Procedural models of European Union politics: Contributions and suggestions for improvement

Introduction

In the past 20 years, procedural modeling has established itself as one of the principal approaches to the study of European Union (EU) institutions and their functioning. Procedural models study how procedures and their reform affect policy and the powers of institutions involved in policy-making. ¹ They represent but one strand in an extensive political–economic literature that takes a rational choice approach to the study of decision-making in the EU.

Most procedural models are game-theoretical and spatial. In such models, political actors care about policies. These policies are represented by points in an n-dimensional policy space. Each dimension then corresponds to a particular policy issue. Usually, political actors are assumed to have Euclidean preferences. That is, each actor has an ideal policy, and his utility declines as policy moves further away from it. The preferences of the actors, the location of the status quo, and the procedures together determine policy in procedural models.

In this paper, we present an overview of the literature on procedural models of EU politics. We study and compare the principal models of the legislative process. We further discuss the main empirical tests of these models. Moreover, we examine models of other procedures and aspects of EU policy-making. Most importantly, we study the models’ contributions to the political–economic literature on EU policy-making, as well as their shortcomings, and discuss avenues for improvement and future research opportunities.

We argue that procedural models have contributed greatly to the understanding of EU legislative procedures, often presented as overly complicated and inscrutable. The literature presents parsimonious models that yield clear conclusions regarding the impact of procedures and preferences on the policies that emerge from the legislative process, the extent of gridlock, and the powers of political actors. Moreover, it has stimulated valuable empirical research. Furthermore, we argue that procedural models can further increase our grasp of the functioning of the EU if applied to other procedures and aspects of EU politics, beyond the legislative process. Finally, we claim that the literature contributes to the understanding of the functioning of bicameral systems and the impact of institutional reform. The models of consultation and codecision are meticulously elaborated and can be considered as standard models of unicameral legislatures with an agenda setter and bicameral legislatures with bargaining between the two chambers, respectively. As the EU continues to move toward bicameralism, the Commission will continue to lose its formal legislative powers and will thus need to derive its powers from other sources.

We do not believe that the literature is built on quicksand, as Hörl et al. (2005) suggest. They present an overview of one decade of procedural models of EU policy-making and question the justification of some of the key assumptions. ² Yet, rather than finding flaws in the models, they identify shortcomings within the empirical tests. Our overview covers two decades of procedural modeling of the EU and reaches different conclusions, as summarized above. We argue that, even though there is now considerable empirical research that tests the procedural models, these tests underestimate the predictive power of the models, as pointed out by Slapin (2014). We further contend that the procedural modeling literature itself has suffered from misinterpretations and misrepresentations of the procedures in the past and would benefit from more accuracy in the modeling in the future.

Models of EU legislative procedures

The earliest procedural models of EU politics were presented by Steunenberg (1994), Tsebelis (1994), and Crombez (1996, 1997a). Steunenberg (1994) and Crombez (1996, 1997a) both analyzed and compared the consultation, cooperation, and codecision procedures, and the powers of the EU’s main institutions – the Council, the European Parliament (EP), and the Commission – under these procedures. ³ In addition, Crombez (1996, 1997a) also studied the assent procedure. ⁴ Tsebelis (1994), by contrast, focused exclusively on the cooperation procedure.

Steunenberg (1994) and Crombez (1996, 1997a) both present one-dimensional spatial models. ⁵ Their models of the consultation procedure, the oldest and until recently most important procedure, are similar, as are their conclusions. They find that the Commission has considerable powers as an agenda setter, because the Council can amend its proposals by unanimity only, whereas a qualified majority suffices for their adoption. Thus, the Commission can successfully propose any policy that a qualified majority prefers to the status quo, provided that no policy is preferred to it by all member states. The latter requirement may represent a binding constraint only if the Commission is a preference outlier, and all member states are to its left (right). The EP has no formal decision-making powers under consultation.

The main difference between Steunenberg’s (1994) and Crombez’s (1996, 1997a) analyses of consultation can be found in their interpretation of the Commission’s rights, and in particular, the question of whether the Commission has gatekeeping rights. Crombez (1996, 1997a) assumes that it has no such rights, because the Treaty on the Functioning of the European Union (TFEU) requires that it formulates a proposal if the Council or EP request one (Art. 225 and 241 TFEU). Under this interpretation, a unanimous Council can adopt policies without the consent of the Commission. However, since the ratification of the Treaty of Lisbon in 2009, the TFEU states that the Commission can respond to Council and EP requests for proposals by informing them of its reasons for not submitting a proposal. To what extent this confers gatekeeping rights upon the Commission remains to be determined: the EP and the Council can take the Commission to the EU Court of Justice if they consider that it fails to fulfill its duties (Art. 265 TFEU).

Steunenberg, by contrast, points at the Commission’s right to alter its proposals as long as the Council has not acted on them, a right that is usually interpreted as including the right to withdraw its proposal. He considers this as an ex-post veto right and models it as a gatekeeping right. Under Steunenberg’s interpretation of the Commission’s rights, a winning coalition necessarily includes the Commission. The Commission is thus more powerful under this interpretation – that is, the equilibrium policy is (weakly) closer to its ideal policy – and there is more potential for gridlock. Nonetheless, Steunenberg’s and Crombez’s policy predictions differ only if the Commission is a preference outlier.

Crombez et al. (2006) discuss the differences between gatekeeping and ex-post veto rights. The Council can anticipate an ex-post Commission veto by approving a policy that the Commission prefers to the status quo. Thus, the Commission does not keep the gate closed if it has an ex-post veto right. Whenever there are policies that the Commission and a qualified majority prefer to the status quo, such a policy becomes EU policy. This would not necessarily be the case if the Commission did indeed have a gatekeeping rather than an ex-post veto right. Crombez et al. (2006) argue that the Commission does not have gatekeeping rights. They also find that giving a gatekeeping right to an institution is not Pareto optimal: every institution is better off if an ex-post veto rather than a gatekeeping right is granted.

Steunenberg’s (1994) and Crombez’s (1996, 1997a) interpretations and models of the codecision procedure, now the most important procedure and renamed the ordinary legislative procedure, differ more than their models of consultation. They both find that the EP acquires an absolute veto right under codecision. This veto right increases the EP’s powers and gridlock.

Crombez (1996, 1997a), however, sees a second important change relative to consultation: the EP and a qualified majority in the Council can together amend Commission proposals in the Conciliation Committee. He argues that the EP gains significant power under codecision at the expense of the Commission and becomes a genuine colegislator with the Council. The Commission faces more restrictions when formulating a proposal than under consultation: to be adopted, its proposal needs to be preferred to the status quo by the EP and a qualified majority in the Council, and no policy should be preferred to it by the EP and a qualified majority. Moreover, there should be no other policy that satisfies these conditions and is preferred by all member states. Whether codecision then leads to more gridlock than consultation depends on the preference configuration: roughly speaking, if the EP is closer to the pivotal member states than the Commission and extreme member states, there is then less gridlock under codecision, as analyzed in more detail by Crombez and Hix (forthcoming).

The 1997 Amsterdam Treaty altered codecision. Crombez (2000a) analyzed the implications of this reform. Under the Amsterdam version of codecision, the status quo prevails if the EP and the Council fail to reach an agreement in the Conciliation Committee, whereas prior to the Amsterdam Treaty, they could confirm the original Commission proposal, possibly as amended by a unanimous Council, in such a situation. Crombez (2000a) shows that as a result of this reform, the Commission lost its formal agenda-setting powers under codecision to the EP and the Council. The goal of the reform was to increase further the power of the EP. Crombez (2000a) concludes that whether the reform achieved its objective depends on the bargaining powers of the EP and member states in the Conciliation Committee.

A third legislative procedure, cooperation, was introduced in the mid-1980s. It was later replaced by codecision and is no longer used. We briefly discuss it in this article because procedural modelers reached vastly different conclusions on the institutions’ powers under this procedure. These disagreements led to a heated academic debate that had the value of raising awareness of the literature on procedural models of EU politics at an early stage of its development, but also hindered its progress.

Under cooperation, the EP acquired a veto right that only a unanimous Council could override and an amendment right. Steunenberg (1994) and Crombez (1996, 1997a) both concluded that the veto right did result in modest extra powers for the EP, compared to the consultation procedure, but that the amendment right was inconsequential because the Commission and Council could ignore the amendments. They found that, when compared to consultation, the Commission’s power decreased somewhat under cooperation, as the EP’s veto right limited the set of policies the Commission could successfully propose. Furthermore, they concluded that the prospects of gridlock rose.

Tsebelis (1994) studied the cooperation procedure and the EP’s agenda-setting powers. He presented a model that excluded the first steps of the procedure and the Commission’s monopoly proposal right, focusing instead on the EP’s amendments and the steps that followed. Contrary to Steunenberg (1994) and Crombez (1996, 1997a), he concluded that the EP gained major powers under cooperation. In particular, he argued that it acquired conditional agenda-setting powers as it was more difficult for the Council to amend EP amendments than to accept them if the Commission agreed with them. Accepting then required the support of a qualified majority, whereas amendments needed unanimity. This was, however, also true for EP opinions under consultation.

Moser (1996, 1997) was one of the first critics of Tsebelis’ analysis. His conclusions were in line with those of Steunenberg (1994) and Crombez (1996, 1997a). He argued that EP amendments were successful only if preferences or the perception of policy consequences changed. Hubschmid and Moser (1997) applied Moser’s analysis to the EU’s decision on car emission standards and reached similar conclusions. Tsebelis (1996) responded to Moser’s criticism by pointing out that Moser assumed complete information, thus ignoring that he presented a complete information model as well.

Garrett (1995), Garrett and Tsebelis (1996), Tsebelis (1997), and Tsebelis and Garrett (1996) extended Tsebelis’ analysis to codecision. They argued, in contrast to Steunenberg’s (1994) and Crombez’s (1996, 1997a) conclusions, that the EP’s powers were reduced under the version of codecision that was used prior to the Amsterdam Treaty, as compared to cooperation, because the Council could revert to the proposal it approved earlier if it failed to come to an agreement with the EP in the Conciliation Committee, thus eliminating the agenda-setting powers Tsebelis ascribed to the EP under cooperation. Scully (1997a) refuted these claims both theoretically and empirically. His theoretical arguments were consistent with Crombez’s (1996, 1997a) analysis. Moreover, he provided empirical evidence that the introduction of codecision resulted in more successful EP amendments. Tsebelis and Garrett (1997), Scully (1997b) and Garrett and Tsebelis (1997) perpetuated the argument.

The controversy was also discussed in a forum article by Crombez et al. (2000). In that article Crombez argued that the EU was moving toward a bicameral system and that the democratic deficit was being reduced (see also Crombez, 2003). Steunenberg questioned these assertions, whereas Corbett expressed a practitioner’s disbelief with Tsebelis’s (1994) conclusions on cooperation and some of the assumptions made by other scholars. Garrett et al. (2001) continued the discussion.

Tsebelis and Garrett (2000) summarized their earlier conclusions and presented an overview of the main legislative procedures and the way they evolved over time. Their conclusions on the version of codecision used since the Amsterdam Treaty were consistent with those of Crombez, as were those of Steunenberg (1997). Thus, procedural modelers reached similar conclusions on the procedures in use today and the controversy ended. The differences that remain are relatively minor and relate to the aforementioned gatekeeping rights of the Commission and the bargaining powers of the EP and Council in the Conciliation Committee. Gatekeeping rights only matter when the Commission is an outlier, however, whereas the bargaining process in the Conciliation Committee is not formalized in the TFEU. Procedural modelers have thus recognized that EU policies depend on the relative bargaining powers of the EP and Council.

This brief overview illustrates the main contributions and shortcomings of the procedural modeling literature. The literature manages to capture the key aspects of complicated legislative procedures, presents relatively simple, understandable models of them, and yields clear predictions in terms of policies, institutions’ powers, and the extent of gridlock. Regretfully, misinterpretations and misrepresentations of the procedures in some of the early models led to heated scholarly debates. These debates raised awareness of the field, but also hindered its progress. Moreover, they distracted modelers and forced them to turn inward and spend their time arguing with one another, rather than to turn outward, link the models to the wider literature of procedural models of politics, and discuss their contributions to it.

Procedural models of EU legislative procedures have indeed contributed to the broader literature on political institutions. As the models show, consultation can be considered as a procedure with an agenda setter and a unicameral legislature, whereas codecision can be regarded as a procedure with a bicameral legislature and bargaining between the two chambers. The theories are meticulously elaborated, often in more detail than those of other political institutions, and can serve as standard models of such procedures. Moreover, the introduction and extension of codecision, and the EU’s move from a unicameral to a bicameral system, represent a unique opportunity for the study of the implications of institutional reform.

The models suggest that as a result of the most recent reforms, introduced by the Lisbon Treaty, and as the EU continues to reform and move further toward bicameralism in the future, the Commission will continue to lose power, and EU policy will not so much be determined by the Commission as agenda setter but rather be the result of bargaining between the Council and the EP. The Commission will thus have to derive its powers in the future from outside the legislative process. One institutional reform that could increase its powers, for example, is the enhanced role of the EP in the Commission appointment process. This could perhaps turn the Commission into the EP’s executive body and thereby politicize the supranational agent, with the executive playing the dominant role, as in many parliamentary democracies.

Empirical tests of procedural models of the legislative process

During the first few years, little empirical research was performed to assess the claims of the early procedural models and compare them to the predictions of other scholarly work. Making such comparisons requires data on the actors’ policy preferences, the location of the status quo, and the policies that emerge from the legislative process. Such data were scarcely available a decade ago. Gabel et al. (2002) lamented this dearth of publicly available data and suggested avenues for further development. Ten years on, there has been considerable improvement.

Having accurate information on actors’ preferences is highly important. Junge and König (2007) show that procedural models are less robust to misspecification of players’ preferences than to misinterpretation of the procedures. Information on preferences can be gathered in a number of ways. First, surveys can be conducted to determine political actors’ policy positions on a range of issues. Castles and Mair (1984) and Huber and Inglehart (1995) used expert surveys to study national party positions on a left–right policy scale, whereas Ray (1999) analyzed their attitudes toward EU integration. Benoit and Laver (2006), Steenbergen and Marks (2007), and Hooghe et al. (2010) conducted more expert surveys. Döring and Manow (2012; see also Döring, 2013) used these data to create ParlGov, a comprehensive dataset on party, parliament, and government positions. Alternatively, voter surveys, such as the Eurobarometer, can be used to locate the parties they support, as did Schneider (1995), Hix and Lord (1997), and König and Hug (2000).

Member-state positions can be calculated by taking a weighted average of the governing parties’ positions. To determine the positions of the Commission (EP), each Commissioner (MEP) can be considered to have the same position as their national party. Political groups in the EP can be thought of as having the positions of their medians. Warntjen et al. (2008) present data on the location of political actors and compare different measures.

In their overview of procedural models, Hörl et al. (2005) criticized what they referred to as key assumptions of the models and suggested that they were built on quicksand. However, they mainly disapproved of the calculations of the actors’ preferences in empirical tests. Specifically, they criticized the use of the actors’ left–right and pro–contra integration preferences to study policy-making on a wide variety of issues. Moreover, they urged scholars to consider intra-institutional decision-making processes rather than assume that actors have the same preferences as their national parties. They thus identified a need for issue or area-specific data on preferences and more attention from theoretical and empirical scholars for decision-making within institutions.

Second, actor positions can be determined by studying party programs. The Manifestos Research Group (Budge et al., 2001) measured parties’ positions on a series of issues by counting the number of positive and negative statements on those issues in the parties’ respective manifestos. Carrubba (2001), for example, used the measures of party positions on EU integration when testing theories of integration. Gabel and Hix (2002) analyzed EP election manifestos.

Third, studies of past voting behavior can be used to locate political actors in the policy space. Hix et al. (2006, 2007) took this approach in their study of the EP. They applied the NOMINATE-algorithm to EP roll call votes to locate MEPs in the policy space. They found that the left–right and pro–contra integration dimensions are the dimensions that matter most in EU politics, a conclusion also reached earlier by Hix (1999). Carrubba et al. (2006, 2008b) and Hug (2010) study the sampling problems in studies of roll-call votes. In a study of Council-voting records, Mattila (2004) also concluded that the EU policy space consists of the two dimensions mentioned. He further found that leftwing governments tend to be more in favor of integration than their rightwing counterparts. For records of the voting behavior of MEPs and member states, we further refer to the VoteWatch.eu database, founded by Hix, Hagemann, Noury, and Frantescu.

Fourth, expert interviews can be conducted to locate political actors’ positions on specific issues, as was done to establish the decision-making in the European Union (DEU) dataset (see Thomson et al., 2006). Zimmer et al. (2005) used this dataset and found that the left–right dimension plays a role in the Council, but that the north–south division is more important, particularly the conflict between net-receivers and net-contributors to the EU budget. Kaeding and Selck (2005) studied the DEU data and concluded that the Commission and the EP are more pro-integration than the Council. Moreover, they argued that whether a country is pro-integration largely depends on if it receives more money from the EU than it pays into it.

Two comprehensive datasets provide information on legislative activity and outcomes: Häge’s (2011) EUPOL dataset, based on the PreLex database, and the dataset put together by König et al. (2006), which is based on the CELEX and PreLEx databases and provides extra information on such aspects of the legislative process as the voting procedure used on a specific issue. The status quo of an issue can be located through expert interviews and document analysis.

Among the first to test and compare the predictions of competing procedural models were König and Pöter (2001). They studied four pieces of legislation that were passed under the cooperation procedure and focused on the EP’s influence on it. They found mixed support for the various models of cooperation discussed above.

The first to present a comprehensive test of procedural and other models of EU decision-making were Thomson et al. (2006). ⁶ They analyzed 66 legislative proposals considered under consultation and codecision, used expert interviews to gather data on policy positions as part of the DEU project, and compared the predictions of procedural models to those of bargaining models. They found that bargaining models outperform procedural models and that among procedural models those that assume that the EP sets the agenda in the Conciliation Committee under codecision best predict policies.

Thomson (2011) studied 125 controversial legislative proposals and also found that bargaining models outperform procedural models. He further concluded that actor positions and winning coalitions vary markedly from one proposal to another. He considered this to be one of the EU’s strengths, as it allows all actors to gain victories in the legislative process. ⁷ He further argued that this strength may also represent a weakness, because voters cannot vote a governing coalition out of office.

The contributions by Thomson et al. (2006) and Thomson (2011) are the most comprehensive in the field, and their conclusions have been widely accepted thus far. However, Slapin (2014) questions the performance of bargaining models and examines whether the tests underestimate the influence of the Commission and EP. He uses the same dataset as Thomson (2011) and finds that the data on the Commission and EP positions suffer more from measurement error than do the measures of the Council positions, because the Commission and EP positions are point estimates, whereas the Council position is the result of estimates of all member-state positions. As a result, the influence of the Commission and EP is underestimated in empirical tests, and procedural models that find these institutions to be powerful thus underperform. Slapin (2014) therefore identifies an important shortcoming of the tests and demonstrates a need for empirical analyses that correct for differences in the measurement errors of the actors’ positions.

Leinaweaver and Thomson (2014), however, disagree and argue that Slapin’s (2014) conclusions result from unrealistic assumptions. Most importantly, they contend that in reality, member-state ideal policies are clustered around a small number of policy positions rather than randomly distributed across the policy space and that measurement error is unlikely to have biased earlier test results under that assumption. They do not conclude that procedural models are wrong, however. In particular, they point out, for example, that the dataset used focuses on issues that were controversial during the legislative process, whereas procedural models apply to issues on which controversy has ebbed away prior to the legislative process and the agenda setter formulates a proposal under perfect information.

Some scholars focused on the powers of the political actors. Selck and Rhinard (2005) studied 70 recent decisions and found that on average, the success of the Council is higher than the successes of the Commission and the EP. They measured success by the average distance between ideal policy and outcome and used interviews to determine the positions of the actors on issues considered under consultation and codecision. In a study of 48 Commission proposals, König and Junge (2009) found that even if member states have veto rights, they often do not use them as a consequence of logrolling and anticipation by the Commission.

Kreppel (1999, 2002), Tsebelis and Kalandrakis (1999), and Tsebelis et al. (2001) studied the success rate of EP amendments and concluded that the EP is influential. Hix (2002) contended that the EP gained significant powers thanks to the Amsterdam Treaty, consistent with the conclusions of the procedural models. He claimed that the EP was able to impose its interpretation of the Maastricht Treaty on the member states. Selck and Steunenberg (2004) studied the EP’s proximity to policy outcomes under consultation and codecision. They found that the EP is close to the outcomes under both procedures and argued that this is the result of its power under codecision, whereas the EP is just lucky under consultation. Kardasheva (2009) found that the EP, through its power to delay legislation, has substantial influence on policy under consultation. König (2008) studied the empowerment of the EP and argued that member states introduced codecision because they prefer the EP to be the agenda setter rather than the better-informed Commission. Under codecision, they can mislead the EP about their preferences.

Using data on the preferences of member states at Inter-Governmental Conferences (IGC), Slapin (2006) found that large states do not have more bargaining power, but that domestic ratification constraints have a larger impact. He also concluded that states that are contra integration outperform states that are not. Schneider and Cederman (1994) noted that countries may use domestic politics to strengthen their bargaining positions. König and Hug (2006) analyzed the role of domestic constraints in the negotiations at the Convention and subsequent IGC. Slapin (2008) argued that veto power in institutional models better explains IGC outcomes than does member-state size in intergovernmental models.

König and Slapin (2004) challenged the notion that national parliaments of member states have little say on EU policy. Conducting research on the Amsterdam IGC, they found that the parliaments of the member states least in favor of integration have considerable decision-making powers and that constrained players with a unanimous national parliament get an outcome closer to their ideal policies.

Other scholars examined the duration of the legislative process, a concept related to gridlock. As mentioned above, Crombez and Hix (forthcoming) predicted that whether gridlock rises with the introduction and extension of codecision depends on the actors’ preferences. Golub (1999) found that successive Treaty reforms did not lead to increases in legislative activity and the speed of decision-making. Schulz and König (2000) argued, by contrast, that the EU was able to streamline its procedures to deal with an expanding legislative agenda. König (2007) analyzed the relative importance of procedures and preferences for the duration of the legislative process. He concluded that European legislative integration is slowing down as a result of the divergence of member-state preferences. Leuffen and Hertz (2008) found that member states anticipated EU enlargement by increasing legislative activity prior to the accession of new member states.

Models of other procedures and aspects of policy-making

As shown above, procedural models of EU legislative procedures are well-established in the literature. Yet, other aspects of EU politics have received less attention. The understanding of the functioning of the EU would greatly benefit from analyses of the special procedures that apply to trade, competition, and the budget, for example. Moreover, our grasp of EU decision-making would improve if procedural modelers paid more attention to decision-making procedures within EU institutions, such as the EP and the Conciliation Committee. In this section, we focus on models of four features of EU politics – delegation to the Commission, Commission appointment, the Court of Justice, and logrolling – to illustrate the contributions that such models can make to the literature.

Steunenberg et al. (1996) analyzed the delegation of the implementation of EU legislation. They found that the specific procedures followed have an important impact on the powers of the Commission. Dimitrova and Steunenberg (2000) studied to what extent national policy convergence depends on characteristics of the EU legislative and implementation processes. Franchino (2000) found that the Commission enjoys considerable discretion in the implementation of EU policies. Later, he showed that the Council delegates more power to the Commission when an act is adopted by qualified majority rather than unanimity, because the majority wants to make sure that countries that oppose the act implement it. He further concluded that delegation to the Commission depends on the level of conflict between the Commission, the member states, and the EP (see Franchino, 2004, 2005).

Crombez (1997b) studied the Commission appointment process. He concluded that the Commission’s ideal policy is largely determined by the member state closest to the status quo. Hug (2003) argued that the Commission cannot be seen as a unitary actor and that the Commission is more pro-integration than the member states. Thomson (2008) found a remarkable concurrence between the Commissioners’ positions and the position of their home countries. Crombez and Hix (2011) concluded that the requirement of EP approval for the appointment of the Commission, introduced by the Maastricht Treaty, had no impact on Commission preferences. However, the Treaty led to policies further away from the status quo due to the introduction of codecision. As a result of the Treaty of Nice, unanimity in the Council for Commission appointment was no longer required. Consequently, a member state could end up in ‘opposition’ during an entire Commission term.

A number of scholars studied the role of the EU Court of Justice, with Carrubba et al. (2008a) finding that it behaves strategically. They argued that it does not issue rulings that it expects to be overruled in a treaty. Stone-Sweet and Brunell (2012) contended, by contrast, that the Court is strongly independent and enhances integration. They considered the high compliance with Court rulings as a sign of its power. In response, Carrubba et al. (2012) argued that the Court does indeed enhance integration but only insofar as it enjoys member state support.

Crombez (2000b) developed a model of logrolling in the EU. He considered three sequential games: Commission appointment, logrolling, and policy-making games. In the logrolling game, the institutions compare the logroll to the equilibrium policy in the absence of a logroll. Under consultation, for example, the set of policies a qualified majority in the Council and a majority of Commissioners prefer to that policy can be approved as a logroll. König and Proksch (2006) also presented a model of vote exchange. They found that all Council members benefit from exchange.

Discussion

During the past 20 years, the procedural modeling literature established itself as one of the standard approaches to the study of EU politics, despite disagreements and controversies in the early years. In this section, we outline its principal contributions and shortcomings and formulate suggestions for improvement.

Its main contributions to the study of EU politics are fourfold. First, the functioning of the EU, the institutions’ roles in policy-making, and the legislative procedures used are often considered as overly complicated and inscrutable. Procedural models have shown that the EU legislative procedures, however intricate and confusing the treaty articles that describe them may be, can be understood quite easily by focusing on the key steps in the procedures. The models have demonstrated that the legislative process is not more complicated than it is in other political systems that have been the object of academic research for much longer than the EU has been. They have made clear, for example, that the steps that precede the Conciliation Committee in the codecision procedure are irrelevant under certain assumptions and that policy under this procedure is determined by bargaining between the two chambers of the EU legislature, the EP and the Council.

Second, the literature offers valuable insights into the impact of procedures and political actors’ preferences on EU policies. It does so by presenting rigorous and parsimonious models that yield clear, testable conclusions. This allows for empirical evaluations of competing theories. EU policies typically depend on three aspects of policy-making: the procedures, the preferences, and the location of the status quo. Testing the procedural models then requires data on the actors’ positions and the locations of the status quo and EU policy.

Third, procedural models have yielded specific conclusions on the institutions’ powers and the extent of gridlock under different procedures. Most importantly, they have shown that the EP has become a genuine colegislator with the Council under codecision. Furthermore, they demonstrate that, as the EU continues to move toward bicameralism, the Commission loses its formal legislative powers and needs to derive its powers from other sources. In addition, the models allow for clear comparisons with the procedures of other countries and organizations, and the institutions’ powers and extent of gridlock therein.

Fourth, the models of the legislative procedures have sparked an extensive theoretical and empirical literature. The early models were followed by theoretical analyses of special procedures and decision-making in specific policy areas. In addition, empirical scholars set out to gather data on preferences and legislative activity, and test the models.

Before addressing some of the shortcomings of the literature, we briefly address two red-herring criticisms that are often raised to cast doubt on the conclusions of procedural models. It is often assumed in such models that the policy space is one-dimensional and that political actors have perfect information. These assumptions are regularly criticized, occasionally to the point that it is claimed that they invalidate the models’ conclusions.

The criticism of one-dimensional models is based on standard spatial models that show that chaos ensues in multidimensional settings (McKelvey, 1976), whereas the median voter’s ideal policy prevails in one-dimensional models (Black, 1958). However, this criticism ignores that institutions and procedures can create the stability we observe in the real world, whether the policy space is one-dimensional or multidimensional. In the EU, the legislative procedures and the internal procedures of the EP, the Council, and the Commission prevent chaos. It would be more accurate to include the institutions’ internal decision-making procedures in multidimensional models rather than represent the EP and the Commission by their respective medians and the Council by two pivotal member states. The internal processes create stability in a multidimensional world. ⁸ Moreover, Drüner (2007) argues that legislative decisions in the EU typically do not involve more than two dimensions and that stability ensues as a result.

In reality, political actors rarely, if ever, have perfect information on the consequences of policies and the preferences of other actors. It is therefore not surprising that perfect information models cannot explain certain empirical regularities, such as successful EP amendments under consultation. Perfect information models do, however, offer the advantage of simplicity and clarity. Assuming that actors have private information or that information becomes available during the process, by contrast, complicates the theoretical analyses, often with little gain in additional insights. Moreover, conclusions are highly dependent on the precise informational assumptions made. For these reasons, perfect information models offer valuable contributions to the literature.

We now point to some of the shortcomings of the literature and suggest avenues for improvement. First, as the controversy over the EP’s powers under cooperation and in the early version of codecision indicates, the literature has suffered from misinterpretations and misrepresentations of the legislative procedures, as well as misunderstandings amongst scholars. As Corbett’s contribution to Crombez et al. (2000) demonstrates, most formal scholars have at times been somewhat careless in their interpretation of the rules. Rather than recognizing their mistakes, they have subsequently dug in and argued their points more strongly. At other times, simplifying assumptions made by some have mistakenly been considered as their principal claims by others. The literature would benefit from more accuracy in the modeling of the formal rules. If procedural models do not manage to accurately represent the main formal rules, there is no point to them. Encouragingly, there has been improvement in this respect.

Second, the legislative procedures have been studied and modeled extensively, but this is less true for other procedures and aspects of policy-making. Some steps have been made to study budgetary policy and the special procedures that apply to it, for example, but more can be done. The internal rules and procedures followed within EU institutions, often considered as unitary actors in formal models, can also be studied more extensively. The study of politics within the EP and the Conciliation Committee could benefit from such analyses, for example.

Third, testing competing procedural models and comparing their performance with the performance of other types of models requires the availability of data on preferences and outcomes. Today, a lot more information is available than was 10 years ago, and great work has been done to test competing models. However, as pointed out by Slapin (2014), there is room for improvement as far as the design of the tests is concerned. ⁹

Fourth, the EU is too often studied in isolation, with no reference or comparison to other political systems. Of course, the EU is unique. It has gone further in terms of integration and the surrendering of sovereignty by its members than any other international institution. Every political institution is unique, however. Scholars of EU politics would contribute more to the formal modeling literature as such if they thought more about what the functioning of the EU teaches us about institutions and politics in general. Similarly, the procedural modeling literature on the EU would benefit from a closer link to the broader political economy literature. The models of consultation and codecision can be considered as standard models of unicameral legislatures with an agenda setter and bicameral legislatures with bargaining between the two chambers, respectively. The literature on procedural models of the EU would benefit if EU scholars emphasized such points and compared their models to models of policy-making elsewhere.

In our opinion, there are many valuable ways to study EU politics. Procedural modeling represents just one important avenue of research to achieve a better understanding of the EU. In the past 20 years, procedural modelers have developed parsimonious models that yield clear, testable conclusions, and, in the process, have stimulated empirical research on EU politics.