuances of every vote. That means we are less likely to find a pattern than in a qualitative-historical approach. A whole range of different motivations and considerations related to the conflict over the power of the Parliament can be picked up by a qualitative approach. When using a correlation-based method, we can only find a positive result if a very specific pattern is present. This makes it a more conservative test of the extent to which conflict about the power of the Parliament affects decision-making in it. Therefore, the results of a quantative test have greater external validity. There are two options for such a quantitative analysis: dimension reduction methods and the dyadic approach.

Dimension Reduction Approach
Scholars who study the stucture of conflict in legislatures often rely on dimensionreduction methods such as Nominal Three-Step Estimation (NOMINATE) developed by Poole and Rosenthal (1984). This method has been applied to the EP (Hix, 2001). It has also been used to study the conflict about parliamentarization in Europe's historical parliaments, showing that it is possible to get a grasp on the contigent and context-specific conflict about the power of parliament using quantitative methods (Häge, 2017;Herrmann and Sieberer, 2018). These methods reduce thousands of votes to a low-dimensional map of the political space. This is a very insightful way to show the main lines of conflict in the Parliament (Van der Veer, 2018, p. 70). The key drawback of these methods that is that they generate only spatial representations. The interpretation is left to the researcher. One could potentially analyse the results of a NOMINATE analysis in a regression analysis. However, methods like NOMINATE give priority to the dominant patterns and smaller patterns are treated as errors (Voeten, 2009). A considerable part of what is left out of the NOMINATE models actually reflects meaningful patterns of disagreement within and between groups (Van der Veer, 2018, pp. 85-6). NOMINATE models overestimate the importance of the left-right dimension, at the cost of factors such as the EU dimension and sharing party groups and government participation (Van der Veer, 2018, pp. 90-1).

Dyadic Approach
To assess the importance of the division over the power of the Parliament in relation to the left-right dimension or the EU integration dimension, one would need to use a regression-based technique. The dyadic approach developed by Van der Veer (2018) builds on the study of co-sponsorship of parliamentary proposals (Alemán et al., 2009). This approach studies parliamentary behaviour as a relational characteristic. In other words, rather than studying whether a legislator votes in favour of or against a proposal, the dyadic approach examines pairs of legislators and the extent to which they vote similarly. Combining this information with external indicators of the ideological distance between legislators on various dimensions allows one to assess the degree to which these ideological distances influences voting, the relative importance of different factors and whether these factors have a statistically significant effect. Moreover, the method allows for control variables such as party and coalition membership. All in all, compared with earlier research, the dyadic approach allows us to get a far more complete image of patterns in parliamentary voting.
The dyadic approach works as follows. The first step is that all possible pairs of legislators are created. Then the degree to which legislators vote similarly (yea-yea, nay-nay or abstain-abstain) is calculated as a proportion of the total number of votes in which both legislators participate. 9 We gathered these data from the VoteWatch (2019) database. The number of cases is nearly equal to half the square of the number of MEPs in that parliament. The voting similarity score can have values between zero (the two MEPs vote completely differently on all votes) and one (the two MEPs vote exactly the same on all votes).

Dependent and Independent Variables
In this study we examine patterns for the fifth to the eighth EP terms. Every case is an MEP-MEP dyad. We include the data per parliamentary term. This means that pairs of MEPs who have been in parliament for many terms appear many times. As external measures of party ideology we use the CHES. 10 We use an item of whether political parties favour or oppose increasing the power of the EP to test our central hypothesis. As controls, we use the pro or anti-EU dimension and the left-right dimension. 11 All CHES dimensions are converted to distances by taking the absolute difference between the position of legislator A and legislator B, recalculated to range from 0 to 1 for each of the three dimensions. As we saw above, the EU integration item and the EP item are strongly correlated, and the same is true for the two distance measures (Pearson's r = 0.77, significant at the P = 0.01 level). A reason not to include highly correlated variables would be that this inflates standard errors; given that our method of data analysis has a very large N, this should be less of an issue than in analyses with a smaller N.
As controls we include binary variables to indicate whether both legislators are member of the same political group, whether they belong to the grand coalition of EPP and S&D, and whether both legislators are members of national parties that govern in their member states during that parliamentary term (Döring and Manow, 2012). All variables are recalculated so their maximum is one and their minimum is zero to ease comparisons of effect sizes. Table S1 in the Appendix shows the descriptive statistics of these variables and Table S2 lists the correlations between them.

Modelling Strategy
We used simple ordinary least squares (OLS) regression. The main advantage of this, besides its simplicity, is that we can compare the strength of different explanatory variables. 12 Because the data on MEPs' ideological positions, membership of national governments and the grand coalition are at the party level, we ran the model with cluster robust standard errors. Given the dyadic structure, we ran the model with two clusters for both parties involved in the dyad (specifically, the ivreg2 procedure in Stata). 9 Pairs of legislators are coded as missing when together they participate in less than 2.5% of all votes during a legislative term. This prevents MEPs who have served very briefly from affecting the study results. 10 We choose the closest CHES wave to the start of the term.

11
The specific question is: 'We now turn to a few questions on the ideological positions of political parties in [country] in [year]. Please tick the box that best describes each party's overall ideology on a scale ranging from 0 (extreme left) to 10 (extreme right)'. 12 One disadvantage of using OLS is the possibility of model predictions outside the zero to one range. This happens in fewer than 2 per cent of the cases.

VI. Results
Above we sketched the process of parliamentarization in national parliaments in Europe and in the EP. The central hypothesis we derived from this discussion is that the extent to which MEPs agree on the levels of power that the Parliament should hold would predict the extent to which they vote the same way. The division between those who favour and those who oppose more power for the Parliament should not be visible only in the decisions about the new treaties, but should be visible in the day-to-day voting in Parliament. Therefore, we look at all votes in the EP to see to what extent this division is visible. We do not expect that this division is stronger than the more well-established left-right and pro/anti-EU dimension, but we expect it to be a third underlying factor. Table 1 shows a regression model with the ideological variables and the control variables. The effect of the distance between parties on the EP dimension is significant, negative and substantial: if one moves from a pair of MEPs who agree completely on this issue to a pair of MEPs who take completely opposite views on it, the share of votes in which both vote the same way decreases by 8 percentage points. This decrease is almost a tenth of the total scale. This result cleary supports our central hypothesis.
The effect of the EU integration dimension is of comparable size to the EP dimension. Moving from a pair of MEPs who agree most on this dimension to a pair of MEPs who agree least decreases the share of votes in which they vote the same by 11 percentage points. It appears to be the case that the question of the power of the Parliament is nearly as good a predictor of how MEPs vote as is the question of the power of the EU itself.
We also include a number of control variables. The strongest predictor is the distance between MEPs on the left-right dimension: if one moves from a pair of MEPs who have the exact same position on the left-right dimension to a pair of MEPs who stand on opposite ends of this dimension, the share of votes in which they vote the same decreases by 25 percentage points. This effect is the strongest of the three ideological dimensions. MEPs who are both from the Grand Coalition are also expected to show higher levels of voting similarity. Indeed, compared with other dyads, S&D and EPP pairs vote in the same way more often (+3 percentage points). We expected that groups that are both in the Council vote together more often. This is indeed the case (+4 percentage points). We also expected that dyads that are in the same political group vote together more often. The share of identical votes of a pair of MEPs who are in the same group is 21 percentage points higher than for a pair that is not in the same group.
In Table S3 in Appendix we delve deeper into the data: first, we include a model with a measure for two MEPs being a member of the same national party. If this has a strong effect, it would indicate that there are important party-level variables that we have omitted. This is not the case. This means that all the similarity that is caused by two MEPs being from the same national party is captured by the other party-level measures. Second, we look at two models that approach the clustering in a different way; one party*term clusters instead of party clusters and one without clusters. The interpretation of the effects does not change. The fourth model employs a beta regression (as we employ a bounded dependent variable), that shows the same patterns as the OLS regression. Next, we look at one model without the EP dimension and one without the other ideological variables. This shows that these items tend to pick up on the same variation: the R 2 is less than 1 per cent lower if the item on the power of the EP is not included. When we remove the EP dimension the effects of both the left-right and the EU integration dimension increase. If we drop both these ideological variables and include only the EP dimension, this becomes a much better predictor. All this indicates that adding the EP dimension does not add significantly to the explanatory power of the model, although it does help to understand what substantial differences underpin both dimensions of the Hix-Lord model.
We also look at whether the EP power dimension matters more during different types of decision-making. The notion is drawn from Kreppel and Tsebelis (1999): those who favour a powerful Parliament band together when votes are more contentious between the Parliament and the Council; that is, when MEPs form one bloc in order to ensure that the Council adopts the common position of the Parliament. We look at this possible effect of contentious votes by paying special attention to votes on legislation, amendments, later readings and final votes. We find a clear interaction effect for the final votes. The EP power dimension matters strongly during final votes and not at all during other votes. This sustains the notion that those who favour a strong Parliament band together during votes that are politically more important. We do not find indications that this kind of interaction is present for amendments, legislative votes or votes in later stages of decision-making. The additional data are available for the seventh Parliament. As a baseline for comparison, we have included a model for the seventh term. This shows a much stronger effect of the EP dimension than in the pooled data, which implies that the effect of the EP dimension is stronger more recently than it was before. All in all, we find considerable evidence that the question of how much power the EP should have plays a role in how MEPs vote.

Conclusion
This study adds to our understanding of the political conflict in the EP as well as our understanding of parliamentarization. The standard Hix-Lord model has up until now focused on the idea that voting in the Parliament reflects substantive disagreements over the policy direction of the EU. This model sees the Parliament as an arena in which parties interact on the basis of their policy preferences. It thus reinforces earlier evidence (Benedetto and Hix, 2007;Kreppel and Tsebelis, 1999;Rittberger, 2003) that this model is incomplete: MEPs also compete and cooperate within the Parliament to enact their preferences on the power of this body. The assumption that the Parliament is an unitary actor that wants more power for itself is incorrect. The Parliament is an arena for this intra-institutional conflict about its own power.
When it comes to parliamentarization the results show two things. Firstly, the fight over the powers of the Parliament is part of the daily politics in this body. The question of how power is distributed within the EU is not just a matter of treaties; it is also part of the dayto-day decision-making in the Union. Secondly, the conflict between those who favour the power of the EP and those who oppose it resembles the conflict in Europe's historic parliaments. The parliamentarization of the EU resembles how many European polities got to where they are now. This study shows that, like those studies on historical parliaments, such a conflict is not just part of the process of constitution-writing, but a continuous and intra-institutional conflict. The establishment of parliamentary government is not just part of modern history but also of contemporary politics in Europe.
One pattern which we unexpectedly found is that the effect of the EP dimension on parliamentary voting is stronger in the seventh term than in the pooled data. Future research may want to delve further into why the issue of how much power the Parliament ought to have matters more in the seventh EP than before. Explanations may look at the composition of the Parliament, at the institutional context or at historical circumstances (Otjes and van der Veer, 2016). One explanation may be the growing contestation of European integration by forces of the radical left and radical right. As we saw above, many parties of the radical left oppose European integration but favour a stronger Parliament. On the other side, conservative parties from Central Europe, such as the Hungarian Fidesz, are more opposed to increasing the power of the Parliament than they are to increasing the power of the EU. It may be the case that when radical left and radical right-wing forces grew in the Parliament the differences between their policy preferences became more visible. Another explanation may be the changing institutional rules: the seventh Parliament is the first after the Lisbon Treaty, which expanded the powers of the Parliament. Ironically, it may be that the conflict over the power of the Parliament became more visible as an intra-institutional conflict because a group of MEPs was satisfied with the Treaty of Lisbon and did not want to extend the power of the Parliament. This may have divided a Parliament that was previously united on expanding its power.
This study contributed methodologically to the study of parliamentarization and the study of voting in the Parliament. Firstly, compared with most existing historical studies of the process of parliamentarization, which underlines the historicity, actorship and contigency of this process, our study takes a quantitative approach. We show that the conflict about parliamentarization can be captured by this approach. This shows that differences in opinion about the power of the Parliament do not affect only votes on treaty changes but also the day-to-day process of decision-making in the Parliament. This is an important test of the idea that the EU is in a constant process of constitution-making. Yet the method we employed also has its limitations. Our method used a one-dimensional indicator to capture conflict about the powers of the Parliament. Therefore, our approach is unlikely to pick up on the true complexity of decision-making in the Parliament. It is notable, however, that even with such a narrow measure we find a significant result. Future research may want to delve further into the process of parliamentarization from a qualitative-historical perspective: to see this process beyond our narrow scope. Such research may pick up on specific conflicts about the power of the Parliament that do not follow the simple measure employed here.
Secondly, in contrast with much of the quantitative literature on parliamentary voting that uses dimension reduction methods, we used the dyadic approach developed by Van der Veer (2018). This method allowed us to sketch a more nuanced picture of parliamentary voting compared with the image of the bidimensional Parliament that prevails in the literature.
Our results have provided a substantive addition to the existing literature on the Parliament, which has long focused on preferences over public policy. Given the lack of the clear opposition-coalition divides that dominate voting in national legislatures, it could be the case that in their votes MEPs only expressed their preferences for public policy. This study shows that EP voting reflects not only differences in preferences about European policy but also intra-institutional differences in preferences about the European policy. Voting in the Parliament is not just about policy and ideal points but also reflects a political struggle about the nature of the body itself: in this sense the Parliament is, to quote Abraham Lincoln, 'a house divided against itself ' (Fehrenbacher, 1960).