The Impact of Enlargement on Legislative Decision Making in the European Union

Transcription

1 The Impact of Enlargement on Legislative Decision Making in the European Union Robert Thomson Department of Political Science Trinity College Dublin Ireland Version: July 30, 2007 Abstract What impact has the 2004 enlargement of the European Union (EU) had on legislative decision making? The present paper answers this question, first, by assessing the nature of the controversies and the alignments of actors on those controversies before and after enlargement. I find substantial variation in actor alignments across issues, both before and after enlargement. However, since enlargement a divide between old and new member states has been evident, particularly on issues about financial subsidies. Second, the paper examines whether enlargement has changed the way in which controversies are resolved. What model of legislative decision making gives the most accurate predictions of decision outcomes? Both before and after enlargement, a simple bargaining model, the Nash Bargaining Solution, gives significantly more accurate predictions of decision outcomes than a procedural model. The paper examines the impact of enlargement with a new dataset on legislative decision making in the EU. The dataset has been assembled through 228 semistructured interviews with key informants. The dataset contains information on the positions taken by EU actors on 229 controversial issues raised by 87 legislative proposals from before and after the 2004 enlargement. Prepared for delivery at the General Conference of the European Consortium for Political Research, Pisa, Italy, 6-8 September, Panel Title: Decision-making in the Council of the EU after enlargement, Thursday 6 September,

2 The Impact of Enlargement on Legislative Decision Making in the European Union Robert Thomson Trinity College Dublin Ten countries joined the group of fifteen European Union (EU) members in May The Union enlarged again to include Bulgaria and Romania in January 2007, bringing the total membership to 27 countries. In addition to increasing the EU s population and expanding its territory, enlargement added a considerable amount of diversity to the EU. Practitioners and academics expressed concern about the possible impact of increased numbers and diversity of decision makers on the EU s capacity to act. Increasing numbers of actors could have made decision making more difficult, introducing a bias toward the status quo (Hosli 1999). Similarly, König and Bräuninger (2004: 421) note that enlargement has the potential for creating gridlock. Zimmer et al. (2005) expect enlargement to have widened the gulf between netcontributors and net-recipients from the EU budget. If the EU s decision-making processes have been affected, this could have profound impacts throughout the EU and beyond, because a substantial proportion of national laws in Europe are mandated by the EU. In addition to warnings from EU specialists, research on collective decision making in other contexts also suggests that increasing the number of decision makers fundamentally alters the way in which groups take decisions. Social psychologists have long known that small and large groups typically have different ways of taking decisions (e.g. Simmel 1902). Small groups are able to reach decision outcomes based on informal contacts. Intense communication and personal contacts facilitate the exchange of information and ideas about alternative options, and the effective imposition of sanctions if social norms are violated. Larger groups, by contrast, are less able to engage in such deliberative processes effectively, and free riding may be more prevalent. Brewer and Kramer (1986: 549), for example, found that individuals were less willing to contribute to public goods when they were part of larger groups. However, other studies have shown that cooperative patterns of behavior are possible in large groups of 20 individuals (e.g. Liebrand 1984). This implies that at least some larger groups maintain cooperative ways of deciding. Is the EU one of these? This paper examines the impact of enlargement using a new dataset with comparable information on decision making before and after the 2004 enlargement, more details of which are given in the following research design. The data on decision making before enlargement have been studied extensively in the volume The European Union Decides (Thomson et al. eds. 2006). The EU-15 data were collected in more than 150 semi-structured interviews with key informants, and consist of EU actors positions on 174 controversial issues from 70 legislative proposals. The data on decision making in the enlarged EU were collected in 78 semi-structured interviews with key informants, and consist of EU actors positions on 55 controversial issues from 17 legislative proposals. The data on the enlarged EU are part of an ongoing study. 1 Together, these data are a unique resource with which to examine the impact of enlargement on the EU s decision-making processes. 1 The data on the enlarged EU were collected by the present author, Javier Arregui (University of Pompeu Fabra, Barcelona), Rory Costello (Trinity College Dublin), and Robin Hertz (University of Zurich). 2

3 This paper examines the impact of enlargement in two respects. First, the analysis addresses the actor alignments found in the European Union. To what extent are there structures in the positions actors take across a range of issues? To answer this question, I apply multi-dimensional scaling, and conduct a detailed analysis of the alignments found on each controversial issue. The findings indicate that although there is some structure in the alignments of actors found across issues, these tendencies are evident in only a minority of controversial issues. This is true of the EU-15 and of the enlarged EU. The nature of the structures in actor alignments has changed since enlargement. In the EU-15, there was a division between Northern and Southern member states on around one third of the issues. Disagreement between Northern and Southern member states tended to be on issues about choices between regulation and free market alternatives. Northern member states generally favored the free market approach. In the enlarged EU, the North-South divide appears to have been superceded by the old-new divide. Disagreement between old and new member states tends to be about issues concerning the level of financial subsidies. Second, the paper examines how decision outcomes are arrived at. The analysis employs a modeling approach. I compare the accuracy of competing predictions of decision outcomes generated by different models of decision making. This type of analysis was performed extensively on the EU-15 data, and is reported in Thomson et al. eds. (2006). That study developed and applied a range of models. Each model purported to capture the essence of the process through which actors positions are transformed into decision outcomes. Some of the models, labeled procedural models, emphasize formal decision-making rules, the procedures that stipulate which actors can introduce and amend legislative proposals, and the levels of support required for proposals to be adopted. Other models, bargaining models, emphasize the informal norms that guide actors behavior. Both before and after enlargement, a simple co-operative bargaining model, the Nash Bargaining Solution or compromise model, gives more accurate predictions of decision outcomes than a procedural model. Research Design By conceptualizing and measuring disagreement spatially, we are able to study both actor alignments and decision making more comparatively and quantitatively than would be possible with thick descriptions of each case. Controversies raised by legislative proposals are conceptualized as issue continua or scales. The proposal on the reform of the sugar sector is one of the proposals examined in this paper. The main issue raised by this proposal was the size of the price cut, which would in effect reduce the EU subsidy for sugar production (Figure 1). At one end of the issue continuum we find the status quo position, in other words no price cut. At the other end of the continuum, we find the alternative of a very large price cut of more than 39 percent. Intermediate positions are placed on positions between these two alternatives on a scale of 0-100, to reflect key informants judgments on the political distances between the alternatives. This way of defining issues enables us to identify the extent to which there are similar actor alignments across a range of issues. This procedure allows us to identify which models give predictions of decision outcomes that are closest to actual decision outcomes. This method of representing controversies spatially has been applied in a range of studies of decision making in national and international politics (see e.g. Bueno de Mesquita 2003; Bueno de Mesquita and 3

4 Stokman eds. 1994). Full details of the research design decisions for the EU-15 study can be found in Thomson and Stokman (2006). <FIGURE 1> The Selection of Legislative Proposals Legislative proposals were selected for the EU-15 study according to the following three criteria summarized in Table 1: the time period involved, the type of legislative procedure followed, and the level of political importance. Regarding the time period, each legislative proposal was on the Council s agenda in the years 1999 and/or Concerning the decision-making procedure, the legislative proposals selected were subject to either the consultation or the codecision procedures. 42 of the 70 legislative proposals, covering 99 of the 174 issues, were subject to the consultation procedure (Table 2); the remaining issues were subject to the codecision procedure. Regarding political importance, the selection was restricted to proposals on which there was an indication of at least some political importance and controversy. Each proposal was mentioned in Agence Europe, a news service covering European affairs. Furthermore, informants had to identify at least one substantive disagreement between at least some of the actors. Given these selection criteria, the policy areas represented most prominently in the selection are agriculture and internal market, each with 14 proposals, although fisheries (7 proposals) and other policy areas are also present. When completing the data collection for the EU-15 study in mid-2002, decision outcomes had been reached on 66 of the 70 legislative proposals (162 of the 174 issues). Similar criteria were used to select legislative proposals for the study of decision making in the enlarged EU. Regarding timing, we selected proposals that were discussed in their entirety after the May 2004 enlargement. Proposals introduced up to December 2005 were included in the study. Therefore, the time period covered refers mainly to EU-25, before the accession of Bulgaria and Romania in January However, many of the proposals were also discussed in 2006 and 2007 and are still pending. So far, we have not obtained information on the positions of Bulgaria and Romania on the controversies raised. Like the EU-15 study, we selected legislative proposal subject to both consultation and codecision procedures. However, unlike the EU-15 study we considered only directives and regulations, excluding decisions which are a more specific type of legislative instrument. As in the EU-15 study, we also considered proposals that were generally considered to be politically important by the policy experts we interviewed. For the EU-25 study, we carried out an initial screening of the codecision legislative proposals using the European Voice news service. European Voice has a somewhat more general readership than Agence Europe among EU practitioners. The effect of this change is therefore to focus the selection somewhat more on high-profile proposals. Nonetheless, the EU-25 selection includes both high-profile cases, such as the new European Regional Development Fund, the working time directive, and sugar sector reform, as well as more day-to-day pieces of legislation, such as rights for airline passengers with reduced mobility and measures concerning the conservation of fisheries. Compared to the EU-15 study, the selection of proposals for the EU-25 study is more evenly distributed across different policy areas, including agriculture (four proposals), fisheries (three proposals), employment (two proposals), and environment (two proposals). The study is still underway. At the time of the 4

5 compilation of the dataset for this paper at the end of May 2007, decision outcomes had been taken on 13 of the 17 proposals (40 of the 55 issues). 2 <TABLE 1> <TABLE 2> <TABLE 3> The Selection of Key Informants For the EU-15 study, at least 150 interviews were held with at least 125 key informants. These interviews lasted an average of one hour and forty minutes. For the EU-25 study 102 interviews were held before the end of May of these referred directly to the legislative proposals I examine here. The remaining 24 referred to legislative proposals on which we do not yet have complete sets of information on the controversial issues and actors positions on those issues. The key informants were selected for their knowledge of the detail of the dossiers under investigation. Usually they were participants. We require detailed information on the positions taken by actors. This means our sources have to be close to the discussions. Individuals with different institutional affiliations were interviewed, as detailed in Table 3. The Commission officials interviewed were usually the individuals responsible for drafting the proposals and monitoring the subsequent discussions. The officials from the permanent representations were usually the responsible desk officers. The individuals from the EP were usually the rapporteurs or their assistants. We held a larger number of interviews on each proposal in the EU-25 study than in the EU-15 study. There are several reasons for this. First, identifying the positions of 25 member states is considerably more time consuming than identifying the positions of 15 member states. Second, part of the EU-25 study involves the collection of information on the positions of actors in the European Parliament. The interviews with informants in the EP provided information included in the following analyses, and also information that will be examined in other publications. Third, in addition to replicating the EU-15 study, we aimed to move beyond it by collecting additional qualitative information on each of the cases. The following analyses focus on the quantitative estimates that can be compared directly with the EU-15 study. The Interviews Each of the controversial issues was represented spatially, in the form of a policy scale ranging from 0 to 100. The decision outcomes favored most by the Commission, each of the member states and the European Parliament were estimated by the experts during the interviews. The informants estimates of the actors positions refer to the 2 The 17 proposals included in the EU-25 study are the Regional Development Fund (COD/2004/167), the working time directive (COD/2004/209), LIFE environmental programme (COD/2004/218), the European Neighbourhood Instrument (COD/2004/219), port services (COD/2004/240), passengers with reduced mobility (COD/2005/007), data retention (COD/2005/182), pensions (COD/2005/214), broadcasting (COD/2005/260), sugar sector reform (CNS/2005/118), waste (COD/2005/281), hake fish conservation (CNS/2003/318), conservation of fish in Biscay area (CNS/2003/327), agricultural fund (CNS/2004/161), CAP financing (CNS/2004/164), fisheries fund (CNS/2004/169), and animal welfare (CNS/2005/009). 5

6 decision outcomes favored most by each of the actors at the time of the introduction of the Commission s proposal. For each issue, the most extreme decision outcomes are located at the ends of the policy scale, at positions 0 and 100. Actors with other positions were placed between these extremes by the key informants to represent their views on the political distances between their positions and each of the extremes. Whenever possible, the key informants identified the reference point. This is the policy alternative that would prevail if the actors were to have failed to reach an agreement. The informants were also asked to estimate the level of importance each of the actors attached to each issue. This level of importance was estimated on a scale of 0-100, whereby a score of zero indicates that the issue was of no importance whatsoever, 50 that it had an average level of importance to the actor concerned, and 100 that the issue could hardly be more important. The relations between the salience scores for different actors are more important than the absolute values of the scores. As with the procedures for estimating actors capabilities and actors positions on controversial issues, the procedure for estimating issue salience was adapted from a widely-used procedure for decision analysis (Bueno de Mesquita 2003: ). When obtaining the judgments on actors positions and the levels of importance they attached to the issues, they were asked to substantiate their judgments extensively. Validity and reliability tests were conducted on the informants judgments. These tests consisted of comparing informants judgments with information from Council documentation, and comparing judgments from different informants (Thomson 2006). These tests show that of all the points of discussion raised in the Council, key informants generally focus on issues that are more controversial, and that are more difficult to resolve. These are exactly the kinds of issues most relevant for our purposes of identifying actor alignments and the relative predictive accuracy of models of decision making. Informants estimates of the positions actors favor most usually match information reported in Council documentation. When they differ, these differences are due to the fact that Council documents do not refer to policy preferences, but to the decision outcomes actors were prepared to accept during the course of the negotiations. In addition, König et al. (2007) compared 31 point estimates provided by these key informants with estimates from informants in the European Parliament and found that 30 match perfectly or almost perfectly. Actor Alignments The main finding concerning actor alignments in the EU, both before and after enlargement, is that there is considerable variation in the positions actors take on different issues. Even on two controversial issues raised by the same legislative proposal, any two actors might share the same position on one of the issues and take different positions on the other issue. To the extent that there is structure in the positions taken by actors, these common tendencies are found on approximately one third of the controversial issues examined. The following sub-section applies multidimensional scaling techniques to explore patterns in the alignments of actors. Scaling techniques are a useful tool for identifying structure in the positions taken by actors to the extent that these exist. The maps produced do not, however, provide insights into the substantive meaning of the structures found. One way of doing so is to identify the nature of the issues on which these structures are evident. Therefore, to supplement the multidimensional scaling analysis, I examine the alignments of actors on each specific issue. This issue-level 6

7 analysis reveals both the prevalence of these common alignments of actors, and the extent to which they occur on certain types of issues. Multidimensional Scaling Analyses The first set of analyses use multidimensional scaling techniques to explore the structures in the alignments of actors to the extent that these exist. Multidimensional scaling techniques provide graphical summaries of objects, in this case EU actors, on the basis of a matrix containing numerical information on the dissimilarities between each pair of objects. The matrices on which the following analyses are performed contain the total distances between each pair of actors, summed over all of the issues in the dataset. Therefore, if two actors take consistently different (similar) positions on many issues, there will be a relatively large (small) number in the cell of the matrix containing the dissimilarity between these two actors. Multidimensional scaling techniques generate a map of the objects that best represents the original data from the dissimilarity matrix. The researcher specifies the number of dimensions that these maps have, and can assess the extent to which the maps give an adequate representation of the dissimilarity matrices by referring to various goodness-of-fit measures. Separate analyses were performed on the EU-15 data and the EU-25 data. The EU-15 data were analyzed extensively using multidimensional scaling by Thomson et al. (2004). Here, I repeat the main findings of that analysis and compare them with the findings from the EU-25 data. I apply individual difference models (Cox and Cox 2000: Chapter 10) to represent the aggregated distances among the EU actors. 3 Figure 2 contains a two-dimensional map of the aggregated distances between the EU-15 member states, the Commission, the European Parliament and the 3 Thomson et al. (2004) used the ALSCAL program to explore the EU-15 data. Here, I apply the PROXSCAL program to explore the EU-25 data. ALSCAL applies Carroll and Chang s (1970) individual distances scaling procedure (INDSCAL). PROXSCAL applies Carroll and Chang s (1972) generalization of their INDSCAL model, called individual distances in orientation scaling (INDIOSCAL). INDIOSCAL is an improvement upon INDSCAL since it allows the spatial representation of the objects to be manipulated to a greater degree by the original dissimilarities between the objects (Cox and Cox 2001: 211). I also examined the EU-25 data using the ALSCAL program, which gave results virtually identical to the PROXSCAL program. In due course, I will re-examine the EU-15 data with PROXSCAL. The matrices on which the following analyses are performed are symmetrical: the distance between actor A and actor B is the same as the distance between actor B and actor A. Thomson et al. (2004) also analyzed asymmetrical matrices, in which the distances were weighted by the levels of salience each actor attached to each issue. The results did not differ substantively. Multidimensional scaling procedures are able to treat the matrices as containing either interval-level information or ordinal-level information. I report the results assuming ordinal-level information, but the results for interval-level assumption are virtually identical. Indifferent actors (actors with no positions on an issue) are coded as having a distance of zero to all other actors. Also, 15 of the 55 issues in the EU-25 dataset are still pending, so we have no information on decision outcomes. In these cases, for the time being the outcome is also coded as having a distance of zero to all actors. I examined alternative ways of dealing with missing positions extensively using the EU-15 data, none of which substantially altered the results. 7

8 reference point. 4 The first dimension, which runs horizontally, separates the reference point to the left, the member states in the middle, and the Commission and European Parliament to the right. The second dimension, which runs vertically, separates the Northern member states to the top, and the Southern member states to the bottom. I refer to this alignment as the North-South alignment for brevity, but note that it is not perfectly correlated with geographic latitude. Finland, for instance, is not at an extreme position in this alignment. The goodness-of-fit measures indicate that this spatial representation summarizes well the distances between the actors in the original matrix of dissimilarities. Two dimensions offer the most appropriate representation: the fit of the one-dimensional solution is considerably poorer, and the threedimensional solution can only improve marginally on the fit of the two dimensional solution. 5 <FIGURE 2> <FIGURE 3> Figure 3 contains a two-dimensional map of the aggregated distances between the EU-25 member states, the Commission, the EP and the reference point. The first dimension, which runs horizontally, again separates the Commission and the reference point. The Commission and the EP do not appear to be as close as they were before enlargement. The second dimension, which runs vertically, separates the new and old member states. Most of the new member states are located at the top of the graph, while the old member states are clustered together at the bottom. Some remnants of the North-South divide are also visible, although this appears to be looser than in the EU-15. Again, the goodness-of-fit measures indicate that the spatial summary provides an accurate summary of the information in the distance matrix. As in the previous scaling analysis, the two-dimensional solution improves upon the fit of the one-dimensional solution, while the three-dimensional solution can hardly improve upon the fit of the two-dimensional solution. 6 Issue-Level Analyses The following issue-level analyses complement the multidimensional scaling analyses. I first examine the frequency of the alignments of actors identified by the multidimensional scaling solutions, as well as the frequency of some alignments not identified in the previous analyses. I then describe the nature of the controversies identified by the key informants, and the extent to which common alignments are found of particular types of issues. These analyses allow us to interpret the alignments more substantively. The information in Table 4 indicates that the actor alignments identified by the multidimensional scaling analyses occur in a substantial minority of controversial issues. Consider first the positions of the supranational institutions, the Commission and the European Parliament. In the EU-15, the Commission was a radical actor, in 4 For presentation purposes, the orientation of the maps was changed so that the Commission and reference point are at position zero on the y-axis, with the reference point to the left and the Commission to the right, and so that the outcome is at position zero on the x-axis. 5 The one-dimensional solution has a Kruskal s stress value of.32, the two-dimensional solution.14, and the three-dimensional solution The one-dimensional solution has an S-stress value of.15, the two-dimensional solution a value of.06, and the three-dimensional solution a value of.04. 8

9 the sense that its position was at the opposite end of the issue continuum to the reference point on 35 percent of the issues. This occurs on 27 percent of the issues in the EU-25 data. The average distance between the Commission and the reference point also dropped by approximately ten scale points from EU-15 to EU-25, suggesting that Commission proposals are closer to the status quo. 7 <TABLE 4> Similarly, the EP s position is at the opposite end of the issue continuum to the reference point on 28 percent of the issues in the EU-15 and 24 percent of the issues in EU-25. We also observe a fall of approximately ten scale points in the average distance between both the EP s positions and the reference points between EU-15 and EU A comparison of Figures 2 and 3 suggests there is less agreement between the Commission and the EP in the EU-25 compared to the EU-15. This is confirmed by the issue-level analysis. We define the Commission and the EP s policy positions as close, if the distance between them is ten scale points or less. We find that the Commission and the EP are close on 33 percent of the issues in EU-15, which falls to 24 percent in EU-25. Similarly, the average distance between the Commission and EP s positions increases from (s.d , n=137) in the EU-15, to (s.d , n=44) in the EU-25 (t=-1.62, p=.11). The multidimensional scaling solutions also identify patterns in the alignments of member states. In the EU-15, there is a clear North-South alignment. This finding is in line with previous research on coalitions in the Council. Elgström et al. s (2001) survey of Swedish civil servants in the Council revealed that coalitions that divide Northern and Southern member states were common. Mattila and Lane s (2001) analysis of voting records in the Council also revealed a difference between the voting behavior of Northern and Southern member states in the Council. Table 4 indicates that there was a significant North-South alignment on 35 percent of the issues in the EU-15. We are less likely to encounter a significant North- South alignment in the EU-25. Northern member states disagree with Southern member states on 24 percent of the issues in the EU-25 dataset. In the EU-25, disagreement between new and old member states is more common (33 percent) than disagreement between Northern and Southern states (24 percent). Other possible alignments of actors were also examined, including disagreement between Left and Right-wing governments. The Left-Right dimension is clearly an important dimension of political contestation in the European Parliament (e.g. Hix et al. 2007: chapter 9). However, the evidence is mixed regarding the prevalence of alignments that distinguish Left-wing and Right-wing governments in the Council. Mattila s (2004) analysis of voting behavior in the Council concludes that Right-wing governments that are not strong supporters of European integration tend to vote against the Council majority more than Left-wing governments. However, Right-wing governments that are strong supporters of European integration do not vote against the majority more than Left-wing governments. 7 The average distance between the Commission and the reference point was (s.d , n=130) in EU-15, and (s.d , n=45) in EU-25 (t=1.47, p =.142). 8 The average distance between the EP and the reference point was (s.d , n=100) in EU-15, and (s.d , n=40) in EU-25 (t=1.63, p=.11). 9

10 By contrast, the present analysis shows that the alignment of member states according to the Left-Right position of their governments is rare (Table 4). 9 The Left- Right positions of states governments correlate significantly with the alignments of member states on only three to four percent of issues. This is about what we would expect to find on the basis of chance alone. The data examined here concern policy positions, while the data on the basis of which there is some evidence of the Left- Right dimension concern voting behavior (Mattila 2004). This may account for the different findings. Right-wing governments that are unsupportive of European integration may wish to signal their ideological positions by voting against the majority relatively frequently. However, as far as their positions on the controversies raised by legislative proposals are concerned, they do not form a cohesive group. I also searched for other possible alignments of actors and found none of these to be particularly common. Thomson et al. (2004) report few significant correlations between the prevalence of pro-european attitudes in each country and states positions on controversial issues. Per capita GDP is obviously strongly correlated with the differences between Northern and Southern member states and with the difference between new and old member states. This does not, however, add any substance to the interpretation of the alignments found. For this, we must identify the nature of the issues on which the alignments occur. <TABLE 5> Table 5 shows the categorization of each of the 239 controversial issues according to three criteria (Thomson et al. 2004: 251). Each criterion refers to the nature of differences between the policy alternatives on each issue. The first is whether the controversy involves choices between different levels of European harmonization. 23 percent of issues meet this criterion in the EU-15, and 35 percent in EU-25. The second criterion is whether the issue involves choices between different levels of regulatory intervention: choices between regulation and free market. This covers just over half of the issues in both the EU-15 and the EU-25. The third criterion is whether the issue involves choices between different levels of financial subsidies. 15 percent of the issues involve levels of financial subsidies in the EU-25, 27 percent in the EU-25. The issue of the minimum size of health warnings on tobacco products is an example of an issue that meets the first two criteria. Higher minimum health warnings result in more harmonization between member states. They also result in more regulation of industry. The level of EU subsidy for sugar production meets the second and third criteria. Lower subsidies do not result directly in less harmonization across the EU, since the same level of subsidies applies throughout. Therefore, it is possible for an issue to be placed both in the regulation category and either the harmonization category or the subsidy category. The alignment found on the first dimension of both spatial analyses, with the 9 In addition to the Left-Right positions of governments as measured by expert judgments (Huber and Inglehart 1995; Benoit and Laver 2006), reported in Table 4, Thomson et al. (2004) also applied estimates of governments Left-Right positions based on election programmes (Budge et al. 2001). These alternative measures gave similar results. For the EU- 15 analysis reported in Table 4, the estimates of the Left-Right positions of the parties in government are weighted by the numbers of ministerial posts each party received. For the EU-25 analysis, I simply took the average of governing parties Left-Right positions. Governments positions on Benoit and Laver s taxes versus spending dimension gave similar results.

11 Commission opposite to the reference point, bears a superficial resemblance to the supranational scenario to which many models of the EU s legislative procedures have been applied (e.g., Tsebelis 1994; Tsebelis and Garrett 2000). In this scenario, the political space is conceived of as a uni-dimensional policy continuum on which the status quo, actors preferences, and the decision outcome can be placed. At the extreme left of this continuum, we find the status quo that is the potential policy outcome that incorporates the least harmonized or integrationist solution. At the extreme right of the continuum, we find the preferences of the Commission and the European Parliament. There supranational institutions are said to be natural allies on this dimension; they both prefer the most extensive harmonization or integration of Member States policies. The resemblance is only superficial, because this alignment is not concentrated in the group of issues concerning choices between more or less harmonization. Considering the EU-15 data for example, of the 60 issues on which the Commission and reference point are at opposite extremes, only 27 percent are harmonization issues. Indeed, this alignment is not concentrated in any particular category of issues. Consequently, Thomson et al. (2004: 256) interpret this alignment in terms of policy change, rather than as part of a particular substantive disagreement. The Commission introduces radical proposals in the knowledge that these proposals are bound to be amended during the course of negotiation and compromise in the Council. <TABLE 6> <TABLE 7> The North-South alignment was particularly visible in the EU-15; it occurred on 35 percent of the issues. The issues on which there was disagreement between Northern and Southern states disproportionately concerned choices between regulation and market-based solutions to policy questions (Table 6). In general, the Northerners supported less regulation than the Southerners. For example, on agricultural subsidies, copyright legislation, and banking regulations Northern states were significantly more liberal than Southern states. The minority of issues on which there was a North-South disagreement with the Northerners supporting more regulation concerned animal welfare and environmental protection. Given the nature of the issues on which North-South disagreements are found, it would not be appropriate to interpret this alignment solely in terms of financial subsidies. In the EU-15, the association between North-South and the broader group of regulation issues is stronger than with the narrower group of subsidy issues (Tables 6). The information in Table 7 indicates that as the North-South alignment has weakened with enlargement, so too has the concentration of this alignment in the group of regulatory issues. Disagreement between old and new member states is disproportionately concentrated on issues about levels of financial subsidies. On issues that do not concern levels of financial subsidy, 25 percent evoke significant disagreements between old and new members. Of the issues that do concern levels of financial subsidies, 53 percent evoke significant disagreements between old and new states. In all cases, the new member states generally favored higher subsidies. Note that disagreement between old and new members is concentrated in this sub-group of subsidy issues, rather than the broader group of regulation versus market issues. In

12 this respect, the old-new alignment in the enlarged EU is qualitatively different from the North-South alignment in the EU-15. Decision Making The data on decision making in the EU-15 have been analyzed extensively with a range of models of legislative decision making (Thomson et al. eds. 2006). Each of these models aims to capture the essence of the process through which actors favored positions at the time of the introduction of the legislative proposals are transformed into decision outcomes at the time of adoption. The models are broadly similar in the sense that they make predictions of decision outcomes given a certain alignment of actors positions. Moreover, they are similar in the sense that they are rational choice institutionalist models. Actors are assumed to be purposeful, in the sense that they strive to achieve decision outcomes in line with their positions subject to constraints. However, the models differ fundamentally from each other in the nature of those constraints. Procedural models focus on the formal rules of the game, for example the treaty rules that stipulate which actors have the right of initiative, and what the required thresholds of support are for proposals to be adopted (e.g. Steunenberg and Selck 2006; König and Proksch 2006). Bargaining models focus on the informal norms that constrain actors during negotiation. Some bargaining models are based on cooperative assumptions, while others are based on non-cooperative assumptions (e.g. Achen 2006a; Arregui et al. 2006). Mixed models combine elements of formal rules and informal bargaining (Bailer and Schneider 2006; Boekhoorn et al. 2006; König and Proksch 2006; Widgrén and Pajala 2006). One of the main findings of the analysis of the EU-15 data was that a simple bargaining model, the Nash Bargaining Solution or compromise model, generates more accurate predictions of decision outcomes than other models. The following recapitulates the findings regarding the application of a procedural model and the Nash Bargaining Solution/compromise model to the EU-15 data. It also applies these models to the data on decision making in the enlarged EU. These two models are chosen because they encapsulate the most contrasting views on the nature of the decision-making process. A Procedural Model The procedural model considered here is König and Proksch s (2006) application of Tsebelis and Garrett s (2000) spatial model of legislative choice in the EU. This procedural model is distinct from other procedural models in that it is applied to each issue separately, rather than to a multi-dimensional space constructed using all of the issues raised by each legislative proposal. There are several advantages to this issueby-issue approach. First, it makes the model comparable with the Nash Bargaining Solution, which is also presented as an issue-by-issue model. Second, in reality issues are often dealt with separately, rather than combined in the working groups. Therefore, the issue-by-issue approach is realistic. Third, the construction of a multidimensional space using the issues requires arbitrary assumptions, such as that the issues are connected orthogonally. Steunenberg and Selck (2006) develop and apply other procedural models, including those that combine separate issues into a multidimensional space. None of these variants generated more accurate predictions.

13 <FIGURE 4> The top part of Figure 4 illustrates the predictions made by the procedural model on issues subject to consultation. The relevant actors are the Commission, with the exclusive right to initiate proposals and the pivotal member state. The alignment of actors in Figure 4, with the Commission to the right, is only for illustrative purposes; it is not part of the model. Under the unanimity requirement, the pivotal member state is the one whose position is at or is closest to the status quo. In Figure 1, the pivotal member state (at position 10) is indifferent between the status quo (position 0), and position 20. Consequently, the model predicts that the Commission will make a successful proposal at position 20 on the issue continuum. Under qualified majority voting (QMV), we must consider both the required QMV threshold and the possibility of an amendment to the Commission s proposal supported by all member states. The pivotal actor (or actors) under QMV is the member state that turns a losing minority into a blocking minority, when counting votes from the state closest to the status quo. In the EU-15, this is the member state that controls the 26th vote, since 62 out of 87 votes were needed for a qualified majority. In the EU-25, this is the member state that controls the 90th vote, since 232 out of 321 votes were needed for a qualified majority, as well as at least 13 member states with 62 percent of the EU population. Under QMV, the Commission can secure a decision outcome up to the point at which the QMV pivot is indifferent between the Commission s proposal and the possible unanimously supported amendment by the Council. This is position 40 on the top of Figure 4. The codecision procedure is a bargaining game between the pivotal member state in the Council and the European Parliament. The model assumes that the Council and the EP have equal power under the codecision procedure. This does not, however, necessarily mean that the outcome is exactly half way between the position of the Council pivot and the EP. The bargaining space ends when either the Council pivot or the EP is indifferent between the outcome and the status quo. Therefore, under QMV, the outcome predicted on the basis of the configuration at the bottom of Figure 4 is 60. Note that this is not half-way between the position of the QMV pivot (position 30) and the EP (position 100). Under the less common rule of codecision combined with unanimity in the Council, when the unanimity pivot is at position 10 and the EP is at position 100, the predicted outcome is 20. The Nash Bargaining Solution / Compromise Model The compromise model s prediction is simply the mean average of the actors policy positions, weighted by the product of their capabilities and the levels of salience they attach to the issue on which the prediction is being made. As a formula: O a n ia i= 1 = n x c s i= 1 i i c s ia ia Where O a is the prediction of the decision outcome on issue a. x ia denotes the position of actor i (from the set of actors, n) on issue a. c i denotes the capabilities of actor i. s ia is the level of salience actor i attaches to issue a.

14 The compromise model was first proposed in this form by Jan Van den Bos (1991) in his study of decision making in the Council of the European Community. When describing the decision-making process this model represents, he emphasized that it takes all positions of Member States into account, weighting these by the resources a Member State can apply during the negotiation and the importance each attaches to the decision at hand (Van den Bos 1991: 176). The compromise model is not concerned with the composition of actors capabilities. That is exogenous to the model. Rather, it is concerned with the transformation of actors positions into decision outcomes, and how the relative capabilities and levels of issue salience affect this transformation. Achen (2006a) greatly improved the theoretical standing of the compromise model. First, he drew parallels between this model and the research traditions of institutional realism in political science and social action theory in sociology. He concluded that [t]his sophisticatedly simple equation neatly summarises much of the previous century s thought about political policy-making (ibid.: 94). Second, Achen proved that if a certain condition is met, the compromise model is a first-order approximation of the Nash Bargaining Solution (Nash 1950). Nash formulated the bargaining solution as an answer to the question of what each actor should get in a situation where they must collaborate for mutual benefit. Informally, the essence of Nash s answer is that it is the decision outcome that minimizes the utility losses of the actors involved. Achen s insight is that if the disagreement outcome is highly undesirable, the compromise model and the Nash Bargaining Solution are one and the same. It is clear that the disagreement outcome is generally highly undesirable in EU decision making, and therefore that the compromise model is an appropriate formula with which to represent the Nash Bargaining Solution in this context (Achen 2006a: 101-3). Close observers of decision making in the EU know that negotiators go to great lengths to avoid breakdown in discussions, even when parts of the legislative proposal are unpopular. Heyes-Renshaw and Wallace refer to this as the imperative of making propositions yesable (2006: 303). The successful application of this norm is also evident in the paucity of no votes in the Council. It is true that on certain controversial issues, including those studied in the dataset examined here, there often member states who would prefer the so-called reference point. The reference point is the decision outcome that would prevail if no decision were taken. However, this reference point does not capture two very important negative consequences of a failure to agree. The first is that other, perhaps uncontroversial parts of the legislative proposal would be lost if no agreement were reached. The second is that breakdowns in the decision-making process are damaging to the long-term relationships among decision makers, a cost that is not worth bearing unless the stakes are extremely high. To apply the compromise model, estimates are also needed of the relative capabilities of the three institutions, the Council, the Commission and the European Parliament (EP), as well as the relative capabilities of the member states. For the analysis of the issues in the EU-15 dataset, I use Shapley-Shubik Index (SSI) scores (Shapley and Shubik 1954; Thomson and Stokman 2006: 48). These scores provide estimates of actors power in decision situations where a vote could be taken before a decision is adopted. To apply this index, a list of all possible permutations of actors is made, the member states, the Commission and the EP. For each possible permutation, the actor that turns a losing coalition into a winning coalition is identified and said to be pivotal. The number of times an actor is pivotal divided by the total number of times all actors are pivotal is that actor s SSI power score. The Commission is not a

15 pivotal actor in the codecision procedure. Therefore, it has an SSI score of zero on all codecision issues. The EP has an SSI score of 45 percent of the total Council under the QMV variant of codecision, and 7 percent of the total Council under the unanimity variant of codecision. The EP is not a pivotal actor under consultation. Therefore, it has an SSI score of zero on all consulation issues. The Commission has an SSI score of 45 percent of the total Council under the QMV variant of consultation, and 7 percent of the total Council under the unanimity variant of consultation. 10 The member states relative SSI scores are very similar to their relative number of qualified majority votes when the QMV rule applies under both codecision and consultation. All member states have equal SSI scores when unanimity is required in the Council. For the EU-25 data, I use improved capability scores that are based partly on information from the EU-15 study. 11 Under codecision combined with QMV, the SSI scores underestimate the capabilities of the Commission, and overestimate the capabilities of the EP. Under consultation combined with QMV, the SSI scores overestimate the capabilities of the Commission, and underestimate the capabilities of the EP. Thomson and Hosli (2006) identified the relative capability scores for these three institutional actors that minimize the prediction errors of the compromise model. 12 These scores are used in the present analysis. Under the codecision procedure, the Commission has 15 percent of the capabilities of the Council as a whole, while the EP has 25 percent of the capabilities of the Council as a whole. Under the consultation procedure combined with QMV in the Council, the Commission has 30 percent of the capabilities of the Council as a whole, while the EP has 15 percent of the capabilities of the Council as a whole. Under consultation combined with unanimity voting in the Council, all capabilities are held by the Council members. For the relative capabilities of the member states in the EU-25, I apply capability estimates based on interviews with key informants. When the unanimity rule applies in the Council, the SSI scores of all member states are equal. This is not realistic. As Heyes-Renshaw and Wallace note, [i]t would be naïve to suggest that some member states do not carry more weight than others in the Council (2006: 252). The estimates by key informants are described and analyzed by Bailer (2004) and Thomson and Stokman (2006: 51). Expert judgments of the relative capabilities of the EU-15 member states, for both the QMV rule and the unanimity rule, are very close to 10 It is assumed that winning coalitions under codecision-qmv consist of either a qualified majority of member states and the EP, or all member states without requiring the EP. Likewise, under consultation-qmv, it is assumed that winning coalitions consist of a qualified majority of member states and the Commission, or all member states without requiring the Commission. Several authors in Thomson et al. eds. (2006) also applied a variant of the Shapley-Shubik Index in which the Commission was a necessary member of all winning coalitions. Models loaded with these alternative SSI scores generated significantly less accurate predictions. 11 Because these improved capability estimates use information from the EU-15 study, it would not be appropriate to use them to assess the relative predictive power of the Nash Bargaining Solution and the procedural model for the EU-15 data. This would unfairly advantage the Nash Bargaining Solution over the procedural model. 12 Thomson and Hosli (2006) identified these weights by systematically applying several thousand alternative combinations of capability scores for the Commission, Council and EP, to the EU-15 data using the compromise model. They then identified which weights were associated with the most accurate predictions. These optimal weights used here for the analysis of the EU-25 data.