JCMS 2006 Volume 44. Number 2. pp. 391-417 Who Has Power in the EU? The Commission, Council and Parliament in Legislative Decisionmaking* ROBERT THOMSON Trinity College, Dublin MADELEINE HOSU Leiden University Abstract What is the relative power of the European Commission, the Council of Ministers and the European Parliament (EP) in the European Union (EU)? Both scholars and practitioners of EU affairs provide different answers to this seemingly straightforward question. In this article, we examine the balance of power among these three actors in the context of legislativedecision-making. We report the results of a small survey among a select group of practitioners of EU affairs. Their judgements on the relative power of the three organizations vary considerably. We distinguish between two contrasting views: a Council-centric view that attributes more power to the Council of Ministers than to the Commission and Parliament, and a supranational view that attributes large amounts of power to the supranational organizations relative to the Council. To test the veracity of these alternative views, we incorporate them into two variants of a simple and testable bargaining model that makes forecasts of decision outcomes, based on information on actors' preferences. The models are then applied to a dataset that includes information on EU actors' policy positions on 162 controversial issues of which the decision outcomes are known. The variant of the bargaining model incorporating the Council-centric view provides significantly more accurate forecasts. We are grateful to the anonymous JCMS referees, and to Raj S. Chari, Genna Gaspare and Frans N. Stokman for comments on earlier versions of this article. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford 0X4 2DQ, UK and 350 Main Street Maiden, MA 02148, USA 392 ROBERT THOMSON AND MADELEINE HOSLI Introduction What is the relative power of the European Commission, the Council of Ministers and the European Parliament (EP) in legislative decision-making in the European Union (EU)? What levels of resources do these actors possess that enable them to exert influence on each other and on the contents of legislation? The balance of power between these actors matters, because it influences the contents of legislation, which in turn affects just about every aspect of economic, social and cultural life in the EU's Member States. Each of the three actors has distinct roles in the legislative decision-making process. The Commission holds the right to initiate legislative proposals, which it is widely acknowledged gives it considerable potential to influence the contents of legislation. However, the Commission's proposals can be amended by, and need to be approved by, the Council of Ministers and, depending on the policy area, Parliament too. The Council and the Parliament hold legislative authority in the EU. The Council, where Member States' governments are represented, is considered by some to be the most powerful institutional actor in the EU. The formal rules of interaction between the Commission, Council and Parliament are specified in the legislative procedure that applies to whichever proposal is being discussed. The choice of procedure depends on the Treaty article on the basis of which the Commission introduces its proposal. Today, the most important and widely used legislative procedures are co-decision and consultation. The specifications of the legislative procedures certainly influence the extent to which each of the actors is able to leave its mark on the contents of legislation. Under the co-decision procedure, the Council and Parliament are essentially equals, whereby the approval of both bodies is required before legislation can be adopted. The co-decision procedure as specified in the Treaty of Amsterdam, the version examined here, provides for the possibility of two parliamentary readings of a legislative proposal. In the event of a protracted disagreement between the Council and the Parliament, a Conciliation Committee is set up, composed of representatives of the Council and Parliament, with a view to formulating a joint text. The Committee must succeed in formulating a joint text and this must be approved by both the Council and Parliament for the legislative proposal to be adopted. Consultation is an older and simpler procedure than co-decision. Under consultation, the Commission proposes and the Council decides; the EP has an advisory function only. The relative power of these three actors has been a central element of many debates on law-making in the EU, even among scholars adopting similar research approaches. For example, there have been debates among formal modellers who have sought to understand legislative politics by examining the constraints 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 393 and opportunities provided by procedural rules. The analyses are based on the premise that these procedures, or formal institutions as they are referred to, are the key factors influencing actors' behaviour (Garrett and Tsebelis, 2001, p. 356). Despite focusing on the same type of institutions, researchers using this approach have reached different conclusions on the effects of successive Treaty revisions on the power of these three actors. Differences in their conclusions are due to the indeterminacy of the formal rules, in the sense that these rules are open to interpretation. For example, some researchers have argued that the co-decision procedure introduced by the Maastricht Treaty reduced the power of the Parliament compared with the co-operation procedure it replaced (Garrett, 1995; Tsebelis, 1997; Tsebelis and Garrett, 2000), while others have argued that co-decision increased Parliament's power (Scully, 1997; Crombez, 2000). Further, a recent review of formal models of the legislative procedures currently used in the EU argued that there are different interpretations of these institutions, each of which has different implications for the relative power of the Commission, Council and Parliament (Selck and Steunenberg, 2004).: For example, in the rules describing the current version of co-decision, although the Commission is charged with introducing the proposal, what happens in the Conciliation Committee is open to interpretation: does the Council make a proposal to the Parliament concerning the content of the final legislation to be adopted, or does Parliament make this proposal? This matters, since the actor able to make the first move has a distinct advantage when it comes to leaving its mark on the legislative proposal just before approval. Researchers who include informal institutions in their analyses also reach different conclusions regarding the balance of power between the Commission, Council and Parliament. In contrast to formal institutions - written rules enforced by third parties - informal institutions are unwritten, consist of norms of behaviour, and are enforced by the actors themselves (Knight, 1992). In their consideration of the balance of power between the Council and the Parliament under co-decision, Farrell and Héritier (2003) argue that informal institutions are of vital importance. They contend that 'the dynamic interaction between formal and informal institutions ... has important consequences for legislative outcomes and the relative decision-making power of European political actors' (Farrell and Héritier, p. 578). Informal institutions are shaped by actors' different time horizons, sensitivities to failure and levels of resources. These factors are said to shift the balance of power in favour of the Parliament. For example, the Parliament's time horizon is longer than the Council's, which is largely defined by the six-monthly presidencies. This gives the Parliament the advantage in negotiating with the Council. Further, Parliament has an advantage in terms of 1 On relative power, formal representation and voting weights in EU institutions, see Hosli and Machover (2004) and Taagepera and Hosli (forthcoming). 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 394 ROBERT THOMSON AND MADELEINE HOSLI resources, since parliamentary committees focus on particular areas of policy, while officials from the Committee of Permanent Representatives (Coreper) in the Council tend to be generalists. Moreover, Farrell and Héritier (2003, p. 593) report that parliamentarians have been expanding their influence under the consultation procedure by using informal rules that have been developed in the practice of the co-decision procedure. With the data presented in this article we contribute to the ongoing debate on the relative influence of the institutional actors in EU decision-making in two respects. First, we describe the views and quantitative judgements of a select group of prominent practitioners of European affairs regarding the balance of power between the Commission, Council and Parliament. Given the variance we find in practitioners' views, it should perhaps not surprise us that academics' views also differ. Second, we incorporate alternative sets of judgements into alternative variants of a simple bargaining model, and apply these to a large number of controversial issues that were the subject of debate in EU legislative decision-making. We find that the views of a minority of practitioners, those who attribute relatively modest levels of power to the Commission and the EP, generate the most accurate predictions. This allows us to infer which judgements on the balance of power are most accurate. This analysis identifies the net effect of formal and informal institutions on the three actors' power, as defined by their ability to shape decision outcomes in legislative decision-making. The results also provide indications, although not precise estimates, of the relative impact of formal and informal institutions in defining actors' power. These analyses do not distinguish between the variety of alternative strategies actors may employ when using their power. Another limitation of the present analyses is that they do not incorporate the role of the European Court of Justice. Instead, we focus here on the three institutional actors who are most relevant in legislative decision-making prior to the adoption of legislative proposals. This article is organized as follows. Section I describes the design of the small survey of practitioners and reports the main findings of the survey. Section II describes and illustrates the model used to test alternative views on the balance of power in the EU against each other. The model is called the compromise model, a computationally simple, but predictively powerful model of political bargaining. Section II also describes the data to which the model is applied. Section III contains the analyses of the data on EU decision-making using the model. It begins by illustrating how the application of the model enables us to differentiate between alternative views on the balance of power among the Commission, Council and Parliament. The analyses in which the two alternative views on the balance of power - referred to as the Council-centric view and the supranational view - are then presented. These inquiries are 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 395 supplemented by exploring a larger number of alternative views on the balance of power, to identify which estimates allow the model to generate the most accurate forecasts. We conclude by drawing inferences on the relative power of the Commission, Council and Parliament in legislative decision-making. I. The Survey The survey on the distribution of capabilities between the Commission, Council and Parliament was fielded among practitioners of European affairs. This small survey was part of a larger project 'Decision-Making in the European Union' (Stokman and Thomson, 2004; Thomson et al, 2006). The respondents were interviewed in person, asked to discuss their views and, where possible, quantify their judgements on the relative power of the three actors. The interviews were held between September 2000 and July 2001. The selection of practitioners and the structure of the interviews are outlined first. Subsequently, the article reports the main findings concerning the arguments they used in support of their judgements, and the numerical estimates they gave. Survey Design The practitioners were selected such that, although the survey was small, the results would represent the views of a range of well-informed individuals. Given the small size of the sample (21) and the distribution of the informants (most were from the Commission or the Council), this survey should not be considered representative. Instead, it aims to explore the opinions of a range of informed practitioners and to generate hypotheses for further investigation. Ten of the 21 informants were officials from the Commission, two from the Council Secretariat, seven from the permanent representations of the Member States in the Council and two from the EP2 Most were selected on the basis of their professional position. They had been working in the EU for many years, and/or had attained a position such that they had a vantage point from which to observe a range of decision situations. For example, two of our informants had headed the Commission's civil service. Most of the civil servants from the permanent representations were responsible for co-ordinating decision-making in Coreper. The interviewees were also selected due to the fact that they displayed a broad knowledge of EU affairs in previous discussions with them.3 2 Practitioners were also asked for their judgements on the distribution of capabilities within the Council. Here, however, we report only on the estimates regarding the relations between the Commission, Council and Parliament. 3 In the larger project on EU decision-making of which the present study is part (see Stokman and Thomson, 2004; Thomson et al., 2006), more than 150 interviews were held with informants on specific Commission proposals. Some of the informants used in the present survey were identified through and selected on the basis of their performance in these interviews. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 396 ROBERT THOMSON AND MADELEINE HOSLI The interviews were semi-structured. The judgements on the relative power of the Commission, Council and Parliament were elicited using a standard request and a general procedure.4 The standard request was worded as follows: Within the policy domain (specify domain) subject to (type of legislative procedure, including Council voting rule), the different stakeholders have different capabilities or amounts of potential to influence decision outcomes. This ability is based on a number of different resources: for example, the formal authority to take decisions, financial resources, information, access to other important stakeholders, leadership of a large number of people etc. Please indicate the capabilities of each stakeholder on a scale from 0 to 100. When responding to this request, the interviewees were asked, first, to rate the capabilities of the three actors - the Commission, Council and EP - relative to each other. These interviews resulted in 36 sets of estimates on the relative capabilities of the Commission, Council and Parliament, 18 relating to the co-decision procedure and 18 relating to the consultation procedure. Some of the informants indicated that their estimates were specific to a particular policy area subject to a particular legislative procedure. For example, one informant provided different estimates for the area of the internal market, subject to co-decision and transport policies, also subject to co-decision. Others made an overall judgement on the distribution of capabilities for decision-making under a given procedure without distinguishing between different policy areas. It was left to the informants themselves to decide whether to concentrate on a legislative procedure in general or to focus more specifically on a policy area. The question was intended to obtain judgements on the capabilities of each actor relative to those held by others. To assist the informants in making these judgements, hypothetical coalitions of actors were posed and compared to others; for example, 'what would happen if the Commission took one position and the Council and EP another on an issue they attached equal importance to?'. Through such comparisons, an attempt was made to gauge the relative weight that should be attributed to the three actors. Although a scale of 0 to 100 was suggested and most informants stayed within this range, they were encouraged to use any numbers they felt comfortable with to represent their judgements; it was emphasized that the relations between the numbers were more important than their absolute values. In most cases, the informants found it easiest to attribute a score of 100 to the actor or actors that in their view held most capabilities and to rate the others relative to this score. Others preferred 4 For a more detailed description of the procedure of comparing coalitions of stakeholders to obtain estimates of their relative capabilities, see Bueno de Mesquita (2000). 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 397 to describe the relations using other numbers: such as Commission 70, Council 80 and EP 60; or 'one third for each of the three institutions'. To make these estimates comparable with each other, they were re-scaled after the interviews such that the Council has a score of 100, and the other two institutional actors are scaled higher or lower than this score. So, for example, Commission 70, Council 80, EP 60 becomes Commission 87.5, Council 100, EP 75. It was also possible for the informants to rate the three institutional actors equally. Survey Results Since the informants had to justify their numerical estimates, the interviews yielded a considerable amount of qualitative argumentation relating to the power relations. These arguments are summarized below. Despite the similarities between the qualitative arguments mentioned during the interviews, the informants provided quite different numerical estimates, as will be discussed later in this section. Box 1 contains a list of the 12 arguments forwarded by at least two of the 21 practitioners consulted. A detailed report was written on each of the 21 interviews, describing the arguments used by the informants when rating the three actors' relative capabilities. The arguments listed in Box 1 were distilled out of the many examples, personal experiences and anecdotes contained in the reports. The most common argument advanced is that the Commission has almost exclusive right to initiate proposals. Of course, the initiating role played by the Commission is a prominent feature of many of the procedural models referred to earlier. However, the informants indicated that the importance of the right to initiate proposals was not confined to the Commission's ability to locate its legislative proposal somewhere in the policy space, as depicted by the procedural models. The right to initiate also gives the Commission the opportunity to frame the proposal in more general terms, by defining the terms in which issues are discussed. Further, the Commission's influence is not restricted to the formulation of the initial proposal. It is also actively involved in the policy discussions in the Council, and in negotiations between the Council and the Parliament. Despite similarities between the practitioners' qualitative arguments, a substantial amount of variation is present in their numerical judgements on the relative capabilities of the Commission, Council and Parliament. The figures in Table 1 summarize the judgements on the power of the three institutional actors. Most of the interviewees preferred to discuss the power relations in the context of a defined policy area subject to a certain procedure (e.g. agriculture subject to consultation and QMV voting), rather than a procedure in general (e.g. all policy areas subject to consultation). 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 398 ROBERT THOMSON AND MADELEINE HOSLI Box 1: Summary of Arguments Relating to the Relative Capabilities of the Commission, Council and Parliament Arguments Regarding the Commission's Power 1. The Commission has the almost exclusive right to initiate and frame proposals (12 reports). 2. The Commission is adept at forging political deals between other actors, both within the Council and between the Council and EP. This provides the Commission with a strong potential to influence the decision outcome (6 reports). 3. It is easier for the Council to accept than to amend a Commission proposal (4 reports). 4. The Commission has the potential to withdraw a proposal before the Council has adopted it, at least under consultation and the first stage of co-decision (2 reports), although this potential is of limited use in practice (2 reports). 5. The Commission has great expertise on the policy areas affected by its proposals (3 reports). 6. Extensive consultation and arbitration prior to the introduction of the proposals give the Commission authority during the discussions with the Council and EP (2 reports). 7. After the release of the proposal, the Commission is usually able to present a consistent line and united front in discussions with the other institutions; this cannot be said for the Council and EP (2 reports). Arguments Regarding the Council's Power 8. The Council derives much of its capabilities from the strength of the national bureaucracies that support the Member State representatives, providing them with expertise (2 reports). 9. In the Conciliation Committee at the end of the co-decision procedure, the Council is generally better organized than the EP and is able to present a consistent policy stance (2 reports). Arguments Regarding Parliament's Power 10. The formal rules of the co-decision procedure provide the EP with a potential comparable to that of the Council to influence decision outcomes (5 reports). 11. The EP's lack of expertise on the technicalities of policy often means that it cannot make full use of the potential it has according to the procedural rules (3 reports). 12. Even under the consultation procedure, the Parliament's opinions are taken into account, particularly by the Commission (5 reports). Note: The number of the 21 interview reports that contained this argument are in parentheses. There is a substantial amount of variation among the estimates, even within sets of estimates that are said to refer to the same policy area, subject to the same legislative procedure. The standard deviations reported in Table 1 are high. Even when the estimates were elicited by the same interviewer, from individuals from the same organizations, large differences are found. For 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 399 Table 1: Informants' Estimates of Capabilities of Commission and Parliament Relative to the Council under Consultation (CNS) and Co-decision (COD) Procedure Policy Area N Commission EP Mean s.d. Mean s.d. CNS Internal market 4 108.3 19.5 41.1 10.3 Agriculture 6 79.5 41.4 35.7 29.5 Fisheries 1 100.0 30.0 Taxation 1 87.5 75.0 Not specific 6 69.2 51.2 21.7 16.0 Total 18 84.1 39.8 34.1 22.7 COD Internal market 7 115.6 37.0 80.4 14.3 Environment 1 140.0 100.0 Health/consumer 1 150.0 80.0 Transport 1 28.6 128.6 Social affairs 1 40.0 60.0 Not specific 7 81.4 29.5 94.3 7.9 Total 18 96.5 42.1 88.0 17.0 Source: Authors' own data. Note: Council is set to a value of 100 for presentation. example, an informant in one of the permanent representations was asked for his judgement on the relative capabilities of the three actors in the policy area of agriculture (subject to the consultation procedure and QMV in the Council), to which he responded: Commission 15, Council 100, EP 3. On the same day, the same interviewer posed the same question to another informant from the same delegation and obtained the estimates Commission 100, Council, 100, EP0. A closer inspection of the estimates reveals that there are two groups of estimates for each legislative procedure, each reflecting a distinguishable view on the relative power of the supranational actors relative to the Council under the consultation and co-decision procedures. We label these views the Council-centric view and the supranational view. Figure 1 shows the frequency distribution of judgements on the Commission's capabilities on issues subject to the consultation procedure, relative to the Council (the Council is given a score of 100 for presentation purposes). In four of the estimates, relatively modest scores are attributed to the Commission, in the range of around 20^10. In the remaining 14 estimates, considerably higher scores are attributed to the Commission, in the range of around 80-140. Similar distinctions can be made regarding estimates of the Commission's capabilities under co-decision, and 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 400 ROBERT THOMSON AND MADELEINE HOSLI regarding the estimates of Parliament's capabilities under consultation and co-decision (see Table 2). Under co-decision, we distinguish between a group of estimates in which the EP is said to have equal or greater weight than the Council, and a group that attributes less capability to the EP than to the Council. Within each of the two sets of estimates, there are still differences between the numerical estimates and the supporting qualitative argumentation. However, the variation is considerably smaller than that found among all the estimates. The practitioners' judgements do not appear to vary systematically by the institution with which they were affiliated. Such a bias might have been expected. It might, for example, have been supposed that civil servants from the permanent representations of the Member States would attribute more capabilities to the Council relative to the Commission and EP. Table 3 provides Figure 1: Distribution of Judgements on Commission's Capabilities Relative to Council at 100 6 - Council-centric view Supranational view w 5 -0 .1 4 -w "5 3 " Í 2-1 -0 1. Hi, l 0 20 40 60 80 100 120 140 Commission's capability score Source: Authors' own data. Table 2: Council-centric and Supranational Views on the Balance of Power Procedure Policy Area Commission EP N Mean s.d. N Mean s.d. CNS Council-centric 4 18.8 UA 4 3.3 4.7 Supranational 14 102.7 18.3 14 42.9 17.1 COD Council-centric 4 38.4 7.0 10 76.3 9.9 Supranational 14 113.2 31.1 8 103.6 10.1 Source: Authors' own data. Note: Estimates of relative capabilities under consultation (CNS) and co-decision (COD) relative to the Council score of 100. © 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 401 Table 3: Informants' Estimates, by their Location, of the Relative Capabilities of Commission and EP Relative to Council under Consultation (CNS) and Co-decision (COD) Procedure Informant's Location N Commission EP Mean s.d. Mean s.d. Commission 10 86.9 36.8 35.0 14.2 Council Secretariat 1 100.0 - 70.0 - Permanent Representation 6 72.1 50.6 23.0 29.9 European Parliament 1 111.1 - 55.6 - Total 18 84.1 39.8 34.1 22.7 Commission 10 104.1 34.1 90.5 15.3 Council Secretariat 1 77.8 - 77.8 - Permanent Representation 3 116.7 35.1 83.3 5.8 European Parliament 4 67.1 62.1 89.6 28.8 Total 18 96.5 42.1 88.04 17.0 Source: Authors' own data. Note: Council is set to a value of 100 for presentation. information on the judgements on the relative power of the actors by the location of the practitioners who provided these estimates. Although the numbers of observations are limited, there do not appear to be large and consistent differences. It might be ventured, however, that officials from the permanent representations hold a somewhat more Council-centric view of the balance of power under the consultation procedure than do other informants. Given the debates among academics on the relative power of the institutional actors in the EU, it should perhaps not be too surprising us that practitioners hold a similar variety of views. In the following section, these alternative views are explored in more depth. II. Comparing Alternative Views on the Balance of Power in the EU To test the Council-centric and supranational views against each other, we load a simple model of political bargaining with these two sets of estimates. This model, the compromise model, generates predictions of decision outcomes on controversial issues dealt with by the EU using three variables: actors' policy positions on these issues, the importance they attach to them, and the actors' relative capabilities. We construct two variants of this model: one with Council-centric estimates of the actors' relative capabilities, and one with 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 402 ROBERT THOMSON AND MADELEINE HOSLI supranational estimates. We then identify which variant of the model generates the most accurate forecasts of decision outcomes in the EU. The first part of this section describes the model. The second part describes the data set on which the analyses are performed. The Compromise Model The compromise model's forecast is simply the mean average of the actors' policy positions, weighted by the product of their capabilities and the levels of salience they attach to the issue. As a formula: / , x c s O = " '" ' '" V" where Oa is the prediction of the decision outcome on issue a generated by the compromise model, xa denotes the position of actor i (from the set of actors, n) on issue a, c. denotes the capabilities of actor i in this decision situation, and s. is the level of salience actor i attaches to issue a. la Van den Bos (1991) proposed the compromise model in his study of decision-making in the Council of the European Community.5 When describing the decision-making process this model represents, he emphasized that it 'takes all positions of Member States into account, weighting these by the resources a Member State can apply during the negotiation and the importance each attaches to the decision at hand' (Van den Bos, 1991, p. 176). The compromise model conceives of power as a 'capability', a potential to influence other actors and decision outcomes; this is a potential that may be used by actors. The extent to which an actor puts its power into effect depends on the level of importance it attaches to the issue at stake. Further, researchers in the tradition with which the compromise model is associated - models of collective decision-making - commonly refer to power as being based on 'resources' (see, for example, Thomson et al., 2003). This term is used to indicate the variety of sources from which actors' power derives, including formal and informal institutions. The resources relevant to bolstering an actor's power depend on the particularities of the decision situation. For example, military might is relevant in the face of an explosive international conflict, but is unlikely to help secure any particular outcome when the reform of the banking sector is being decided on. The compromise model is not concerned with the composition of actors' power; this is exogenous to the model. Rather, it is concerned with the transformation 5 Achen (2006) provides a more extensive discussion of the theory behind the compromise model, and proves that it is a first-order approximation of the generalized Nash bargaining solution. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 403 of actors' positions into decision outcomes and how their relative power affects this transformation. The compromise model has been tested in a variety of contexts against supposedly more sophisticated models of decision-making and has performed well. In one study, for example, the compromise model was tested against two models of bargaining in the Council of Ministers of the European Community: a non-co-operative conflict model and a co-operative exchange model (Bueno de Mesquita and Stokman, 1994). The forecasts of decision outcomes generated by the compromise model were not statistically distinguishable from those of the more complex bargaining models. In a more recent study, the compromise model was tested against a broader range of rational choice institutionalist explanations of legislative decision-making in the EU (Stokman and Thomson, 2004; Thomson et al, 2006). Many of the more complex models generated less accurate predictions than the compromise model and none generated more accurate forecasts. Therefore, although the compromise model, like any other model, does not produce perfectly accurate predictions, it does seem to have relatively high predictive power. Here, we take advantage of this predictive power by using the model to identify which estimates of the balance of power among the EU's institutional actors are associated with the most accurate predictions of decision outcomes. A Dataset on Legislative Decision-making in the EU The dataset to which the variants of the compromise model are applied contains information on the policy positions of EU level actors in the period 1999-2001: the Commission, each of the then 15 Member States and the EP.6 The information refers to controversial issues, 162 in total, raised by 66 legislative proposals that had been introduced by the Commission. The legislative proposals were subject to either the consultation procedure or the co-decision procedure. The information was collected using key informants, backed up wherever possible by documentation. The selection of these legislative proposals was based on three criteria: the type of legislative procedure followed, the time period involved and the level of political importance. Concerning the decision-making procedure, the legislative proposals selected were subject to either the consultation or the co-decision procedures. These are the most commonly used legislative procedures in the EU, and these were the procedures in which we wished to examine the interaction of the three institutions.7 Forty of the 66 Commission proposals, 6 For more details of this dataset, see Stokman and Thomson (2004), Thomson et al. (2004, 2006). 7 Note that we selected only legislative proposals whose decision-making procedure did not change after the Treaty of Amsterdam came into effect in May 1999, which made changes to the co-decision procedure. This avoids the results of the study being contaminated by changes in the procedures during the course 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 404 ROBERT THOMSON AND MADELEINE HOSLI containing 94 of the 162 issues, were subject to the consultation procedure; the rest were subject to the co-decision procedure. With regard to the time period followed, the proposals had to be pending in 1999 or 2000. Finally, the selection includes proposals on which there was an indication of at least some minimum level of political importance and controversy. The legislative proposals had to be reported on in Agence Europe, a news service that covers EU affairs, and the informants had to be able to identify at least one issue on which there was a substantive disagreement between at least some of the actors. Of course, issues on which all actors take basically the same position do not provide an opportunity to test alternative views on the relative power of the actors. We used key informant interviews backed up wherever possible by documentation to collect information on the controversial issues raised by these proposals, actors' policy positions on these issues and the levels of importance they attached to them. On average, between two and three interviews were held on each of the 66 Commission proposals selected. The interviews were held with key informants who had detailed knowledge of the decision-making situation in question. These were usually civil servants from the Commission or permanent representations, who were involved in the discussions. Outsiders simply did not have access to the detailed information required for these analyses. The validity and reliability of the experts' judgements is examined in detail elsewhere (Thomson, 2006). The validity of the informants' judgements was assessed by comparing these to information from documentation, minutes of Council meetings and publicly available inter-institutional documents. Informants were selective in the issues they told us about; they tended to identify issues that were most controversial, that interested most actors and that were most difficult to resolve. These are precisely the kind of issues we are most interested in. The reliability of the judgements was assessed by comparing the judgements of different interviewees. Differences were found between interviewees in the style in which they represented issues. For instance, some broke issues down into a number of detailed points, while others summarized detail in smaller number of more broadly defined dimensions. However, there was agreement in terms of the substance of the policy questions informants identified as controversial and the general policy stances of the actors. Each of the controversial issues within each of the legislative proposals was described spatially. With the help of key informants, the actors and the policy alternatives they favoured were located on policy scales to reflect the political distances between them. In addition to placing the actors on these policy scales to reflect their most favoured decision outcomes, the experts were asked to of the decision-making process. The Amsterdam version of the co-decision procedure applied to all the proposals decided on by co-decision. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 405 estimate the level of importance each of the actors attached to each of these issues. This level of importance was estimated on a scale of 0-100, whereby a score of zero indicates that the issue was of no importance whatsoever, 50 that it had an 'average' level of importance to the actor concerned, and 100 that the issue could hardly be more important. The relations between the salience scores for different actors are more important than the absolute value of the scores. When obtaining the judgements on actors' positions and the levels of importance they attached to the issues, they were asked to substantiate their judgements extensively. An example of this way of representing a controversial issue is provided by an issue raised by a set of proposals introduced by the Commission in 1996 concerning the revision of certain directives on food standards.8 The Commission's aim was to simplify these directives, so that they covered only essential requirements, thereby facilitating the free movement of such products in the internal market. An issue of particular relevance to our concern with the balance of power among the institutional actors was about the establishment of the comitology procedure for the future updating of these directives (Figure 2). The establishment of an extensive comitology procedure would have meant that the directives could have been updated on the advice of a committee of experts; there would be no need for the Commission to introduce new legislative proposals for the approval of the Council to update the directives. This was a controversial issue both within the Council and between the Council and the Commission. At the right end of the policy scale used to describe this issue, the Commission favoured the introduction of an extensive comitology procedure for updating these directives. At the left end of the policy scale, the UK and the three Nordic countries favoured the requirement of new legislation before changing the directives in the future. Other Member States took more intermediate positions, favouring more or less extensive forms of comitology than the UK or the Commission. This was clearly an issue that divided the Council and the Commission, since it affected the future relations between the two actors on these policy questions. Some Council members were less wary of giving a comitology committee say over an area in which the Council approval was required. Smaller Member States tended to attach less importance to limiting the scope of the comitology committee. In the end, a comitology procedure described as having 'substantial scope' was instituted. This outcome corresponded with position 60 on the issue scale used to depict the debate. This exercise was performed for 162 issues from the 66 selected B These directives were passed into law in 2001. They included Directive 2001/110/EC relating to honey and Directive 2001/112/EC on fruit juices. All were subject to consultation with the EP. Directive (2000/36/ EC) on chocolate was introduced at the same time, but was discussed separately; it was subject to the co-decision procedure. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd Figure 2: The Issue of Comitology in Food Directives. Directive 2001/110/EC Relating to Honey (Salience Scores in Parentheses) How extensive should the comitology procedure for updating the directive be? Prediction: Prediction: Council-centric supranational model model 51 79 UK, DK, FIN, Sw (50) COM (100) 0 20 D 0: status quo: 20: limited to no comitology/ technical parts new legislation of the required for directives updating 60: outcome: comitology with substantial scope 100: very extensive comitology procedure WHO HAS POWER IN THE EU? 407 Commission proposals. The policy scales were standardized so that the end points are 0 and 100. III. Analysis We now turn to the analysis of the data on legislative decision-making using the compromise model. This section first illustrates how the analyses are performed, and presents the results of the comparison of the two views on the balance of power between the three actors. These analyses are then supplemented with a more exploratory analysis that examines a range of capability scores and identifies those associated with the most accurate forecasts of decision outcomes. The Council-centric and Supranational Views Compared Two variants of the compromise model were created: one with the Council-centric estimates of the balance of power between the Commission, Council and Parliament, and one with the supranational estimates.9 The errors of these two variants of the compromise model in predicting decision outcomes are then compared. Consider the example of the comitology issue depicted in Figure 2. The Commission proposal in which this issue was raised was subject to the consultation procedure. The Council-centric view on the balance of power under consultation attributes a power score of 19 to the Commission and 3 to the EP, relative to the Council's score of 100 (Table 2). When these weights are used in the compromise model, it generates the prediction that the decision outcome will correspond with position 51 on the policy scale, nine points to the left of the actual decision outcome. The supranational view on the balance of power attributes a score of 103 to the Commission and 43 to the EP, relative to the Council's score of 100. When these supranational weights are applied, the model's prediction of the decision outcome lies at point 79 on the scale, 19 points to the right of the actual decision outcome. In this case, the Council-centric estimates generate more accurate predictions than do the supranational ones. In this particular case, the EP was said not to have taken a position. Nevertheless, the case still allows us to test alternative views on the relative power of the Commission and the Council. ' In these analyses the following scores were attributed to the 15 Member States (standardized so that the scores add to 100): A, 3.92; B, 5.94; DK, 4.40; FIN, 4.01; F, 12.09; D, 11.25; IRL, 3.70; I, 9.52; L, 2.50; NL, 6.57; P, 5.14; E, 10.02; SW, 5.00; UK, 11.95. These estimates were obtained from a survey of 23 experts (including most of the experts consulted for scores reported in this article). There was a much higher level of stability in the estimates of the capabilities of the Member States, compared with those of the institutions. The Member States' capabilities correlate highly with the number of votes they hold under qualified majority voting in the Council. In a separate article, the expert judgements on the Member State capabilities are examined. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 408 ROBERT THOMSON AND MADELEINE HOSLI Table 4 contains the results of a comparison of the Council-centric and the supranational variants of the compromise model. For both procedures, the capability scores reflecting the Council-centric view generate more accurate predictions of decision outcomes than do the scores reflecting the supranational view. On the 94 issues subject to consultation, the predictions of the Council-centric variant of the model have a lower mean average absolute error than the supranational variant (21.9 compared to 26.1). A statistical comparison of the errors indicates that the Council-centric compromise model's predictions are significantly more accurate. The same pattern is found for the issues subject to the co-decision procedure, although the difference between the two variants is smaller in terms of the mean average error of the models' predictions (25.6 compared to 27.5). Nevertheless, the Council-centric model is still significantly more accurate than the supranational model on the issues subject to co-decision. The Best-fitting Power Estimates This section supplements the previous analyses by exploring which combinations of capability scores for the Commission, Council and Parliament allow the compromise model to generate the most accurate predictions. We explore all logically possible combinations of power scores and identify the combination that fits best. It may be the case, for example, that some combination of capability scores within or outside the range between the Council-centric and supranational views allows the compromise model to generate more accurate forecasts. The results of these exploratory analyses Table 4: Errors of Council-centric Compromise Model Compared with Errors of the Supranational Compromise Model on Issues Subject to Consultation (CNS) and Co-decision (COD) CNS ■(n = 94) COD (n = 68) Council-centric average error 21.9 25.6 Supranational average error 26.1 27.5 Council-centric prediction Better than 54 46 Worse than 36 22 Equal to 4 0 Supranational prediction Z (p) 2.42 ■(P-- = 0.015) 2.10 (p = 0.036) Source: Authors' own data. Note: Wilcoxon signed rank test used to compare the accuracy of the predictions. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 409 suggest that, with regard to issues subject to the consultation procedure and unanimity voting in the Council, even the Council-centric view overestimates the power of the Commission and the EP; the same appears to be the case regarding the co-decision procedure. With regard to issues subject to the consultation procedure and qualified majority voting in the Council, the optimal weights for the supranational actors lie between the Council-centric and the supranational views. The compromise model was applied to each of the issues in the data set using different relative capability scores for the Commission, the Council and the EP. First, the total power score of the Council was set to 100.10 Then, the power scores of the Commission and the EP were varied (independently of each other) from 0 to 200 relative to the Council at 100. The power scores of the Commission and the EP were varied in steps of 5. This meant that we first held the EP's power at 0, and gradually increased the power of the Commission from 0 to 200 in increments of 5. We then increased the power of the EP to 5, and again gradually increased the power of the Commission from 0 to 200 in increments of 5. We followed this procedure until we had given the EP a power score of 200. Therefore, on each issue 1,681 variants of the compromise model were applied. This allows us to identify which combination of power scores allows the compromise model to generate the most accurate forecasts of actual decision outcomes in the EU. Following this procedure, the relative power of the Commission and the Council was varied on the comitology issue depicted in Figure 2. As mentioned above, the EP played no role in the discussions on this issue; it was a dispute between the Commission and the Council. Therefore, in this particular case, only the power score of the Commission was varied. When the Commission is given a power score of zero relative to the Council's 100, the compromise model generates a prediction of 30 on the scale. This is quite far (30 scale points) to the left of the actual outcome, corresponding to a much less extensive form of comitology than was actually passed into law. When the Commission is given a power score equal to that of the Council total (corresponding approximately to the supranational view), the compromise model generates a prediction at point 79 on the scale, to the right of the actual outcome. The compromise model generates the most accurate prediction when it is assumed that the Commission has a power score of 30 relative to the Council's 100. This method of deriving empirical estimates of the power balance is not able to distinguish between different assumptions about the actors' relative power 10 As in the previous analyses in Section IV, experts' judgements on the relative capabilities of the Member States in the Council were used. We also experimented with other estimates of the distribution of capabilities within the Council based on the Shapley-Shubik voting power index. These alternative estimates did not substantially change the results. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 410 ROBERT THOMSON AND MADELEINE HOSLI when either of the following two situations occurs. The first is, as in the above example, when one of the institutional actors does not take a position. This is most common for the EP under the consultation procedure. Second, when an actor takes a position at the centre of the configuration of other actors and this position corresponds with the actual outcome, this method is unable to distinguish between alternative assumptions. For example, suppose only the Council members and the EP are involved in the negotiations on a particular issue. The EP takes a position at the centre of the policy scale, and has Council members of equal effective power to either side of it. Suppose, further, that the EP's position corresponds with the actual outcome. In such a situation, it is impossible to say whether the EP's position corresponds with the actual outcome because it is powerful or simply lucky (Barry, 1970). In this case, regardless of the assumptions we apply about the relative power of the three actors, the compromise model would generate the same prediction. In such a situation, a score of 200 for the EP relative to the Council would generate exactly the same prediction as a score of zero for the EP relative to the Council. In other words, some alignments of actors allow us to differentiate more than others between different assumptions about the balance of power. Nonetheless, this method does not lead us to conclude that an actor is powerful simply because the decision outcome is close to its preferred outcome. The issues were first divided into those subject to the consultation and co-decision procedures, since the relative capabilities of the Commission, Council and Parliament were expected to differ between them. In particular, we expect the Commission to have more power under consultation than under co-decision. The EP's power score is expected to be higher under co-decision than under consultation. In addition, a distinction can be made between two variants of the consultation procedure: one in which the Council must take decisions by qualified majority voting, and a second under which unanimity is required.11 Although the informants did not distinguish explicitly between these two variants of the consultation procedure, the relative power of the three actors may be expected to vary between them. Unanimity is usually confined to policy areas that are particularly sensitive areas of national interest, or that fall outside the core competencies of the EU, such as justice and home affairs and taxation. As such, we would expect the Commission and the EP to play less prominent roles on such issues. We turn first to the distribution of capabilities between the three actors on the 55 controversial issues subject to consultation with the Parliament and qualified majority voting in the Council. The best-fitting power scores for these issues 11 Co-decision is usually coupled with qualified majority voting in the Council. Some Treaty articles require unanimity in the Council in combination with co-decision. However, this is exceptional, and such issues do not feature prominently in our selection; only 12 issues required unanimity and co-decision. The article is not, therefore, able to examine this distinction in the analyses. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 411 Figure 3: Relative Capabilities of EU Actors: Consultation QMV, 55 Issues Minimum error when Commission is 30 and EP 15 relative to the Council's 100 Error Commission are 30 for the Commission and 15 for the EP, relative to the Council at 100. These 55 controversial issues were raised in the discussions on 22 legislative proposals. These proposals were concentrated in the policy areas of agriculture (11 proposals) and fisheries (7 proposals), as defined by the Council sector in which they were discussed.12 Figure 3 shows the average errors in prediction that result from different assumptions on the distribution of capabilities between the three actors. Figure 3 shows that the average error of the compromise model, displayed in the vertical axis, is relatively high when it is assumed that both the Commission and the EP have no power relative to the Council. The front left of Figure 3 shows that the average error of the compromise model is then 26.7 scale points when it is assumed that the Commission and the EP have no power relative to the Council. The average error falls sharply as higher power scores are attributed to the Commission and the EP relative to the Council. The errors reach a minimum (of 23.0) when the Commission is given a power score of 30, and the EP 15, relative to the Council's score of 100. Higher capability scores for the Commission and EP relative to the Council result in increases in the model's error. 12 The other areas covered are employment (1 legislative proposal), internal market (2 proposals) and industry (1 proposal). Note that, although these proposals were discussed in the sectoral councils, they need not have been discussed by ministers; they could also have been resolved at the level of working groups or in Coreper. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 412 ROBERT THOMSON AND MADELEINE HOSLI Figure 4: Relative Capabilities of EU Actors: Consultation Unanimity, 39 Issues Minimum error when Commission 0 and EP 0 relative to Council's 100 Commission For issues subject to consultation with the EP and unanimous voting in the Council of Ministers, the best-fitting power scores are those that attribute all power to the Council, and none to the Commission and EP (Figure 4). There are 39 such issues in the dataset that arose from 18 legislative proposals. The Council of Ministers configurations that dealt with most of these were the Justice and Hone Affairs Council (five proposals), the Economic and Financial Affairs Council (Ecofin; another five proposals), and the General Affairs Council (four proposals), where the Member States are represented by their foreign policy ministers. Figure 4 shows that attributing weight to the Commission and EP relative to the Council results in less accurate predictions of decision outcomes. Table 5: Capability Scores that Minimize the Errors of the Compromise Model (No. of Issues on which the Analyses are Based) Procedure Commission Relative EP Relative to the Council at 100 to the Council at 100 Consultation QMV issues (55) 30 15 Consultation unanimity issues (39) 0 0 Co-decision issues (68) 15 25 Source: Authors' own data. Note: Capabilities of Commission and EP relative to a Council of 100 that minimize the mean average error of the compromise model. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 413 Table 5 reports the scores representing the balance of power between the three institutional actors that are associated with the most accurate forecasts on issues subject to co-decision. There are 68 issues subject to co-decision between the Council and the EP in our dataset. These 68 issues were raised during the discussions on 26 Commission proposals. Eleven of these legislative proposals were discussed in the Internal Market Council, and four were discussed in the Transport Council.13 Under co-decision, the best estimates of balance of power attribute 15 to the Commission, 100 to the Council and 25 to the Parliament. The Parliament's capabilities under co-decision are somewhat higher than under consultation. The best-fitting capability score for the Commission under co-decision is between that of its scores under the unanimity and the qualified majority variants of consultation. Conclusions The findings reported in our article have four implications. The first and most obvious pertains to our understanding of the balance of power between the three institutional actors in the time period considered here. The results indicate clearly that the Council-centric view on the balance of power among the European Commission, Council and Parliament is the more accurate depiction of recent legislative decision-making. As such, the findings support accounts of the EU decision-making in which Member States' interests are seen as important in defining decision outcomes. It would be inaccurate to dismiss this as obvious. Most of the practitioners we interviewed attributed much more power to the supranational institutions, the Commission and Parliament, than our findings suggest is appropriate. Furthermore, many analyses of the EU's formal decision-making procedures suggest that the Parliament is essentially on an equal footing with the Council under co-decision (for a recent review see Selck and Steunenberg 2004). This does not, however, imply that actors other than the Council are marginal. The Commission and Parliament - and also a range of other actors, including domestic and transnational interest groups - may affect the policy positions that governments advocate in negotiations within the Council, thereby exerting indirect influence on decision outcomes. Further, the activities of the supranational actors may feed back into the process of domestic preference formation (Stone Sweet and Sandholtz, 1997). The results indicate that the Commission and Parliament have substantial weight in the decision-making process, even though those weights are far less 13 Other policy areas covered under co-decision are: agriculture (3 proposals, particularly relating to internal market aspects of agriculture), culture (1), development (1), Ecofin (1), energy (1), general affairs (2), health(l) and social affairs (1). 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 414 ROBERT THOMSON AND MADELEINE HOSLI than that of the Council. Even the Council-centric view tested above attributes considerable power to the Commission and Parliament. This becomes apparent when the relations between the actors are depicted numerically. With regard to the qualified majority variant of the consultation procedure, our best estimate is that the Commission has a power score equal to the sum of two or three large Member States; the same is true of the Parliament under the co-decision procedure. Under the QMV variant of consultation, where the Parliament's opinions are non-binding, its opinions nevertheless appear to affect the contents of policy. Here, the weight of the Parliament is similar to that of a large Member State. The second implication is that the formal procedures or 'rules of the game' have a considerable impact on the power of each of the institutional actors. This finding should surely be welcomed given the considerable amount of academic investment devoted to the study of these formal procedures. Substantial differences were found in the balance of power between the three institutional actors, depending on the legislative procedure under consideration. In general, when controversial issues need to be resolved by the consultation procedure and unanimity voting in the Council, the preferences of the Commission and Parliament are of limited relevance in determining the final policy outcomes. In our analyses, this refers to issues in the areas of justice and home affairs, taxation and foreign policy. It come as no surprise that the supranational actors' preferences play less of a role in these policy domains, since European policy in these areas is still in its infancy. In the qualified majority variant of the consultation procedure, the Commission and Parliament have more potential to influence the contents of the decision outcomes, but still nowhere near the level of the Council as a whole. Compared with the qualified majority voting variant of the consultation procedure, the Commission appears to be weaker and the Parliament stronger under the co-decision procedure. The third implication pertains to the variability of the actors' influence across the range of issues considered. This should encourage us to pursue a more detailed specification of the conditions under which the Commission, Council and Parliament have more or less power to shape decision outcomes. The variation in actors' power is evident not only across different decisionmaking procedures, such as consultation and co-decision, which have been studied extensively; it is also present across different issues subject to the same decision-making procedures. Many of our key informants pointed to the importance of differences in power across policy areas. In the same vein, it would be possible to examine the relative power of the institutional actors across fundamentally different types of issues. Burns' (2005) study on the influence of Parliament has started to explore this line of research. Using a case study approach, Burns found that Parliament has more influence on regulatory 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 415 issues than on distributive issues. The dataset used here allows these and similar hypotheses to be tested in a more quantitative fashion. The fourth and final implication is that explanations of actors' power in EU decision-making should consider, and preferably integrate, both formal and informal institutions. This article has not addressed the effects of different kinds of institutions directly; instead it focused on describing the resulting balance of power between the Commission, Council and Parliament. Nevertheless, it is clear that the power relations identified here are not the result of formal institutions alone. For example, legalistic analyses of the co-decision procedure may attribute equal power to the Parliament and the Council. However, the results presented here indicate that the Parliament's power is much lower in practice. Some of the practitioners indicated that the Parliament's position is weakened by the lack of technical policy expertise among MEPs compared with the Council, whose Member State representatives are supported by large national bureaucracies. Furthermore, legalistic analysis of the consultation procedure may attribute less power to the Parliament than it has in practice. Some of the practitioners interviewed in this study indicated that Parliament's power under consultation derives in part from the receptivity of the Commission to its views. The necessity of integrating formal and informal institutions when analysing power in the European Union becomes ever more pertinent in an EU of 25 Member States. One analysis of the Nice Treaty reforms suggests that the changes to the formal decision rules - particularly the new triple majority requirement in the Council - may have shifted the balance of power in favour of the Council (Tsebelis andYataganas, 2002). Similarly, Hosli (1998) claimed that enlargement in combination with maintenance of the QMV decision threshold could shift power towards the Council in the EU's inter-institutional setting. However, enlargement will undoubtedly also bring changes to the informal sources of power. Some of our informants argued that the Council's power derived partly from its ability to present a united front in its dealings with the Parliament in the Conciliation Committee. Given a more heterogeneous set of preferences in the Council of 25 Member States, it is questionable whether the Council will maintain its pre-eminent position in the future. Therefore, possible trends towards increased intergovernmentalism could be partially counterbalanced. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd 416 ROBERT THOMSON AND MADELEINE HOSLI Correspondence: Robert Thomson Department of Political Science Trinity College, Dublin 2-3 College Green Dublin 2, Ireland email: thomsor@tcd.ie References Achen, C.H. (2006) 'Institutional Realism and Bargaining Models'. In Thomson, R., Stokman, F.N., Achen, C.H. and König T. (eds) The European Union Decides (Cambridge: Cambridge University Press). Barry, B. (1970) Sociologists, Economists and Democracy (London: Macmillan). Bueno de Mesquita, B. (2000) Principles of International Politics: People's Power, Preferences and Perceptions (Washington DC: CQ Press). Bueno de Mesquita, B. and Stokman, F.N. (eds) (1994) European Community Decision-making: Models Applications and Comparisons (New Haven: Yale University Press). Burns, C. (2005) 'Who Pays? Who Gains? How do Costs and Benefits Shape the Policy Influence of the European Parliament?'. Journal of Common Market Studies, Vol. 43, No. 3, pp. 485-505. Crombez, C. (2000) 'Codecision: Towards a Bicameral European Union'. European Union Politics, Vol. 1, No. 3, pp. 363-8. Farrell, H. and Héritier, A. (2003) 'Formal and Informal Institutions Under Codecision: Continuous Constitution-Building in Europe'. Governance, Vol. 16, No. 4, pp. 577-600. Garrett, G. (1995) 'From the Luxembourg Compromise to Codecision: Decision-making in the European Union'. Electoral Studies, Vol. 50, pp. 289-308. Garrett, G. andTsebelis, G. (2001) 'Understanding Better the EU Legislative Process'. European Union Politics, Vol. 2, pp. 353-61. Hosli, M.O. (1998) Tnsitutions and Intergovernmentalism in the European Union'. Paper presented at the ECPR Joint Sessions, Warwick, 23-8 March. Hosli, M.O. and Machover, M. (2004) 'The Nice Treaty and Voting Rules in the Council: A Reply to Moberg (2002)'. Journal of Common Market Studies, Vol. 42, No. 3, pp. 497-521. Knight, J. (1992) Institutions and Social Conflict (New York: Cambridge University Press). Scully, R.M. (1997) 'The European ParliamentandCo-decision: ARejoinder to Tsebelis and Garrett'. Journal of Legislative Studies, Vol. 3, pp. 93-103. Selek, TJ. and Steunenberg B. (2004) 'Between Power and Luck: The European Parliament in the EU Legislative Process'. European Union Politics, Vol. 5, No. 1, pp. 25^6. 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd WHO HAS POWER IN THE EU? 417 Stokman, F.N. and Thomson, R. (eds) (2004) 'Special Issue: Winners and Losers of EU Decision-making'. European Union Politics, Vol. 5, No. 1. Stone Sweet, A. and Sandholtz, W. (1997) 'European Integration and Supranational Governance'. Journal of European Public Policy, Vol. 4, No. 3, pp. 297-317. Taagepera, R. and Hosli, M.O. (forthcoming) 'National Representation in International Organizations: The Seat Allocation Model Implicit in the European Union Council and Parliament', Political Studies. Thomson, R. (2006) 'Comparison of Expert Judgements with Each Other and with Information from Council Documentation'. In Thomson, R., Stokman, FN., Achen, C.H. and König, T. (eds) The European Union Decides (Cambridge: Cambridge University Press). Thomson, R., Boerefijn, J. and Stokman, F.N. (2004) Actor Alignments in European Union Decision-making'. European Journal of Political Research, Vol. 43, No. 2, pp. 237-61. Thomson, R., Torenvlied R. and Stokman, FN. (eds) (2003) 'Special Issue: Models of Collective Decision-Making'. Rationality & Society, Vol. 15, No. 1. Thomson, R., Stokman, F.N., Achen, C.H. and König, T. (eds) (2006) The European Union Decides (Cambridge: Cambridge University Press). Tsebelis, G. (1997) 'Maastricht and the Democratic Deficit'. Aussenwirtschaft, Vol. 52, pp. 29-56. Tsebelis, G. and Garrett, G. (2000) 'Legislative Politics in the European Union'. European Union Politics, Vol. 1, No. 1, pp. 9-36. Tsebelis, G. and Yataganas, X. (2002) 'Veto Players and Decision-making in the EU After Nice: Policy Stability and Bureaucratic/Judicial Discretion'. Journal of Common Market Studies, Vol. 40, No. 2, pp. 283-307. Van den Bos, J.M.M. (1991) Dutch EC Policy Making: A Model Guided Approach to Coordination and Negotiation (Amsterdam: Thesis). 8 2006 The Author(s) Journal compilation © 2006 Blackwell Publishing Ltd