The European Community (EC) considers itself a community of law.... integration is perceived primarily as "integration through law" (Cappelletti et... Dirk Peters
Transcription
The European Community (EC) considers itself a community of law.... integration is perceived primarily as "integration through law" (Cappelletti et... Dirk Peters
Dirk Peters Making European Law Work: Why Member States (Fail to) Implement EC Directives1 - Draft 1 Introduction The European Community (EC) considers itself a community of law. European integration is perceived primarily as "integration through law" (Cappelletti et al. 1986). This attempt to realize European integration by way of creating common legal norms has resulted in an often lamented proliferation of legal acts issued by the EC2. However, issuing legal acts alone does not advance European integration. These regulations and directives need to be put into practice by member states if they are to have any effect at all. National implementation of EC law, therefore, is a crucial aspect of European integration. Yet, it also is an aspect that has attracted only little attention by European policy makers and scholars of European integration alike. Especially the question whether there are patterns of implementation across functional domains (is there systematic variation in implementation rates between member states?) is a seriously underresearched issue. Peter M. Haas has highlighted this desideratum in an article for the Journal of European Public Policy some time ago and suggested ways of filling this gap (Haas 1998). In particular, he proposed to use general theories of International Relations and Comparative Politics stressing state agency to develop explanations for the implementation behavior of EU member states. In this article I pick up on these suggestions and develop two models to explain variation in implementation rates of EU member states. In two respects, however, I will depart from the path charted by Haas. First, my approach is more narrow than the research program outlined by Haas, because I will confine myself to the exploration of rationalist arguments about implementation behavior. I exclude constructivist approaches mainly for practical reasons (i.e. limitations of time and space) and not to suggest that they do not have to offer some valuable insights into the issue at hand. This more focused approach enables me to elaborate the explanatory models in more detail. Thus, second, I will move beyond outlining avenues for future research and instead 1 I am grateful for numerous helpful comments on earlier drafts of this article by Peter Mayer, Volker Rittberger and Wolfgang Wagner, as well as by Rainer Baumann, Tanja Börzel, Andreas Hasenclever, Volkhart Heinrich, Klemens Käppeler, Annette Lutz, Manuela Reichle, Regina Stegmüller and Klaus Stodick. 2 The European Community (EC) is the first pillar of the European Union (EU). Because all binding legal acts are issued by organs of the EC I will employ the terms "EC law," "EC directives" etc. throughout, but speak of the "EU" when I refer to the organization as a whole. 2 elaborate the models to the point where I am able to derive hypotheses which can be tested empirically. Hence, in what follows, I will develop and preliminarily test two rationalist explanations for the national variation in the implementation rates of EC diretives. The first model, which will be outlined in section 4, is based on the assumptions of the rational institutionalist school of thought in International Relations theory. Two hypotheses are derived from this model which link a state’s rational self-interest to its implementation behavior. The second model combines elements of liberal International Relations theory and of Comparative Politics. It opens up the "black box" of the state – national decisions about implementation or non-implementation of EC law are viewed as the outcome of complex decision-making processes in which a variety of societal actors are involved. From this model two hypotheses will be derived which link national decision-making procedures to a state's implementation rate (section 5). Both models are tentatively tested in section 6. Due to limitations of available reliable data, I will restrict the test to the legal implementation of EC directives, i.e. to the transposition of EC directives into national law. It will turn out that both a state's capacity to comply with EC law and the role of domestic veto players are crucial for explaining differences in implementation behavior. Section 7 summarizes the findings and outlines their policy implications. Before the models are developed, however, I will first briefly describe how the implementation of EC law works at all and also briefly review the literature on the implementation of EC directives in sections 2 and 3. 2 The Implementation of EC Directives While EC policies are decided on in "Brussels", i.e. at European level, they are put into practice only at the national level. Consider an attempt to improve air quality in EU member states. The Commission may draft a clean-air directive providing for cuts in the emission of certain pollutants. After some intra-EC bargaining, this directive may be adopted by EC organs (the Council and the European Parliament). However, to make this policy work the EC depends crucially on national actors: First, the member states need to transpose the directive into national law; i.e. the national parliament or some other national institution has to pass a law that contains essentially the rules laid down in the directive (transposition or legal implementation).3 Second, this national law has to be applied by the national administration (administrative implementation). Third, individuals need to comply with the law, i.e. producers of pollutants have to actually cut 3 In the EC context, legal implementation is unique to directives. All other binding legal acts issued by the EC (regulations and decisions) take direct effect and need not be transposed into national law (see Article 248 [189] TEC). 3 their emissions (compliance). Only then there is, finally, a chance for an actual reduction in the overall emission of pollutants and thus for clean air (effectiveness).4 While the responsibility for implementation lies primarily at the national level, the failure to implement EC law can have important repercussions on the European level. It can "cause discontinuities in the Community’s legal system, create asymmetries among states, and undermine the effects of Community policies." (Mendrinou 1996: 19 n. 9). In order to avoid such detrimental repercussions, the Commission watches over legal and administrative implementation and attempts to ensure that member states fulfill their obligations. Yet it is difficult for the Commission to monitor implementation efforts, let alone to actually influence implementation behavior. The only phase which can reliably be monitored is legal implementation, because member states have a duty to report on the measures they have taken to transpose a directive into national law (Nugent 1999: 134). Failures at administrative implementation are much harder to detect because there are no national reports on administrative implementation and the Commission – due to its limited resources – is by no means able to effectively ensure the correct administrative application of more than one thousand directives in 15 member states (From/Stava 1993: 62-64; Nugent 1999: 137). Moreover, once it has detected implementation failure, the Commission can hardly make the respective member state change its behavior because the Commission's ability to impose sanctions is quite limited. Only since the Treaty on the European Union (TEU) took effect in 1993, the European Court of Justice (ECJ) has the possibility to impose fines on recalcitrant member states.5 3 Research on Non-Implementation Despite the importance implementation has for European integration, there has been little effort to systematically identify the causes of non-implementation in EU member states. To be sure, there is a considerable number of case studies on the implementation of individual EC directives in one or more member states (e.g. Brenner 1996, Edwards, V. 1996, Nettesheim 1999, Schiøler Sørensen 1996). However, most of these studies are concerned with the legal aspects of implementation only. They describe in great detail the transposition of a single directive in one or more member states pointing out factors that made implementation in this special case easy or difficult. They do not claim to explain the implementation decision, that is to uncover the causes of the implementation 4 This distinction of the several phases of implementation is borrowed from Andresen et al. (1995: 27). For details of the infringement procedure which is laid down in Article 226 [169] TEC, see Nugent (1999: 134-138) and Oppermann (1991: 630-635). For the "new" possibility to impose fines on member states, see Article 228 [171] para. 2 TEC. 5 4 or non-implementation of EC directives in general.6 A second group of studies attempts to find such causes by comparing the implementation of several EC directives or the process of implementation in general in EU member states. These studies, however, tend to be either rather eclectic, summing up potential causes of implementation failure and paying less attention to systematically linking causes and effects (Mosley 1996; Siedentopf/Ziller 1988; Schwarze et al. 1990); or to focus their attention on a single cause, e.g. involvement of national parliaments in the formulation of directives (Martin 1995). Only Peter Haas (1998) has made an attempt to systematically formulate coherent theoretical models of implementation behavior, especially compliance with EC directives. However, he does not formulate testable hypotheses, but only discusses "possible elements for a research program on understanding patterns of compliance with EU legislation, focusing on manipulable variables which may influence a state’s decision to comply" (Haas 1998: 18). Hence, non-implementation of EC law has not yet been systematically explained. By developing and (preliminarily) testing two models of implementation I attempt to contribute to such an explanation. 4 Rational Institutionalism 4.1 Basic Assumptions The first of the two explanatory models presented here is based on the rational institutionalist school of thought in International Relations Theory, which was influenced most importantly by the work of Robert Keohane (1984, 1989a,b).7 Rational institutionalists regard states as unitary, rational actors which act according to calculations of costs and benefits. State behavior reflects primarily the conditions set by the international system, while the internal characteristics of a state are of less importance (Hasenclever et al. 1997: 24-26; Keohane 1984: 25f., 1989a: 1f., 7f.; Slaughter Burley 1993: 218). Moreover, rational institutionalism posits that states (or at least "advanced market-economy countries" (Keohane 1984: 6) which exist in zones of "complex interdependence" and thus do not have to fear attack by their neighbors (Keohane/Nye 2001: 23f.)) are rational egoists: They strive for absolute gains and do not 6 The only exception from this pattern which has come to my attention is Duina (1997), who attempts to show the effect of existing domestic institutions on the probability that a directive is implemented by analyzing the implementation of the equal pay directive in three countries. 7 Rational institutionalism also goes by the name of "neoliberal institutionalism" (Keohane 1989a). To avoid confusion with the liberal school of thought, introduced in the next section, which rests on a slightly different notion of "liberalism", I employ the term "rational institutionalism", which reflects the basic assumptions of this school of thought more clearly. 5 care about the gains of other states (Hasenclever et al. 1997: 29; Keohane 1984: 27; Krasner 1982: 195). That states care only about their own gains and not about those of others facilitates cooperation in situations in which states can realize joint gains. But this does not imply that states will cooperate in every situation in which joint gains would be attainable. Rather there are "problematic social situations" (Raub/Voss 1986: 86-88) in which cooperation is hampered by uncertainty about the other states’ behavior. Every state fears that the others could choose to act in a way that would – in combination with its own behavior – prevent it from realizing any gains at all (Keohane 1989b: 64). Therefore, a state will join the cooperation only when it can be certain (at least to some degree) about the other states' behavior. 4.2 Cooperation in Problematic Social Situations The most important problematic social situations in international politics are coordination games ("dilemmas of common aversions") and prisoners' dilemma games ("dilemmas of common interests") (Stein 1982: 304-311; Snidal 1985a: 926-937).8 A typical example of a coordination game in the EC context is the standardization of industrial goods, for instance of certain industrial components. All states whose industries import and export those components are interested in setting some common standard for their size or quality in order to facilitate their exchange. Nonetheless, it might be difficult for them to agree on one such common standard, because every state will try to get that standard accepted as the Community standard which its own domestic industry already applies. If it succeeds its industry will avoid the cost of adapting to a new standard.9 Yet once the states have agreed on a common standard every state that does not implement the agreement will only harm itself. Its domestic industry would be able neither to export the components it produces nor to import components it needs. Therefore, agreement in coordination games, once established, is self-enforcing. Every 8 Some authors (e.g. Zürn 1992; Martin 1993) regard two more situations as relevant for international cooperation: assurance games and suasion games. Assurance games, however, are problematic social situations only under conditions that arguably do not hold for the EU context (at least one actor is uncertain about the others' preference ordering or at least one actor does not expect the others to act rationally) (Hasenclever et al. 1997: 50). In suasion games the uncoordinated outcome is so robust that some authors do not consider them problematic social situations at all (Stein 1982: 304; Taylor 1987: 40). The probability that common rules are agreed upon in such instances is extremely low (Zürn 1992: 212). Therefore it is safe to assume that only a very small fraction of EC directives, if any, regard suasion games. I will therefore exclude assurance and suasion games from discussion. 9 To be sure, this is only a special case of a coordination game, namely a coordination game 'with distributional conflict' (Zürn 1992: 191). There are also coordination games without distributional conflicts (for example the standardization of a new component which is not yet produced in any of the states involved and whose standardization therefore will cause no costs of adaptation for any of them). In those situations the only difficulty for the participating states is to pick one common standard among several possible ones. 6 rational state will implement it. "Thus there is no problem here of policing and compliance." (Stein 1982: 314; see also Hasenclever et al. 1997: 48; Zangl 1994: 292; Zürn 1992: 180). This does not hold, however, for prisoner’s dilemma (PD) situations. Often international environmental problems can be modeled as PD games. Take, for instance, a clean-air directive designed to cut the emission of certain air pollutants. In this case, even after agreement on the directive has been reached, there remain incentives for every state to defect from cooperation, i.e. not to implement the directive. A state that does not force its domestic industry to cut its emissions while the other countries do (i.e. a "free rider"), realizes several advantages over the other states. First of all, the air in the concerning state will be cleaner than before, because less emissions from neighboring countries will cross its borders. At the same time, the state (or its domestic industry) does not have to adjust its behavior. Thus it can avoid costs while realizing benefits. Moreover, it achieves an advantage over the other states because it has less demanding environmental laws and therefore may gain in attractiveness as a location for industry. All in all, this makes for strong incentives not to implement the directive. International institutions, that is "persistent and connected sets of rules (formal and informal) that prescribe behavioral roles, constrain activity, and shape expectations," (Keohane 1989a: 3f.) may in this case be helpful to ensure implementation, because they can alter the incentive structure for states in PD situations to some extent. For instance, they may make a state's defection visible to the others, thus causing at least reputational losses for the defecting state or maybe even the collapse of the cooperation altogether, leaving the state concerned even worse off than before. However, while international institutions might be able to reduce the incentives for defection, they cannot eliminate them altogether, because an international institution eliminating all incentives for defection would be quite costly to create. Rather, rational actors will attempt to design an institution in a way that optimizes the ratio of costs and benefits. It might be perfectly rational not to eliminate all possible incentives for all possible defectors, but to tolerate a certain degree of non-cooperative behavior (Chayes/Chayes 1993: 197-204; Young 1979: 113-115). In the case of the EC, rules indeed do leave considerable room for incentives to defect. While defection might be noticed by the Commission and maybe even reprehended by the ECJ, the defecting state does not have to fear material sanctions (at least there were no such sanctions until the TEU came into force). This makes for considerable incentives not to implement directives regulating cooperation in a PD situation. 7 4.3 Hypotheses about the Implementation of EC Directives We have seen that rational institutionalism suggests that states have incentives not to implement directives only in PD situations. As we shall see shortly, the incentives for non-implementation are not the same across all EU member states. There are countries with strong incentives for non-implementation and countries for whom these incentives are weaker. Those countries for which the incentives are stronger should have lower implementation rates than the other states. Thus, by identifying differences in the incentives to implement PD directives we will be able to derive hypotheses about crossnational differences in implementation rates. Implementing as well as not implementing EC directives causes costs. These costs determine how strong the incentives for a member state are to implement or not to implement a directive. We can distinguish at least four types of costs pertaining to the (non-)implementation of EC directives. Consider once again a clean-air directive as an example: First, a state that implements the directive has to bear the cost of ensuring that the production of the concerning pollutants is cut (costs of implementation). Second, by implementing the directive it abandons the opportunity to spend the resources needed for implementation in a different way, e.g. for national policy programs (opportunity costs of implementation). A state that does not implement the directive, however, runs the risk of being punished for its deviant behavior (costs of non-implementation). Moreover, it risks that the other states, recognizing that they are exploited by a free rider, will also stop cooperating. The state would then have to do without the gains it would have realized when implementing the directive, for example clean air (opportunity costs of non-implementation). Do these different kinds of costs create different implementation incentives for different states, thus explaining differences in implementation rates? (1) Costs of Non-Implementation Until 1992 there were no material sanctions linked to the non-implementation of EC directives. However, non-implementation caused immaterial costs, especially reputational losses. At least the failure to transpose a directive into national law was almost certain to be detected and made public by the Commission. Therefore, a state could hardly afford generally not to implement EC directives, lest it wanted to earn a reputation as "non-implementer" and thereby jeopardize the willingness of the other states to cooperate in the future (see Keohane 1984: 105f.). The TEU, which came into force in 1993, gives the ECJ the possibility to impose fines on member states which attempt to evade implementation of EC legislation. Since a 8 state’s capabilities enter into the calculation of the fine10, the risk of being sentenced by the ECJ deters both small and large member states to the same degree. Therefore, no hypothesis about differences in the implementation behavior of member states can be derived from the costs of non-implementation. These costs create roughly the same incentives for all member states. Since the costs were raised considerably after 1992, one would ceteris paribus expect the implementation rates of all states to go up after 1992. (2) Costs of Implementation The costs of implementation do vary from state to state. They depend primarily on a state’s capacity to implement. It seems fair to assume that affluent states (states with a high per capita income) in general possess a higher capacity to implement directives than less affluent states. Less affluent countries often lack an efficient public administration and tend to have relatively low standards of regulation (Chayes/Chayes 1993: 194; Jacobsen/Weiss 1995: 127; Mosley 1996: 28). While affluent countries may be able to rely on existing bureaucratic resources and build on existing regulations, poorer countries need to "catch up" first, which means they have to bear higher costs to implement the same rules. Higher costs mean stronger incentives to avoid these costs by not implementing directives. Thus, we arrive at the first hypothesis about differences in the implementation rates of EC member states. Hypothesis 1: The less affluent a country is, the lower is its implementation rate with respect to PD directives.11 (3) Opportunity Costs of Implementation Implementation causes opportunity costs because the resources invested in the implementation of a directive cannot be used to other ends, e.g. for domestic policy programs. Non-implementation of PD directives, in contrast, enables a state to both realize gains from implementation by others (e.g. its air will be less polluted) and to spend its money on domestic policy programs, thus further enhancing its national welfare. 10 This is ensured by the "factor n" by which the fine is multiplied. This factor takes into account a country’s GDP and its number of votes in the Council (which is a measure for the size of its population) (Official Journal of the European Communities, C 63 (28 Feb 1997)). This "factor n" is 1 for Luxembourg and 26.4 for Germany; that is for the same infringement Germany would have to pay a fine 26.4 times higher than that of Luxembourg. 11 The ceteris-paribus clause applies to all hypotheses. 9 This possibility to gain twice by non-implementation is more attractive for poorer countries than for affluent ones; because the gain in welfare that can be realized through the non-implementation of directives is more attractive for poor than for affluent countries. This is due to the law of diminishing marginal utility (see Marshall 1961: 93): The more an actor already possesses of a good, the less her utility will increase when she gains an additional unit of the good (i.e. the seventh piece of cake that a person eats will increase her utility less than the first one she had that day). In terms of utility, therefore, an affluent state will gain less from not implementing an EC directive than a poor one. Thus, looking at the opportunity costs of implementation we can again state the expectation that poorer countries will have lower implementation rates with regard to PD directives than more affluent countries (Hypothesis 1). (4) Opportunity Costs of Non-Implementation It will be useful for a state to avoid the cost and opportunity costs of implementing a PD directive only if the other states continue to cooperate despite its own defection. If the other states retreated to defective behavior themselves the original defector would be even worse off than before: Its air would be as polluted as before and in addition the state would have gained a reputation as "non-implementer." This would be even worse than if the state had implemented the directive and taken on the costs and opportunity costs of implementation.12 The risk that a state’s defection leads to the collapse of cooperation and thus to high opportunity costs varies from state to state. Some states can anticipate that their own defection, if detected, would lead the other states to retreat to defective behavior, too. For other states it is fair to assume that their own defection – albeit it will not be welcomed by the others – will not result in the collapse of cooperation. This is due to the fact that in a multi-person PD situation not all actors have to cooperate for the cooperation to be profitable for everyone. Rather "there is some number, k, greater than 1, such that if individuals numbering k or more choose their unpreferred alternative [i.e. cooperation, D.P.] and the rest do not, those who do are better off than if they had all chosen their preferred alternatives [i.e. defection, D.P.], but if they number less than k this is not true." (Schelling 1978: 218) Thus, in a multi-person PD situation cooperative behavior of k actors (the cooperation of a "k-group") is enough to ensure a gain in utility for all actors. Even if there are free riders it is rational for the members of the k-group to continue their cooperation. The k-group, therefore, is "the smallest disciplined group that, though resentful of free riders, can be profitable for those who join (though more profitable for those who stay out)." (Schelling 1978: 218) 12 Otherwise it would have been irrational to agree to the directive in the first place. 10 Snidal (1985b) has demonstrated that this game-theoretic definition can be applied to international politics. He also showed that the composition of the k-group depends on the distribution of capabilities among the states concerned. The basic insight is that there are states that can make substantial contributions to successful cooperation and states that can contribute only little to the overall success. In an extreme case the k-"group" may be comprised by a single state. Proponents of hegemonic stability theory, for instance, argue that in the 1950s and 1960s U.S. capabilities in the world financial system were so superior compared to those of other states that it was rational for the U.S. to supply world trade with certain public goods (capital flow or markets for manufactured goods, for instance) regardless of whether the other states cooperated or not (Kindleberger 1988: 189). Although by free riding other states were able to realize higher gains than the U.S., the utility for the U.S. was still higher than in the case of U.S. protectionism. If there is no hegemon in the system, the same effect can be produced by the cooperation of a small group of big states.13 Snidal (1985b: 604-612) argues that after the decline of U.S. hegemony in the 1970s, Germany and Japan joined the k-group. By cooperating to supply the world market with certain public goods, these three large states were able to produce such large gains that their cooperation was rewarding for them even if all other states had not cooperated but acted as free riders. Incentives to cooperate are distributed asymmetrically between members and nonmembers of the k-group. Non-members can retreat to defective behavior with relatively low risk. Even when they drop out it remains rational for the members of the k-group to continue cooperation. By not cooperating, non-members of the k-group can realize their most preferred outcome in PD situations: They defect while the others cooperate. Therefore, they have high incentives to defect. At the same time, members of the kgroup cannot afford to defect from cooperation. If they did so the cost of cooperation would rise for the remaining members of the k-group and, by definition, these costs would exceed the benefits from cooperation for these states. Being rational actors, they too would stop cooperating and cooperation would collapse. Hence, the opportunity costs of defection for every single member of the k-group are extremely high. Which states are members of the k-group with respect to PD directives in the EC? The exact composition of the k-group will depend on the problem a directive deals with. 13 In theory, this effect can be produced by the cooperation of a large number of small states, too. However, the smaller the group of states that need to cooperate the more likely the k-group will actually form. Hence, the smallest possible k-group is the k-group most likely to form (see Olson 1965). This is the group composed of those states which by themselves can make the largest contribution to cooperation. 11 With respect to the clean-air directive cited above those states that are the biggest polluters in absolute terms would have to be considered members of the k-group. In the context of this article it appears justified to assume that primarily those states with large economic capabilities will be members of the k-group. In general, they will make a particularly important contribution to cooperation in PD situations because all EC directives concern the economic realm. Countries which are economically less powerful will make less important contributions and therefore tend not to be members of the kgroup. The stronger a state is economically, the more likely it will be a member of the kgroup with respect to the implementation of PD directives; and the higher will be its incentives to implement those directives. Hypothesis 2: The larger a state’s economic capabilities are, the higher is its implementation rate with respect to PD directives.14 Note that what has been said about the opportunity costs of non-implementation implies that only states which are not members of the k-group can have sufficient incentives not to implement a PD directive. Non-implementation of a PD directive can increase a state’s utility only if the state succeeds in free riding, that is if the other states implement the directive nonetheless. The derivation of hypothesis 2 implied that countries with large economic capabilities cannot reasonably expect to be successful free riders, because if free riding of a member of the k-group is detected, cooperation will break down. Thus, members of the k-group never can have sufficient incentives not to implement PD directives, even if the absolute costs of implementation might be rather high.15 Hence, the differences in implementation incentives that form the basis of hypothesis 1 are relevant only for those states which are not members of the k-group. Consequently, we can formulate a hierarchy of the independent variables in the institutionalist model: Non-implementation is attractive at all only for those states with relatively small economic capabilities (hypothesis 2). Among these small states the 14 Although hypotheses 1 and 2 seem quite similar they must, nonetheless, be distinguished. The first one rests on the distinction between poor and affluent countries, the second one on the distinction between countries with large and those with small economic capabilities. These distinctions are not identical. There are, for instance small affluent countries on the one hand and small poor countries on the other, Luxembourg being an example for the first, Portugal for the second category. These states will face different incentives according to hypothesis 1: While the costs of implementation and thus incentives for defection will tend to be high for poor Portugal, they will be considerably lower for the significantly more affluent state of Luxembourg. However, both small countries are unlikely to be members of the k-group and therefore both countries also have strong incentives not to implement directives according to hypotheses 2. Only both hypotheses together capture the rational institutionalist model of implementation. 15 The costs of implementation will, by definition, not exceed the gains to be reaped from cooperation. Otherwise it would not have been rational to agree to the directive in the first place. 12 incentives to defect are most pronounced for non-affluent states (hypothesis 1). We will have to keep this hierarchy in mind when testing the model in section 6. 5 Liberalism The institutionalist model introduced in section 4 treats states as unitary, rational actors which consciously decide in favor of or against the implementation of a directive. To treat the state as a unitary actor and as a "black box" whose internal structure can be abstracted from has been a dominant approach in International Relations for decades (Moravcsik 1997: 514). Yet for several years scholars of International Relations have criticized this approach and demanded that a state’s internal characteristics be taken into account when its behavior in the international system or outcomes at the international level are to be accounted for (e.g. Haggard 1991; Milner 1992; Moravcsik 1993: 11-14). Recently, this criticism has become common sense: "Lately no word – apart from 'rethinking' – seems to appear more often in international relations titles than 'domestic.'" (Jacobsen 1996: 93). The second model of implementation to be presented here takes this criticism into account and opens up the "black box" of the state. This liberal model regards implementation not as the result of the cost-benefit calculation of rational states but as the outcome of a more or less complex decision-making process in which a number of societal actors participate with their individual preferences. In order to account for the overall implementation behavior of a state by reference to its internal decision-making procedures we need to single out those characteristics of the decision-making process which influence the implementation decision across all directives. Therefore, in the remainder of this section I will develop a simple model of the decision-making process from which I will then derive two hypotheses about the implementation rates of EC member states. 5.1 Veto Players and Decision-Making The basis of my model is provided by the veto-player approach which George Tsebelis (1995a, 1995b) developed to make domestic decision-making processes comparable across democratic countries and which has been applied successfully to different issues in comparative politics (e.g. Hallerberg/Basinger 1998; Radaelli 1997; Tsebelis 1999).16 Since the model will later be tested with regard to legal implementation I will focus on 16 The approach has also been utilized in the International Relations literature, for instance by Helen Milner (1997: 14, 106f.). 13 legislative decision-making, but the model and the hypotheses will be applicable to other phases of the implementation process as well. Let us start out from a quite simple linear model of the decision-making process. The process basically consists of three phases: In phase one, a delegation participates for the member state in the negotiations about a directive on the European level. In phase two, after a directive was agreed upon, the government of the member state decides whether to introduce a bill into the domestic legislative process in order to transpose the EC directive into national law. In phase three, the actual legislative decision-making takes place, i.e. several actors have to consent to the bill. All actors whose consent is necessary (in any phase) in order for the directive to be implemented will be called veto players: "A veto player is an individual or collective actor whose agreement is required for a policy decision." (Tsebelis 1995a: 293) Two kinds of veto players can be distinguished: institutional and partisan veto players (Tsebelis 1995a: 302-305). Institutional veto players are those whose veto power is laid down in the constitution and who are the main actors in the final phase of the decisionmaking process. Depending on the constitution, they include upper and lower houses of parliament, and/or presidents with legislative powers.17 In contrast, partisan veto players are the main actors in what I have labeled phase two of the decision-making process: the parties which comprise the government. Those parties do not have formal, but they do have de facto veto power. This de facto veto power is in part due to the fact that the government is a gatekeeper in the implementation process. It is the government that decides which directives are introduced into the formal legislative, and hence into the implementation process. To be sure, in West European democracies not only the government can introduce parliamentary bills. But again, de facto most bills in these democracies are formulated by the government (Tsebelis 1995b: 95f.); and it is certainly plausible to assume that all bills which implement international agreements are introduced by the government, which has been responsible for negotiating the international agreement in the first place, and not by members of parliament. Hence, due to its gatekeeping function, government is a veto player in the implementation process. But it is not only the government as a whole that possesses this informal veto power, but every single party inside a government coalition. This is obvious for "minimum winning coalitions" (Riker 1962: 40 et passim), i.e. for coalitions which would lose their parliamentary majority if a single member left the coalition. 17 In the legislative process there can be additional veto players, for instance courts or referenda. Even special requirements regarding the majority with which a bill has to be passed can be regarded as veto "players" (Tsebelis 1995a: 307). I will ignore these veto players here because they play at best a minor role with regard to the implementation of EC directives. 14 Here, it is vital for the coalition to ensure the assent of every member to every policy decision. Otherwise they would run the risk that the outvoted member would leave the coalition, dissatisfied because – albeit itself a part of government – it could not influence government policy. This would imply the loss of the parliamentary majority for the remaining government members and thus probably the loss of governmental power, a loss they will be eager to avoid. But also government coalitions which are not minimum winning coalitions (i.e. oversized and minority coalitions) will seek to ensure the assent of all their members to every policy decision. Obviously, in these cases a coalition is not forged simply to win a majority in parliament. There must be additional political reasons that prompt the parties to form such coalitions. Oversized government coalitions, for instance, are a typical feature of consociational democracies (Lijphart 1977: 25-36). They can be regarded as expression of a societal consensus that government should be formed by a majority as broad as possible (Kaiser 1997: 427). Hence, they are not formed for the larger parties to be able to outvote the smaller coalition partners at will but to ensure that government policies are backed by a majority as large as possible. Therefore, even small members of an oversized government coalition will have veto power. This also holds for minority coalitions. Even in such a coalition it would not make sense to include a party in the government coalition only to outvote it later with regard to specific policy decisions. Therefore, all parties in government coalitions can be regarded partisan veto players (Tsebelis 1995a: 304). Of course, the legislative decision-making process is not only influenced by institutional and partisan veto players. Also a host of societal actors attempt to influence the legislative decision formally and informally. They might indeed be very important for specific policy decisions. Nonetheless, I will exclude them from discussion here for two reasons. First, every societal actor will become active only with regard to specific issues. These issue-specific activities can be excluded from an explanation of the general trend in the implementation behavior of a country. Second, even with regard to single policy decisions it would be difficult to decide which societal actors to treat as veto players according to the definition, that is to judge which societal actors actually could prevent a policy decision. Therefore I will include only those actors which have actual veto power in every single (legislative) implementation decision, that is institutional and partisan veto players. "The assumption is that while the number of veto players may vary by issue or over time, these variations will cancel each other out when applied across several issues for sufficiently long periods of time." (Tsebelis 1995a: 308) To sum up it can be said that the liberal approach regards implementation as the result of a decision-making process that involves several domestic actors. A directive, which has been negotiated on the European level with the participation of a delegation from 15 each member state, will be implemented in a state if and only if all veto players give their assent to implementation. In the legislative process, which determines legal implementation, institutional and partisan veto players can be distinguished. 5.2 Deriving Hypotheses Now that we have a simple model of decision-making we can start to derive hypotheses about the results the decision-making process will produce. The implementation decision will be determined by the preferences of the veto players. For the sake of simplicity I will combine all veto players which hold identical preferences into a single one, because in decision-making it does not matter how many veto players hold a certain preference.18 One can assume that there are veto players which indeed will hold (nearly) identical preferences across (nearly) all implementation decisions. For instance, one might assume that in general a majority government will be supported by the parliamentary majority (Beyme 1982: 369). In this case it would not make sense to regard parliament as an additional veto player; because in general parliament will agree to a decision to which the members of the government coalition have already agreed. In these cases Tsebelis’ "absorption rule" (Tsebelis 1995a: 309-311) applies: Institutional veto players with identical majorities are treated as one institutional veto player. Furthermore, an institutional veto player will be regarded as a distinctive veto player only if the parties comprising government cannot determine its decisions, i.e. if the government parties have no majority in this institution. Partisan veto players, however, will in general be treated as distinctive veto players, because only in very rare cases will several parties hold nearly identical preferences over a broad range of issues (see also Tsebelis 1995a: 306f.). Hence, in all that follows I will take into account only those veto players which hold different preferences. One last concept is needed to derive hypotheses from the liberal model: the concept of the "win set." This concept, which is taken from rational choice theory, was introduced into the International Relations literature by Robert Putnam (1988). With regard to an international negotiation the win set of a state participating in these negotiations is the set of those international agreements which would gain the necessary majority for ratification in the state concerned (Putnam 1988: 437). To translate this for our purposes: With regard to the negotiations on an EC directive the win set of a member state is the set of all possible agreements (directives) that would be implemented in that 18 As said above, the implementation decision is not a majority decision – it is a unanimous decision to which all veto players have to agree. Thus the number of veto players having a certain preference will not matter as long as this preference is represented at least by one actor. 16 state, i.e. that would gain the support of all veto players.19 Consider an agriculture directive: Germany’s win set with respect to this directive includes all possible results of the negotiations that would gain the support of all German government parties plus the support of the German upper and lower chamber of parliament provided the government parties do not have a majority there. Now we have finally reached the point where we can derive hypotheses from the model: A directive will be implemented in a member state if and only if it is part of the win set of the respective member state. The probability that it is part of the win set is the higher (a) the larger the win set is; and (b) the better the national delegation negotiating the directive at EC level is informed about the contents of the domestic win set. I will briefly discuss this and formulate two hypotheses. 5.2.1 Size of the Win Set The larger the win set is, the more directives it contains. Hence, the larger the win set the more likely a directive that has been agreed upon in the Council is part of this win set; that is the more likely is the implementation of the directive by the member state (see also Putnam 1988: 439). Therefore, we need to identify those factors which influence the size of the win set across all directives in order to formulate hypotheses about the implementation rates of EU member states. Tsebelis (1995a) considers two variables that influence the size of the win set: the distance between the preferences of the veto players; and the number of veto players.20 The closer the preferences of different veto players are to one another, the easier it is to come to an agreement; the further they lie apart, the harder it is to reach agreement. Therefore, the win set becomes larger, the smaller the distance between the preferences of the veto players is. The distance between the veto players’ preferences, however, depends to a large extent on issue-specific factors. Two veto players may strongly disagree with regard to an agriculture directive but have very similar preferences with regard to an equal-pay directive. Since this variable varies with respect to different 19 For a formal definition of the win set, see Shepsle/Weingast (1987: 90); but note that they regard ratification as a majority decision whereas in our case the assent of all veto players is needed. 20 Tsebelis also discusses a third variable, namely party cohesion. He makes it very clear, that party cohesion is not identical with party discipline (which can be measured by voting patterns in parliament, for instance) (Tsebelis 1995a: 311f.), but he does not give any hint as to how to operationalize or measure party cohesion. I do not see any possibility to operationalize this variable and, therefore, exclude it from discussion. 17 issues, it can be excluded from a discussion of those variables that determine the general trend of the implementation behavior of EU member states. Yet the second factor influencing the size of the win set, namely the number of veto players, is important for the problem under investigation. The more veto players there are, the smaller is the win set. For the agreements that n+1 veto players can arrive at, can only be a subset of what n veto players could agree on (Tsebelis 1995a: 297). Or to put it differently: If three actors A, B, and C have to agree on a common denominator, they can only agree on a subset of those positions A and B could agree on. Since the number of institutional and partisan veto players is not issue-specific, we can directly derive the first liberal hypothesis: Hypothesis 3: The larger the number of veto players is in a country, the lower is its implementation rate. 5.2.2 Information about the Domestic Win Set Hypothesis 3 could remain the only liberal hypothesis if the bargaining of a directive at the European level on one hand and national implementation on the other were two independent processes. Yet the bargaining and the implementation of a directive are not completely independent. Rather, bargaining is a two-level process in which the national delegation serves as a link between the international and the domestic level (Putnam 1988: 434). The delegation attempts to influence the content of the directive with a view to what can be implemented at the domestic level. The delegation’s aim is to formulate the directive so that it is contained in the domestic win set. The probability that the delegation succeeds depends not only on the size of the win set; but also on how well the delegation is informed about the contents of the win set. The better the delegation is informed about the preferences of the veto players (which determine the contents of the win set), the better it is able to bargain in such a way that the resulting directive will be contained in the domestic win set. Lisa Martin (1995), for instance, has shown that the fact that the Danish parliament is consulted before the Danish delegation in the Council takes its position is conducive to Danish legal implementation rates. This institutional mechanism enables the Danish government to anticipate domestic opposition and to take it into account when it engages into bargaining at the European level. The literature on implementation in general regards it as conducive to implementation when those who are responsible for implementing a 18 policy decision can participate already in the making of the respective policy decision (see Andresen et al. 1995: 54; Mosley 1996: 26). Hypothesis 4: The better the national delegation of a state is informed about the preferences of the domestic veto players, the higher is the implementation rate of this state. 5.3 Putting the Hypotheses Together The final step in building a liberal model of implementation is to clarify how the two independent variables interact to determine the implementation behavior of a state. They are not completely independent. Consider a country in which the national delegation possesses perfect information about the preferences of all national veto players. Since this delegation knows the exact contents of the win set it will agree only to such directives in the Council that actually lie within the win set; i.e. to directives whose national implementation is guaranteed. The size of the win set does not matter. Consider now a country in which the delegation has no information whatsoever about the national veto players’ preferences. In this case it is completely due to chance whether the directive passed at the European level lies within the national win set. The larger the win set, the more likely the directive is contained within the win set. In this case the size of the win set is all that matters for the probability of implementation. Hence, how well the national delegation is informed about the veto players’ preferences has on one hand an independent effect on the probability of implementation (the better the delegation is informed, the more likely is implementation). On the other hand, it conditions the effect which the number of veto players has (the better the delegation is informed the less the number of veto players matters). Two expectations result: (a) The better the national delegation in a country is informed about the national veto players’ preferences, the higher is this country’s implementation rate (hypothesis 4). (b) The number of veto players matters only in those countries in which the national delegation is not well-informed about the veto players’ preferences. Among these countries implementation rates will vary according to hypothesis 3. 6 Preliminary Test of the Hypotheses 6.1 Dependent Variable Let us now turn to a preliminary test of the hypotheses derived in the preceding sections. I will use data on the legal implementation rates of EC countries in the three-year period 19 immediately following 1989 (1990 to 1992). Due to limitations of space I can only briefly give some reasons as to why I choose these data to test the hypotheses. I will focus on legal implementation because these records are the only reliable data readily available, as noted in section 2. Furthermore, I will pool the implementation rates of three years, a period long enough to exclude singular events that might distort a country’s implementation rate in a single year. Both models assume that a state has to implement only directives to which it itself has explicitly agreed.21 This makes it necessary to focus on the period immediately preceding the entry into force of the Maastricht Treaty. Before the TEU came into force the Council of Ministers dominated decision-making in the EC. Directives were passed according to the "consultation" and "cooperaton" procedures (see Nugent 1999: 50, 360-366; Weindel 1996: 75-78). In both procedures it is the Council of Ministers that makes the final decision. The European Parliament (EP) is consulted and might even ask for changes in the text of a directive, but eventually the Council can overrule the EP’s decision. Thus, until 1992 the Council and through the Council the governments of the member states were the most important actors when it came to deciding about EC directives. In most cases directives were passed unanimously22 which implies, in turn, that every member state was able to influence to a considerable extent the contents of the directives which it had to implement later on (Edwards, G. 1996: 130). This changed after 1992, because the Maastricht Treaty extended qualified majority voting in the Council to a wider range of issues and also extended the rights of the EP which received veto powers with regard to some issues (see Article 251 [189b] TEC). Therefore, until 1992 EC directives can be regarded as intergovernmental agreements, whereas their supranational character became more prominent afterwards. Since both models assume that national delegations are able to influence the contents of a directive decisively, it is appropriate to test the hypotheses with data that excludes the directives passed after 1992. Although the national implementation rates from 1993 to 1997 cannot be used for rigorous hypothesis testing, I will not completely disregard this data, but use it as additional evidence in section 7. The institutionalist hypotheses 1 and 2 apply only to PD directives. Therefore it may seem necessary to look not at overall implementation rates, but only at the 21 In the institutionalist model this assumption makes it possible to exclude suasion games from discussion. In the liberal model the underlying assumption is that the national delegation can co-decide on the contents of a directive on the international level. 22 Only if the national delegations were not able to find a consensus neither in the working groups nor in the Committee of Permanent Representatives nor in the Council of Ministers itself and if the directive was concerned with one of those (few) issues on which the Council according to the TEC could decide by way of qualified majority voting, a vote was taken in the Council (Hayes-Renshaw/Wallace 1997: 48; Nugent 1999: 167-170). 20 implementation of PD directives to adequately test these hypotheses. But how could one single out PD directives from the more than one thousand directives that were in force at the beginning of the 1990s? Fortunately, it is possible to avoid this demanding task and to preliminarily test institutionalism using data on the implementation of all directives without having to single out PD directives. Rational institutionalism expects variation in implementation rates only with regard to PD directives. It expects no variation in the implementation of coordination directives, because the incentives to implement those directives are overwhelming for every state. Thus, from a rational institutionalist viewpoint, overall variation in implementation rates can only be due to variation in the implementation of PD directives. Thus we can use overall implementation rates to assess differences in the implementation of PD directives.23 6.2 Independent Variables There are two independent variables in each model, which have to be operationalized in order to test the hypotheses: affluence and economic capabilities for rational institutionalism; the number of veto players and the information the national delegation has about the domestic win set in the liberal model. For the operationalization of the institutionalist variables I rely on indicators that are widely used in the literature. I measure a country’s affluence by reference to its GDP per capita, its economic capabilities by reference to its GDP. The operationalization of the liberal independent variables is somewhat more complex. I count as a veto player: every party in government; every individual or collective actor with absolute veto power in the law-making process who is not a member or whose majority are not members of the parties in government. To determine how well the delegation is informed about the win set I look at national decision-making procedures: 23 The effect that can be detected with regard to overall implementation rates, however, will be smaller than the effect the independent variables have on the implementation of PD directives (see King et al. 1994: 130-132). Consider a simple numerical example: Two states A and B have to implement 100 directives, 50 coordination directives and 50 PD directives. Both of them implement all 50 coordination directives. But while state A also implements all 50 PD directives, poorer and smaller B implements only 25 of these directives, i.e. only 50% of those implemented by A. Yet with regard to overall implementation rates, A (100 directives implemented) has implemented only 25% more directives than B (75 directives implemented). The smaller the proportion of PD directives in the sample, the more pronounced is this effect. To ensure that an effect can be detected at all, there must be more than just a negligible few PD directives among all EC directives. From a rational institutionalist viewpoint there is reason to believe that this is indeed the case. PD is regarded as one of the most important situation structures in international cooperation. Robert Keohane’s After Hegemony (1984) is based on this situation type and a number of other rational institutionalist studies of international cooperation are based solely on PD (see Grieco 1988b: 604; Hasenclever et al. 1997: 109f.). PD is especially well-suited to model problems of collective goods and free trade (Stein 1982: 307f.), two issue areas which are particularly prominent in the EC context. 21 The delegation has "comprehensive information" about the win set if there are national institutions which make it possible for all veto players to voice their preferences with regard to a pending EC decision before the national delegation formulates its position for the negotiations in Brussels; it has "some information" if there are institutions in which a majority of veto players can voice their preferences; otherwise it has "no information". 6.3 Test The data for the independent and dependent variables are reproduced in detail in the appendix. For the preliminary test of the hypotheses I use simple tables to compare the values of the independent and dependent variables. To assess the independent effect of each variable, I will additionally use a multiple linear regression. 6.3.1 Rational Institutionalism Table 1 contains the data for a test of the institutionalist hypotheses. According to the model the countries ranking highest in the table (those with a large share of the EC GDP) should have the highest implementation rates, because they will most likely belong to the k-group. Countries ranking lower should have lower implementation rates; within this latter group those countries with the lowest per capita income should have the lowest implementation rates, according to hypothesis 1. There are two cases that clearly do not conform with institutionalist expectations: Italy and Denmark. Italy has large economic capabilities compared to those of the other EC countries and it also has a fairly large GDP per capita but its implementation rate is exceptionally low. Thus, Italy’s implementation behavior cannot be accounted for by the institutionalist model. On the other hand, Denmark has the highest implementation rate, although it is one of the smallest EC countries. One could argue that its large per capita income contributes to its high implementation rate; however this cannot explain why Denmark’s implementation rate is so much higher than that of every other small country and even higher than those of the four big countries of the EC. Excluding these two "outliers" for the moment, the pattern in the table conforms fairly well with the expectations of hypothesis 2. The large countries have the highest implementation rates (averaging 93.6%), while those of the small countries are consistently lower (averaging 89.5%).24 The expectations of hypothesis 1, however, are hardly confirmed. To be sure, the two small countries with the lowest GDP per capita 24 η² = .530, p = .017. 22 (Portugal and Greece) do have the lowest implementation rates. But there is no consistent pattern for the small countries with higher GDP per capita. Table 1: Rational Institutionalist Variables and Implementation Rates Country Share of EC GDP per capita Implementation GDP (Purchasing rate Power Parities per capita) GER 26.40% 16,732 92.7% FRA 19.00% 16,951 93.8% ITA 17.69% 15,743 82.4% UK 15.52% 14,896 94.2% SPA 8.15% 11,734 91.9% NET 4.57% 15,499 92.1% BEL 3.14% 16,335 90.1% DEN 2.05% 16,172 96.5% GRE 1.38% 9,208 87.5% POR 1.20% 9,451 86.2% IRE .74% 11,555 90.3% LUX .17% 22,903 88.5% avg. implementation rate of "big countries" (GER, FRA, ITA, UK): 90.8% avg. implementation rate of "small countries": 90.4% Thus, this more or less informal evaluation of the institutionalist model suggests that a country’s economic capabilities do have an effect on its implementation rate, while its affluence does not have a visible effect. That is, it is most important whether a country belongs to the k-group or not. Italy’s and Denmark’s implementation behavior cannot be accounted for by the institutionalist model. 6.3.2 Liberalism Table 2 contains the data for the test of the liberal model. The expectation is that the countries ranking highest in the table (national delegation is well-informed about the win set) have the highest implementation rates, those ranking lowest have the lowest rates and those in the middle take the middle ground. Within the first group, the number of veto players should have no visible effect on implementation rates, whereas among the countries at the bottom of the table those with higher numbers of veto players should 23 have clearly lower implementation rates. For the countries in the middle group ("some information"), some effect of the number of veto players should be visible, too. Table 2: Liberal Variables and Implementation Rates25 Country Information about the National Win Set Number of Veto Players Implementation Rate DEN comprehensive 3.3 96.5% UK comprehensive 1 94.2% FRA comprehensive 2.6 93.8% GER comprehensive 2.9 92.7% NET comprehensive 2 92.1% BEL comprehensive 4.6 90.1% POR comprehensive 1.6 86.2% SPA some 2 91.9% IRE some 3 90.3% GRE none 2.1 87.5% ITA none 4.4 82.4% means comprehensive 92.2% some 91.1% none 85.0% (Note: Luxembourg has 2 veto players. Since data on how well the national delegation is informed about the domestic win set is not available, Luxembourg was excluded from the test of the liberal model.) The only exceptional case with respect to the liberal model is Portugal. In Portugal the national delegation is well-informed about the national win set, and there are even only very few national veto players, but the Portuguese implementation rate is one of the lowest in the Community. However, even if Portugal is not excluded, the overall "fit" of the liberal model is quite astounding: Those countries in which the national delegation is well-informed about the national veto players’ preferences, on average, have the highest implementation rates. Those in which the delegation has no information about the win set, clearly have the lowest implementation rates, and those in which there is only some information about the national veto players’ preferences take the middle ground.26 25 I have discussed the values for the variable "information about the national win set" at length elsewhere (Peters 1999: 66-89). My main sources included: Grottanelli de Santi (1992), Ioakimidis (1994), Laffan (1991), Nehring (1992), Norton (1996), Rambaud (1993), Rometsch/Wessels (1996), and Sørensen/Væver (1992). For details on the number of veto players, see appendix. 26 η² = .509, p = .058. 24 Furthermore, in the countries belonging to the first group, the number of veto players seems not to systematically affect implementation rates. In the other groups, those countries with most veto players have the lowest implementation rates. Of course, this cannot be regarded as a final proof of the liberal hypotheses. The number of cases in the categories "some information" and "no information about the national win set" is simply to small. But the evidence for the liberal model is nonetheless more convincing than that for the institutionalist model. 6.3.3 Both models For a final assessment of the explanatory power of the variables I enter them into a multiple regression. The multiple regression, however, does not reflect the hierarchical relations between the independent variables as specified in the different models. Moreover, the results can hardly be statistically significant due to the small number of cases (n = 12) and the large number of independent variables. Therefore the regression cannot provide a conclusive test of the models either, but it will serve to further strengthen or weaken our confidence in the explanatory power of the two models. The results of the regression are reported below.27 Unstandardized Coefficient B Std. Error (Constant) 83.932 6.528 Number of veto players -.430 1.038 Comprehensive information (dummy) 6.903 Some information (dummy) Log GDP 27 Standardized Coefficient Beta t Significanc ep 12.857 .000 -.119 -.414 .693 2.931 .863 2.355 .057 6.133 3.677 .615 1.668 .146 1.004 2.118 .137 .474 .652 To reduce the number of independent variables, I exclude GDP per capita from the regression. According to the models this should be the least important of the independent variables. I use log(GDP) as variable representing a country’s economic capabilities, because the relationship between a country’s GDP and its implementation rate is non-linear as a scatterplot reveals which is not reproduced in this paper. 25 The results of the regression once again highlight the importance which national decision-making procedures have: A country in which the national delegation is wellinformed about the national veto players’ preferences has an implementation rate which is on average 6.9 percentage points higher than that of a country in which the delegation has no information about the win set. This effect is the only statistically significant effect in the regression. As predicted by the liberal model, the number of veto players has (a) a much weaker, and (b) a negative effect on a country’s implementation rate. Not surprisingly, this effect has no statistical significance in the regression. The effect of GDP also is statistically insignificant. This underlines that the liberal variables are more important for explaining overall implementation rates than those of the institutionalist model. It does not imply, however, that the institutionalist model completely fails at explaining implementation behavior. Note that, as mentioned above, the strategy to test those hypotheses which apply only to PD directives using data on all directives produces statistical results that might seriously underestimate the actual effect that the independent variables have on the implementation of PD directives. There is additional evidence which leads one to suspect that both models contribute to an explanation of implementation behavior. Those countries which are outliers for one model are explained, but extreme cases in the other model. This suggests that an extreme value on one of the independent variables overshadows the effect that the other independent variables have. Consider Italy: This country has an exceptionally low implementation rate although its economic capabilities and its per capita income would lead us to expect a fairly high implementation rate. From the liberal viewpoint the extremely low implementation rate is hardly surprising: In Italy the national delegation has no information about the national win set and on top of that, Italy has one of the highest numbers of veto players. The exceptionally high implementation rate of Denmark, which is the second country not accounted for by institutionalism, is also explicable by reference to the liberal model. Denmark is the only country in which the national delegation not only is informed about the national veto players' preferences but in which these veto players have actual veto power before a directive is negotiated at the EC level. The Folketing, Denmark's parliament28, has a European Affairs Committee with exceptional institutional powers. The government has to inform the committee about all directives which are to be negotiated at the European level. In the committee the responsible minister explains the government position and takes notice of the committee's position. Although de jure the committee can formulate a binding mandate for the negotiations only in special cases (Sørensen/Væver 1992: 5), the Danish government usually follows the position of the committee (Arter 1996: 111). Since 28 The Danish parliament acts as a veto player with respect to the implementation of EC directives because during the period covered here Danish governments were minority governments. 26 coordination among members of government is also very strong in Denmark (Pedersen 1996: 199-202) all national veto players have de facto veto power not only when it comes to implementing a directive but already when the Danish position on a directive in the Council is formulated. Thus, not only is the Danish delegation well-informed about the domestic win set, it is even obliged to negotiate in such a way that the resulting directive will fall within this win set. This way high implementation rates are guaranteed from a liberal point of view. Therefore, in the Danish case, as in the Italian case, an extreme value on the liberal independent variables overrules the effect of the other independent variables, thus making Denmark an outlier for the institutionalist model. Finally, Portugal cannot be accounted for by the liberal model: Although the national delegation is well-informed about the win set (and although there are only few veto players in Portugal), Portugal has a very low implementation rate. This in turn can be explained by the institutionalist model. Portugal has both a very small GDP and a very small GDP per capita relative to the other EC member states. This very unfavorable economic situation might overrule the effect of the relatively favorable structures of national decision-making and account for the low Portuguese implementation rate. Taken together, there is evidence that both rational institutionalism and liberalism account for the implementation behavior of EC member states. 7 Summary of Findings and Policy Implications Why then do EC member states fail to implement EC directives? Interests and structures of national decision-making account for non-implementation. To some degree nonimplementation may be explained by a state’s interests as a voluntary decision of the state against the implementation of directives due to a rational calculation of costs and benefits. Especially for smaller states (that is, for those states which are likely not to belong to the k-group) there may be incentives to act as free riders with regard to certain directives. However, the better part of non-implementation can be explained as "nonvoluntary defection" (Putnam 1988: 438f.): The national delegation negotiates a directive with the intention to actually have it implemented later by domestic actors. However, national veto players refuse to implement the directive because it does not reflect their preferences adequately. This gap between the intentions of the national delegation and the decision of national veto players is most often due to the fact that the national delegation is not appropriately informed about national veto players’ preferences before it takes its stance in the negotiations at EC level. Where the national delegation has no comprehensive information about the veto players’ preferences, a high 27 number of veto players has additional detrimental effects on national implementation rates. Now that (some of) the causes of non-implementation have been anlayzed we can return to the opening question: How can implementation rates be improved? Since legal implementation rates are higher than or only little below 90% some might ask why one should aim at increasing implementation rates at all? But to infer from high legal implementation rates that there is no need for a further improvement of the implementation of EC legislation would be premature. EC organs have repeatedly pointed to the need to improve implementation. Furthermore, one can assume that high legal implementation rates mask much lower administrative implementation and compliance rates. Since both models presented here are general enough to apply to all phases of the implementation process, it seems worthwhile to translate the findings into some policy advice. According to the results presented here, there are basically two ways of improving implementation rates: by changing states’ cost-benefit calculations and by changing the institutional structures of the domestic decision-making process. For the EC institutions, which can hardly influence national decision-making structures, it will be most rewarding to focus on the first of these two possibilities, that is to influence national cost-benefit calculations. It is true that the most important variable affecting national calculations of costs and benefits, namely membership in the k-group which is determined by a country’s economic capabilities in relation to those of other EU countries, cannot be changed by European institutions. The clear differences in the economic capabilities of the member states cannot be completely leveled. Thus, there will always be incentives for smaller states to choose free-riding strategies with respect to PD directives. However, the EC can counter these incentives by raising the cost of non-implementation. Indeed, this was the main strategy employed in the past years. In Maastricht the possibility to impose fines on states which did not fulfill their implementation obligations was introduced into the TEC. Thus the costs for nonimplementation were raised. If states are sensitive to costs, as the institutionalist model suggests, implementation rates should have gone up since then. Moreover, the institutionalist model would lead us to expect that the rise was especially pronounced in those countries which had had large incentives for free-riding before, that is for small states not belonging to the k-group. For the members of the k-group the incentives for non-implementation have always been very low so that they will be less affected by a rise in the costs of non-implementation. The empirical evidence supports these 28 expectations. There are only two states whose implementation rates did not rise after the Maastricht Treaty had entered into force, namely Germany and France, both members of the k-group.29 Excluding Italy once again, which as we have seen above is a special case for the institutionalist model, those countries with the lowest rise in implementation rates after 1992 are the United Kingdom (the smallest of the three "big" states considered here) and Spain (the largest of the "small" states). The smaller states all show considerably larger rises in their implementation rates. This may serve as further evidence for the institutionalist model. It also illustrates that the strategy of raising the costs of non-implementation appears to have been successful. The Commission and the ECJ now seem to have a more effective means at their hands to ensure that EU member states stick to their obligations. Yet the results of this study also suggest that such a European strategy can only be partially successful, because a state’s utility function is only one factor among many which influence its implementation behavior. Domestic decision-making procedures have an even more important effect on implementation rates. To be sure, there are ways in which European institutions could counter detrimental effects of unfavorable domestic decision-making procedures. It would be possible, for instance, to directly include national veto players’ preferences in the formulation of directives effectively bypassing national delegations. For instance, there might be hearings with national interest groups (which are important veto players in the compliance phase of implementation) before the Commission formulates the draft directive. Also the EP could serve as a channel for national societal veto players to influence the European decision-making process, at least for those subject matters where the EP has an effective co-decision power. The participation of the Committee of Regions in the European decision-making process could have positive effects on implementation, too, because the regions are in many countries important veto players during the phase of administrative implementation. In these ways, European institutions could help to adjust the contents of directives to the national veto players’ preferences. This can be expected to have positive effects on national implementation rates. But the most obvious way of including national veto players into decision-making would be via national decision-making procedures that is by giving the veto players a say already during the formulation of national positions in the Council. This cannot be accomplished by EC institutions but is up to the member states themselves. In several member states there have indeed been considerable changes in this respect. Especially national parliaments were able to increase their leverage over the formulation of national positions in the Council (see Hayes-Renshaw/Wallace 1997: 226, and the 29 See appendix for the exact figures. 29 contributions to Norton 1996). While parliaments themselves are veto players only in some special cases (namely with respect to legal implementation if the government has no parliamentary majority), they also serve as access channels to the national decisionmaking process for interest groups and other societal actors, so that a better integration of parliaments into national decision-making prior to negotiations at EC level will also serve to improve compliance rates. The liberal model suggests that it is most important to provide the national delegation with comprehensive information about national veto players’ preferences before negotiations start in Brussels. If this can be achieved the number of veto players will lose its importance and implementation rates will improve considerably even if the number of veto players in a state is large. This will also serve to address the "democratic deficit" in the Union. By improving the integration of those affected by policy decisions into the formulation of these decisions (be it through European or through national channels) democratic control over policy is improved.30 Hence, there is a multitude of possibilities to improve implementation rates and thus to make European policies work better. Particularly promising are the use of material sanctions against non-implementers and better incorporation of those who are responsible for implementing a policy decision into the making of this decision. Note, however, that there is an important trade-off with respect to the latter option. The more veto players are involved in the formulation of a policy decision, the more difficult it is to reach agreement about this decision. That is, by involving the veto players in the formulation of a decision the implementation problem is shifted to the formulation phase. The more a delegation is restricted to a certain position in the negotiations by the domestic veto players, the less flexible it will be on the European level and the harder it will be to strike a bargain. Denmark, which is the only member state in which parliament can actually co-decide about the national position in the Council, is a case in point. The Danish delegation is often the least flexible one in Council negotiations and it often explains its inflexibility by reference to the domestic decision-making procedures which put firm constraints on it (Hayes-Renshaw/Wallace 1997: 232). One can easily imagine that decision-making in the Council would become virtually impossible if all member states duplicated the Danish decision-making procedures. Thus, there is a trade-off between improving implementation and deteriorating efficiency of decision-making in the Council. The EC and its member states have to strike a careful balance between a functioning European bargaining system on the one 30 This is noted by Lisa Martin (1995) with respect to the access to European decision-making through national parliaments. 30 hand and acceptable implementation rates on the other. If they prefer to improve implementation rates it might be necessary to accept that agreement on further progress of European integration will become more difficult to reach. 31 Appendix Implementation Rates Legal Implementation Rates 1990-1997 (Percent) BEL DEN FRA GER GRE IRE ITA LUX NET POR SPA UK 1990 91.7 96.6 93.5 95.3 84.5 91.2 81.7 90.1 93.1 84.0 93.5 94.6 1991 88.0 97.0 95.0 92.9 89.6 88.6 76.5 86.9 90.0 85.5 92.0 94.7 1992 90.5 96.0 93.0 89.9 88.3 91.1 89.1 88.4 93.1 89.2 90.2 93.2 1993 90.7 95.4 89.7 88.9 88.1 88.7 88.9 90.7 92.4 89.2 90.1 92.3 1994 89.7 97.6 92.3 91.0 86.7 91.9 88.4 93.7 93.7 89.4 91.2 96.5 1995 89.5 97.9 92.7 92.9 89.8 92.8 88.7 94.2 97.2 90.0 93.1 95.1 1996 92.4 98.2 91.8 93.5 91.1 93.0 89.6 93.4 97.3 91.8 94.7 94.2 Sources: EC Commission 1991-1998 Change in Implementation Rates Change Implementation Implementation Rate Rate 1993/96 as 1990-1992 1993-1996 compared to 1990/92 (%) (%) BEL DEN FRA GER GRE IRE ITA LUX NET POR SPA UK 90.1 96.5 93.8 92.7 87.5 90.3 82.4 88.5 92.1 86.2 91.9 94.2 90.6 97.3 91.6 91.6 88.9 91.6 88.9 93.0 95.1 90.1 92.3 94.5 .57% .77% -2.34% -1.23% 1.67% 1.44% 7.85% 5.14% 3.34% 4.49% .41% .38% Change 1993/96 as compared to 1990/92 (Percentage Points) .5 .7 -2.2 -1.1 1.5 1.3 6.5 4.5 3.1 3.9 .4 .4 1997 91.8 97.0 93.6 93.6 92.8 94.1 92.4 94.2 96.4 93.5 95.1 94.7 32 Data for the Independent Variables of the Institutionalist Model Gross Domestic Product (in current prices, billion ECU) BEL DEN FRA GER GRE IRE ITA LUX NET POR SPA UK 1990 152.6 101.7 941.5 1182.2 65.3 35.9 861.2 8.1 223.4 53.1 387.5 769.6 1991 160.6 104.7 971.7 1391.5 72.1 37.5 931.1 8.8 234.8 61.8 427.6 818.7 1992 171.7 109.6 1022.1 1522.3 75.6 40.4 941.7 9.8 248.9 71.1 446.0 809.3 1993 180.8 115.2 1066.8 1630.9 78.6 41.5 842.0 11.0 267.3 70.1 408.6 805.9 1994 193.6 123.1 1122.6 1725.3 82.5 45.4 855.6 12.3 284.0 71.5 407.1 857.1 1995 205.9 132.5 1176.2 1845.2 87.4 49.2 831.4 13.3 302.5 76.9 428.1 841.6 1996 208.4 137.7 1217.5 1854.0 96.8 55.3 956.5 13.8 309.2 81.9 462.7 902.5 Source: Eurostat 1997: 208 Gross Domestic Product per Capita (Purchasing Power Standards per capita) BEL DEN FRA GER GRE IRE ITA LUX NET POR SPA UK 1990 15,313 15,302 16,114 17,046 8,539 10,583 14,886 21,197 14,832 8,709 10,966 14,527 1991 16,293 16,474 17,163 16,071 9,249 11,532 15,895 22,934 15,521 9,568 12,031 14,708 Source: Eurostat 1997: 210 1992 17,399 16,739 17,575 17,080 9,837 12,551 16,449 24,577 16,144 10,076 12,206 15,453 1993 18,025 17,775 17,307 17,142 10,223 13,191 16,126 26,103 16,510 10,596 12,354 15,705 1994 18,929 19,049 17,920 18,325 10,799 14,706 17,059 28,074 17,447 11,197 12,668 16,406 1995 19,337 20,023 18,544 19,063 11,322 16,022 18,304 29,134 18,473 11,585 13,230 16,516 1996 20,424 20,889 19,307 19,656 11,773 18,106 19,075 30,639 18,990 12,252 13,971 17,958 33 Data for the Independent Variables of the Liberal Model (1990-1992) Country government coalition (cabinet) government government majority in majority in lower house? upper house? BEL until Aug 1991: CVP, PSC, SP, PS, VU since Sep 1991 CVP, PSC, SP, PS until Nov 1990: KF, V, RV since Dec 1990: KF, V until May 1991 PS May 1991-Apr 1992: PS, GE since Apr 1992: PS, MRG since 1982: CDU/CSU, FDP CDU/CSU, FDP yes number of number of partisan additional veto players institutional veto players 5 0 yes 4 0 no 3 1 no 2 1 no 1 1 no 2 1 no 2 1 until Sep 1990: 2 no Oct 1990 until 2 Dec 1990: yes 1 CDU/CSU, FDP yes 1 until Mar 1990: ND, PASOK, KP since Apr 1990: ND since Jul 1989 FF, PD Jul 1989-Apr 1991 DC, PSI, PRI, PSDI, PLI since Apr 1991: DC, PSI, PSDI, PLI since Jul 1989: CSV, LSAP since Nov 1989: CDA, PvdA until Sep 1991: PPD/PSD, CDS yes since Jan 1991: 2 no 3 0 no 1 1 no 2 1 3 4.4 DEN FRA GER GRE IRE ITA LUX NET POR SPA UK yes yes number of veto players 1990-1992 (avg.) 4.6 3.3 2.6 2.9 0 2.1 yes yes 5 0 yes yes 4 0 yes 2 0 2 yes 2 0 2 yes 2 0 1.6 1 0 1 1 2 1 0 1 since Oct 1991: yes PPD/PSD since Nov 1989: no PSOE since 1979: yes CP no upper house upper house has no absolute veto power Sources: European Political Data Yearbook 1992, 1993, 1994; where necessary, data was amended by reference to Ismayr (1997). 34 References Commission of the European Communities: 1991: Eighth Annual Report on Monitoring the Application of Community Law (1990), KOM(91) 321 final. Commission of the European Communities: 1992: Ninth Annual Report on Monitoring the Application of Community Law (1991), KOM(92) 136 final. Commission of the European Communities: 1993: Tenth Annual Report on Monitoring the Application of Community Law (1992), KOM(93) 320 final. Commission of the European Communities: 1994: Eleventh Annual Report on Monitoring the Application of Community Law (1993), KOM(94) 500 final. Commission of the European Communities: 1995: Twelfth Annual Report on Monitoring the Application of Community Law (1994), KOM(95) 500 final. Commission of the European Communities: 1996: Thirteenth Annual Report on Monitoring the Application of Community Law (1995), KOM(96) 600 final. Commission of the European Communities: 1991: Fourteenth Annual Report on Monitoring the Application of Community Law (1996), KOM(97) 299 final. Commission of the European Communities: 1998: Fifteenth Annual Report on Monitoring the Application of Community Law (1997), KOM(98) 317 final. European Political Data Yearbook 1992-1994. Eurostat 1997: Eurostat-Jahrbuch: Europa im Blick der Statistik, Luxemburg: Amt für Amtliche Veröffentlichungen der Europäischen Gemeinschaften. Andresen, Steinar, Jon Birger Skjærseth and Jørgen Wettestad 1995: Regime, the State and Society: Analyzing the Implementation of International Environmental Commitments (Working Paper, WP-95-43), Laxenburg: IIASA. Arter, David 1996: The Folketing and Denmark's 'European Policy': The Case of an 'Authorising Assembly', in: Philip Norton (ed.): National Parliaments and the European Union, London: Frank Cass, 110-123. Beyme, Klaus von 1982: Parteien in westlichen Demokratien, Munich: Piper. Brenner, Michael 1996: Die Umsetzung der Richtlinien über öffentliche Aufträge in Deutschland, in: Jürgen Schwarze (ed.): Das öffentliche Auftragswesen in der EG, Baden-Baden: Nomos, 23-45. Chayes, Abram and Antonia Handler Chayes 1993: On Compliance, in: International Organization 47: 2, 175-205. 35 Duina, Francesco 1997: Explaining Legal Implementation in the European Union, in: International Journal of the Sociology of Law 25: 2, 155-179. Edwards, Geoffrey 1996: National Sovereignty vs Integration? The Council of Ministers, in: Jeremy J. Richardson (ed.): European Union: Power and Policy-Making, London: Routledge, 127-147. Edwards, Vanessa 1996: The Insider Dealing Directive and Its Implementation in the United Kingdom, in: Maastricht Journal of European and Comparative Law 3: 3, 287313. From, Johan and Per Stava 1993: Implementation of Community Law: The Last Stronghold of National Control?, in: Svein S. Andersen and Kjell A. Eliassen (eds.): Making Policy in Europe. The Europeification of National Policy-Making, London: Sage, 55-67. Grieco, Joseph M. 1988a: Anarchy and the Limits of Cooperation. A Realist Critique of the Newest Liberal Institutionalism, in: International Organization 42: 3, 485-507. Grieco, Joseph M. 1988b: Realist Theory and the Problem of International Cooperation. Analysis with an Amended Prisoner’s Dilemma, in: Journal of Politics 50: 3, 600-624. Grottanelli de Santi, Giovanni 1992: The Impact of EC Integration on the Italian Form of Government, in: Francesco Francioni (ed.): Italy and EC Membership Evaluated, London: Pinter, 179-190. Haas, Peter M. 1998: Compliance with EU Directives: Insights from International Relations and Comparative Politics, in: Journal of European Public Policy 5: 1, 17-37. Haggard, Stephen 1991: Structuralism and Its Critics: Recent Progress in International Relations Theory, in: Emanuel Adler and Beverly Crawford (eds.): Progress in Postwar International Relations, New York: Columbia University Press, 403-437. Hallerberg, Mark and Scott Basinger 1998: Internationalization and Changes in Tax Policy in OECD Countries: The Importance of Domestic Veto Players, in: Comparative Political Studies 31: 3, 321-352. Hasenclever, Andreas, Peter Mayer and Volker Rittberger 1997: Theories of International Regimes, Cambridge: Cambridge University Press. Hayes-Renshaw, Fiona and Helen Wallace 1997: The Council of Ministers, Basingstoke: Macmillan. Ioakimidis, Panayotis C. 1994: The EC and the Greek Political System: An Overview, in: Panos Kazakos and Panayotis C. Ioakimidis (eds.): Greece and EC Membership Evaluated, London: Pinter, 139-153. Ismayr, Wolfgang (ed.) 1997: Die politischen Systeme Westeuropas, Opladen: Leske und Budrich. 36 Jacobsen, John Kurt 1996: Are All Politics Domestic? Perspectives on the Integration of Comparative Politics and International Relations Theories, in: Comparative Politics 29: 1, 93-115. Jacobson, Harold K. and Edith Brown Weiss 1995: Strengthening Compliance with International Environmental Accords: Preliminary Observations from a Collaborative Project, in: Global Governance 1: 2, 119-148. Kaiser, Andre 1997: Types of Democracy: From Classical to New Institutionalism, in: Journal of Theoretical Politics 9: 4, 419-444. Keohane, Robert O. 1984: After Hegemony. Cooperation and Discord in the World Political Economy, Princeton, NJ: Princeton University Press. Keohane, Robert O. 1989a: Neoliberal Institutionalism. A Perspective on World Politics, in: Robert O. Keohane: International Institutions and State Power: Essays in International Relations Theory, Boulder, Colo.: Westview Press, 1-20. Keohane, Robert O. 1989b: Theory of World Politics. Structural Realism and Beyond, in: Robert O. Keohane: International Institutions and State Power. Essays in International Relations Theory, Boulder, Colo.: Westview, 35-73. Keohane, Robert O. and Joseph S. Nye 2001: Power and Interdependence, 3rd ed., Harper Collins. Kindleberger, Charles P. 1988: Dominance and Leadership in the International Economy: Exploitation, Public Goods, and Free Riders, in: Charles P. Kindleberger (ed.): The International Economic Order: Essays on Financial Crisis and International Public Goods, New York: Harvester Wheatsheaf, 185-195. King, Gary, Robert O. Keohane and Sidney Verba 1994: Designing Social Inquiry. Scientific Inference in Qualitative Research, Princeton, N.J.: Princeton University Press. Krasner, Stephen D. 1982: Structural Causes and Regime Consequences, in: Stephen D. Krasner (ed.): International Regimes, Special Issue of International Organization 36: 2, 185-205. Laffan, Brigid 1991: The Political System Assessed, in: Patrick Keatinge (ed.): Ireland and EC Membership Evaluated, London: Pinter, 206-208. Lijphart, Arend 1977: Democracy in Plural Societies: A Comparative Exploration, New Haven/London: Yale University Press. Marshall, Alfred 1961: Principles of Economics, 9th ed., vol. 1: Text, London: Macmillan. Martin, Lisa L. 1993: The Rational State Choice of Multilateralism, in: John Gerard Ruggie (ed.): Multilateralism Matters: The Theory and Practice of an International Form, New York: Columbia University Press, 91-121. 37 Martin, Lisa 1995: The Influence of National Parliaments on European Integration, in: Barry Eichengreen and Jeffrey Frieden (eds.): Politics and Institutions in an Integrated Europe, Berlin: Springer, 65-92. Mendrinou, Maria 1996: Non-Compliance and the European Commission’s Role in Integration, in: Journal of European Public Policy 3: 1, 1-22. Milner, Helen V. 1992: International Theories of Cooperation. Strengths and Weaknesses, in: World Politics 44: 3, 466-496. Milner, Helen V. 1997: Interests, Institutions, and Information: Domestic Politics and International Relations, Princeton: Princeton University Press. Moravcsik, Andrew 1993: Integrating International and Domestic Theories of International Bargaining, in: Peter B. Evans, Harold K. Jacobson and Robert D. Putnam (eds.): Double-edged Diplomacy. International Bargaining and Domestic Politics, Berkeley, Cal.: University of California Press, 3-42. Moravcsik, Andrew 1997: Taking Preferences Seriously: A Liberal Theory of International Politics, in: International Organization 51: 4, 513-553. Mosley, Layna 1996: State Compliance and the Legitimacy of International Law: The Case of the European Community, paper presented at the 37th Annual Meeting of the International Studies Association, San Diego, California, April 16-21. Nehring, Niels Jørgen 1992: Parliamentary Control of the Executive, in: Lise Lyck (ed.): Denmark and EC Membership Evaluated, London: Pinter, 77-81. Nettesheim, Martin 1999: Die mitgliedstaatliche Durchführung von EG-Richtlinien. Eine Untersuchung am Bespiel der Luftqualitätsrahmenrichtlinie, Berlin: Duncker & Humblot. Norton, Philip (ed.) 1996: National Parliaments and the European Union, London: Cass. Nugent, Neill 1999: The Government and Politics of the European Union, 4th ed., Durham: Duke University Press.. Olson, Mancur 1965: The Logic of Collective Action: Public Goods and the Theory of Groups, Cambridge, Mass.: Harvard University Press. Oppermann, Thomas 1991: Europarecht. Ein Studienbuch, Munich: Beck. Pedersen, Thomas 1996: Denmark, in: Dietrich Rometsch and Wolfgang Wessels (eds.): The European Union and Member States: Towards Institutional Fusion?, Manchester: Manchester University Press, 197-215. Peters, Dirk 1999: Die Implementation von EG-Richtlinien im Ländervergleich: Drei Erklärungsmodelle für das Implementationsverhalten der EG-Staaten, Magisterarbeit, Institut für Politikwissenschaft, Universität Tübingen. 38 Putnam, Robert D. 1988: Diplomacy and Domestic Politics: The Logic of Two-Level Games, in: International Organization 42: 3, 427-460. Radaelli, Claudio M. 1997: How Does Europeanization Produce Domestic Policy Change? Corporate Tax Policy in Italy and the United Kingdom, in: Comparative Political Studies 30: 5, 553-575. Rambaud, Patrick 1993: The French Legal Order and Community Law, in: FrançoisGeorges Dreyfus, Jacques Morizet and Max Peyrard (eds.): France and EC Membership Evaluated, London: Pinter, 175-184. Raub, Werner and Thomas Voss 1986: Conditions for Cooperation in Problematic Social Situations, in: Andreas Diekmann and Peter Mitter (eds.): Paradoxical Effects of Social Behavior: Essays in Honor of Anatol Rapoport, Heidelberg: Physica, 85-103. Riker, William H. 1962: A Theory of Political Coalitions, New Haven, Conn.: Yale University Press. Rometsch, Dietrich and Wolfgang Wessels (eds.): The European Union and Member States: Towards Institutional Fusion?, Manchester: Manchester University Press. Schelling, Thomas C. 1978: Hockey Helmets, Daylight Savings, and Other Binary Choices, in: Thomas C. Schelling (ed.): Micromotives and Macrobehavior, New York: W. W. Norton, 211-244. Schiøler Sørensen, Vibeke 1996: An Analysis of the Danish Implementation of the EC Remedies Directive, in: Public Procurement Law Review 5: 5, 186-214. Schwarze, Jürgen et al. 1990: The 1992 Challenge at the National Level: A CommunityWide Research Project on the Realization and Implementation by National Governments and Business of the Internal Market Programme, Baden-Baden: Nomos. Shepsle, Kenneth A. and Barry R. Weingast 1987: The Institutional Foundations of Committee Power, in: American Political Science Review 81: 1, 85-104. Siedentopf, Heinrich and Jacques Ziller (eds.) 1988: Making European Policies Work: The Implementation of Community Legislation in the Member States, London: Sage. Slaughter Burley, Anne-Marie 1993: International Law and International Relations Theory: A Dual Agenda, in: American Journal of International Law 87: 1, 205-239. Snidal, Duncan 1985a: Coordination Versus Prisoners' Dilemma: Implications for International Cooperation and Regimes, in: American Political Science Review 79: 4, 923-942. Snidal, Duncan 1985b: The Limits of Hegemonic Stability Theory, in: International Organization 39: 4, 579-614. 39 Sørensen, Henning and Ole Væver 1992: State, Society and Democracy and the Effect of EC, in: Lise Lyck (ed.): Denmark and EC Membership Evaluated, London: Pinter, 325. Stein, Arthur 1982: Coordination and Collaboration: Regimes in an Anarchic World, in: Stephen D. Krasner (ed.): International Regimes, Special Issue of International Organization 36: 2, 299-324. Taylor, Michael 1987: The Possibility of Cooperation, Cambridge: Cambridge University Press. Tsebelis, George 1995a: Decision Making in Political Systems: Veto Players in Presidentialism, Parliamentarianism, Multicameralism, and Multipartyism, in: British Journal of Political Science 25: 3, 289-325. Tsebelis, George 1995b: Veto Players and Law Production in Parliamentary Democracies, in: Herbert Döring (ed.): Parliaments and Majority Rule in Western Europe, Frankfurt/M.: Campus, 83-111. Tsebelis, George 1999: Veto Players and Law Production in Parliamentary Democracies: An Empirical Analysis, in: American Political Science Review 93: 3, 591-608. Young, Oran R. 1979: Compliance and Public Authority, Washington, D.C.: Resources for the Future. Zangl, Bernhard 1994: Politik auf zwei Ebenen. Hypothesen zur Bildung internationaler Regime, in: Zeitschrift für Internationale Beziehungen 1: 2, 279-312. Zürn, Michael 1992: Interessen und Institutionen in der internationalen Politik. Grundlegung und Anwendung des situationsstrukturellen Ansatzes, Opladen: Leske und Budrich.