The Institutional Foundations of Intergovernmentalism and Supranationalism in the European Union George Tsebelis and Geoffrey Garrett We present a unified model of the politics of the European Union (EU). We focus on the effects of the EU's changing treaty base—from the founding Rome Treaty (ratified in 1958) to the Single European Act (SEA, 1987), the Maastricht Treaty on European Union (1993), and the Amsterdam Treaty (1999)—on the relations among its three supranational institutions—the Commission of the European Communities, the European Court of Justice, and the European Parliament—and between these actors and the intergovernmental Council of Ministers. We conceive of these institutions in terms of the roles they perform in the three core functions of the modern state: to legislate and formulate policy (legislative branch), to administer and implement policy (executive branch), and to interpret policy and adjudicate disputes (judicial branch). The Council and the Parliament are predominantly legislative institutions. The Council directly represents the national governments of the member states. Council support is necessary for the passage of all EU legislation. Since 1987, the Council has increasingly made decisions under "qualified majority voting" (QMV) rather than under unanimity voting.1 The citizens of Europe have directly elected the Parliament since 1979. The Parliament makes decisions by absolute majority at the Earlier versions of this article were presented at Harvard University, the Institution Juan March, Yale Law School, and the ECSA conference in Pittsburgh. Special thanks to Karen Alter, Robert Cooter, Simon Hix, Simon Hug, Torben Iversen, Lisa Martin, two anonymous reviewers, and the editors of IO for many insightful comments. Tsebelis would like to acknowledge financial support provided by the National Science Foundation (grant no. SBR 9511485). 1. While observers note that most Council decisions still seem to be made by "consensus" (that is, unanimity), the formal rule of qualified majority no doubt affects the dynamics of this consensus building. Golub 1999. QMV is a voting rule in which the votes of member governments are weighted and in which roughly five-sevenths of these weighted votes are required for passage. In the current fifteen-member EU, the following weights are assigned: France, Germany, Italy and the United Kingdom, ten votes each; Spain, eight votes; Belgium, Greece, the Netherlands, and Portugal, five votes each; Sweden and Austria, four votes; Denmark, Finland, and Ireland, three votes each; and Luxembourg, two votes. Sixty-two votes (of the eighty-seven possible) constitute a qualified majority. International Organization 55, 2, Spring 2001, pp. 357-390 © 2001 by The IO Foundation and the Massachusetts Institute of Technology 358 International Organization end of all legislative procedures.2 The Parliament had no effective influence over legislation until the ratification of the SEA in 1987. But with the ratification of the Amsterdam Treaty on 1 May 1999, the Parliament became a coequal with the Council in what is effectively a bicameral EU legislature for all policy areas covered by the reformed "codecision" procedure (Amsterdam Treaty, Art. 189b).3 Under this new regime, new legislation requires the support of both a qualified majority in the Council and an absolute majority in the Parliament. The Court is the EU's judicial branch. Its mandate is to interpret the EU's treaty base and secondary legislation passed pursuant to the treaties in the arbitration of conflicts among EU institutions and among these institutions, member states, and citizens. The Court has been remarkably effective in the past forty years, successfully "constitutionalizing" the EU's treaty base, claiming wide powers of judicial review over this would-be constitution, and exercising judicial activism in the interpretation of secondary legislation. Finally, the Commission fulfills two discrete functions in the EU: It is both a legislator with a monopoly on the drafting of bills and the bureaucracy charged with implementing legislation.4 Most commentators consider the Commission to have been the prime mover behind the reinvigoration of European integration in the mid-1980s. Our central contention is that the balance of power among these four institutions has changed considerably since the signing of the Rome Treaty. We analyze these changes in terms of the evolution of the EU's legislative regime and its impact on the discretion of the Commission to implement policy and of the Court to adjudicate policy disputes. The connection between legislation and discretion is straightforward. The more difficult it is for new legislation to be passed (for example, because of higher voting thresholds or more veto players), the more discretion bureaucracies and courts have to move policy outcomes closer to their own preferences.5 The history of European integration can be divided into three epochs. First, the Luxembourg compromise period (1958-87) was characterized by legislative grid- 2. The Parliament currently has 626 members: Germany has 99; France, Italy, and the United Kingdom, 87 each; Spain, 64; the Netherlands, 31; Belgium, Greece, and Portugal, 25 each; Sweden, 22; Austria, 21; Denmark and Finland, 16 each; Ireland, 15; and Luxembourg, 6. 3. This is not, of course, to ignore other areas of increasing Parliament authority, such as blocking the appointment of new Commissions. 4. The Commission does not have gate-keeping power. It must make legislative proposals when requested to do so by the Council and, since Maastricht, by the Parliament as well. 5. This general approach to linking the legislative, executive, and judicial branches of government is not new. The argument can be found in McCubbins, Noll, and Weingast 1989; Tsebelis 1995; Hammond 1996; Hammond and Miller 1987; and Hammond and Knott 1999. As far as we know the point has not been disputed with respect to courts; for comparative corroboration, see Cooter and Ginsburg 1996; and Alivizatos 1995. For preliminary attempts to apply the framework to legal politics in the EU, see Cooter and Drexl 1994; Bednar, Ferejohn, and Garrett 1996; and Alter 1998. With respect to bureaucracies, some of the literature makes the point that in order to avoid bureaucratic autonomy legislatures will try to write more restrictive laws, threaten to use the courts, or apply other restrictions. See Ferejohn and Shipan 1990; Moe 1990; and McCubbins, Noll, and Weingast 1987. Of course, disagreeing legislators may or may not be able to write such laws to restrict bureaucrats. In the context of the EU, administrative oversight is provided by various "comitology" procedures we discuss later. Intergovernmentalism and Supranationalism in the EU 359 lock in the Council. In this period the Council was an ineffective collective institution, with the system of national vetoes protecting the sovereignty of member states. In turn, the unanimity-voting requirement in the Council greatly mitigated the legislative power of the Commission (because its agenda-setting power was negligible). But the Commission was doubly hamstrung under the Luxembourg compromise because the small volume of legislation produced by the Council gave the Commission scant opportunities to exercise its (potentially extensive) bureaucratic discretion to implement policy afforded by unanimity voting. In contrast, legislative gridlock in the Council facilitated Court activism because only treaty revisions could rein in the Court. The freedom of the Court to interpret the Rome Treaty was thus the primary force propelling European integration during the Luxembourg compromise. The second epoch of European integration began when the SEA was ratified. In this period the Council became a more effective legislative institution, at the cost of national sovereignty—individual governments that could no longer veto legislation of which they disapproved. The Court's discretion to interpret secondary legislation was curtailed by the move from unanimity to QMV in the Council (though its discretion in constitutional interpretation was unaffected because treaty revisions require unanimity among member states). The change in Council voting rules also gave the Commission agenda-setting power (though this power was shared with the Parliament under the "cooperation" procedure). Moreover, the proliferation of EU legislation associated with the internal market program (the "1992" agenda) gave the Commission many more opportunities to affect outcomes through policy implementation than was the case under the Luxembourg compromise. Thus the effective removal of national vetoes in the Council rendered the Commission the prime mover behind European integration in the decade following the ratification of the SEA—so long as its legislative proposals respected the preferences of the pivotal members of the Council under QMV and the Parliament (under the cooperation procedure). The origins of Europe's third and current epoch lie in the Maastricht Treaty, and these foundations were cemented at Amsterdam. The Parliament is now a powerful legislator, coequal with the Council under the reformed codecision procedure. In contrast, the Commission's legislative agenda-setting powers are far more limited than they were in the immediate post-SEA era. But empowering the Parliament in a bicameral legislature has increased the probability of gridlock between it and the Council. Consequently, the discretionary space available to the Commission to implement policy and to the Court to adjudicate disputes has increased. In this current epoch, therefore, all four of the EU's major institutions play important roles that are reminiscent of those of legislatures (the Council and the Parliament), bureaucracies (the Commission), and legal systems (the Court) in national polities with bicameral legislatures (such as Germany). Portions of these epochs may seem familiar to some readers. Court-watchers agree that the Luxembourg compromise represented the halcyon days of legal 360 International Organization activism in Europe.6 They have also noted, though mostly in passing, that the Court's influence seemed to wane after the ratification of the SEA.7 We offer a simple explanation for this change and make predictions about the role of the Court after the Amsterdam Treaty. Commission-enthusiasts will likely accept our rendering of the post-SEA period, but we know of no other arguments that explain the politics of the period as clearly as we do. We are also unaware of other analyses of the post-Amsterdam period that are as comprehensive as ours. More generally, we believe that too much of the voluminous literature on the day-to-day operations of the contemporary EU couches generalized descriptions of political processes in vague and imprecise theoretical arguments inspired by Ernst Haas's neofunction-alism.8 In contrast, we draw sharp distinctions both among the influences of the EU's different institutions and across the epochs in the evolution of European integration. Moreover, our analysis has important implications for the politics of institutional choice. We agree with Andrew Moravcsik that the EU's changing treaty base has had a marked impact on the process of European integration.9 We stress, however, that understanding the consequences of these treaties necessitates first analyzing how they will affect the interactions among the EU's major institutions.10 This position, in turn, requires our assuming that the member governments understand what they are doing when they sign treaties. The issue of intentionalityhas been keenly debated. For some analysts, the law of unintended consequences has been a central fact of treaty revisions11. Others suggest that governments pay so much attention to treaty revisions that they are the least likely aspects of European integration to be subject to this purported law.12Although the member governments have sometimes made mistakes in the details of their institutional innovations, we show that often these mistakes have subsequently been eradicated, as they have, for example, in the codecision procedure. Thus we believe that the institutional interactions we analyze in this article generally reflect the collective will of the member governments concerning their desired trajectory for the evolution of the EU. Our analysis is organized as follows. We first outline what we consider the strengths and weaknesses of current research into institutional interactions among the different branches of government in the EU. We then offer a general model for understanding the relationships among legislation, policy implementation, and legal adjudication. We then move on to more specific arguments about legislative politics in the EU and analyze their effects on the discretion of the Commission and Court. 6. Weiler 1991. 7. Mattli and Slaughter 1998, 205. 8. Haas 1958. 9. Moravcsik 1998. 10. Moravcsik does pay attention to some of the details of the EU's institutions but only in terms of the credible commitments to integration that they represent. 11. Sandholtz and Zysman 1989. 12. Pierson 1996. Intergovernmentalism and Supranationalism in the EU 361 We conclude by discussing the implications of our analysis for European integration research in general. Existing Research on Institutional Interactions in the EU Scholarship on institutional interactions in the EU has proliferated in recent years, particularly among those who focus on supranational actors in the Haasian tradition. The most ambitious work in this vein tends to make two fundamental claims: first, that supranational institutions play a powerful role in the everyday operation of the EU; and, second, that this process is self-reinforcing, invariably creating "more Europe." Rather than trying to summarize the voluminous literature, we focus on three influential studies. First, Alec Stone Sweet and Wayne Sandholtz analyze the dynamics of what they term "supranational governance" in the EU.13 Second, Anne-Marie Burley and Walter Mattli provide a "neofunctionalistjunsprudence" of European integration.14 Finally, Gary Marks, Liesbet Hooghe, and Kermit Blank discuss what they call "multilevel governance" in the EU.15 Stone Sweet and Sandholtz explicitly seek to distance themselves from classical neofunctionalism. They "problematize the notion . . . that integration is the process by which the EC gradually but comprehensively replaces the nation-state in all its functions."16 Nonetheless, the similarities between Stone Sweet and Sandholtz and Ernst Haas are apparent when it comes to analyzing the integration dynamic (rather than predicting its endpoint): "We view . . . decision-making as embedded in processes that are provoked and sustained by the expansion of transnational society, the pro-integrative activities of supranational organizations, and the growing density of supranational rules.....These processes gradually, but inevitably, reduce the capacity of the member states to control outcomes."17 What is the process by which this transfer of authority from the nation-state to the EU takes place? Here, Stone Sweet and Sandholtz are quite vague. Their argument appears to have two elements. The first is purely functional: "The expansion of transnational exchange, and the associated push to substitute supranational for national rules, generates pressure on the EC's organizations to act."18 It would be hard to disagree with this proposition. Even Moravcsik in his intergovernmentalist accounts contends that the preferences of multinational firms and exporters for larger markets, pan-European regulation, and exchange-rate stability have been prime movers behind European integration since World War II.19 The second facet of Stone Sweet and Sandholtz's argument supplements this demand-side logic with 13. Stone Sweet and Sandholtz 1997. 14. Burley and Mattli 1993. 15. Marks, Hooghe, and Blank 1996. 16. Stone Sweet and Sandholtz 1997, 299. 17. Ibid., 300. 18. Ibid., 299. 19. Moravcsik 1998. 362 International Organization a claim about the self-reinforcing dynamic of supranational activity: "As European rules emerge and are clarified and as European organizations become arenas for politics, what is specifically supranational shapes the context for subsequent interactions. . . . This creates the 'loop' of institutionalization. Developments in EC rules delineate the contours of future policy debates as well as the normative and organizational terms in which they will be decided."20 Unfortunately, Stone Sweet and Sandholtz do not probe deeper into this "loop of institutionalization." This is where we believe our analysis can help make more precise such plausible but vague intuitions (that seem quite similar to Haasian "spillovers"). Of course, this precision may come at a cost to Stone Sweet and Sandholtz's agenda. We highlight the importance of the EU's treaty base to the way the EU operates on an everyday basis. And it seems that the treaty revisions the member governments have undertaken have been motivated by a relatively clear and consistent set of principles about what the EU polity should look like. We now turn to Burley and Mattli. Unlike Stone Sweet and Sandholtz, Burley and Mattli are happy to acknowledge their neofunctional heritage.21 Indeed, they claim that "the legal integration of the [European] community corresponds remarkably closely to the original neofunctionalist model developed by Ernst Haas."22 According to Burley and Mattli, the Court has been able to promote European integration throughout its existence by insisting that it is only implementing the law, as opposed to playing politics. This "mask" of formal legalism allows the Court to "shield" its judgments from political retaliation, even when governments disapprove of these rulings: "The margin of insulation necessary to promote integration depends on a minimal degree of fidelity to both substantive law and the methodological constraints imposed by legal reasoning. In a word, the staunch insistence on legal realities as distinct from political realities may in fact be a potent political tool."23 They also consider the Commission to be an important partner for the Court when it comes to furthering European legal integration. Their reasoning is quite similar to the mask and shield metaphor for the Court. Burley and Mattli believe that the Commission is a powerful actor precisely because it has a reputation for being an impartial provider of information and expertise that is above the political fray: "From the Court's . . . perspective . . . the chief advantage of following the Commission is the 'advantage of objectivity.' ... In neofunctionalist terms, the Court's reliance on what Pescatore characterizes as 'well-founded information and balanced legal evaluations,' as 'source material for the Court's decisions' allows it to cast itself as nonpolitical by contrasting the neutrality and objectivity of its 20. Stone Sweet and Sandholtz 1997, 311. 21. Burley and Mattli tempered their position in Mattli and Slaughter 1998. Nonetheless, their original article has been very influential because of its clear statement of a neofunctional jurisprudence of the EU. 22. Burley and Mattli 1993, 43. 23. Ibid., 44. Intergovernmentalism and Supranationalism in the EU 363 decision-making process with the partisan political agendas of the parties before it."24 There is thus a clear analytic difference between Burley and Mattli's position and Stone Sweet and Sandholtz's. Stone Sweet and Sandholtz's argument has no microfoundations; they are content with macro-level assertions about the loop of institutionalization as the mechanism by which supranational actors promote European integration. Burley and Mattli, in contrast, do have a micro logic—the Court (and the Commission) can deflect criticism from member governments that it is acting "politically" by asserting that it is only doing its job. But there is a tension at the core of Burley and Mattli's argument. The Court and Commission are able to further the integration agenda because they can always credibly claim that they are only doing their jobs, impartially and apolitically. In essence, their argument requires that member governments cannot discriminate between actions by the Court and Commission that are consistent with their mandates and those that are not. We consider this to be a heroic assumption. Burley and Mattli (and most other observers) confidently claim that it is patently clear that both the Commission and the Court are pro-integration, supranational entrepreneurs that stretch their authority as far as they can to further their own agendas. What is readily apparent to these scholars, however, supposedly eludes the member governments—though, presumably, they have a greater interest in monitoring the Commission and the Court than do academics! A more prudent assumption would be that the EU's member governments understand quite well (at least most of the time) the preferences of the Commission and Court as well as the consequences of these preferences as these institutions willingly carry out their mandates (as written in the EU's treaties or its secondary legislation). The governments have strong efficiency incentives to delegate authority for implementing and adjudicating policy, but this delegation may still give the Commission and the Court real influence over the course of European integration. This is the standard principal-agent approach we adopt in subsequent sections. Finally, Marks, Hooghe, and Blank pay more attention to the details of the interrelationships among the EU's supranational institutions.25 In particular, they explicitly include the Council in their analysis as a potentially cooperative actor in the EU game, in contrast with the adversarial "us versus them" attitude to the Council implied by Stone Sweet and Sandholtz and Burley and Mattli. Rather than proposing a parsimonious theory, however, Marks, Hooghe, and Blank's primary concern is to describe how the EU operates day to day. They recognize that the Commission's effective agenda-setting power is shaped by its formal interactions with other EU institutions: "The European Council, the Council, and the European Parliament have each succeeded in circumscribing the Commission's formal monopoly of initiative more narrowly.. . . Agenda-setting is 24. Ibid., 71. 25. Marks, Hooghe, and Blank 1996. They also highlight the effects of European integration on subnational politics. 364 International Organization now a shared and contested competence among the four European institutions, rather than monopolized by one actor."26 But rather than analyzing in detail the changing location of agenda setting under the EU's different legislative procedures, they simply note that "the Council is locked in a complex relationship of cooperation and contestation with the two other institutions [Commission and Parliament]." How, then, can one understand the Commission's legislative power in these complex relationships? At this key point, Marks, Hooghe, and Blank fall back on an approach inspired by neofunctionalism: "The Commission has considerable leverage, but it is conditional, not absolute. It depends on its capacity to nurture and use diverse contacts, its ability to anticipate and mediate demands, its decisional efficiency, and the unique expertise it derives from its role as think-tank of the European Union."27 Indeed, they conclude "the Commission's power is predominantly soft in that it is exercised by subtle influence rather than sanction."28 Our approach is more straightforward. We assess the (changing) legislative role of the Commission. Where its formal powers are strong, subtle influence is unnecessary. Where it is formally weak, subtle influence may be all the Commission has. Surely the first step should be to delineate precisely the Commission's formal position? We react in a similar way to Marks, Hooghe, and Blank's analyses of the discretion afforded to the Commission in policy implementation. They point to the limitations of the "comitology" rules that are designed to allow the Council and Parliament to oversee how the Commission implements legislation: "At first sight, comitology seems to give state executives control over the Commission's actions in genuine principal-agent fashion. But the relationship between state actors and European institutions is more complex. Comitology is weakest in precisely those areas where the Commission has extensive executive powers, e.g., in competition policy, state aids, agriculture, commercial policy, and the internal market. Here, the Commission has significant space for autonomous action."29 Our analysis is different. We explain how the rules that govern the passage of legislation not only affect the laws the Commission implements but also, and more importantly, how they influence the discretion in implementation available to the Commission. Rather than viewing implementation and legislation as separate facets of the EU polity, we integrate them. Finally, Marks, Hooghe, and Blank address the role of the Court in the following way: "Court rulings have been pivotal in shaping European integration. However, the ECJ depends on other actors to force issues on the European political agenda and condone its interpretations. Legislators . . . may always reverse the course set by the Court by changing the law or by altering the Treaties. In other words, the ECJ is no different from the Council, Commission, or European Parliament in that it is locked 26. Marks, Hooghe, and Blank 1996, 358. 27. Ibid., 359. 28. Ibid., 366. 29. Ibid., 367. Intergovernmentalism and Supranationalism in the EU 365 in mutual dependence with other actors." We do not disagree with this statement. The notion that the Court's behavior is affected by the reactions it anticipates from the EU's legislative branch is at the core of our understanding of legal politics in the EU. Indeed, it is the last piece in our integrated model of the EU's political system. More broadly, we view Marks, Hooghe, and Blank's description of Europe's multilevel polity as a stimulus for further analysis. Can we delineate precisely the interactions among the Commission, Council, Court, and Parliament that characterize the contemporary EU? We take up this challenge in the remainder of the article. Legislation, Implementation, and Adjudication In this section we provide a framework for analyzing institutional interactions in the EU. Our basic model focuses on the impact of legislative rules on the discretion of bureaucrats to implement policy or judges to adjudicate statutory disputes in ways that further their own preferences. This simple model allows us to characterize political dynamics under the Luxembourg compromise. We then add the empirically relevant details of the post-SEA period in subsequent sections. The Model The EU's political system has three institutional components.31 The "legislature" writes laws but delegates authority to implement them. The "bureaucracy" writes detailed regulations designed to implement legislation and monitors compliance with them. The "judiciary" adjudicates disputes over legislation and compliance. We view the EU's political system as one in which, through the Rome Treaty and its subsequent revisions, the member governments have delegated implementation and adjudication powers to the Commission and the Court, respectively. They have done so because such delegation is more efficient than trying to write detailed legislative rules to govern the myriad contingencies that arise in the day-to-day operation of the EU32. But as in all principal-agent relationships, this delegation of power creates a problem for the legislative branch: its agents may not carry out the intent of the legislation. In fact, these agents have a significant level of autonomy in their decisions (in Giandomenico Majone's words, they have a "fiduciary relationship" with their principals).33 30. Ibid., 370. 31. A fourth actor, the presidency, would have to be introduced for cases such as the United States. Although some pundits argued for the creation of an EU president during the Delors Commission, it is clear that the parliamentary model was entrenched at Maastricht and Amsterdam. 32. Garrett and Weingast 1993. 33. Majone 1999. We generally assume here that national governments are bound to follow Court decisions unless they are overturned by new legislation. For a discussion of the conditions under which this is likely, see Garrett 1992 and 1995a; and Garrett, Kelemen, and Schulz 1998. 366 International Organization How will this complicated relationship, in which the principal delegates significant autonomy to its agents, play out? How can the principal be confident that the agents will not overstep their mandates, and how can the agents know that the principal will not interfere with decisions delegated to them? In other words, how do the relevant actors resolve the joint problems of the scope of delegation and the credibility of member states' commitments to it? Our answer to all these questions is the same: these problem are solved by structuring the institutions governing the various relationships. However, the way the EU has chosen to organize its institutions has changed considerably over its history, significantly affecting the power relationships among them. For clarity of exposition in this section, we employ two simplifying—but empirically inaccurate—assumptions: that the Council is the EU's sole legislator and that the Commission's only role is to implement legislation. We also assume that the Council can monitor at low cost the behavior of its agents—the Commission and the Court—and that it can also cheaply pass new legislation if it is unhappy with the behavior of these agents. The Commission and the Court are both assumed to prefer that the legislative branch not overturn their actions because this would harm their reputations for acting (more or less) impartially. Under these conditions, the rules by which legislation is generated will have a marked impact on how legislation is implemented by the Commission and the Court. In our model the bureaucracy or the judiciary moves first: it selects how to translate existing legislation into political outcomes. If the legislature disagrees with this choice, it overrules the bureaucracy or the judiciary. Consider the following setup (Figure 1). There are seven governments in the Council whose preferences can be arrayed on a straight line from less to more European integration. The governments in the Council select Commissioners and Justices, but they cannot remove them from office when they behave in ways disapproved of by the Council.34 The Council can, however, react to Commission and Court behavior by passing new legislation. The preferences of the Commission and the Court may be different from those of any member of the Council. All actors have Euclidean preferences and know each other's ideal points and the structure of the legislation-discretion game. Consider a scenario in which the Council's legislative decision rule is unanimity (such as when the legislation is a treaty, was being considered under the Luxembourg compromise, or is a present-day policy domain where unanimity still applies). The Council can act unanimously to pass new legislation to alter any status quo—which we will interpret here as the relevant piece of legislative implementation by the Commission or judicial interpretation by the Court—so as to bring it within the Pareto set of the Council (that is, the interval 1-7). Conversely, the unanimity requirement renders any policy outcome within the Pareto set (including 34. In reality, of course, the appointments of commissioners and justices are for renewable fixed periods (four and six years, respectively). But the threat of not being reappointed seems not to act as a powerful constraint on their behavior. Intergovernmentalism and Supranationalism in the EU 367 Discretion for Court and Commission to implement policy without legislative override under unanimity Discretion under QMV No discretion under simple majority 12 3 4 5 6 7 Council members: 1, 2,..., 7 FIGURE 1. Legislation, adjudication, and administration policy as implemented by the Commission or adjudicated by the Court) invulnerable to being overturned by new legislation. Thus under a unanimity legislative rule, the Commission and Court have considerable policy discretion. If, as is the common assumption in the literature, the Commission and Court are both more integrationist than any government in the Council, we would expect these actors to implement outcomes just to the left of Council member 7. Things change considerably, however, if the threshold required for the passage of new legislation is reduced. The EU's QMV rules can be best approximated as a five-sevenths majority rule. In this case, members 3 and 5 are the legislative pivots in the Council. Legislation disapproved of by either cannot pass. Thus under QMV any status quo (produced by bureaucratic implementation or judicial interpretation) that is outside the 3-5 interval would be overturned by the Council, but any outcome within this interval would be invulnerable. Under QMV, therefore, pro-integration entrepreneurs in the Commission and Court could still use their discretion to generate outcomes at the right-hand end of this 3-5 interval, but any such outcome would be less integrationist than under unanimity. Indeed, if the Council were to alter its decision rule from QMV to simple majority (four-sevenths in this case), the Commission and Court would have no discretion—so long as the policy space were truly one-dimensional. This is the median voter theorem, and it helps explain the relative weakness of courts and bureaucracies in countries with unicameral majoritarian (Westminster) systems. In the EU instance, the Council would choose policy at the ideal point of member 4. Any effort by the Commission and Court to move this outcome would be defeated by passing new legislation to reaffirm the preferences of the median voter.35 35. In the case of the Court, of course, this legislation-discretion game does not apply to "constitutional" cases, where irrespective of the legislative rule, the Court is only formally constrained by the unanimity requirement among member governments for amending the EU's treaty base. 368 International Organization Two potential criticisms of our approach merit further consideration. First, the Council will ex ante restrict the discretion of bureaucrats.36 Second, Council members would not alter policies to bring them within their Pareto set because in so doing they would risk having them moved closer to the ideal point of the Commission in the implementation phase.37 Note that both objections concern the restrictions on bureaucratic discretion, not on statutory interpretation. With respect to the first criticism, the Council may well want to restrict the discretion of the Commission in implementation (we later discuss the EU's comi-tology procedures). However, it is also likely that the Council would have a harder time agreeing on such restrictions unanimously than by QMV, or by simple majority. The decision-making rule thus affects the discretion of the Commission in the manner we describe whether or not it is applied directly (through legislation about policy) or indirectly (through legislation governing how policy is to be implemented). Whether the Council can overcome internal disagreements and impose restrictions on the implementation behavior of the Commission is an empirical matter, not a theoretical one. The second criticism concerns an extreme strategy by Council members that may not always be available to them: not passing legislation because bureaucrats may subsequently exercise discretion in implementing it. This strategy is not always available because legislation on the subject may already exist—in which case bureaucrats will be able to implement it (and therefore exercise whatever discretion they have). The strategy is extreme because legislators create laws that they anticipate will be in effect for a relatively long time and hence will likely be implemented by different bureaucrats with different ideal points. It seems an overreaction not to produce legislation just because some bureaucrats might abuse the powers delegated to them. The EU has selected creative methods for monitoring the implementation of legislation, known under the general name of "comitology." The Commission's decisions about how to implement legislation are reviewed by different committees composed of representatives of the member governments. There are three different kinds of committees: advisory, management, and regulatory. Advisory committees decide by simple majority whether the Commission's behavior has been consistent with the intent of the legislation; they cannot veto a Commission decision. Management committees decide by qualified majority; if they disapprove of the Commission's behavior, they send the issue to the Council. Regulatory committees send the issue to the Council unless they agree with the Commission's implementation by qualified majority.38 In the remainder of this article we deal implicitly with two scenarios. The first involves the bulk of EU legislation (80 percent for the 1987-95 period), in which comitology is not used. The second involves roughly 40 percent of the remaining 36. See McCubbins, Noll, and Weingast 1987; and Moe 1990. 37. This objection was brought to our attention by an anonymous reviewer. 38. See Hix 1999, 41-45; Pollack 1997; and Steunenberg, Koboldt, and Schmidtchen 1996. Intergovernmentalism and Supranationalism in the EU 369 legislation, in which comitology has been invoked but where Council oversight of policy implementation is purely advisory.39 In both scenarios, passing new secondary legislation is the only effective way to rein in Commission implementation activism.40 The Effects of Legislative Unanimity Having clarified these points, we now turn to the implications of our model for the areas of EU activity governed by unanimity voting among member states. First, given that the EU's treaty base can only be modified by the unanimous agreement of the member governments, we would expect that the Court would have considerable latitude in interpreting "constitutional issues" (aided and abetted by the Commission's bringing cases to the Court). This would be all the more true were we to introduce more realism into the model by acknowledging that there are significant costs (time and money, at a minimum) to convening intergovernmental conferences to revise the treaties. The same argument also suggests that the Court had considerable latitude in statutory interpretation under the Luxembourg compromise as well. However, gridlock in the Council meant that relatively little secondary legislation was passed in this period. Thus the Court's discretionary activism was likely to be most apparent in constitutional cases.41 Second, during the Luxembourg compromise the Commission had considerable discretion to implement secondary legislation. As mentioned earlier, however, the Council did not produce much legislation in this period, and as a result the preconditions for Commission activism were not present. Third, for the areas of EU policy still subject to unanimity voting in the Council today—such as citizenship and immigration, foreign and security policy, international agreements, and policing and taxation—considerable discretion would seem to be available to the Commission and Court.42 Again, however, it should be noted that this theoretical power in the case of the 39. Dogan 1997,45. 40. In the other two types of oversight—which empirically only account for just over 10 percent of legislation—the Commission can exploit any disagreements between the different members of the committees or between committees and the Council, or among members of the Council. To overrule Commission implementation, the comitology committees must first send the matter to the Council, and then a qualified majority in the Council must vote in favor of overturning the Commission. 41. Of course, the member governments have an alternative to revising the treaty base when they disapprove of constitutional Court decisions—they can simply not abide by these decisions. Garrett, Kelemen, and Schulz 1998. In fact, there is evidence that other institutions consider seriously the power of member countries to ignore decisions and to tailor their decisions to avoid this risk. For example, Kilroy argues that the Court's decision making is very sensitive to the expressed positions of coalitions of member states and to their voting weights in the Council. Kilroy 1999. She provides aggregate evidence that Court decisions tend not to go against the expressed opinions of certain coalitions of member states and uses cases to show where the Court altered its jurisprudence in response to public statements by government leaders from the member states). 42. For a complete description of the use of unanimity and QMV-based legislative rules in the various areas of policy subject to EU legislation, see Hix 1999, 366-75. 370 International Organization Commission is contingent on the existence of relevant secondary legislation— and in these sensitive political areas the Council has been reluctant to act. The Court, in contrast, could certainly try to "constitutionalize" some of these issues since they are subject to various treaty articles. The Court took this course in the 1960s when it sought to harmonize indirect taxation, and more recently when it entered immigration debates by casting immigration as a human rights issue. But the Court, fearing radical rollbacks of its power, in general seems reluctant to act in areas that the member states feel are at the core of their sovereignty—a legitimacy constraint in the view of Mattli and Slaughter.43 Hence one should not overstate the scope for judicial discretion in legislative areas that remain subject to unanimity voting in the Council. Fourth, unanimity voting has two effects on the Council. On the one hand, it respects the sovereignty of individual member states. On the other hand, it cripples the Council as a collective actor. Given that European integration entails the pooling of national sovereignty, it is not surprising that member governments have been unwilling to remove the unanimity constraint in areas of great salience in national politics. Over time, however, the clear trend has been away from unanimity rules as member governments have come to embrace effective collective decision making in increasing numbers of policy areas. There is, of course, a flip side to this analysis of the effects of unanimous legislative rules. The move to QMV since the mid-1980s has "unblocked" collective decision making in the Council. But our analysis also implies that the increased use of QMV should have reduced the discretionary power of the Commission and the Court. Joseph Weiler's groundbreaking analysis of the history of European legal integration posits a clear inverse relationship between the Court's activism and legislative activism.44 He suggests that the Court carried the burden of furthering integration when the governments were shackled by the Luxembourg compromise but that the Council picked up the ball after the SEA. There also seems to be some sentiment among EU-watchers that the power of the Commission declined in the 1990s (even before the ouster of the Santer Commission). Analyzing the interaction between legislation and discretion since the mid-1980s, however, requires considerable attention because the move to QMV has not been the only important legislative innovation in the EU. In particular, the legislative role of the Parliament increased from the SEA to the Amsterdam Treaty. Moreover, in this section we have ignored the legislative role of the Commission. Having an understanding of both is essential to analyzing the interactions among the legislative, bureaucratic, and judicial branches of the contemporary EU. We now turn to this complex reality. 43. Mattli and Slaughter 1998. 44. Weiler 1991. Intergovernmentalism and Supranationalism in the EU 371 Legislative Politics in the post-SEA Period We have thus far discussed the dynamics of legislative discretion in the EU by making a gross distinction between unanimity voting and QMV rules. The EU has employed, however, a variety of complicated legislative procedures involving QMV. Differences among these procedures have important implications for legislative dynamics, and for the policy discretion of the Commission and the Court. We now analyze the EU's changing legislative environment from the Luxembourg compromise to the Amsterdam Treaty.45 We then move on in the next section to the impact of the changing legislative environment on the policy discretion available to the Commission to implement legislation and the Court to adjudicate disputes over it. The Procedures The SEA radically changed the EU's legislative procedures. Much of the day-to-day legislative agenda was invigorated by the Council's applying QMV to the issues originally intended in the Rome Treaty and to additional policy areas adumbrated in the treaty. The ambit of QMV was subsequently expanded further at Maastricht and again at Amsterdam. The broader institutional environment in which QMV is embedded, however, varies significantly by policy areas and the specific legislative procedures applied to them. Under the "consultation" procedure (which was written into the Rome Treaty), Commission proposals become law if they are accepted by a qualified majority of Council members. A unanimous Council can amend Commission proposals (this also applies to the other QMV-based procedures). The "cooperation" procedure (now Art. 189c, Maastricht Treaty) was introduced in the SEA to govern the "1992" agenda, but the internal market was subsequently moved under "codecision" at Maastricht. Today, cooperation applies to areas such as social policy, implementation of regional funds, research and technological development, and some environmental issues. The most important institutional feature of cooperation was to give the Parliament its first substantive legislative role. The Parliament may amend Commission proposals. If the Commission accepts these amendments, they are presented to the Council, which can accept them under QMV or amend them unanimously. The Parliament can also reject proposals that can only be overridden by an agreement between the Commission and a unanimous Council. The codecision procedure was added to the EU's legislative arsenal at Maastricht (Article 189b, Maastricht Treaty), covering not only the internal market but also new policy domains such as education, culture, public health, and consumer protection. This initial version of codecision differed from cooperation in two ways. 45. For more detailed description and analysis of legislative politics in the EU, see Tsebelis and Garrett 2000. 372 International Organization First, the Council could not reject Parliament amendments accepted by the Commission; instead it had to request a conciliation committee (composed of all members of the Council and an equal number of Parliament members) to discuss such amendments. Second, if the conciliation committee could not agree to a joint text, the Council could reaffirm its prior common position, possibly with amendments proposed by the Parliament. This Council proposal became law unless an absolute majority of Parliament members vetoed it. The codecision procedure was modified in the Amsterdam Treaty (Art. 189b as amended). Additional policy areas were brought under its aegis,46 but the procedure itself was also changed. Under the reformed codecision procedure, the conciliation committee is the last stage of the legislative game. The proposed legislation lapses if the representatives of the Council and Parliament cannot agree on a joint text (Art. 189b(6), Amsterdam Treaty); that is, the member governments decided to remove the last two stages of the original codecision procedure—the Council's final proposal to the Parliament, and Parliament's decision whether to reject it. The Parliament Two statements about the legislative powers of the Parliament under the EU's QMV rules are unlikely to be controversial. First, prior to the passage of the SEA and the creation of the cooperation procedure, the Parliament had scant legislative influence, even after its direct election in 1979 and the Court's "isoglucose" decision that the Parliament had to be consulted before new laws could be passed. When the Council decides by unanimity or when the consultation procedure applies, the Parliament's influence is limited to the threat of delaying legislation, not unlike the House of Lords in the United Kingdom.47 Second, at the other end of the spectrum, the Parliament is a true coequal legislator with the Council for policies governed by the reformed codecision procedure. In this case, new legislation can only be passed if a qualified majority in the Council and an absolute majority of Parliament members present support it. We cannot be more precise about where on the Council-Parliament contract curve legislation will be passed because there are no institutional constraints on bargaining in the conciliation committee. Nonetheless, using any of the standard models (Nash, Rubinstein, or Baron and Ferejohn), one might expect outcomes to "split the difference."48 This is a long way from the pre-cooperation environment. Consequently, the empowering of the Parliament as a legislator is a key institutional development in the modern history of European integration. 46. These include gender equality (Art. 119), administration of the European Social Fund (Art. 125), health and safety (Arts. 129 & 129a), some aspects of environmental policy (various sections of Art. 130), and fraud (Art. 209a). The treaty brought many other issues under codecision, but subject to unanimity in the Council. 47. Tsebelis and Money 1997. 48. See Baron and Ferejohn 1987; Nash 1950; and Rubinstein 1982. Intergovernmentalism and Supranationalism in the EU 373 Even more complicated and contentious are the intermediate cases of parliamentary influence: cooperation and the Maastricht version of codecision. Under the cooperation procedure, the Parliament is a "conditional agenda setter."49 The initial codecision procedure took away this power from the Parliament but replaced it with an unconditional veto—new legislation could not be passed over the Parliament's opposition. The conventional wisdom is that this was a good trade for the Parliament.50 We have argued, however, that the Parliament is often more influential over integration policies when equipped with conditional agenda setting than with veto powers.51 Let us briefly present our argument. Conditional agenda-setting power exists only under certain conditions: if there is a proposal that makes a qualified majority of the Council better off than any unanimous decision, if there is an absolute majority in the Parliament to support it, and if the Commission adopts it. But when these conditions are met, conditional agenda setting gives the Parliament more influence over legislation because it permits the Parliament to select among different alternatives the one it likes the most; veto power simply enables the Parliament to reject the options it does not like. Consequently, the impact of exchanging conditional agenda setting (cooperation) for unconditional veto (codecision, Maastricht-style) varies with the relationship between the Parliament's preferences and those of members of the Commission and the Council. When the Parliament and the Commission are more integrationist than any member of the Council, and when the members of the Council do not have identical positions, the Parliament can exercise more influence over EU integration policies under cooperation than under the initial version of codecision.52 But the intricacies of legislative dynamics under cooperation and Maastricht codecision should not detract from our fundamental point here. The trajectory of European integration is clear when we compare the oldest legislative procedure under QMV—consultation, in which the Parliament's only role was advisory—to the codecision procedure written into the Amsterdam Treaty—in which the Parliament is a coequal legislator with the Council. The member states have collectively 49. Tsebelis 1994 and 1997. 50. See Corbett, Jacobs, and Shackleton 1995; Crombez 1996; and Scully 1997. 51. See Garrett and Tsebelis 1996; Tsebelis and Garrett 2000; and Garrett and Tsebelis forthcoming. 52. Tsebelis and his colleagues examine the contending positions on the legislative politics of cooperation and Maastricht codecision with reference to legislation in the 1989-94 period to which the Parliament added amendments (over 5,000 cases). Tsebelis et al. forthcoming. They make three points: First, the overall acceptance rate of Parliament amendments is higher under codecision than under cooperation. This finding is consistent with the Parliament data and the expectations of EU observers. Second, controlling for one of the prerequisites for conditional agenda setting under cooperation (acceptance by the Commission), the Parliament's influence over subsequent Council decisions (the rate at which it accepted Parliament amendments) is higher than the overall acceptance rate under codecision. This finding is consistent with arguments made in Garrett and Tsebelis 1996; and Tsebelis and Garrett 2000. Third, if one controls for acceptance of Parliament amendments by the Commission, there is no difference in acceptance rates between cooperation and Maastricht codecision. This finding is novel and leaves the policy expectations of Garrett and Tsebelis untested, because one would have to subtract from the data that present no difference between procedures the cases of institutional decisions where the Council is expected to have unanimous opinions. 374 International Organization chosen in successive revisions of the Rome Treaty to upgrade the legislative power of the Parliament to the point where today the EU looks very much like a traditional national bicameral legislature. The Commission In important respects the recent history of the Commission as a legislative actor in the EU is a mirror image of the history of the Parliament. Under the consultation procedure, the Commission has considerable influence over legislative outcomes because its right to make proposals allows it to set the Council's agenda. In other words, from all the potential outcomes that would generate QMV support in the Council, the Commission can choose the proposal most preferred by Commission members (or more precisely, that which also makes the pivotal member of the Council indifferent to what could be achieved unanimously). Under cooperation, the Commission has to share agenda-setting power with the Parliament. Although the Commission can still initiate legislation, the Parliament can amend proposals, and the Commission must review these amendments before sending them to the Council. The Council reacts to legislative proposals under the same rules: QMV to accept, and unanimity to amend. Thus, a coalition of the Commission and the Parliament (where it exists) can select among the different proposals that would generate QMV support in the Council. In marked contrast, the Commission's role under both versions of codecision is effectively limited to that played by traditional national bureaucracies. The Commission writes the initial drafts of bills, but in the final stage a coalition between a qualified majority in the Council and an absolute majority in the Parliament and Council can overrule the Commission and amend a bill. Thus the evolution of the EU's legislative regime from consultation to codecision (under the Amsterdam Treaty) has substantially reduced the legislative powers of the Commission.53 53. Not all observers agree with this analysis. The conventional view of cooperation is that even though the Commission formally is forced to share its agenda-setting powers with the Parliament, in fact the Commission's influence over legislation is identical to that under consultation. See Crombez 1996; and Moser 1996. Tsebelis and his colleagues have tested this proposition empirically and shown that the power of the Commission was, all else equal, higher under consultation than under cooperation, and higher in cooperation than under the first version of codecision. Tsebelis et al. forthcoming. Under cooperation, the Commission's opinion was respected (that is, included in the final legislation) 85 percent of the time (88 percent when it rejected a Parliament amendment, 83 percent when it supported it), whereas under the Maastricht codecision procedure this figure dropped to 70 percent (67 percent when it rejected an Parliament amendment and 73 percent when it supported it). It should be noted, however, that these percentages vary greatly over time. With respect to cooperation, the Commission became more influential over time, reaching a zenith during the period when most internal market measures were adopted. In contrast, the Commission's influence over the Maastricht codecision procedure deteriorated over time. Intergovernmentalism and Supranationalism in the EU 375 The Council The study of the legislative role of the Council points out significant differences between written rules and their application in the EU. For example, as we have already said, although QMV was written into the Rome Treaty with the expressed intention of its coming into force in 1966, this transition from unanimity to QMV was blocked, de facto, for twenty years by the Luxembourg compromise. As a result, it was not until the SEA that observers began to focus on the strategic consequences of QMV. In fact, the SEA introduced two distinct institutions that affected the legislative role of the Council: the actual application of QMV and the cooperation procedure. Empirical studies of decision making in the Council even after the SEA indicate frequent recourse to unanimity voting in cases where the treaties called for QMV.54 We address this point explicitly because unanimity voting in the Council would significantly alter the results of our analysis. For example, under unanimity the Parliament and the Commission would lose their agenda-setting power in the cooperation and consultation procedures (since the Council could accept or modify their proposals under the same rule). However, actual accounts of legislative dynamics in the EU point to significant differences between the Luxembourg compromise period and current practices. Most importantly, although official Council votes are often finally reported as unanimous, informal votes to assess the situation relied invariably on QMV.55 Consider the following statement from the European Christian Democratic Parties group in the Parliament: Informal votes are often held which reveal whether a qualified majority exists. If it does, the Council Presidency may simply say that a decision will be deemed to be taken unless anyone objects. Equally, within the Council, significant efforts are often made to secure the widest measure of agreement, to accommodate states, which may not be able to stop a proposal being adopted, but have strong concerns about particular points. Everyone has a vested interest in a divisive vote being avoided. As a result, few formal votes occur. However, without the certainty that a vote can be taken at the end of the day, there would be very little impulsion towards agreement.56 The Maastrict Treaty changed the players controlling the agenda and making the final decision. Under the initial form of codecision, agenda control was given to the Council. In the last stage of the game, if a joint Council-Parliament committee could not reach an agreement, the Council could make a proposal to the Parliament. An absolute majority was required for the Parliament to reject this proposal. This ability of the Council to make the final offer under codecision was eliminated in the Amsterdam Treaty. At present, if there is no agreement in the joint 54. See Hix 1999; and Golub 1999. 55. See Peters 1992, 84-85; Dinan 1994, 254; and Miller 1995. 56. Cited in Tsebelis 1996, 840. 376 International Organization committee, the legislation lapses. This change in the legislative role of the Council, at least in practice, seems to have predated the formal treaty revision. The Maastricht Treaty specified that the Council could make a proposal to the Parliament and force it to take it or leave it, but from the outset the Parliament strenuously resisted the Council's efforts to behave that way. Indeed, the Parliament introduced its own internal rule (Rule 78) stipulating that, in the event the conciliation committee failed to reach an agreement, the Parliament would first request that the Council's proposal be withdrawn. If the Council and the Commission did not agree to this request, the Parliament would place the proposal for a vote without modifications (with the intention of defeating it). In 1994 the Parliament was able to carry out the threat implicit in Rule 78 by not supporting legislation originating from the Council. In 1998 the Council decided against reintroducing its common position on a directive on investment firms and credit institutions.57 Summary In this section we have made three simple points about the trajectory of the EU's legislative regime since the mid-1980s. First, the Parliament's powers have been markedly upgraded since its direct election in 1979—first in the SEA and finally in the Amsterdam Treaty. The latter placed it on an equal footing with the Council with respect to legislation governed by the reformed codecision procedure. Second, the Commission's legislative role has been reduced over the same period. As a result of these two developments, the EU has become a more democratic institution (because members of the Parliament are directly elected by citizens in the member states, whereas Commissioners are not). Third, the EU has also become more democratic in a majoritarian sense at the intergovernmental level because the ever increasing use of QMV has eroded the ability of individual countries to hold up new legislation. Bureaucratic Implementation and Statutory Interpretation Under QMV-based Legislative Procedures Let us now change gears to analyze the ability of the Commission and the Court to exercise discretion in how they implement and interpret legislation passed under the EU's various QMV-based legislative procedures. We have demonstrated that discretion is positively associated with the range of outcomes that are invulnerable to relegislation. We apply this basic insight to the complexities of the contemporary EU.58 57. Hix 2000. 58. We also believe that the Court's power likely received a new boost in the Maastricht Treaty, and that this may be reflected in more active constitutional jurisprudence into the near future. In addition to the codecision procedure, the Maastricht Treaty also introduced the principle of subsidiarity as the one that should govern all activity in the EU. But the treaty did nothing more than state the criterion that Intergovernmentalism and Supranationalism in the EU 377 ö o t» Commission Parliament Council Left-Right FIGURE 2. EU institutions in two dimensions The Model To analyze the effects of QMV-based legislative procedures on the discretionary power of the Commission and Court, we present a more realistic model of policy dynamics in the EU than the simplistic version presented in Figure 1. We use a two-dimensional policy space (see Figure 2) for at least two reasons. On the one hand, many results from one-dimensional spatial models do not hold in policy spaces of higher dimensions. For example, there is almost never a median voter in the Council when preferences are arrayed on two dimensions.59 On the other hand, many important policy disputes in the contemporary EU appear to take place in a two-dimensional issue space—one dimension describes actors' preferences for more or less regional integration, and the other is more akin to a traditional left-right cleavage (most notably on regulatory matters).60 The locations of the actors in Figure 2 represent plausible general preference configurations in these two dimensions.61 In both dimensions the Council and the Parliament are likely to be the more "extreme" actors, whereas the Commission is likely to be positioned somewhere in-between them. On the left-right dimension the decisions should be taken at the most appropriate level. What this level is will vary on a case-by-case basis, and it will be up to the Court to determine whether legislation in member states and at the EU level satisfy this criterion. The implications of subsidiarity, however, are beyond the scope of this article. 59. The exception to this rule is when preferences are perfectly symmetrical around a single point. 60. Kreppel and Tsebelis 1999. 61. In the graphic the first dimension is left-right, and the second is integration. 378 International Organization Commission is more likely to be closer to the national governments that appoint the commissioners; on the integration dimension, however, the Commission and the Parliament are more likely to be allied as pro-Europe actors.62 What emerges from these assumptions is that the locations of the three actors represent the corners of a triangle. Theoretically, this is the most general representation of all cases in which the three actors can have any position with respect to each other—except where two of them have identical positions, or where one of them is located exactly on a straight line connecting the central points of the other two. It should be emphasized, therefore, that the analytic thrust of our analysis would hold regardless of the relative position of actors. Let us now rotate Figure 2 by 45 degrees (for presentational purposes only), and incorporate the fact that all three institutional actors are in fact multimember bodies deciding by simple, absolute, or qualified majorities (Figure 3). We array the preferences of a Parliament made up of nine members to characterize what is a de facto supermajority threshold for voting in the Parliament under the absolute-majority requirements for passage in the second reading of legislative bills. We incorporate this restriction into our model by requiring a majority higher than five-ninths for a bill to be adopted. As a result, in Figure 3 there is no majority to the left of line P1P5, no majority above line P3P8, and so on.63 As a result of this de facto supermajoritarian requirement, there are some points located centrally in the Parliament that cannot be defeated by the required qualified majority. We use a three-member Commission deciding by majority of its members (two out of the three) since this is the formal decision rule for the College of Commissioners. Finally, we again analyze a seven-member Council where five of its members represent the required qualified majority for decision making. The central feature of Figure 3 is its description of the "core" of the EU's legislative institutions under the various QMV-based legislative procedures. The core of a legislative rule is the set of outcomes that cannot be overruled by applying that rule. For our purposes, the concept of the core has a vital role in the legislation-discretion game. The core of the EU's different legislative procedures describes the discretionary space available to the Commission to implement legislation, and to the Court to adjudicate it.64 The propositions we derive generalize to more than two dimensions, so long as the core exists.65 Note that we also assume that the outcome of legislative interactions—in the long run—will select points 62. The justification for these preferences is elaborated in more detail in Tsebelis and Garrett 2000. 63. Since there are only five points in the specified directions and the requirement is more than five-ninths of the votes. 64. It should be remembered at this point that our analysis of the Court's discretion under QMV-based legislative procedures applies to cases in which the Court interprets secondary legislation. Where the Court is interpreting the EU's treaty base directly, the only formal recourse of member governments is to act unanimously to rein in their discretion (through a treaty revision). 65. The core formally ceases to exist if one sufficiently increases the dimensionality of the policy space (except under unanimity). In such cases, Tsebelis generates propositions similar to those presented here on the basis of legislative "veto players." Tsebelis 1995. Intergovernmentalism and Supranationalism in the EU 379 MX^/A Core of unanimity voting in Council MWffi Core of consultation procedure ^>^y1 Core of cooperation procedure I I Core of codecision I and II procedures FIGURE 3. The core of EU legislative procedures inside the core. Indeed, no matter what the decision-making rule is, some point inside the core can always defeat any point outside the core. Thus, in equilibrium, we would expect the legislative status quo to be inside the core, even if at particular times the actors cannot agree to such a Pareto-improving move. Let us begin by briefly reinterpreting political dynamics where the Council decides by unanimity. In such cases, a unanimous Council is required for a change of the legislative status quo. Any point inside the CI . . . C7 heptagon cannot be modified by unanimity because at least one member of the Council would object to any change in the status quo. The hatched area in Figure 3 (regardless of its shade) is thus the core of unanimity-based legislative procedures (and for treaty revisions). Turning to discretion, the Commission and the Court could therefore effectively implement or interpret a given piece of legislation (the status quo) in any way they wish—so long as the ensuing policy outcome remains within the core. This would be true even if the Commission's implementation or the Court's interpretation were inconsistent with the Council's intent when it passed the legislation. 380 International Organization The final observation we should make concerns the spatial location of actors. Obviously, preference convergence (for example, if CI to C7 were clustered more tightly under unanimity, or if the distances between Council, Parliament, and Commission shrank) would reduce the core and hence the scope of discretion in implementation and adjudication as well. Increasing heterogeneity would have the opposite effect. In the context of the EU, adding new members to the EU might be expected to increase heterogeneity in some cases (the southern accessions and, in the future, those from Eastern Europe66) but decrease it in others (Austria, Finland, and Sweden, on many issues). Others analysts argue that the preferences of existing actors have converged over time.67 Moreover, there might be reasons to expect the distance among the institutions to be reduced—for example, if citizens come to hold their Parliament members more accountable (and then vote the same way in national and Parliament elections). For our purposes, however, it is more interesting to hold preferences constant and analyze differences in the cores of EU legislation—and hence the scope for bureaucratic and judicial discretion—in terms of the procedures used to aggregate the preferences of legislative actors. Consultation and Cooperation Under the consultation and cooperation procedures, legislation can pass in two ways. A decision can be made by agreement of the relevant actors or by unanimity in the Council (acting alone). The "relevant actors" in this case are a qualified majority of the Council and a majority of the Commission. Under the cooperation procedure, an absolute majority of the Parliament must be added to this list. We have already calculated the unanimity core of the Council. What constraints does the alternative rule (agreement of Commission for consultation, or of the Commission and the Parliament for cooperation) impose on policy discretion? We concentrate on the cooperation procedure because of the additional complexities generated by the Parliament's participation in legislation. Recall that we are assuming that the absolute majority requirement in the Parliament creates a de facto supermajority threshold of more than five-ninths. In Figure 3 the five-ninths core of the Parliament can be identified by connecting each Parliament member with another so that three other members are on one side of the line, and the other four members are on the other side. Such lines are the pairs P1P5, P1P6, P2P6, P2P7, and so on. These lines define a nine-sided polygon inside PI . . . P9. This is the Parliament's core under absolute majority. We will call this specific set of outcomes the "five-ninths Parliament core." It is obvious that the Parliament cannot modify anything located in that core—even if it could act alone—without the support of the Council or the Commission. The reason is that there is a majority of more than five-ninths against moving away from any particular point of this nine-sided 66. Bednar, Ferejohn, and Garrett 1996. 67. Moravcsik 1998. Intergovernmentalism and Supranationalism in the EU 381 polygon. Similarly, there is a core for the Council when it decides by five-sevenths QMV. As Figure 3 indicates (and for reasons similar to those for the Parliament) this "QMV core" is a heptagon located inside CI . . . C7. The lightly shaded area of Figure 3— connecting what turns out to be the decisive Commissioner (point 1 in the figure) with the extreme points of the Parliament's five-ninths core and the Council's QMV core—is thus the core of legislation requiring a qualified majority in the Council, an absolute majority in the Parliament, and a simple majority in the Commission. But this is not the core of the cooperation procedure, because a unanimous Council can also pass legislation. The core of cooperation is thus defined as the intersection of the unanimity core of the Council (the hatched area) and the inter-institutional core (the shaded area). In Figure 3 the crosshatched area denotes this cooperation core. Note that this area is always smaller than the Council's unanimity core (which defines the room for policy discretion under the Luxembourg compromise, treaty revisions, and legislation still subject to unanimity voting). It is easy to calculate the consultation core, which is simply a subset of the cooperation core—since the salient difference between the two procedures is that the agreement of the Parliament is not required. This consultation core is represented in Figure 3 by the heavily crosshatched area (regardless of shade). If the Commission or the Court wants to make a decision that will not be overruled under the cooperation procedure, either one can implement and interpret legislation anywhere within the crosshatched area. The size of this area, of course, depends on the positions of the Commission and the Parliament in relation to the Council (and the cohesion of individual actors' preferences in these institutions). If, for example, the Commission were located close to P3, the core would shrink. One may think that given the selection mechanism for the Commission (which requires approval by both the Council and the Parliament) this would be the most realistic position of the three actors most of the time. The core would expand, however, if the Council were located between the Commission and the Parliament. Again, our argument covers all the relative positions of actors, so we do not need to specify the conditions under which any particular configuration would occur. Codecision The major characteristic of both versions of codecision is that at the end of the legislative game, an agreement by a qualified majority of the Council and an absolute majority of the Parliament can overrule other actors. In particular, they can bypass the Commission. Recall that we have argued that the legislative influence of the Council and Parliament is different under the original and reformed versions of the procedure. Under the Maastricht rules, the Council had effective agenda-setting power; under the post-Amsterdam rules, the Parliament is the Council's true coequal. The difference in procedures, however, does not affect the size and location of the legislative core and, in turn, the discretion of the Commission as an administrator and the Court as a statutory interpreter. 382 International Organization Indeed, under both the Maastricht and the Amsterdam version of codecision (II), the agreement of the Council and the Parliament is the only mechanism that generates legislation. Consequently, the heavily shaded area of Figure 3 that connects the five-ninths Parliament core and the five-sevenths Council core represents the core of codecision II. The greater the policy differences between the Council and the Parliament (and the greater the preference dispersion inside these institutions), the greater the size of the core, and hence the greater the discretion available to the Commission to implement policy and the Court to interpret the law. Until now, most analysts have believed that the distance between Council members and Parliament members is significant. But this may change over time, either because European citizens come to hold their Parliament members more accountable or because party organizations linking both institutions become more pronounced.68 If and when either of these changes takes place, the codecision core would shrink—and with it the policy discretion of the Commission and Court. Figure 3 shows that the core of the codecision procedure is likely to be larger than the cooperation core but this may not always be the case. If the five-ninths core of the Parliament were located inside the five-sevenths core of the Council (that is, if the Parliament and the Council had very similar positions), the cores of the cooperation and codecision II procedures would be similarly sized. Nonetheless, to describe the amount of discretion available to the Commission and Court as a concertina is generally reasonable. Under the Luxembourg compromise, the discretionary space was large (the unanimity set of the Council). The core shrank appreciably with the activation of consultation, increased with cooperation, and has most likely expanded considerably again since the introduction of co-decision. Summary The Luxembourg compromise imposed unanimity decision making in the Council and created a large core of legislation that could not be overturned. As a result, the jurisdiction of the Court expanded, and to a large extent the Court became the EU's de facto policymaker (the Commission could not play an important role as an administrator because of the small amount of EU legislation in this period). The revival of the consultation procedure and the creation of cooperation in the mid-1980s reduced significantly the size of the legislative core, curtailing the role of the Court as an adjudicator and the Commission as a bureaucrat (although the Commission had a significantly higher volume of legislation to administer). Similarly, codecision I reduced the size of the core even further. The codecision II procedure increased the size of the EU's legislative core of invulnerable legislation because it requires agreement of the Council and the Parliament for new bills to be passed. As a result, the roles of the Commission (as 68. Tsebelis and Garrett 2000. Intergovernmentalism and Supranationalism in the EU 383 an administrator) and the Court (as an adjudicator) have increased. This expansion of discretion is not a necessary feature of codecision; it relies on a divergence of preferences between the Council and the Parliament. This has always been the case until now, but we may witness some convergence as citizens come to hold their Parliament members more accountable. Linking Institutional Consequences to Institutional Choices We have provided a theoretical overview of the interactions among the four principal EU institutions—the Council, Commission, Court, and Parliament—from the signing of the Rome Treaty in the late 1950s to the ratification of the Amsterdam Treaty in 1999. We have made four basic points. First, the Court was the prime mover behind European integration under the Luxembourg compromise because the member governments were gridlocked. It was extremely hard for them to pass any new secondary legislation unanimously, and political conditions made treaty revisions unlikely. In this environment the Court could act to further legal integration knowing that political efforts to rein in this activism were unlikely. The Commission, however, was a less effective pro-integration entrepreneur under the Luxembourg compromise. Even though unanimity voting in the Council potentially gave the Commission considerable discretion in policy implementation, there were few pieces of secondary legislation to implement. Second, the move to QMV in Council decision making in the late 1980s reduced the discretionary implementation space available to supranational entrepreneurs. The lower threshold in the Council for passing new legislation, all else equal, should have resulted in less pro-integration activism by the Commission and the Court. Prominent scholars have suggested that this hypothesis is correct when applied to the Court,69 but it is harder to find research arguing that the Commission's administrative power declined. Our analysis suggests one reason QMV did not affect the Commission's discretionary space as strongly as it did the Court's. Although the discretionary space available to both supranational actors decreased, the proliferation of secondary legislation in this period increased the number of issue-areas in which the Commission could exercise its (spatially reduced) discretion. But the reason most commentators give for the Commission's resurgence after the mid-1980s involves its agenda-setting powers. We disagree with this assertion. Third, the Commission's agenda-setting power under QMV-based procedures has declined steadily in the past decade while the legislative role of the Parliament has increased. The Commission can significantly influence the course of legislation under the consultation procedure (in which the Parliament plays no effective role). But its agenda-setting power is shared with the Parliament under cooperation, was further eroded by the initiation of codecision at Maastricht, and all but eliminated 69. See Mattli and Slaughter 1998; and Weiler 1991. 384 International Organization under the Amsterdam version of codecision. At the same time, the Parliament's legislative role has increased to the point where it is the Council's coequal under the reformed codecision procedure. Finally, if (as seems likely) the reformed codecision procedure becomes the legislative norm in the EU, the discretionary space available to the Commission and Court for pro-integration activism in policy implementation and statutory interpretation will increase again in coming years. Under this truly bicameral procedure, the effective constraint on supranational activism will be the extent to which pivotal Parliament members continue to prefer more integration than the pivotal governments in the Council. The greater this legislative gridlock interval, the greater the Commission's and the Court's discretion. As citizens come to realize the Parliament's powers, however, they may come to demand that their Parliament members act more as their delegates than as pro-integrationist coconspirators with the Commission and Court. If and when this happens, the Commission's and Court's discretionary powers under codecision will decrease. But this is for the future.70 We conclude by locating our institutionalist analysis in the context of the dominant approaches to European integration—intergovernmentalismand suprana-tionalism. We begin by explaining why we consider formal institutional interactions to be so important. In any human interactions, there are three necessary concepts: the players (individual or collective), their strategies (which jointly determine the outcomes of their interactions), and the payoffs (which they receive at the end of their interactions). In game theory these three concepts are sufficient for the description of any game. Most attention typically focuses on strategies, which depend on the sequence of moves that define the game, the set of choices, and the information players have when they are called upon to move. These constraints that affect the strategies of actors are all provided by "institutions." Formal institutions play numerous roles in the process of European integration. They specify what is permitted and what is not—for example, the treaties specify that environmental issues today (but not in the 1960s) are within the jurisdiction of the EU. Institutions specify that legislation must start with the introduction of a draft of a directive or a regulation by the Commission to the Parliament and ends with the approval by the Council (cooperation) or by both the Parliament and the Council (reformed codecision). Institutions also determine the available choices of actors— for example, if the Parliament wants to move a paragraph from one point in a bill to another, it must introduce two amendments (one deleting the original text, and the other reintroducing it in the new position). Since institutions determine the sequence of moves, the choices of actors, and the information they control, different institutional structures affect the strategies of actors and hence the outcomes of their interactions. Consequently, institutions can be studied as independent variables (as we have done here) to see how they 70. This scenario is analyzed in Tsebelis and Garrett 2000. Intergovernmentalism and Supranationalism in the EU 385 influence outcomes, or as dependent variables to see how particular institutions are chosen. At the risk of oversimplifying, we suggest that intergovernmentalists who focus on treaty bargaining view the EU's institutional structure as the dependent variable. Moreover, they conceive of this structure in general terms—such as Moravcsik's focus on EU institutions as credible commitments to integration—rather than analyzing the detailed interactions among the EU's four primary institutions and their likely effects on policy.71 Supranationalists, in contrast, view the EU's institutions as actors, not dependent variables: the Commission, Court, and Parliament undertake actions that affect the direction of European integration. But as we have shown in this article, supranationalists tend to rely on general neofunctionalist concepts (modernized renderings of "spillovers") and eschew analysis of the strategies available to different actors and the constraints under which they operate. In other words, they do not analyze institutions as generators of particular equilibrium outcomes, which supranationalists consider too unpredictable and contingent to merit close attention. Viewed in this light, it is not surprising that the debate between intergovernmentalism and supranationalism seems little closer to resolution today than when Stanley Hoffmann famously criticized Ernst Haas more than thirty years ago.72 Supranationalism and intergovernmentalism differ not only in the importance they attach to member governments in the process of integration but also in how they treat the EU's institutions. Our institutional approach demonstrates that intergovernmentalists' laser-like focus on treaties requires a prior study of the everyday realities these treaties generate (or are likely to generate) in the EU, and that supranationalists' focus on the study of these realities requires microfoundations and structure. In our view these three major streams of research can be distinguished along two dimensions (see Figure 4). The focus in the first dimension is solely on interactions among member governments as defining the integration process. Here, our institutional approach is closer to supranationalism than to intergovernmentalism It avoids the—inappropriately—myopic focus of intergovernmental analyses on treaty revisions by paying close attention to the multitude of clearly important directives, regulations, and Court decisions that influence the course of European integration from day to day. The focus in the second dimension concerns whether the course of European integration is the product of intentional choices by (and strategic interactions among) the relevant actors. For supranationalists influenced by Haas, the law of unintended consequences is an article of faith—it is what spillovers are all about. For intergovernmentalists, in marked contrast, the governments that sign treaties are not only in the driver's seat but also know exactly where they are going. 71. Moravcsik 1998. 72. Hoffmann 1966. 386 International Organization Intergovernmentalism Supranationalism Institutionalism Are governments the only (important) actors? Unintended consequences? Yes No No No Yes No (under complete information) FIGURE 4. Three approaches to European integration Our position on this issue is more qualified. If actors operate under complete information (that is, they know all relevant information about each other), they will design institutions that best promote their preferences—subject to the constraint that every other actor will behave similarly. Nonetheless, even under conditions of complete information, our institutional analysis suggests a different type of research on treaty bargaining than is typical in intergovernmentalism Intergovernmentalism treats the EU's institutional structure as a dependent variable; it is the product of treaty bargaining. We have demonstrated, however, that it is simply impossible to analyze institutional choice without first understanding institutional consequences. The fact that intergovernmentalists typically eschew "institutions as independent variables" analysis significantly lessens their ability to understand institutional choice. Consider two examples. First, most analyses of the reinvigoration of European integration in the mid-1980s debate the impetus for completing the internal market. To us this focus on stated policy objectives—the three hundred directives in the Commission's White Paper—neglects the fact that the commitment to complete the internal market, in the form of the four freedoms, was already written into the Rome Treaty.73 What was new about the mid-1980s was the collective decision of the governments to end the Luxembourg compromise. Moreover, the member states decided in the SEA that the general objective of completing the internal market be translated into detailed legislation using a new procedure, cooperation, in which the Parliament was an active participant. As we have shown, the use of cooperation not only affected the types of legislation that could be passed but also affected the discretion available to the Commission and the Court in implementing and interpreting this legislation. The second example is the Maastricht Treaty. Not surprisingly, perhaps, most analysts have focused on the decision to move toward monetary union. But this focus has deflected attention from what we consider to be the most important political reform in the treaty—the creation of the codecision procedure (to govern 73. It is true, of course, that Delors chose to emphasize internal market reforms over, say monetary or political union, because he thought it was more likely to be supported by all governments. Intergovernmentalism and Supranationalism in the EU 387 internal market legislation, among other things). We have spent a good deal of time here analyzing the significant and far-reaching implications of codecision for legislation, administration, and legal adjudication in the EU. The point of both examples is simple but important. The study of institutional consequences is logically prior to the study of institutional choice. Institutions determine how policy objectives will be translated into political outcomes. Even if intergovernmentalists are right to assume that treaty bargaining takes place under complete information, the fact that they pay more attention to stated policy objectives rather than the institutions created to implement them is a serious weakness in their mode of analysis. But how appropriate is the complete-information assumption for treaty bargaining? Our position is in-between the black-and-white positions of the supranation-alists and the intergovernmentalists. The complete-information assumption is a strict one. In our analyses of legislative politics, for example, we believe the assumption is only appropriate in the final stages of the EU's complex procedures.74 With respect to implementation and adjudication, the fact that the Barber protocol was written into the Maastricht Treaty to countermand a Court decision is good evidence that the Court does not always accurately predict the reactions of member governments.75 In the case of treaty bargaining, the threshold for complete information is even higher—because the governments are making decisions that will have long chains of effects into the indefinite future. If they do not know all relevant information about each other, or if they operate under cognitive pressures that restrict their ability to behave perfectly rationally, or if they expect with some probability that shocks in the political environment will change the endowments of other actors, the strict complete-information assumption is unlikely to be very helpful. But as an empirical matter, it is worth asking how much of the evolution of the EU since the mid-1980s was anticipated by the member governments during the treaty-making processes, and how much has been unintended. If one focuses on debates about reducing the democratic deficit through the reform of the EU's legislative procedures—in our view one of the most important features of European integration in the past twenty years—the balance seems to tip in favor of the complete-information assumption. By and large, the institutional modifications introduced by the SEA and the Maastricht and Amsterdam Treaties had the intended effect of reducing the Commission's role and increasing the Parliament's.76 We cannot say, however, that every new constitutional innovation was completely understood by all players at the moment of its introduction. For example, Tsebelis and Amie Kreppel trace the history of conditional agenda setting and find that not all the participants in the Rome Treaty understood its significance (although 74. Tsebelis and Garrett 2000. 75. Garrett, Kelemen, and Schultz 1998. 76. We leave to others the question of why the member governments wanted to empower the Parliament. 388 International Organization the major actors certainly did). By the time of the Amsterdam Treaty, however, they had enough information to modify the details of the codecision procedure and make the Parliament a coequal legislator with the Council; this may have been their intent all along, since presumably the procedure was created at Maastricht with this purpose in mind. In sum, we believe that the purported "law of unintended consequences" has empirically been riddled with many more exceptions than most commentators on European integration suggest. Thus focusing on the formal institutional interactions in the EU not only allows us to explain how the EU has operated in different epochs but also gives us important insights into how the member governments have decided to pool their sovereignty in the integration process. References Alivizatos, Nicos C. 1995. Judges as Veto Players. Id Parliaments and Majority Rule in Western Europe, edited by Herbert Döring, 566-89. New York: St. Martin's Press. Alter, Karen J. 1998. "Who Are the Masters of the Treaty?" European Governments and the European Court of Justice. International Organization 52 (1): 121-47. Baron, David P., and John A. Ferejohn. 1989. Bargaining in Legislatures. American Political Science Review 83 (4): 1181-1206. Bednar, Jenna, John Ferejohn, and Geoffrey Garrett. 1996. The Politics of European Federalism. International Review of Law and Economics 16 (3):279-94. Burley, Anne-Marie, and Walter Mattli. 1993. Europe Before the Court. International Organization 47 (l):41-76. Cooter, Robert, and Josef Drexl. 1994. The Logic of Power in the Emerging European Constitution. International Review of Law and Economics 14 (3):307-26. Cooter, Robert D., and Tom Ginsburg. 1996. Comparative Judicial Discretion: An Empirical Test of Economic Models. International Review of Law and Economics 16 (3):295-313. Corbett, Richard, Francis Jacobs, and Michael Shackleton. 1995. The European Parliament. 3d ed. London: Cartermill. Crombez, Christophe. 1996. Legislative Procedures in the European Community. British Journal of Political Science 26 (2): 199-228. Dinan, Desmond. 1994. Ever Closer Union? Boulder, Colo.: Lynne Rienner. Dogan, Rhys. 1997. Comitology: Little Procedures with Big Implications. West European Politics 20 (3):31-60. Ferejohn, John, and Charles Shipan. 1990. Congressional Influence on Bureaucracy. Journal of Law, Economics, and Organization 6:1-20. Garrett, Geoffrey. 1992. International Cooperation and Institutional Choice: The European Community's Internal Market. International Organization 46 (2):533-60. ---------. 1995 a. The Politics of Legal Integration in the European Union. International Organization 49 (1):171-81. ---------. 1995b. From The Luxembourg Compromise to Codecision: Decision Making in the European Union. Electoral Studies 14 (3):289-308. Garrett, Geoffrey, and George Tsebelis. 1996. An Institutional Critique of Intergovernmentalism. International Organization 50 (2):269-300. 77. Tsebelis and Kreppel 1997. Intergovernmentalism and Supranationalism in the EU 389 ---------. Forthcoming. Understanding Better the EU Legislative Process. European Union Politics. Garrett, Geoffrey, and Barry R. Weingast. 1993. Ideas, Interests, and Institutions: Constructing the European Community's Internal Market. In Ideas and Foreign Policy, edited by Judith Goldstein and Robert Keohane, 173-206. Ithaca, N.Y.: Cornell University Press. Garrett, Geoffrey, R. Daniel Kelemen, and Heiner Schulz. 1998. The European Court of Justice, National Governments, and Legal Integration in the European Union. International Organization 52 (1):149— 76. Golub, Jonathan. 1999. In the Shadow of the Vote? Decision Making in the European Community. International Organization 53 (4):733-64. Haas, Ernst B. 1958. The Uniting of Europe. Stanford, Calif.: Stanford University Press. Hammond, Thomas H. 1996. Formal Theory and the Institutions of Governance. Governance 9 (2): 107-85. Hammond, Thomas H., and Jack H. Knott. 1999. Political Institutions, Public Management, and Policy Choice. Journal of Public Administration Research and Theory 9:33-85. Hammond, Thomas H., and Gary J. Miller. 1987. The Core of the Constitution. American Political Science Review 81 (4):1155-74. Hoffmann, Stanley. 1966. Obstinate or Obsolete? The Fate of the Nation-State and the Case of Western Europe. Daedalus 95 (3):862-915. Hix, Simon. 1999. The Political System of the European Union. London: Macmillan. ---------. 2000. Constitutional Agenda-Setting Through Discretion in Rule Interpretation: Why the European Parliament Won at Amsterdam. Paper presented at the 96th Annual Meeting of the American Political Science Association, September, Washington, D.C. Kilroy, Bernadette. 1999. Integration Through the Law: Court and Governments in the EU. Ph.D. diss., University of California, Los Angeles. Kreppel, Amie, and George Tsebelis. 1999. Coalition Formation in the European Parliament. Comparative Political Studies 32 (8):933-66. Majone, Giandomenico. 1999. The Coasean Problem of Democratic Theory: Nonmajoritarian Institutions and Political Transaction Costs. Paper presented at the European Communities Studies Association meetings, Pittsburgh, Pa. Marks, Gary, Liesbet Hooghe, and Kermit Blank. 1996. European Integration from the 1980s: State-Centric Versus Multi-level Governance. Journal of Common Market Studies 34 (3):341-78. Mattli, Walter, and Anne-Marie Slaughter. 1998. Revisiting the European Court of Justice. International Organization 52 (1): 177-210. McCubbins, Matthew D., Roger G. Noll, and Barry R. Weingast. 1987. Administrative Procedures as Instruments of Political Control. Journal of Law, Economics, and Organization 3 (2):243-77. ---------. 1989. Structure and Process, Politics and Policy: Administrative Arrangements and the Political Control of Agencies. Virginia Law Review 75 (2):431-82. Miller, Gary. 1995. Post-Maastricht Legislative Procedures: Is the Council "Institutionally Challenged?" Paper presented at the 4th International Conference of the European Community Studies Association, Charleston, S.C. Moe, Terry M. 1990. Political Institutions: The Neglected Side of the Story. Journal of Law, Economics, and Organization 6:213-53. Moravcsik, Andrew. 1998. The Choice for Europe. Ithaca, N.Y.: Cornell University Press. Moser, Peter. 1996. The European Parliament as a Conditional Agenda Setter: What Are the Conditions? American Political Science Review 90 (4):834-38. Nash, John. 1950. The Bargaining Problem. Econometrica 18 (2):155-62. Peters, B. Guy. 1992. Bureaucratic Politics and the Institutions of the European Community. In Euro-Politics, edited by Alberta Sbragia, 75-122. Washington, D.C: Brookings Institution. Pierson, Paul. 1996. The Path to European Integration. Comparative Political Studies 29 (2):123-63. Pollack, Mark A. 1997. Delegation, Agency, and Agenda Setting in the European Community. International Organization 51 (1):99—134. Rubinstein, Ariel. 1982. Perfect Equilibrium in a Bargaining Model. Econometrica 50 (1):97—110. 390 International Organization Sandholtz, Wayne. 1992. Hi-Tech Europe: The Politics of International Cooperation. Berkeley: University of California Press. Sandholtz, Wayne, and John Zysman. 1989. 1992: Recasting the European Bargain. World Politics 42 (1):95-128. Scully, Roger M. 1997. The European Parliament and the Co-Decision Procedure: A Re-Assessment. Journal of Legislative Studies 3 (3):58-73. Steunenberg, Bernard, Christian Koboldt, and Dieter Schmidtchen. 1996. Policy Making, Comitology, and the Balance of Power in the European Union. International Review of Law and Economics 16:329-44. Stone Sweet, Alec, and Wayne Sandholtz. 1997. European Integration and Supranational Governance. Journal of European Public Policy 4 (3):297-317. Tsebelis, George. 1994. The Power of the European Parliament as a Conditional Agenda-Setter. American Political Science Review 88 (1): 128-42. ---------. 1995. Decision Making in Political Systems. British Journal of Political Science 25 (3):289- 326. ---------. 1996. More on the European Parliament as a Conditional Agenda Setter: Response to Moser. American Political Science Review 90:839-44. Tsebelis, George, and Geoffrey Garrett. 2000. Legislative Politics in the European Union. European Union Politics 1 (1):9—36. Tsebelis, George, and Amie Kreppel. 1998. The History of Conditional Agenda Setting in European Institutions. European Journal of Political Research 33 (1):41—71. Tsebelis, George, and Jeanette Money. 1997. Bicameralism. New York: Cambridge University Press. Tsebelis, George, Christian B. Jensen, Anastassios Kalandrakis, and Amie Kreppel. Forthcoming. Legislative Procedures in the European Union: An Empirical Analysis. British Journal of Political Science. Weiler, Joseph H. H. 1991. The Transformation of Europe. Yale Law Journal 100 (8):2403-83.