Political conflict within and between the European Parliament and Council of Ministers

Transcription

1 Political conflict within and between the European Parliament and Council of Ministers Rory Costello PhD Candidate, Trinity College Dublin Paper to be presented at the ECPR Joint Sessions, Rennes, April 2008 Abstract Research to date on the political space occupied by legislative actors in the EU has tended to examine one or other of the legislative institutions (the European Parliament (EP) and Council of Ministers) in isolation. Consequently, while there is an emerging consensus regarding the nature of competition within the EP and the Council, much less is known about competition between these institutions. Furthermore, it is difficult to reconcile the standard accounts of the political space of the EP and the Council into one coherent picture. A more complete account of legislative decision making in the EU requires information on the internal politics of both institutions taken together. This paper takes a step in this direction by describing political conflict within and between these two institutions in relation to 41 recently negotiated legislative issues. I find that the types of issues which divide member states are different from the types of issues that divide political groups in the Parliament. Furthermore, the alignments of actors in the two institutions suggest that the Parliament is systematically to the left of the Council in several policy areas. 1

2 Political conflict within and between the European Parliament and Council of Ministers The European Parliament (EP) and Council of Ministers form the bicameral legislature of the EU. These two chambers share decision making power in policy areas covered by the codecision procedure, where they must both agree to the final text of legislation proposed by the Commission. In other policy areas covered by the consultation procedure, final decision making power rests with the Council, and the EP is restricted to an advisory role; yet even here there is evidence to suggest that both chambers influence decision outcomes (e.g. Thomson and Hosli 2006). In any bicameral system, it is important to understand the differences between the two chambers. Indeed, the question of the relative power of the two chambers is only important when there are clear differences in the policy goals of the two chambers. Research on the EU, however, has tended to examine the political space of each chamber in isolation. Consequently, while there is an emerging consensus regarding the nature of competition within the EP and the Council, much less is known about competition between them. Furthermore, it is difficult to reconcile the standard accounts of the political space of the EP and the Council into one coherent picture. In this paper I examine competition in both chambers simultaneously, using a new dataset on legislative decision making. The dataset contains information on the preferences of actors within both the EP and the Council of Ministers in relation to 41 2

3 controversial issues from 15 recent legislative proposals 1. The preference of the Commission relative to other actors is not examined here. An exploratory approach to data analysis is adopted. The goal is to identify patterns of conflict within and between the Council and EP, and to describe the observed differences between actors in policy terms. The findings will subsequently be developed as hypothesis and tested using an alternative data source. This paper is organised as follows. The next section reviews the existing literature on the EU political space. The data and research design for the empirical analysis is then described, and the results presented. The paper concludes by bringing the findings together to give an overall account of political conflict across the two legislative chambers of the EU. The EU Political Space Political conflict in a bicameral system can occur within and between the chambers. Whether or not the inter-chamber divide emerges as a significant dimension of conflict generally depends on how the members of each chamber are selected (Tsebelis and Money, 1997, p.44). Where selection procedures are similar, the chambers will tend to have congruent composition and similar preference structures; in which case the most important political disagreements will generally take place within either chamber. Differing selection procedures for each chamber will tend to result in incongruent 1 The data were collected by the present author, Robert Thomson (Trinity College Dublin) and Javier Arregui (University of Pompeu Fabra, Barcelona). 3

4 composition and distinct preference structures. In this case, the inter-chamber negotiations become an important part of the political process. In the EU, there are significant differences in the composition of the EP and the Council of Ministers. The Council is made up of one representative from the national government of each member state, while the EP consists of several directly elected members from each state. Consequently, the EP is a much more diverse institution, encompassing a much broader range of interests. However, the geographical distribution in each chamber is comparable. Under certain legislative procedures, the Council representatives are allocated voting weights regressively proportional to country size; MEP seats are distributed on a similar basis. The situation is complicated by the fact that elections to the EP are controlled by national, rather than European, political parties. This means that member state governments (who are represented in the Council) are in a position to influence the behaviour of many MEPs, through the threat of de-selection. Furthermore, the chambers of the EU are not dominated by standing partisan majorities, as is the case in most other bicameral legislatures. Unlike parliamentary systems, there is no requirement for a governing majority to form in either chamber; and unlike two-party presidential systems, such as the US, no one party has an overall majority in either chamber. Rather, different coalitions are free to form within each chamber on different issues. For this reason, the extent of division between the chambers is likely to vary according to the nature of the issue at hand, and the particular internal alignment formed within each chamber. These alignments can take place along national or ideological lines. 4

5 National divisions National differences may emerge between actors within one or both of the EU legislative chambers for a number of reasons. Some scholars have argued that divisions exist between actors from wealthy and poor member states (e.g. Selck 2004); others see political divisions in the EU as reflecting cultural differences between regions (e.g. Elgström et al 2001). What is common to these scenarios is that the EU actors behave in predictable ways according to their nationality, and not their political party affiliation. If this is the case in both the EP and the Council, then differences between the chambers are likely to be small, given the comparable geographical representation in each institution. The majority of empirical studies on actor alignments in the Council have found evidence of divisions between geographical regions, in particular between the North and South among the 15 old member states. Several studies have employed the Decision Making in the European Union (DEU) dataset (Thomson et al 2006), using a variety of dimension-reducing techniques (e.g. Thomson et al 2004, Kaeding and Selck 2005, Zimmer et al 2005). A North/South dimension has emerged clearly from each of these studies, and has been interpreted as either a division over resources between the largely Northern net-contributors to the EU budget and the largely Southern net-recipients (e.g. Zimmer et al 2005), or as a divide between Northerners supporting free market policies and Southerners preferring greater levels of market regulation (Thomson et al 2004). These findings are supported by Mattila and Lane (2001), who find a North/South divide from roll call votes in the Council. 5

6 There is also some evidence of a new geographical divide, between East and West in the enlarged EU. Thomson (2007) analyses alignments in the Council in the postenlargement period. In addition to the North-South divide among the 15 old member states, he finds a striking division between the old and new member states, particularly for issues concerning financial subsidies. These alignments were found to vary considerably across issues. Naurin (2007) examines cooperation among Council working groups, and finds similar North/South and East/West patterns. However, he attributes these alignments to cultural rather than economic factors. He argues that the alignments are too consistent across policy fields to be due solely to shared economic interests. Given that the key cultural divisions in the EU largely coincide with divisions between rich and poor member states, it is difficult to test these competing explanations conclusively. In general, there is much less evidence of national/geographical divisions in the EP than the Council. For example, Hix et al (2007) conduct an extensive analysis of rollcall votes in the EP between 1979 and Using this data, they measure the cohesion of party groups and of national groups in the EP. They find that the party groups have become more cohesive over time, and are more cohesive than national groups of MEPs (p.93). This is despite the considerable levels of diversity within the party groups, as demonstrated by survey data (e.g. Thomassen et al 2004, McElroy and Benoit ). The analysis of roll call votes from the EP has come in for some criticism, however (e.g. Carrubba et al 2006, Hug 2006). Roll-calls are only recorded in the EP on the request of a party group or a group of at least 32 MEPs, and tend to occur for only about one-third of votes. There are strong reasons to suspect that roll-calls are requested 6

7 for strategic reasons, for example to demonstrate group cohesion or to embarrass opponents, and cannot therefore be taken as a representative sample of all votes (Carrubba et al 2006). It is possible that the cohesion of party groups is overestimated by roll call analyses. Ideological Divisions One prominent strand in the literature assumes that the main dimension of conflict in the EU relates to differences over the levels of integration sought by different actors (Tsebelis and Garrett 2000). EU integration has been largely driven by a vision of regulated capitalism, a vision which is opposed by extreme parties on both the left and the right. Hence, the EU integration dimension is often considered to form an inverted U-curve relative to the left-right dimension: extreme parties on both the left and the right tend to be anti-integration, while centre-left and centre-right parties are pro-integration (Hooghe et al 2002). However, institutional differences may also come to the fore when it comes to debates over levels of integration. According to Tsebelis and Garret (2000), the Commission and the EP can be characterised as unitary actors in favour of integration, as integration increases the power of supranational institutions. While this may exaggerate the level of cohesion in the EP on this dimension, it may well be the case that EP actors are systematically more in favour of integration than the Council. Apart from the vested interests MEPs may have in deeper integration, it is commonly argued that permanent EU 7

8 politicians become socialised over time in favour of integration (e.g. Christiansen et al 1999). While this argument is made for all EU politicians, it is especially relevant to those in the supranational institutions, who are more disconnected from domestic politics. Consequently, we may find clear differences between the chambers on an integration dimension, regardless of their composition. Another strand in the literature emphases the importance of the ideological lef/right dimension. As Hix (1999) points out, where EU legislation is reforming an existing law rather than creating a new one, the question of more or less integration may not come into play. Rather, competition is likely to emerge between actors who are ideologically disposed towards high levels of regulation and redistribution, and those who prioritise free market competition. The selection process for the EP and the Council gives further credence to this argument. Elections to the EP are second-order national elections (e.g. Marsh 1998), controlled by national political parties. Similarly, membership of the Council is determined by national party politics. If it is the case that EU politics has come to reflect the left/right structure of domestic politics, then the degree to which the EP and the Council disagree will depend on the composition of each chamber at that particular time. Kreppel and Tsebelis (1999) argue that conflict between the EP and the Council during the 3 rd Parliament ( ) can be understood in these terms, with right-wing governments in power in most member states, and the EP dominated by the left. During the period of relevance to this study, the largest group in the EP was the centre-right EPP, while the Council was also dominated by right wing parties. Therefore, assuming the EPP held a dominant position in majority formation in 8

9 the EP, we might expect a left/right alignment to result in relatively little disagreement between the chambers. There is considerable evidence for the importance of both of these ideological dimensions within the EP, based on a variety of sources. Using roll-call data, Hix et al (2007) find that there is one main dimension of voting, which they interpret (based on the average position of MEPs from each national party) as left-right. The ordering of the party groups for the most recent parliamentary period analysed ( ) is found to be as follows: GUE (United Left) and the Greens the furthest left; PES centre left; ALDE in the centre; and the conservative EPP along with the nationalist and anti-european groupings (UEN and Independence/Democracy) on the right. A second, orthogonal dimension is also found to be significant, but much weaker. This is interpreted as an EU integration dimension, with the pro-integration PES, EPP and ALDE separate from the anti-integration GUE, Greens and nationalists. According to Hix (1999), this dimension has not been incorporated into the main left-right structure of party competition, because social classes and political groups are internally divided on the issue of integration. Hence, the main groups do not distinguish between themselves on this issue, and take a pro-integration stance. This second dimension is also found to be associated with institutional interests, defined as having party affiliations with Council or Commission members. As mentioned above, the analysis of roll call votes from the EP has come in for some criticism, and it is possible that the left-right structure of alignments in the EP is overestimated by roll call analyses. Nevertheless, similar conclusions are reached by McElroy and Benoit (2007) using expert survey data. Experts were asked to locate the 9

10 policy position of each party group in relation to eight substantive policy dimensions. The authors find that the differences between the party groups can be captured by two underlying dimensions: a socio-economic left-right dimension, and a European integration dimension. Furthermore, of the eight original policy dimensions, the expert respondents deemed the economic and EU dimensions to be the most salient. Gabel and Hix (2004) examine the content of the manifestos issued by the main EP party groups for each election between 1979 and The results indicate that the traditional socio-economic left/right dimension determines the relative position of party groups across most issues, including matters related to economic integration. The authors conclude that a uni-dimensional model provides a good description of party group competition in the EP. Finally, Thomassen et al (2004) use data from a survey of MEPs, in which respondents gave their opinion on a wide range of issues. Differences between MEPs across issues are found to reflect three underlying dimensions: left/right, pro/antiintegration, and a traditional/libertarian or new politics dimension which divides MEPs in terms of their attitudes on issues such as immigration, moral issues and law and order. There is little evidence of an ideological structure in the Council alignments. The position of member states on the left/right dimension, based on the party competition of the governments, was compared to observed alignments in the DEU dataset by Thomson et al (2004), Zimmer et al (2005), and Kaeding and Selck (2005) without any significant results. Similarly, Hayes-Renshaw et al (2006) find no evidence of left/right alignments on the basis of roll-call results. Mattila (2004) presents one of the few empirical results to support a left/right division in the Council. Using roll-call data, he estimates the probability of voting against the Council majority, and finds that left-wing governments 10

11 are less likely to do so. However, the vast majority of decisions in the Council are made without recourse to formal votes, and even when voting does occur, it represents the final stage of a long negotiation process. Thus, it cannot be taken as a measure of actors ideal points (Hayes-Renshaw et al 2006). To summarise, evidence regarding the shape and content of the EU political space points to a number of trends. To the extent that any structure can be found to competition within the Council, it tends to be geographically rather than ideologically based, with a North/South division among the old member states, and an East/West division in the enlarged EU. This division may reflect cultural or economic differences. Competition in the EP tends to take place between party groups along the left/right dimension, with some evidence also of a pro-/anti-integration dimension. However, given that existing studies do not examine competition within each chamber simultaneously, it is difficult to interpret exactly how these trends impact on competition between the chambers. While there is evidence that the EP and the Council do take distinct positions (e.g. Thomson et al 2004), the nature of these differences remains unclear. Research Design Data The data analysed here contains information on the preferences of actors in both the EP and the Council in relation to 41 controversial issues arising from 15 recently negotiated 11

12 legislative proposals. The information was collected through semi-structured interviews with participants in the decision making process. The preferences of legislative actors are represented spatially, along a one-dimensional scale for each issue. The approach to the data collection draws heavily on the DEU project, which collected information on the initial positions of member states, the EP and the Commission in relation to a wide range of controversial legislative issues negotiated prior to enlargement (see Thomson et al 2006). However, the data used here contains information on a much wider range of actors than the DEU data. Rather than treating the EP as a unitary actor, information is included on the positions taken by legislative actors within the EP. Furthermore, this dataset was collected in the post-enlargement period, and information is included on the positions of 25 member states as opposed to the 15 in the DEU data. As the data collection was primarily carried out through interviews, only a limited number of proposals could be included. Given this restriction, the aim was to select the most relevant proposals for the purposes of studying political conflict. The selection process for the proposals to be included was as follows. Firstly, all proposals introduced by the Commission between July 2004 and December 2005 were identified. These represent the first generation of proposals to be negotiated in their entirety following the enlargement to 25 member states in May The selection was then narrowed down to politically salient proposals. First, proposals for decisions were excluded, as these are directed towards particular member states or legal entities, and tend to have limited political importance. Second, of the remaining proposals, only those that were mentioned 12

13 in the newspaper European Voice were chosen 2. The final selection includes 15 proposals for legislation, listed in Appendix 2 according to legislative procedure and policy area. How representative are these 15 proposals for the purposes of this research? The selection bias in favour of salient proposals is certainly likely to exaggerate the extent to which actors disagree. For many of the excluded proposals, the subject matter was technical (such as the proposal on frequency bands for radio paging) or routine (such as the proposal to facilitate visa applications for participants in the Winter Olympics), and are unlikely to have given rise to differences between political actors. However, this does not mean that there is any systematic bias in terms of the relative level of disagreement between different actors, or groups of actors. The selection of politically salient proposals in which there was at least one point of controversy does not mean that only cases where the EP and Council disagreed were selected. On some issues, the two chambers were in direct conflict, while on others there were disagreements within one or both chambers, but not between them. The majority of these proposals were decided under the codecision procedure, with the remainder under the consultation procedure. The 15 proposals were drawn from a range of policy areas, coming under the responsibility of 9 different EP committees. Experts involved in the decision making were identified in relation to each proposal, and a total of 134 interviews were held. Interviewees were chosen from each institution: officials from the permanent representations of the member states in the Council (77 interviews); officials from the Commission (19 interviews); and MEPs and 2 A total of 22 proposals met these criteria, but a further 7 had to be excluded. Three were excluded as negotiations had not started at the time of interviews; two more were excluded from quantitative analyses as no differences existed between actors; and a further two were excluded as the relevant experts from the EP were unavailable for interview. 13

14 officials from the EP (38 interviews). The interviews took place on several field trips to Brussels between July 2006 and March Experts were first asked to identify the controversial issues that the proposal raised. The 15 selected proposals gave rise to 41 controversial issues that were recognised by experts from both the EP and the Council. Experts were asked to identify the decision outcome favoured most by the internal actors in the Council and the EP at the time the proposal was introduced, along with the position adopted by the EP in its first reading opinion, the position of the Commission, the status quo, and the final outcome. These positions were represented on a scale of 0-100, with the most extreme positions at either end point. Finally, experts were asked to estimate the level of importance each of the actors attached to each issue, on a scale of Within the Council, the internal actors for which information was gathered are the 25 member states (Bulgaria and Romania had not yet joined when these proposals were introduced). Identifying the appropriate level of analysis in the EP was more problematic. There were a total of 732 members of the EP at the time of this study, drawn from 119 different national parties. These MEPs joined one of 7 transnational political groups in the EP. Clearly, experts could not be asked to locate the position of each MEP, or even each national party. However, focusing solely on the positions of the party groups would miss important information on the internal divisions within these groups. The policy pursued here was to first ask the experts to identify the position of the party groups 3. Then, for each group, the experts were asked to identify the positions of any national factions which deviated from the official group line. Appendix 1 lists the 25 3 While there were 7 party groups during this period, information was collected on just 6. The smallest group, Independence/Democracy, which includes just 3% of MEPs, was excluded. 14

15 member states and 6 EP party groups included in the analysis, along with the abbreviations used throughout the remainder of this paper. The proposal on waste will be used to illustrate the data. This proposal was introduced by the Commission because existing regulations, which set standards for waste management and waste reduction, had given rise to some uncertainties regarding how these standards could be met. The principle uncertainty concerned the distinction between recovery operations and disposal' operations, and the new proposal sought to resolve this question. Four controversial issues were raised by this proposal, one of which is described in Figure 1. This issue related to incineration, and whether or not and in what circumstances this could be classed as a recovery operation. The positions of the various actors are found above the line (EP party groups in italics), while the meaning of these positions and their numerical value on the scale are found below the line. The positions of national parties within the EP are not represented here for reasons of clarity. [FIGURE 1] At one extreme in this controversy, represented as position 0 on the scale, we find the Green group and the PSE (socialist) group. These actors did not want incineration to be classed as recovery under any circumstances. This was also the position that prevailed in the EP first reading. At the opposite end of the scale, the French and Italian governments in the Council wanted to classify most incineration as recovery. Several intermediate positions also existed. The original Commission proposal outlined a formula for calculating the energy-efficiency level of incinerators, and a threshold which must be 15

16 met before they can be classed as recovery operations. This was placed by the experts as position 60 on the scale, as it is slightly closer to the French and Italian position than the EP position. A number of actors were in favour of this proposal, including several EP party groups and several member states. At position 40 we find a group comprised mostly of new member states. Many new member states were worried about the affect that having incinerators classed as recovery operations would have on their burgeoning recycling schemes, which they have been struggling to get off the ground. Therefore, they sought more restrictive criteria that would be more difficult to meet. Finally, the Swedish govenrment wanted to have extremely restrictive criteria, which would rule out all but the most energy-efficient of incinerators. This is given as position 20 on the scale. One further point to note about this issue is that the EP came out in favour of position 0, even though the majority of groups were in favour of position 60. This can be explained with reference to internal divisions in the party groups. The positions given in Figure 1 are their official positions of the party groups. However, EPP and ALDE were internally divided, with many MEPs from these groups supporting the position 0. Internal defections by national factions are not included in Figure 1 for reasons of clarity, but they are included in the dataset. Method A range of exploratory data techniques are employed to describe conflict within and between the two chambers using this dataset. Given this inductive approach, confirmatory statistics are not be used to test hypotheses. However, the results of the 16

17 exploratory analysis will allow hypotheses to be developed and tested subsequently. Four separate aspects of the EU political space are analysed: the extent of division between the EP and the Council; the nature of differences between these chambers; the extent of division within the two chambers; and the alignment of actors within each chamber. While the positions of actors in the data are contained on an interval scale, the data is treated as ordinal where possible to increase the reliability of the findings. For ease of interpretation and presentation, the 41 issues are grouped into 7 categories. For each issue, the interview notes were consulted to identify the primary reason for disagreement amongst actors. On this basis, seven different categories of controversy were identified, and the end-points of each issue scale were labelled accordingly. The issue scales in each category were rotated so as to make them directly comparable. The seven types of controversy are as follows: 1. Funding. These are issues that involved disagreements between those who wanted harmonised and centralised rules for allocating and spending EU funds (position 0), and those that sought greater flexibility in the allocation and spending of funds (position 100). Six issues, from three different proposals, fall into this category. One example concerned funding allocations under the Financial Instrument for the Environment. At one end of the scale for this issue (position 0) were actors who wanted funding to be administered to individual projects by the EU (position 0), and those that wanted funding to be divided between member states, who would then decide on allocations to projects (position 100). 2. Harmonisation. The disagreements in these issues relate to whether member states should all be required to implement the same measures (position 0), or whether 17

18 member states are given discretion over when or how to implement measures (position 100). This category comprises seven issues from six different proposals. For example, one issue concerned the possibility of opting out of arrangements under the Working Time directive. At position 0 on this scale were those who opposed any opt-out arrangements, and at position 100 were those who wanted to include an optout provision. 3. Market Regulation. Here, disagreements were between actors who wanted more extensive market regulation/market intervention (position 0), and those who preferred free market solutions (position 100). Eight issues from four proposals fall into this category. An example concerns the regulation of the sugar market, with actors at one end of the scale seeking to maintain a high artificial price for sugar (position 0), and actors at the other end of the scale calling for a large price cut (position 100). 4. Civil Liberties. This category involves disputes between those who wanted to protect or enhance citizen s civil liberties (position 0), and those who sought to improve law enforcement capabilities (position 100). Seven issues from two proposals are included in this category. For example, one issue involved a disagreement over access to visa information collected at the EU borders, with one end of the scale representing those who believed that access to this information should be highly restricted (position 0), and the other end of the scale representing those who wanted decentralised access to be granted to all law enforcement agencies (position 100). 5. Environment. The five issues in this category, drawn from two proposals, relate to disagreements between actors who sought strict environmental regulations and 18

19 ambitious targets (position 0), and those who argued for more relaxed targets and regulations (position 100). For example, on the issue concerning the classification of incinerators (see Figure 1), one end of the scale (position 0) represents those who did not want incineration to be classified as a recovery operation under any circumstances (position 0), and the opposite end represents those who wanted all incinerators meeting some basic criteria to be given this classification (position 100). 6. Interests. This category involves issues where disagreements centred on the representation of interests from competing groups. On one side are those who put the case of consumers and employees (position 0), and on the other side are those who defended employers and business interests (position 100). Six issues from four proposals fall into this category. One issue in this category involved the acquisition of pensions. At one end of the scale were those who wanted pensions to be easily acquired from the start of employment (position 0), and at the other end were those who wanted pensions to be restricted to workers over a certain age and after a certain period of time (position 100). 7. Institutional control. Two issues in the dataset concern disputes over the extent to which the EP would be involved in the monitoring and updating of the relevant legislation. Position 0 represents the extensive involvement of the EP; Position 100 represents limited EP involvement. Appendix 3 provides a complete listing of the issues and their categorisation. 19

20 Analysis Confirmatory analysis of issue categorisation The categorisation of issues adopted here provides a means of interpreting the observed patterns in the data. The reliability of these interpretations depends on the extent to which the categorisation actually fits the data, with similar patterns across issues in each category. Confirmatory factor analysis, using the principal-components approach, is used to test this. The results are summarised in Table 1. A total of 99 actors are included in this analysis: 25 member states, six EP party groups, and 68 EP national parties (from the three largest party groups). The figures represent the proportion of variation across issues that can be explained by a common factor. Comparing the results for each category with the results for all issues taken together gives an indication of whether or not the categorisation groups issues that are not only similar in terms of the type of disagreement involved, but also in terms of the data. Where a high proportion of variation among issues within a category is explained by a common factor (compared to the proportion of variation among all issues explained by a common factor), this indicates that the issues in each category are relatively similar in terms the position of actors. In general, issues within particular categories do appear to have a higher level of association compared to the issues in general. [TABLE 1] 20

21 The best fit is for the institutional category, with 82% of the variation across issues explained by a single factor. This is not surprising given the nature of these issues, which tend to polarise the chambers. There is also a particularly high level of common variance among the Civil Liberties issues and the Environment issues. The other categories do not perform as well, but are still an improvement on the uncategorised data. These results will be discussed in the context of the main analyses. Extent of division between the chambers To provide a graphical summary of the relative proximity of actors in both chambers across all issues, I use nonmetric multidimensional scaling (MDS). MDS is one of a number of data reduction and data visualisation techniques that map the underlying structure of multivariate data. Distances between objects are represented in a low dimensional space, so that the distances between the points in the space match the original dissimilarities as well as possible (Cox and Cox, 2000, p1). The procedure for calculating the MDS solution was as follows. Firstly, the average absolute distance between each pair of actors across all 41 issues was calculated, creating a distance matrix 4. The actors included in this matrix are the 25 member states in the Council and the 6 EP party groups, along with the position of the EP first reading. The national parties in the EP are not included here, as this would greatly increase the 4 The matrix includes one observation for each pair, excluding pairs of the same actor (i.e. a half matrix without diagonals). For some issues, the position of certain actors was recorded as missing, as the interviewees were not aware of them having taken any position. When one or both actors in a pair were missing, these were excluded from the calculation. 21

22 number of points to be represented, without contributing substantially to the objectives of the analysis, which is to identify the extent to which the chambers disagree. National party divisions are returned to later. This distance matrix provides a way of ranking different pairs of actors in terms of their similarity. Nonmetric MDS then gives each actor a position on an n-dimensional map, so that the distances between actors on the map have the same order as the original ranking. The map can identify clusters within the data, and is therefore a suitable way to identify the main divisions among actors. The first step in estimating an MDS model is to determine the number of dimensions to use in mapping the proximities of observations. This is determined here by the Kruskal stress value, which measures the goodness of fit of the model for various dimensions. A rule of thumb suggested by Kruskal (quoted in Lattin et al 2003, p227) is that a stress value of 0.20 should be seen as a poor fit; a value of 0.10 indicates a fair fit, while a stress value of 0.05 is a good fit. Here, the stress value for a one-dimensional solution is 0.20; for two dimensions it is 0.15; and for three dimensions it is In order to reach a stress value that would approach a good fit, five dimensions would be required. This corresponds to the view that the EU political space is complex and multidimensional. Given the difficulties associated with representing this number of dimensions, I have chosen to present a three-dimensional solution. While this does not capture all of the variation in the distance matrix, it does provide a fair fit, and is sufficient for identifying the main clusters in the data. Figures 2 and 3 present this three-dimensional MDS map. [FIGURES 2 & 3] 22

23 A number of patterns are apparent here. There is a clear division between the member states and the EP party groups along the first dimension, with the conservative/nationalist UEN closest to the member states, followed by the EPP. The position of the EP in its first reading is closer to the PSE and ALDE positions than the EPP position. Within the Council, there is a broad north/south pattern among the old member states (dimension 2), and there is also a noticeable division between old and new member states (dimension 3). As well as graphically representing the overall proximities between actors, the division between the chambers is examined separately for each individual issue, and for each category of issue. Division between the chambers is measured in two ways. Firstly, the positions of actors in each chamber are compared. Secondly, the two chambers are treated as unitary actors, and the distances between them are examined. For this, a point estimate is required for the position of each chamber on each issue. This is relatively straight forward for the EP, as the dataset includes the position taken by the EP in its first reading on each issue. It is reasonable to assume that this position is close to what the EP would decide if it was it acting alone: the EP is the first to respond to the Commission s proposal, and at this point the Council has generally not communicated a position to the EP. However, it is not possible to include a position for the Council in the same way. This is because the Council always responds to the EP, and so the formal position taken by the Council in its first reading (the Common Position ) tends to incorporate any concessions it is prepared to give to the EP. Consequently, the Council Common Position can not be taken as the preference of the Council, but rather as a first step 23

24 towards an inter-chamber compromise. Therefore, I estimate the Council preference based on the preferences of its internal actors, according to the compromise model (Thomson et al 2006). This a weighted average of the position taken by each member state, where the weights are based on the relative power of the member states, measured according to the Shapley-Shubik index (from Bräuniger and König 2005), and the salience attached to the issue by the member state. Table 2 presents the division between the chambers for each category of issue. The first row gives the average distance between the chambers (on the 100 point scale). The second row compares the positions of EP Groups and the member states using a nonparametric equality of means test (the Wilcoxon rank-sum test); the figures represent the number of issues in each category for which this test was significant at the 95% level, indicating that the positions of EP groups differed from the positions of member states. [TABLE 2] The results show that the two chambers were distinct across a large proportion of the 41 issues, and in most categories of issues. Not surprisingly, the greatest average difference between the chambers appears for issues concerning institutional control. For three other categories of issue there is also both a large average distance between the chambers and a majority of issues for which the positions of actors in each chamber were significantly different. These are Civil Liberties, Environment and Interests. For the Funding and Harmonisation categories, there is also a large average distance between the chambers, but only a minority of issues show significant differences between 24

25 the EP Groups and member states. This corresponds to a scenario whereby there is diversity in both chambers and the positions of actors in each chamber overlap, but the majority position in either chamber differs. The category with the smallest average difference between the chambers is Market Regulation. This category also has the least number of issues on which significant differences were found between the EP groups and the member states. The nature of divisions between the chambers To understand the observed differences between the chambers, the positions of actors are related to the policy context of the issues. As discussed, the issue scales are rotated to make them comparable, and the end points provided with a policy label for each category of issue. Using these rotated scales, it is possible to interpret the relative position of actors in policy terms. Table 3 presents the rank ordering of positions for the 25 member state governments, and the 6 EP groups, and the EP first reading. The ranking is based on the average positions of actors for each category of issue. For example, for the seven Harmonisation issues, the average position of the GUE group was 17 on the 100 point scale. This was the lowest average position, and so it is ranked first. At the other end of the ranking is Lithuania, with an average score of 93. The mid-point on the scale is marked by a bold line here, France was just below the mid-point (average position of 46). 25

26 [TABLE 3] A cursory glance at this table shows that the EP and the EP groups (in bold) tend to have lower average scores than the member states. For each of these categories, low scores indicate positions that are either to the left or that are pro-integration. For disagreements over funding, most EP groups, and the EP as a whole through its first reading, tended to favour centralised rules for allocating and spending EU money. Most of the main contributors to the EU budget (such as Germany, France, and the UK) were also on this side of the scale. The new member states, on the other hand, tended to push for greater flexibility in spending. The UEN is also on this side of the scale, perhaps due to the fact that it is MEPs from the new member states comprise 59% of the UEN membership, with the Polish Law and Justice party by far the largest single grouping. Alternatively, it may reflect the nationalist policy platform of this group. For the second category of issues, most EP groups and the EP as a whole were on average more supportive of harmonisation than most member states. Some Southern member states (and Belgium) are also towards this end of the scale. Again, the new member states are towards the opposite end of the scale, and the UEN is an outlier in the EP. Disagreements over the extent of market regulation generated much more of an overlap between EP Groups and member states. This corresponds to the earlier finding that the positions of actors from the two chambers were not significantly different for most of these issues. Within the EP, the liberals are towards the free market end of the scale, as expected. Several new member states, along with France and the UK, are also 26

27 on this side. The EPP occupies a position just below the median actor, while the other EP groups are further to the left. The EP first reading position is on average between the positions of the EPP and ALDE for these issues, suggesting an alliance between these groups. A clear inter-institutional divide emerges on the Civil Liberties issues. Here, it is the Council that is relatively unified, with all member states favouring greater law enforcement capabilities. No member state has an average score of less than 50 for these issues. The EP groups were divided along social policy lines, with the two conservative groups (the UEN and EPP) taking positions close to the member states. Here, the EP position (close to the PSE and ALDE) suggests that a different majority coalition may exist for these types of issues. The majority of groups, representing a majority of MEPs, are on the opposite end of the scale. This reflects the large average distance between the chambers (Table 2). The very high level of inter-issue variation explained by a common factor (Table 1) indicates that this pattern is consistent across issues. For disagreements on environmental policy, the EP groups and the EP position were all on the green end of the spectrum. This is perhaps unsurprising, given that environmental policy is an area in which the EP has traditionally sought to make an impact. Again, most member states are on the opposite end of the scale, which corresponds to the relatively large average distance between the chambers and the high number of issues for which there are significant differences between the actors in each chamber (Table 2). Somewhat more diversity appears in the EP and the Council for issues involving employer- versus employee- and industry- versus consumer-interests. A left-wing and 27

28 liberal majority of groups again occupy one end of the scale, along with some of the old member states. The UEN and EPP are on the opposite side of the scale, while the new member states and Germany are the furthest to the right. Finally, the ordering of actors for issues concerning institutional power is quite straightforward, with all the EP groups on one side and all the member states on the other. While the number of cases in this category is small, it does suggest that even the nationalist groups such as the UEN favour an increased role for the EP in monitoring and revising legislation. Extent of division within the chambers To measure the internal division in each chamber, the standard deviation of actors positions is calculated. Figure 4 summarises the standard deviations of actors in the EP and the Council on the issues from each category. The unit of analysis for the EP is the individual MEP, and for the Council the unit of analysis is the member state vote 5. By breaking down the actors in this way, the results take into account the actors relative size. [FIGURE 4] 5 For the EP, the information in the original data is at the level of national parties and party groups. The data contains one estimate per issue for the Green, Gue and Uen groups. For this analysis, these positions were repeated according to the number of MEPs in each group. For the EPP, PES and Alde, the data contains estimates for each national party. Here, these positions are repeated according to the number of MEPs in each national party. For the Council, the information in the data is at the level of member states. In this analysis each member states position was repeated according to their voting weight. 28

29 The EP shows consistently high levels of internal division for the Civil Liberties, Interests and Market Regulation issues. For the Council, on the other hand, the civil liberties issues produced the greatest level of cohesion, along with issues concerning institutional power. For most issues, the Council was divided, particularly for issues in the Harmonisation, Market Regulation and Funding categories. Within the EP, two types of division can be distinguished: division between party groups, and division within party groups. Table 4 examines the level of division within the three main party groups. The positions of individual MEPs from each of these groups (265 MEPs from the EPP group; 200 from the PSE group and 90 from the ALDE group) are compared with the official positions of their party groups for all issues. The R 2 statistic is reported, which represents the proportion of variation in MEP positions accounted for by the positions of their party groups. For reliability, the non-parametric Spearman s Rho is also reported, although the meaning of this coefficient is more difficult to interpret. The two tests show comparable results, with the EPP being the most divided party, and the PSE the least divided. The table also gives the percentage of MEP positions that were in agreement with the position of their party groups. [TABLE 4] The results suggest that 58% of the overall variation in these MEPs positions is accounted for by the positions of their groups. For MEPs from the EPP, the figure is 52%; for the PSE it is 64%; and for ALDE it is 59%. The percentage of variation not accounted for by the party group positions is due to divisions between national parties 29