Abstract
Resolving controversies with DEU data
In many respects the publication of Robert Thomson’s (2011) new book Resolving Controversy in the European Union (hereafter RCEU) can be seen as a culmination of the Decision-Making in the European Union (DEU) project that started more than a decade ago. Thomson’s book is a successful attempt to use the DEU data to provide a systematic and cohesive analysis of how the European Union manages to convert competing policy demands into legislative outputs in its decision-making process. Thus, the aim of the book is ambitious: to provide a comprehensive analytical examination of the way decision-making works in the European Union.
The goal of my evaluation in this forum article is twofold. Robert Thomson has been one of the key members of the DEU project from the beginning and the DEU data set forms an important part of Thomson’s empirical analysis. Therefore, in the first part of this article I present an evaluation of the significance of the DEU project for EU studies in general. Then I proceed to discuss the new book in more detail. I will focus on Thomson’s general approach and main findings, trying to highlight why this book is such an important addition to our knowledge of how the EU works.
The DEU project
In its general approach the DEU project continued and expanded the work that Bueno de Mesquita and Stokman (1994) presented in their edited book European Community Decision Making. In this project, the authors used quantitative data and formal rational choice models to predict the outcomes of controversial decisions in four issue domains of EU decision-making. In one of the book’s opening chapters, Pierce (1994: 14) writes about the limitations of their study and notes that future studies on the representativeness and democratic accountability of the EU need more comprehensive and systematic empirical data. Then he lists what such a data set would ideally include. This list comprises, among other things, estimates of the positions of the EU member states on issues to be decided by the EU institutions and the salience of these issues to the member state governments. In many ways the DEU project 10 years later was a direct answer to Pierce’s call.
The original DEU project started in 1998 and the research group collected data on controversial Commission proposals that were discussed in the Council in the period January 1999 – December 2000. The data cover 66 legislative proposals introduced or pending in 1999–2000. Within these 66 proposals, 162 controversial issues were identified. The actual data collection was based on semi-structured interviews with EU decision-making experts, mostly civil servants working in member state representations, the Commission, the European Parliament (EP) and the Council Secretariat. In addition, some interest groups representatives were interviewed. Altogether, at least 150 interviews with 125 experts were conducted (Thomson and Stokman, 2006). The experts provided information on actor positions (member states and EU institutions) on the selected issues, the level of salience each of the actors attached to each of the issues, and estimates of actor capabilities (that is, their power).
The main results of the DEU project were presented in a special issue of European Union Politics (Stokman and Thomson, 2004) and in the book The European Union Decides edited by Thomson et al. (2006). In addition to these two edited volumes, the DEU data have also been used in dozens of articles and other publications and consequently their importance for EU studies has been remarkable. Probably the best-known applications of these data are the various formal models used to explain negotiation dynamics in EU decision-making (Thomson et al., 2006). The usefulness of the DEU data set is not, however, limited to these modelling applications. The same data have been used in numerous other studies, often with more empirically oriented starting points. As a result, the DEU data have increased our knowledge of the EU decision-making political space and coalition patterns (Kaeding and Selck, 2005; Thomson, 2009; Zimmer et al., 2005), member state success in EU negotiations (Bailer, 2004; Selck and Kuipers, 2005) and power of the Council presidency (Schalk et al., 2007; Thomson, 2008; Warntjen, 2008), to mention just a few studies. Furthermore, the DEU data have been used to explain member states’ strategic negotiating positions (Bailer, 2011), the role of salience in negotiations (Schneider et al., 2010) and possible ‘vote selling’ in the Council (Golub, 2012).
The evaluation of such an extensive research effort as the DEU project is not a simple task. One, albeit limited, way to assess the importance of any research project is to look at the number of citations its main publications have received in the wider political science community. According to my small and rather unscientific study, the book The European Union Decides (Thomson et al., 2006) – that is, the main single book result from the DEU project (at least before Thomson’s new book) – has been cited as a single volume 184 times since its publication and the individual chapters of the edited volume have received 341 citations altogether. 1 Furthermore, the articles in the special issue using the DEU data published in this journal in 2004 (Stokman and Thomson, 2004) have received 337 citations. Without a comparison point it is not straightforward to interpret the significance of these numbers, but I think it is safe to say that the DEU project has had a major impact in the field of EU studies.
Of course, a project of this magnitude has attracted a fair amount of criticism as well. This criticism can be divided into two strands. First, there are critical evaluations related to the theoretical approach, the methodological approach or the results of the DEU project as a whole. The second strand of criticism is more constructive in the sense that it does not question the whole DEU approach as such but is more ‘technical’ in nature, relating mostly to theoretical modelling aspects, data validity or reliability considerations.
The first kind of criticism is often related to the substantive results that the project delivered. For a long time the field of EU studies was dominated by the case-study approach, which is still very strong among EU scholars. Pahre (2005: 119) notes that, because the DEU researchers treat the issues in their data set ‘merely as sources of data and not as cases, these … [studies] … will not persuade the case-study community that formal theory makes useful contributions’. He may be correct in his assessment, but his wish to persuade the case-study community is perhaps a little bit too idealistic and far-fetched. Finding common ground is a difficult, perhaps even impossible, task given the current variety in theoretical and methodological approaches in the field of EU studies.
One recurring theme in criticism of the DEU project is that the formal modelling approach, despite its theoretical clarity and parsimony (or perhaps because of it), has not really produced interesting new results about EU decision-making. For example, John (2008: 111–12) notes in her review of The European Union Decides: ‘[t]he conclusion of the models, that bargaining and compromise are central for EU political decision making, and procedural models based entirely on legal rules are neither sufficient for explanations, nor do they surprise researchers, observers or practitioners. Hence, the main added value of this volume should rather be seen in its contribution to the debate on theoretical modelling and research.’
Similar but somewhat harsher criticism is offered by Heisenberg (2008), who is a strong supporter of qualitative, more case-oriented approaches to EU studies. 2 She acknowledges the usefulness of the hard empirical work behind the project’s results but then goes on to argue that ‘the quantification of preferences that comprises the [DEU] dataset does not add significant value to the overall enterprise of understanding the Council … The conclusions do not suggest any radically new insights into Council behaviour’ (Heisenberg, 2008: 274). According to Heisenberg, the reason for the inability of the modelling approach to produce new results is related to its failure to acknowledge the ‘robust exceptionalism in Council of Ministers’ decision making norms’ (2008: 261). Thus, the argument is that formal models are too simplified versions of reality to capture the real nuances of day-to-day decision-making in Brussels.
Is the criticism that the DEU project has really failed to produce any useful new results justified? I don’t think so. Nonetheless, I must admit that it was somewhat disappointing that one of main the results in The European Union Decides was that the simplest models based on policy position averages produce more accurate results than the more computationally complex ones. 3 It is true that previous studies, mostly based on qualitative case studies, have shown that bargaining and attempts to accommodate all views are defining features of EU decision-making. What has been lacking, however, are more general studies with a larger number of cases and a systematic analytical approach that would verify these observations. The DEU project has been able to fill this gap. In addition, the criticism has almost always focused only on studies that use the formal modelling approach to analyse EU decision-making. As I mentioned earlier, there have been numerous other studies using the DEU data that have applied other, often more empirical approaches to produce new and invaluable information about the way the EU makes its decisions.
The second group of critics is more understanding towards the DEU project as a whole. These critics call attention to theoretical, conceptual or data problems within the project but usually with the aim of improving the models or data collection in similar future projects. A good example of this kind of constructive criticism is the article by Bueno de Mesquita (2004) where he discusses, among other things, measurement problems of the member state policy positions as well as salience and how these problems limit what kinds of model can be applied to the data and how these problems may affect results. He also discusses problems related to systematic comparison of different model outcomes. Others have pointed out how the models could be further improved (Hörl et al., 2005; König, 2005; Sullivan and Selck, 2007) or how the misspecification of actor preferences and the exclusion of actors’ saliencies can seriously distort the results (Junge and König, 2007). The importance of salience in EU decision-making is also emphasized by Warntjen (2012).
I really do not have much to add to this second type of criticism. I do think that most of the formal models employing the DEU data understate the importance of issue linkages in the bargaining process. It may be the case that ‘vote trading’ or ‘log rolling’ at the proposal level are not as frequent as, for example, Mattila and Lane (2001) more than a decade ago believed (see Golub, 2012). However, it is more difficult for me to believe that actors bargaining on the controversial issues within an individual legislative proposal would not engage in some sort of give-and-take process to reach a compromise. I also think that research based on the DEU data too often overlooks the ‘exogenous’ changes in actor preferences during the bargaining process that may occur, for example, when there is a member state government change following a national election (see Miklin, 2009). Furthermore, in the future I would like to see probabilistic models instead of deterministic ones applied to the data (see Achen, 2006a: 271–2) because then we would not have to settle for point estimates to compare the accuracy of competing models. This would probably also make analysing the robustness of the models easier. As far as I know, there has not yet been a systematic analysis of how well the models perform when the inherent reliability problems of the interview data are taken into account, although Junge and König’s (2007) article is a good starting point in this regard. These are important critical observations but their overall significance is probably quite limited. The most important thing is that the DEU researchers have taken this kind of ‘friendly’ criticism into account in their more recent work and adjusted their models accordingly.
Resolving Controversy in the European Union
The main goal of Robert Thomson’s new book is ambitious and wide-ranging: to examine how the contemporary EU system of legislative decision-making works (RCEU: 5). Following Easton’s systems theory logic, Thomson divides his book into three sections: inputs, processes and outputs. In the first section, inputs or competing policy demands (as Thomson also calls them) are scrutinized. This analysis concerns questions such as how the member states differ in their policy demands, whether these differences form consistent patterns over several policy sectors and, most importantly, whether these differences can be systematically explained by factors related to actors’ interests. He also pays attention to how member states’ policy demands deviate from those of the two other main actors in the EU policy process: the European Commission and the EP. The second section analyses processes, which refer to the mechanisms by which policy demands are transformed into policy outputs. Here Thomson relies mainly on the modelling approach that has also been applied in several previous studies that partly use the same DEU data (Stokman and Thomson, 2004; Thomson et al., 2006). Finally, in the outputs section Thomson turns to analysis of the contents of EU laws. Do outputs systematically reflect some actors’ policy demands more than others’, that is, are there systematic winners or losers in the EU’s decision-making process? In this last section, Thomson also studies the extent of delegation of implementation to the Commission or, alternatively, to national authorities in member states.
With the exception of his division of book chapters into three main sections of inputs, processes and outputs, Thomson does not apply the Eastonian systems theory any further. He briefly discusses other alternative theoretical approaches, such as neofunctionalism and intergovernmentalism, but then points out that these approaches are not well suited for the research questions he is interested in. Thomson chooses rational choice institutionalism as his theoretical approach, which assumes purposeful actors trying to achieve their policy goals within the formal and informal constraints imposed by the EU’s decision-making system. The relevant actors in Thomson’s book are the member states, the Commission and the EP, which all share important roles in the EU’s legislative process. In individual chapters, Thomson relies also on theories from comparative politics and legislative behaviour, such as informational and distributive theories of committee systems when analysing the EP’s decision-making, and transaction costs theory when analysing delegation.
As his empirical data, Thomson uses the DEU data (Thomson et al., 2006), which he has extended with his colleagues to cover the post-enlargement period to be able to analyse, among other things, whether and how the increase in the number of member states has affected the EU’s decision-making. The post-enlargement data include 56 proposals that were discussed and introduced in the Council after the accession of the 10 new member states in 2004. The latest of these were introduced in July 2008, meaning that in some cases Bulgaria and Romania are also included in the analysis. In total, the data set used in the book contains 125 legislative proposals that consist of 331 issues.
Collection of the new post-enlargement data followed closely the procedures used to collect the original DEU data. There are, however, some differences. First, the new proposals have on average a somewhat higher profile in public than the ones selected for the original DEU data (RCEU: 30). Second, the interview process has changed (RCEU: 35). In the original DEU interviews, each expert was allowed freely to define the issues within selected proposals. In the post-2004 study, the interviewers compiled a set of estimates of the issue dimension and actors’ positions from one expert and then presented this to another expert in the next meeting. The next expert was then asked if she agreed with the estimates from the previous expert. This procedure obviously holds the risk of the priming effect distorting the expert answers. The second expert may consciously or unconsciously follow the example given by the first expert and give different answers than she would have given without prior knowledge of the previous expert’s answers. Unfortunately, Thomson does not discuss how these changes in data collection could potentially affect the results of his comparisons of the pre- and post-enlargement decision-making processes.
Thomson starts the empirical part of his book with multidimensional scaling (MDS) analyses of actors’ policy positions. The results do not deviate much from the results we have seen in previous similar studies. In the pre-enlargement period, northern member states take different positions from the southern member states, and the results from the post-2004 period show evidence of a new–old member state division. Thomson interprets this new division to reflect different positions between older and new member states on levels of integration, the strength of regulation and levels of subsidies. Basically, new member states favour less integration, lighter regulation and higher subsidies than the older member states. In a more elaborate analysis, Thomson tests which factors can explain the observed actor alignments in the MDS analysis. The results show that member states with similar net EU budgetary positions and similar levels of wealth (GDP per capita) are more likely to have similar policy positions. He also shows that governments’ left–right positions are mostly uncorrelated with member state positions in the political space, indicating that left–right ideology does not play a major part in Council decision-making.
The main overall conclusion is that there is a great deal of variation in the positions taken by the main EU actors. Thomson interprets this observation as one of the factors that make the EU a successful political system because a lack of stable coalitions of member states ensures that there are no clear winners or losers in EU decision-making. On the other hand, Thomson sees that this lack of structure limits the capacity of EU citizens to ‘exert visible control over EU policies’ (RCEU: 76). He does not elaborate this point much further here, but my interpretation is that stable blocs of member states in the Council would lead to a situation where citizens could somehow use their voice in elections ‘to make or break’ these coalitions. However, it is difficult for me to see how this could work in practice, since member state representatives in the Council are elected indirectly in national elections, not in Europe-wide elections.
In the next three chapters, Thomson turns to the explanatory analysis of the policy positions of the Commission, the EP and member states. He discovers that the Commission’s policy positions are not affected by the responsible commissioners’ party political affiliations but the Commission’s position is biased towards the position of the home member state of the responsible commissioner, especially under qualified majority rule. This leads Thomson to conclude that the Commission is clearly not a unitary actor. When analysing the EP’s policy positions, Thomson finds that they are generally in line with those of the median Member of the EP. Thomson’s results also show that rapporteurs from small party groups and small member states bias the EP’s opinions more than rapporteurs from large party groups or large member states. Unfortunately, this somewhat counterintuitive observation is not discussed very comprehensively in the text. Finally, when analysing member state policy positions, Thomson finds that domestic economic and political interests exert significant effects, although the magnitudes of these effects are mostly quite small.
Thomson’s analysis of policy positions is based on dyadic data. For example, when Thomson analyses member state positions, the dependent variable is the distance in policy positions between pairs of member states. Likewise, the explanatory variables are measured as differences between pairs of member states. These kinds of data present numerous technical challenges because the observations are obviously not independent from each other. Thomson’s solution is to use multilevel models where the observations are hierarchically nested and sometimes also cross-classified. The application of these kinds of model sometimes involves difficult choices between various alternative model specifications. I am sure that Thomson’s choices are the most appropriate for his data setup, but I would have been interested in reading a more thorough explanation of how the actual analyses were performed.
The second section of the book analyses processes, that is, the way in which actors’ policy positions are transformed into decision outcomes. Thomson starts with formal models, following the example set by The European Union Decides (Thomson et al., 2006). He compares the predictive accuracy of one procedural model and four bargaining models. Given the results from earlier modelling applications, the outcomes are not very surprising. Two of the models clearly outperform the rest: a simple mean average of actor positions or a salience-weighted mean of these positions. The results lead Thomson to two conclusions: first, actors prefer to resolve controversies by adopting proposals in at least some form; and, second, the enlargement has not fundamentally changed the way in which the EU resolves controversies.
Next, Thomson proceeds to examine the relative power of EU institutions in a quite innovative way. To do this he uses the best-performing model from the previous chapter but with an important modification: he includes actor capabilities in his model. Member states’ power is measured with the Banzhaf voting power index. 4 Thomson proceeds by calculating different model predictions while varying the power scores assigned to the Commission and the EP. The solution that minimizes the distance of the model predictions from the actual outcomes shows the appropriate power scores for the Commission and the EP. The best-fitting power scores indicate that the Commission and the EP together have up to about 30 percent of the Council’s power. Of course, the validity of this result is based on the assumption that the model used in the analysis is an adequate representation of the EU bargaining process. It is important to bear in mind that even the best-performing model predicts outcomes that deviate on average more than 23 points from the actual outcomes on a scale from 0 to 100.
The third chapter in the section of processes, examines the relative power of member states. Thomson makes alternative hypotheses about power differences between groups of member states and then applies the model again to see which power distribution delivers the most accurate predictions. His main result is that differences between the predictions of equal and regressive power distributions are quite small, indicating, contrary to most expert opinions, that bigger member states with more Council votes are really not that much more powerful than smaller member states with fewer votes. According to Thomson’s interpretation, this result is related to the fact that there are few issues on which large member states systematically take positions that differ from those of smaller member states. If large member states do not agree among themselves, their weight does not have any observable implications for actual decision outcomes.
In the final empirical section, Thomson turns to the analysis of decision outcomes. Are there systematic differences among actors between their policy positions and decision outcomes? Or, to put the same question in a different way, are there clearly distinguishable winners and losers in EU decision-making? Thomson’s answer is clear: although there are small combined differences between actor positions and outcomes, they are not systematically biased in favour of the policy positions of certain actors. However, decision outcomes are closer to member states that attach higher levels of salience to the issues in question. Also, Council presidency matters: outcomes are closer to the position of the member state holding the presidency when the final decision is made. Importantly, Thomson also observes that this effect of the Council presidency has declined since the 2004 and 2007 enlargements.
The final empirical chapter turns to the question of delegation. What affects the choice of delegating the implementation of EU decisions either to national member state authorities or to the Commission? The results show that expertise matters. Delegation to member states is more likely when decisions concern matters on which national authorities have particular expertise. Furthermore, the average level of policy position polarization has grown with the Eastern enlargement. This in turn explains why the likelihood of delegating discretionary power to member states has increased since the enlargement.
In the concluding chapter, Thomson relates his empirical findings to the discussion on the EU’s democratic deficit. He does not see increasing political competition as a solution to the EU’s problems, because this would damage the basis for the success of the EU. The lack of stable coalitions and the absence of left–right politics in the Council ensure that there are no consistent winners or losers among the member states. According to Thomson, increased political competition would lead to a situation in which some member states would be forced onto the losing side on a broad range of issues, which would be neither feasible nor wise (RCEU: 291). At the end of the book, Thomson offers some ideas on how to reform the EU. He starts by pointing out that the EU has reached a constitutional equilibrium that is unlikely to change in the foreseeable future. I am not quite so sure about this conclusion. The eurozone crisis has again brought about a discussion on the need for treaty revisions so that the EU can tackle current economic problems. However, most of the reform suggestions (for example, increasing transparency in Council decision-making, strengthening the EP’s role, suspending EP sessions in Strasbourg) offered by Thomson sound very sensible, although not always directly related to his empirical results.
Conclusions
In many respects the DEU project has led EU studies to a new level. It was the first project of its scale to analyse the EU’s decision-making system with a systematic rational choice approach. The project is often too narrowly connected only to analysis of the Council and, in particular, to the application of formal models in Council decision-making. The base of the project was formed by the data collection effort, which provided an extremely useful data set that has since been used in several studies, ranging from analysis of the EU’s political space to more concrete analyses of the effects of the Council presidency or commissioners’ nationality on decision outcomes, to give just some examples. In this way the DEU project and the accompanying data set have enormously enriched our knowledge of EU decision-making.
Robert Thomson’s new book is, at least partly, also a result of the original DEU project. However, unlike the previous books and articles using the DEU data, Thomson’s study is a comprehensive analysis of the EU decision-making process, starting with policy inputs from the member states and ending with delegation of the responsibility for implementation of EU decisions. Furthermore, the study covers a wide time period, which allows Thomson to systematically compare decision-making in the EU15 and in the enlarged Union. On the whole, Thomson’s goal of covering the most important aspects of EU decision-making is challenging and ambitious, but he succeeds in delivering a wide variety of new and convincing insights into EU decision-making. This is why the book deserves a very wide readership.