Legislative negotiations in the European Parliament

Abstract

In order to capitalize on its increasing legislative powers, the European Parliament (EP) has had to specialize internally. The scholarly literature has predominantly studied the structural manifestations of specialization, such as the allocation of committee seats among parliamentarians. This article sheds light on a second selection process: participation in legislative negotiations. For every given legislative dossier, parliamentarians have to decide whether and how strongly to participate in the negotiations. The literature on office allocation contains assumptions about the role of office-holders in negotiations. The first aim of this article is to test the validity of these assumptions. The second aim is to apply theories used in studies on office allocation to negotiations and gain insight into the following questions: Are actors with extreme preferences more active than others? What is the role of experts? How do party politics play out? Although we find effects of preferences and expertise on negotiations, the evidence is most compatible with the view that negotiations serve EP party groups to form and exchange policy positions.

Keywords

committees European Parliament negotiations party groups social network analysis

Introduction

The evolution of the legislative powers of the European Parliament (EP) constitutes one of the most remarkable trends in the history of the European Union (EU). Its constitutional empowerment, however, has put the EP under increasing pressure to become an efficient actor in EU decision-making. This has been necessary for the EP to capitalize on the potential influence of its elevated legislative role (e.g. Bowler and Farrell, 1995), and to diffuse concerns over the efficiency of EU policy-making (e.g. Golub, 1999; Schultz and König, 2000). The EP has been able to meet these challenges for two general reasons: internal legislative specialization (Bowler and Farrell, 1995), and structuring by political parties (Hix et al., 2007). This article contributes to the debate that dates back to the mid-1990s.

Legislative specialization can be understood as the institution of a division of labour: most decision-making processes actually involve only a sub-set of the members of the European Parliament (MEPs) so that there is potential for specialization. Specialization is visible in organizational choices such as the creation of a system of standing committees that deal with legislation before the plenary makes its decision (Bowler and Farrell, 1995). However, it is also a fact of the actual negotiations unfolding around policy choices the EP faces. Only a fraction of the 736 MEPs is active in any given case of legislative negotiations. Hence, while specialization has been important for the EP, it effectively means that a sub-set of the legislature gains great influence over policy choices on any given issue.

Most current research focuses on the structural manifestations of legislative specialization. In particular, scholars have analysed the composition of parliamentary committees and the allocation of rapporteurship positions to MEPs (e.g. Benedetto, 2005; Hausemer, 2006; Hoyland, 2006; Kaeding, 2004; Mamadouh and Raunio, 2003; McElroy, 2006; Whitaker, 2005; Yordanova, 2009). Inspired by theories of legislative organization developed in studies of the United States (US) Congress (in particular, Cox, 2006; Cox and McCubbins, 1993; Krehbiel, 1991; Krehbiel, 2004; Shepsle and Weingast, 1987; Weingast and Marshall, 1988), the question guiding most analyses is whether or not specialization adds policy influence over outcomes to actors with preferences far away from the parliamentary average (henceforth preference outliers) or experts, or whether it is well controlled by political parties. The reasoning is that committee members and rapporteurs play a more pronounced role in legislative choices in their policy area or issue. Hence, if, for instance, these actors have outlying preferences, policy choices risk reflecting this bias.

Less attention is paid to a second specialization process. For every given legislative dossier, all parliamentarians have to decide whether and how strongly to engage in the negotiations. As discussed below, negotiations are to some extent pre-structured by institutional choices, such as the allocation of committee positions. However, there are many legislative dossiers, and not all committee members can be active all the time. Moreover, outsiders may decide to prioritize one legislative process in a given committee over the work of their formal committee assignment. Simply put, specialization does not end with the allocation of formal offices. It extends to the individual choice to engage in legislative negotiations on a particular issue. The aim of this article is to take a first step towards understanding the nature and consequences of this selection process.

The analysis is divided according to two related issues. First, the literature on office allocation contains assumptions about the role of office-holders in negotiations. For instance, rapporteurs are said to be central actors. We assess the validity of assumptions about the effects of institutional positions. However, institutional pre-structuring is consistent with different interpretations. For instance, even if committee membership were to fully overlap with participation in negotiations, committee member behaviour might still be driven by the pursuit of outlying preferences, or experts might be more active than others. Moreover, it is equally possible that some committee members do not participate in negotiations while some MEPs from other committees do. Thus, second, we draw on the aforementioned theories of legislative specialization to develop expectations about the participation and behaviour of parliamentarians in negotiations. Overall, then, this article speaks to the existing literature on specialization, on the one hand, by testing some of its basic assumptions and, on the other, by testing commonly used theories from another perspective. In brief, this article asks what role institutional position, preferences, expertise and party affiliation play in legislative negotiations.

Perhaps the most important benefit of studying legislative negotiations is that it sheds light on dynamic elements of legislative specialization. Analyses of committee assignments necessarily rely on observations of who sits on a committee and who does not. But here further questions come to mind: How active are these people in the legislative process? Who cooperates with whom? Do leading actors take a different position in the sub-group of negotiators than ordinary parliamentarians? Such behavioural patterns can be made visible by looking at the structure of negotiations. To do so, this article employs concepts and techniques used in network analysis. These tools can, for instance, assess whether a given actor is at the centre of negotiation activity by comparing the number of interactions of all actors in the negotiations. Using exponential random graph models, it is possible to estimate the effects of individual characteristics of parliamentarians on various aspects of their negotiation activity: How frequently do they approach other actors and how often are they approached? Are they likely to cooperate with actors with whom they share attributes?

Legislative negotiations

In general terms, legislative negotiations encompass all interactions of MEPs after the allocation of a legislative dossier to a committee and the assignment of formal positions, and prior to the final parliamentary vote. Taking a narrower view, this article defines negotiations as the process in which parliamentarians identify and approach potential cooperation partners and are in turn identified and approached for this purpose. The sum of all individual decisions of parliamentarians results in a network of relations among negotiators. We, thus, focus on the relational structure of negotiations.

The analysis of legislative negotiations thus defined raises two questions. First, who is active in the first place? This refers to the set of actors participating in the negotiation network. For instance, one could find disproportionate numbers of preference outliers or many experts involved in the negotiations. Second, how do the participants behave? This question attempts to understand who approaches many actors, who is often approached, and who cooperates. For instance, two parliamentarians from the same party group could have a higher probability of cooperating than those from differing party groups.

These two questions are silent on other important elements of negotiations. They disregard the substance of cooperation between any two actors. The kind of information exchanged and the type of agreements made fall outside the focus of this article. Even so, analysing the structure of negotiations allows for a more transparent and systematic study of the patterns of cooperation than approaches more focused on the substance of exchanges among parliamentarians (e.g. Benedetto, 2005; Judge and Earnshaw, 2011; Settembri and Neuhold, 2009). Moreover, a systematic mapping of relations among actors is required to analyse how actor attributes affect their cooperation choices in negotiations.

In line with the comments made in the introduction, the theoretical part that follows is divided according to two issues. We first spell out the assumptions of the existing literature as to the effects of prior institutional choices on negotiations. The main interest in this case is not in the theoretical justification of these assumptions but in whether they are relevant empirically. To test their validity, the assumptions have to be made explicit. In theoretical terms, institutional pre-structuring is consistent with different perspectives on legislative politics. Since we also seek to put informational, distributional and partisan perspectives on specialization to a test, we then further discuss what these theories expect regarding MEP participation and behaviour in negotiations.

The effects of prior institutional choices

Previous studies voice a number of expectations about the effect of prior institutional choices on legislation negotiations. The general message is that committee members, rapporteurs and shadow rapporteurs will play an important role in negotiations. Virtually all of the aforementioned studies and observations of practitioners identify committees as the main legislative arena of the EP and committee membership as an important opportunity to influence policy-making (see, for instance, Corbett et al., 2007). Rapporteurships, in particular, are referred to as important positions that provide the opportunity for parliamentarians to represent their constituencies or parties (e.g. Hausemer, 2006). There are also studies indicating that negotiations centre around the rapporteur, who has reason to approach others in order to sound out what kind of report is acceptable and who is approached by actors seeking to influence the draft report (e.g. Benedetto, 2005). Although less empirical research exists on the role of shadow rapporteurs, practitioners indicate that they are often the main interlocuteurs of the rapporteur. Together these actors may even constitute informal sub-committees in legislative negotiations (Corbett et al., 2007: 141). The literature has also suggested that group coordinators and committee chairs are important in negotiations (e.g. Neuhold, 2001). Finally, it has been argued that parliamentarians seek to self-select into committees they value (Yordanova, 2009). If they do so in order to participate in policy-making, it is reasonable to assume that they will more often be involved in legislative negotiations than MEPs who opted for other committees. These arguments are important to encourage the study of office allocation in the EP. However, they make assumptions about the implications of office allocation for legislative negotiations. The validity of these assumptions can be examined empirically on the basis of the following institutional hypotheses.

H1: Committee membership (1) overlaps with the set of negotiation participants and (2) enhances the activity of participants.

H2: Holding a position in addition to committee membership (rapporteurship, shadow rapporteurship, coordinator or chair) enhances the activity and popularity of MEPs.

H3: The group of (shadow) rapporteurs constitutes the centre of the negotiations.

The hypotheses focus on the most common empirical claims advanced to justify the study of office allocation. However, it is likely that they do not exhaustively explain the patterns of participation and behaviour in negotiations. Moreover, they are consistent with different perspectives on legislative specialization to which we turn now.

Distributional, informational and partisan perspectives on legislative negotiations

The theories presented in this section have the common starting point that participation and cooperation in legislative negotiations reflect parliamentarians’ pursuing their individual goals. Although the conventional opinion is that office, vote and policy considerations drive parliamentarians' behaviour, it is likely that policy motivations are most important in the context at hand. The very reason to have legislative negotiations is the ability of the EP to affect a European Commission policy proposal. Moreover, any individual negotiation process is unlikely to impact significantly on MEPs re-election ambitions. The electoral connection of the EP to voters is weak already and is likely to be even less present in the minds of parliamentarians when they decide how to engage in any specific negotiation. There is also no easy route to higher executive office at the EU level. On the basis of shared basic assumptions, distributional, informational and partisan theories stress different elements of parliamentary politics and can be elaborated on to yield expectations about legislative negotiations. We present these perspectives consecutively.

A distributional perspective

The distributional perspective arises from the study of the US Congressional committee system (e.g. Shepsle and Weingast, 1987; Weingast and Marshall, 1988). The committee system is perceived to be a log-rolling arrangement across policy areas, stabilizing exchanges among groups of preference outliers (cf. Shepsle and Weingast, 1994: 155). A given homogeneous group of parliamentarians concedes influence over one policy area in exchange for added influence in its jurisdiction of choice. The overall result of committee specialization is an over-representation of preference outliers by jurisdiction, as a consequence of individual self-selection processes. Parliamentarians would then distribute gains in ways different from what the median legislator would accept on any given issue.

While the distributional account of Congressional politics emphasizes the idea that re-election seeking legislators pursue particularistic gains for their constituents, EP literature more generally suggests that specialization gives added influence to actors with outlying policy preferences as indicated, for instance, through affiliation with a particular interest group (Yordanova, 2009). The general argument is that committee specialization leaves room for self-selection that parliamentarians motivated by outlying policy preferences exploit. The result is an over-representation of MEPs with outlying preferences in committees.

The basic idea of the distributional account can be applied to the context of legislative negotiations. Since the choice to participate in negotiations is not formally regulated, preference outliers can self-select, motivated by the pursuit of their policy goals. In fact, since outliers cannot expect to have their particularistic concerns adopted by other actors, they must raise their concerns in the negotiations. Outliers, thus, are expected to self-select to negotiations and be more active than other actors.

One important caveat must be added. The distributional account expects the aforementioned processes to be one-sided. That is, they only work on one side of the policy space. Two-sided self-selection is at odds with the basic distributional idea that specialization benefits homogeneous groups of preference outliers (cf. Krehbiel, 1991). Additional activity or a higher likelihood of cooperation among two outliers does certainly indicate that outlying preferences play a role in negotiations, but this is likely to result in distributional bias only if it is a one-sided phenomenon. Otherwise, opposing camps can balance each others' pursuit of outlying preferences, as has long been argued in the case of committee assignments (e.g. Bowler and Farrell, 1995). We can thus formulate the following distributional hypothesis:

H4: Outlying policy preferences on one side of the policy space (1) are over-represented in the group of participants and (2) enhance the activity of participants.

An informational perspective

The informational account (Krehbiel, 1991) differs from the distributional account both regarding the individual self-selection process and the collective outcome of this process. As the distributional account, it has been used primarily to study the US Congressional committee system. At the individual level, the informational perspective envisages self-selection subject to majority approval. Although preference outliers have incentives to try to self-select to committees, seat allocation is subject to majority approval, and no majority of legislators approves of being permanently exposed to biased committee work. The majority grants added influence that comes with specialization only in exchange for informational gains. If legislators appear able to reduce the uncertainty of all parliamentarians over the relationship of their policy choices and the outcomes they intend to achieve, they may be granted additional influence in the form of a committee seat. This reasoning leads to the expectation that one will find experts or potential experts in committees. There may be outliers, but, in order to avoid bias, the majority only approves of heterogeneous groups from different sides of the policy space. This precaution reduces the risk of committee members exploiting informational advantages to mislead the majority.

Stated in this way, informational theory cannot be applied directly to legislative negotiations because the set of participants is not subject to majority approval. However, the unregulated self-selection process may, in fact, result in patterns of participation as envisaged in informational theory and, thus, in collective outcomes resembling its predictions in the context of committee seat allocation. It can be questioned whether self-selection to participate in negotiations results in a homogeneous group of preferences outliers. On any given policy issue, there can be outliers in different directions. For instance, on environmental issues some MEPs could prefer measures that do not interfere with industry interests, while others could pursue very strict environmental protection. Moreover, parliamentarians who care about a given issue do not have to be outliers. MEPs are likely to opt for participation in negotiations that are salient for them. Salience is conceptually distinct from policy positions. An actor can have a highly salient position in the middle of the policy space, for instance. If motivated by the pursuit of policy goals, parliamentarians that attribute salience to a given dossier will choose to engage in the negotiations, regardless of their specific policy position. They are also likely to be more active to promote their position than actors who care less about an issue. One would, thus, expect self-selection to occur. However, this should lead to a heterogeneous group of participants for whom the issue is salient but who have diverse policy preferences. Krehbiel (1991: 124) points out that actors with high salience are more likely to invest the time required to make well-informed policy choices. One-sided self-selection and activity, as predicted in the distributional perspective, are at odds with the informational view.

Moreover, MEPs with the motivation to affect policy could also be actors with particular experience in parliament or issue areas. Informational theory emphasizes the contribution of experts to the quality of parliamentary policy-making. While the majority cannot actively encourage experts to participate, there is also nothing preventing them from self-selecting to do so in their area of concern. Compared to laymen, experts are likely to have stronger views on the policy in question giving them additional motivation to participate in the negotiations. On this basis, it is expected that they are also over-represented and more active in legislative negotiations. Whether experts are more popular than other parliamentarians is ambiguous. On the one hand, they have something to offer to other actors, namely their expertise. On the other hand, they could also be less susceptible to outside influence and, thus, less relevant for actors who seek to promote their policy positions. Based on these arguments, we state the following informational hypotheses:

H5: MEPs that attribute high salience to the negotiated issue are (1) over-represented in the group of participants and (2) more active.

H6: Experts are (1) over-represented in the group of participants and (2) are more active.

A partisan perspective

The partisan perspective is based on the idea that legislators form party groups in parliament because by doing so they can better achieve their goals (Aldrich, 1995; Cox and McCubbins, 1993; Hix et al., 2007). In particular, parliamentarians promote their policy goals through party groups in two related ways (Müller, 2000, 312–7).

On the one hand, party groups reduce transaction costs in negotiations. Instead of identifying new collaborators in every legislative negotiation process, parliamentarians act in relatively stable and cohesive groups. Fellow party members are their primary cooperation partners. On the other hand, this strategy only works if groups are indeed able to stabilize legislative exchange among the members across many issues and over time. A party leadership, endowed with disciplinary instruments to enforce common party positions, provides the institutional solution to this problem.

Studies of the EP suggest that party groups have become increasingly cohesive in their voting behaviour (Hix et al., 2007). One instrument to maintain voting cohesion is the allocation of rewards and sanctions in the form of legislative offices. In the context of the study of committees, this reasoning has created the expectation that party leaders control the allocation of positions and reward loyal group members with assignments to committees and rapporteurships. Recent evidence suggests that this is the case (Hausemer, 2006; Yordanova, 2011).

We suggest that in the negotiations of particular legislative dossiers, more attention should be paid to the formation of a common party group position. Loyalty or disloyalty can be sanctioned only once a party has established a clear substantive position. This position is unlikely to exist in detail on every legislative issue confronting the EP because group members cannot predict and settle all possible conflicts over such issues in advance. Moreover, EP party groups are more heterogeneous than national party groups because they are composed of parliamentarians from numerous national parties (e.g. McElroy and Benoit, 2010). What comes into focus, then, is the ability of party groups to reduce transaction costs in the negotiations by clarifying a set of preferred collaborators for each parliamentarian. Party groups do not affect the set of participants in negotiations but rather the behaviour of group members who participate. The informality of negotiations renders it unlikely that the party leadership is in a position to decide on participation. This choice is rather governed by parliamentarians individual motivations or prior institutional choices. The question is whether the existence of party groups constrains parliamentarian behaviour after the decision to participate. Instead of facing sanctions for failing to adhere to a group line that may not yet exist, parliamentarians may face sanctions for failing to cooperate with party group colleagues in the formation of a group line.

These arguments have numerous implications for the way negotiations unfold. One would expect actors with diverging policy positions to engage in intra-party negotiations over a common group position. Failure to do so might trigger the group leadership to impose sanctions. The common position, once formed, would be accepted by all participants from a party group, again under the threat of sanctions. The group would then negotiate as a bloc with other parties. The result of the negotiations would be evaluated against the status quo and the common ideal position, and party group members would vote together accordingly, again facing sanctions from the leadership for disloyalty. This article can test only the implications relating to patterns of cooperation among negotiators. The substance of intra-party exchange and positions finally adopted by party group members cannot be analysed.

From a partisan perspective, we expect to see intra-party cooperation and inter-party exchange via lead negotiators. First, if party group members indeed form a common party position, they have to interact to identify the range of positions and negotiate a common line. Therefore, actors from the same party group will cooperate in the negotiations. Second, while it is difficult to establish from relational data whether party groups act as blocs in negotiations with other groups, one implication is that there are lead negotiators who are central within their party group and who act as a bridge to other groups. At the least, one can say that such actors are aware of the positions within their group and those of other groups. This finding is not sufficient to say that parties act as blocs, but it is a strong indication. Moreover, if all members of a party group were to engage in individual negotiations with all members of another group, maintaining a common group line would be difficult. The lead party negotiators are likely to be those with positions allocated to them prior to the negotiations, either the (shadow) rapporteur, or coordinator, or possibly both.

One particularity of party politics in the EP must be kept in mind: the EP party groups are composed of national party delegations. National party delegations have been found to affect the allocation of offices and parliamentarians' voting behaviour (e.g. Coman, 2009; Whitaker, 2005). It can, thus, be expected that MEPs do not only cooperate within the EP groups, but also within national party delegations. Therefore, we should observe cooperation among parliamentarians from the same national party delegation or from parties with increasingly close policy positions. This process does not conflict with cooperation within the EP party groups because national party delegations with similar policy positions tend to join the same EP group (McElroy and Benoit, 2010). The partisan expectations can be summarized as follows:

H7: Shared EP party group membership enhances the probability of cooperation among two actors.

H8: Proximity of the policy positions of two actors' national parties enhances their probability of cooperation.

H9: Important EP party group actors are central within the group and act as a bridge to other groups.

Research design

This article is based on the study of one negotiation process over the Directive on Ambient Air Quality and Cleaner Air for Europe (COM(2005)447). The case selection was based on a number of considerations. First, we chose from the population of co-decision dossiers since the EP has the greatest legislative impact here. Second, the case was selected so as to be typical of EP politics on a number of variables. With a directive, dealt within the Environment Committee (ENVI), both the most common instrument and the most active committee were chosen. The committee voting majority, 78 percent in the first reading and 89 percent in the second, is within one standard deviation of the average majority in the ENVI of 90 percent (cf. Settembri and Neuhold, 2009). In plenary, the report was adopted by the grand coalition, which is still the most common coalition according to recent data (Votewatch.eu, 2010), although not as prominent as it used to be. The directive is not typical in the sense that it went into second reading. However, it is not clear how this might effect negotiations within the EP. The dossier constituted the second report for the rapporteur in the ENVI. Most rapporteurs have between one and four reports during their parliamentary career.¹ Finally, to ensure that the actors in the negotiations would be able to recall with whom they cooperated, the case was selected from within six months before data collection started. Overall, the case selection strikes a balance between being able to collect valid data, on the one hand, and increasing analytical leverage, on the other. The data was collected through interviews between January and March 2008. A snowball sampling strategy was used, with the rapporteur and the shadow rapporteurs as the initial seeds. They were asked the following question:

With whom in the European Parliament did you cooperate on the Directive on Ambient Air Quality and Cleaner Air for Europe?

All MEPs mentioned in this first round were asked the same question. The procedure, widely used in sociology and epidemiology, was then re-iterated until no new names were mentioned anymore. In total, 55 MEPs (or their assistants) were interviewed. In the study of social networks, this article's approach is the standard data collection strategy (see Salganik and Heckathorn, 2004). However, the procedure has been criticized because of potential biases. Most discussions of bias in snowball sampling is related to designs in which the sampling is discontinued before the ‘snowball stops rolling’ (see Frank, 1977; Goodman, 1961; Snijders, 1992). However, in cases where the snowball is allowed to run its course, the sampling procedure has been shown to be superior to standard survey design when collecting network data (Doreian and Woodard, 1992). Moreover, it is possible that the data collection is inclined towards highly active actors in the network (in the EP case possibly the rapporteur and shadows) that most actors can presumably remember but who cannot remember all actors with whom they dealt. The data we collected distinguishes outgoing activity (out-degrees, in the terminology of network analysis, referring to the parliamentarians mentioned by an interviewee) and incoming activity (in-degrees, referring to the number of times others mention a parliamentarian). There could be a downward bias of the out-degree and upward bias of the in-degree estimate of a popular actor. However, given the considerable number of parliamentarians other than the rapporteur that were mentioned, incomplete memories of interviewees do not seem to constitute a very severe problem. One could also argue that the choice of initial respondents affects the results. Yet, the snowball sampling procedure has been shown to follow the same logic as a Markov chain (Heckathorn, 1997), i.e. at any given state the sampling procedure is only influenced by the prior state. Thus from round 3 and onwards the sampling is no longer influenced by the initial seeds.

The data and operationalizations of our independent variables are as follows. First, dummy variables identify the rapporteur, shadow rapporteurs, committee members, EP party group affiliation and party group coordinators. In order to measure policy preferences, we rely on the data on interest group affiliation introduced by Yordanova (2009). We also use the policy positions of the national parties of MEPs on a pro-/anti-environment scale as reported in Benoit and Laver (2006). They measure policy preferences on a 1–20 scale using expert surveys. A score of 1 means a party favours the environment at the expense of the economy and 20 constitutes the other extreme. On the basis of this variable, we identify policy outliers as MEPs whose national party position is more than two standard deviations above or below the mean environmental policy position in the EP. Salience is also measured based on the national party position as identified in the expert survey of Benoit and Laver. For expertise, we record whether a parliamentarian was a member of the previous EP. We also construct an index ranking parliamentarians without previous experience, with previous EP experience, and with previous EP experience and previous ENVI membership. In our view, it is not possible to make a clear link between university education (which for many MEPs took place decades ago) and expertise. Since it is widely discussed in the study of the EP, the models presented later also include a control for the left–right position of MEPs, again based on data from Benoit and Laver. The hypotheses and associated variables are summarized in Table 1.

Table 1.

Hypotheses and variables: the table shows the variable associated with each hypothesis, the operationalization of the variable and the key summary statistics for each variable

Hypothesis	Variable	Operationalization	Source	Max	Min	Central tendency	Standard deviation
H1	ENVI membership	Binary variable coding 1 for ENVI membership and 0 for no ENVI membership.	Yordanova (2009)	1	0	1	–
H2	Institutional position	Binary variable coding 1 if an MEP is a shadow, rapporteur or coordinator and 0 if an MEP is none of the above.	Own data	1	0	0	–
H4	Outlier	Categorical variable coding 1 if an MEP's national party is two standard deviations above the mean environmental policy position in the EP (an industry outlier). The variable codes 2 if an MEP's national party is two standard deviations below the mean environmental policy position in the EP (a green outlier). If an MEP's national party is within two standard deviations of the mean environmental position in the EP it is coded as a 0 (no outlier).	Benoit and Laver (2006)	2	0	0	–
H4	Green interest group affiliation	A dummy variable coding for whether an MEP has an affiliation with a green interest group	Yordanova (2009)	1	0	0	–
H4	Industry interest group affiliation	A dummy variable coding for whether an MEP has an affiliation with an industry interest group.	Yordanova (2009)	1	0	0	–
H5	Environmental salience	The salience attached by an MEP's national party to environmental matters.	Benoit and Laver (2006)	18.3	7.35	11.64	2.31
H6	Expertise	An ordinal variable coding 0 if an MEP has no previous EP experience, 1 if an MEP was a member of the EP in a prior session, and 2 if an MEP was a member of the EP and ENVI in a prior session.	Yordanova (2009)	2	0	0	–
H7	EP party groups	Categorical variable identifying each MEP's EP party group.	Own data	–	–	–	–
H8	Environmental position	The position of each MEP's national party on a 20 point scale, where 1 is strongly prioritizing the environment over industry, and 20 is strongly prioritizing industry over environment.	Benoit and Laver (2006)	16.77	2.27	10.99	3.8
–	Left/Right	The position of each MEP's national party on a 20 point left/right scale where 1 is the most left and 20 is the most right.	Benoit and Laver (2006)	16.38	3	10.15	4.02

Analysis

Descriptive evidence

We begin with a set of descriptive analyses. As a first perspective on participation in negotiations, Figure 1 compares the network and the EP as a whole. Panels a and b show the environmental position and salience of the national parties of MEPs; panels c and d show green and industry group affiliations; and panel e shows MEPs with and without experience in the previous EP. In all cases, the differences between the EP as a whole and the group of negotiators are very small and best interpreted to be coincidental. Self-selection to participate in the negotiations studied here does not reflect the distributional bias recently found for committee seat allocation (Yordanova, 2009). It neither reflects informational biases. The panels do not capture that a majority of the MEPs in the negotiations are members of the ENVI (35 out of 56). However, the overlap is far from perfect due to the absence (or presence) of a substantial number of committee insiders (or outsiders). Thus H1 on the overlap and activity enhancing effect of committee membership is the only hypotheses on participation that finds some, albeit imperfect, support.

Figure 1.

Comparing the network to the European Parliament: the panels show the distribution of important variables in the network compared to the European Parliament.

Figure 2 contains first insights into how actors who participate are placed in the negotiations and in relation to each other. This sheds further light on the institutional hypotheses dealing with the centrality of negotiators (H3 and H9). Is there a core group of central negotiators? If so, how do central actors relate to their party fellows and to each other?

Figure 2.

Analysis of centrality: panel a shows the distribution of three different centrality measures, while panel b shows the network visualized with the Fruchterman–Reingold force-directed algorithm.

Panel a presents three measures used in network analysis to assess the centrality of actors: degree, betweenness and eigenvector centrality. Each of these measures captures a different aspect of what it means to be a central actor. Degree is often equated with a raw activity indicator. Betweenness is a measure of how often an actor is on a path between two other actors. Finally, eigenvector centrality captures how well connected an actor is to other well-connected actors.² Although these measures emphasize different aspects of an actor's position in the network, they converge when the network resembles a so-called ‘network star’ (see Wasserman and Faust, 1994) – that is, one central actor with ties to a set of other actors who are, however, not connected to each other. In this case, all actors except one have minimal centrality scores. In other words, in a strongly centralized network with few dominant actors, the three centrality measures should resemble each other. As we can see in Figure 2a this is indeed the case. The distribution of all three measures is strongly skewed to the left showing that a few actors in the network are the most active, occupy the roles of interlocuteurs for the other actors, and share ties with each other.

Figure 2b adds a visual representation of the network of negotiators. The Fruchterman–Reingold force-directed algorithm (Fruchterman and Reingold, 1991) was used to place the nodes. The nodes are scaled by their eigenvalue centrality with lines representing ties, arrows the direction of ties, colours recording party affiliation and labels identifying rapporteurs, shadows and coordinators. From visual inspection, the insights from panel a appear to be supported. Several actors are central while the majority has only few links to other actors. In addition, the central actors, indeed, seem to be those with institutional positions. At first glance, it also appears that the central actors are more strongly connected to their fellow party members than to others, indicating that they might act as bridges from their party group to others.

It is possible to investigate further whether the rapporteurs are at the centre of negotiations, and whether there is a process of intra-party coordination and inter-party exchange via the rapporteurs (or other central actors). The first question can be assessed by analysing the so-called clique structure of the network that reveals whether there is one core group and a periphery or, for instance, none or several cores. There are several approaches to detecting core–periphery structures in networks (Borgatti and Everett, 1999; Everett and Borgatti, 1999). In our study the most appropriate approach is not to assume that one core necessarily must exist, but that there can be multiple cores in the network. A measure like this has been developed by Everett and Borgatti (1999). Essentially, the measure identifies highly clustered regions in the network, and then examines the peripheries of these clusters. This allows for a network to have either one or more central clusters and actors to belong to these clusters to different degrees. The measure can be reduced to a simple fraction of how many ties an actor has to a clustered region, compared to how many ties are needed to be a full member of that region. It can be written as

(1)

where r is the number of realized ties to a clustered region, and q is the number of ties needed to be a full member of that region. Thus an actor who is a full member of every clustered region will score the maximum of 1, while an actor who is not in the periphery of any clustered region will score a 0. To assess the degree membership overlap between the different clusters, it is recommended to conduct a correspondence analysis of the matrix that we get from calculating the coreness scores for multiple clusters (see Everett and Borgatti, 1999: 404). Figure 3 shows the results.

Figure 3.

Correspondence analysis: the plot shows closeness of actors as a function of their membership of clusters (the arrows). The closer an actor is to the centre (the intersection of the dotted lines) the more of a member is that actor of all clusters.

The cores have been placed along standard coordinates, while the MEPs have been placed according to their principal coordinates. Arrows go from the origin towards the cores. The points have been scaled according to their mass (see Faust, 2005). The two dimensions plotted account for 87.2 percent of the variation in the core/periphery matrix, thus the depiction gives a fairly good image of the relations between actors and cores. Figure 3 contains only a sub-set of the actors in the network. In the core/periphery analysis, many MEPs came out with a score of 0, that is they were not part of any core in the network and they were dropped from the correspondence analysis. The actors at the centre of the plot are those who have a high degree of membership in several of the cores. These actors can be considered to be the core players. Actors 1, 2, 3, 4, 10, 12 and 13 are the shadows and the rapporteur. They are all present in the plot, and the three most central actors are all shadows. The fact that out of 55 MEPs only 24 are depicted in the plot illustrates that this is a network where the majority of the actors play only a very marginal role. Furthermore, the fact that the most central actors are all shadows tells us that this is a very hierarchical network with a small core of central actors that dominate the negotiations. This analysis supports H3 in the sense that (shadow) rapporteurs are all in the core group of negotiators and three shadows are most central. Studies of rapporteurship allocation seem to be correct in assuming that these are important players. Yet, the correspondence analysis also shows that other MEPs play a role in the negotiations. The models below examine further factors that shape negotiations. This analysis does not reveal whether rapporteurs or other actors are involved in a process of intra-party coordination and inter-party exchange (H9). To obtain a final piece of descriptive evidence, we split the network into five subgraphs corresponding to the party group networks.³ This allows us to examine in more detail whether actors who are central in the network as a whole are also central in their party groups, adding to the visual impression obtained from panel b in Figure 2. Figure 4 presents the subgraphs. From a purely visual inspection, one observes that very few party actors are disconnected from their colleagues. Moreover, at least ALDE, Greens/EFA GUE/NGL and the PES appear to be very centralized.⁴ Within the EPP-ED group, the network ties of different actors seem to be more evenly distributed.

Figure 4.

Party subgraphs: the panels show the subgraphs for the different party groups. IND/DEM and UEN are excluded.

The graphs in Figure 4 can be analysed more systematically based on the measures reported in Table 2. The first column contains the party group label. The second column shows a measure of graph centralization based on in-degree centrality. The measure is a simple sum of deviations:

(2)

where max(C_v) is the node with the highest centrality score and C_i is the centrality score of node i. The measure is normalized according to the maximum obtainable centrality score on a network of a given size. We then calculate the individual in-degree scores, and report the highest score. The institutional role of the MEP with the highest score is reported, as well as the MEP's betweenness ranking in the full network. From a partisan perspective, one expects intra-party activity concentrated on one actor who acts as a bridge to other groups, thus having a high overall betweenness centrality. The central actor is likely to be the (shadow) rapporteur.

Table 2.

Partisan perspective: the table shows the centralization of the different party subgraphs, the normalized in-degree score for the most popular actor, the institutional role of that actor and the actors betweenness ranking in the whole network

Party	Graph centralization	Highest in-degree	Institutional role	Betweenness rank
ALDE	0.81	0.75	Rapporteur	6
VERTS/ALE	0.56	0.5	Shadow/coordinator	5
PES	0.49	0.26	Shadow	1
EPP-ED	0.40	0.26	Shadow	4
GUE/NGL	0.89	0.75	Shadow/coordinator	8

The visual impressions from Figure 4 are only partly confirmed. ALDE and GUE/NGL are, indeed, highly centralized and VERTS/ALE still exceeds the 0.5 mark. The PES and EPP-ED are further away from the maximum obtainable centralization score. However, the implication is not that these networks are poorly connected internally, but that there are several actors with similarly high centrality scores, thus depressing the deviation sum. In line with this, the most central actors in the PES and EPP-ED have lower in-degree centrality scores than in the other groups. However, the most central actors of all groups also rank high as to their betweenness centrality score. The most central party actors are those with the role as (shadow) rapporteur. In two cases this coincides with the position of group coordinator. Overall, Figure 4 and Table 2 constitute support for H9. The centres of intra-party activity are also key interlocuteurs in the overall negotiations. Moreover, party group members are directly or indirectly connected to each other with very few exceptions. However, the EPP-ED in particular is internally less hierarchical than expected.

In summary, the descriptive analyses indicate the following. First, there is support for assumptions in the literature on office allocation as to the effects of prior institutional choices. A majority of the negotiators comes from the ENVI committee and the (shadow) rapporteurs are central actors. However, one should note that the committee chairman was not involved in the negotiations at all. Moreover, the institutional effects are not perfect. There are more actors in the negotiations and in the core than institutional positions alone would suggest. Evidence on participation in the negotiations neither supports the distributional nor the informational perspective. There is no noteworthy over-representation of preference outliers, experts or experienced MEPs. Key interlocuteurs in the overall negotiations are also the centre of activity within their respective party groups.⁵ Thus, the descriptive evidence, overall, supports the assumptions of the literature on office allocation and, beyond this, is most compatible with the partisan perspective on legislative specialization. However, we have only examined a sub-set of the hypotheses. The remaining expectations require modelling the effects of actor attributes on the negotiation behaviour of parliamentarians.

Modelling the network

In order to shed further light on the different theoretical perspectives on legislative negotiations, we turn to exponential random graph (ERG) models. They allow us to estimate the effect of actor and network characteristics on the outgoing (how many other actors they approach) and incoming (how many actors approach them) activity, and the effect of a given shared characteristic of any two actors on their probability of forming a connection. The philosophy behind the ERG modelling approach is that there are local social processes that generate relations between actors (and determine the shape of a social network). Furthermore, it is claimed that these processes depend on the surrounding social environment. Thus, ERG models explicitly model the conditional effect of micro-processes on the macro level (Robins et al., 2007: 177). Hence, the dependent variable in the models is the network, and we are modelling the effect of actor covariates in creating the observed network structure.

The modelling of networks presents a number of challenges for standard statistical methods. First, there is the question of statistical dependence among observations. For practically all real life networks it is not possible to assume that the tie from one actor to another is independent of all other actors and their ties. In order to deal with this, we postulate what kind of dependencies might exist in the network through the specification of dependence graphs. A dependence graph is a subgraph that captures a certain type of dependency in a network. In the models presented below, we have, among others, controlled for the asymmetric structure of the network. Thus, we have specified a series of dependence graph, each representing a certain type of dependency. In the model, we add a network statistic equaling the number of relations in the network that corresponds to the given dependency. This allows us to control for social tendencies in the network, when estimating the effects of our independent variables. The Hammersley–Clifford theorem (Wasserman and Robins, 2005) establishes that a probability model for a network depends only on the components of the dependence graphs, and it yields the now well-known general form of the model:

(3)

where θ is a vector of model parameters and z(x) is a vector of network statistics. k(θ) is a normalizing constant, which guarantees that the distribution is a proper probability distribution (i.e. that it sums up to one over all possible realizations).⁶ The model equation can be rewritten as a conditional logit model:

(4)

The term $X_{i j}^{c}$ denotes all other dyads than X_ij in the network, and δz(x)represents the change in the covariate. The model is conditional in the sense that the effects are conditioned on $X_{i j}^{c}$ , i.e. the effects reflect the change from $X_{i j}^{c}$ to X_ij. Thus, if forming a tie increases z(x) by 1, the log-odds of this tie actually occurring is equal to θ, ceteris paribus. Hence the vector of coefficients can be read as the effect of switching a specific tie on. This form of the equation also illustrates that the θ vector is the conditional log-odds of forming a given tie. In order to establish a baseline for our models, we first estimate the standard Erdös–Rényi model. This model is based on the assumption that each tie has an equal probability of forming, i.e. a tie is not dependent on the rest of the network. To see whether the dependencies we include in our models are valid, the second model consists only of the dependencies. If the dependence assumptions are valid, this model should represent a significant increase in model fit compared to the Erdös–Rényi model. We have estimated four substantial models, one for each set of hypotheses developed in the theory section. Since we are dealing with a directed network in which ties can be incoming and outgoing, there are three possible parameters for each of our actor attribute variables, summarized in Table 1, namely a parameter for their in-degree (i.e. how popular an actor with a given characteristic is), and their out-degree (i.e. how active an actor with a given characteristic is). Finally, we include a homophily effect of actors that share a given attribute. For categorical variables the homophily effect takes the shape of a matching parameter that captures the tendency for actors that share a given trait to be connected. For continuous measure the homophily effect is captured by a differentiated attractiveness parameter that captures the chance of a tie forming between two MEPs as the difference between them increases. A negative effect here means that an increase in distance decreases the probability that two actors are connected.⁷

Table 3 shows the results from several ERG models. We control for dependencies in the data by adding parameters for the mean nodal degree,⁸ asymmetric ties, the geometrically weighted dyadic shared partners (GWDSP) distribution and the geometrically weighted edge-wise shared partner (GWESP) distribution. The two latter dependencies summarize the distribution of the number of common neighbours between actors i and j where the tie between the actors is allowed to vary (dyadic shared partners) or is fixed at 1 (edge-wise shared partners) (Hunter, 2007: 223). Thus the GWDSP summarizes the distribution of two paths, while the GWESP summarizes the triadic distribution. Each measure includes a weighting parameter, which we have fixed at predetermined values (Hunter, 2007: 219).

Table 3.

Exponential random graph models

	Erdös–Rényi	Dependencies	Institutional model	Distributional model	Partisan model	Informational model	Full model
Edges	−2.94*
	(0.08)
Mean degree		−65.98*	−79.31*	−71.40*	−81.42*	−113.64*	−182.68*
		(1.01)	(2.27)	(0.88)	(1.09)	(1.35)	(1.77)
Asymmetric		−1.61*	−1.35*	−1.51*	−1.36*	−1.59*	−1.58*
		(0.06)	(0.10)	(0.09)	(0.05)	(0.05)	(0.06)
GWDSP (λ = 0.8)		0.04*	0.04*	0.07*	0.05*	0.06*	0.09*
		(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
GWESP (λ = 0)		0.42*	0.15*	0.34*	0.32*	0.25*	0.05
		(0.02)	(0.03)	(0.02)	(0.02)	(0.03)	(0.03)
Left–right (incoming)			−0.06*	−0.08*	−0.04*	−0.03*	−0.08*
			(0.01)	(0.00)	(0.00)	(0.00)	(0.00)
Left–right (outgoing)			0.08*	0.11*	0.08*	0.07*	0.14*
			(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Left–right (homophily)			−0.25*	−0.20*	−0.14*	−0.21*	−0.10*
			(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
ENVI (incoming)			0.42*				1.00*
			(0.06)				(0.05)
ENVI (outgoing)			−1.42*				−3.13*
			(0.07)				(0.05)
ENVI (homophily)			1.32*				1.67*
			(0.07)				(0.05)
Institutional position (incoming)			1.41*				2.09*
			(0.09)				(0.06)
Institutional position (outgoing)			0.67*				0.86*
			(0.19)				(0.08)
Institutional position (homophily)			0.00				0.14*
			(0.07)				(0.06)
Outlier (incoming)				−0.02			−0.12
				(0.06)			(0.07)
Outlier (outgoing)				0.40*			0.52*
				(0.08)			(0.07)
Outlier (homophily)				−0.37*			0.19*
				(0.08)			(0.05)
Green interest group (incoming)				0.59*			0.12
				(0.09)			(0.10)
Green interest group (outgoing)				−0.68*			0.10
				(0.15)			(0.11)
Green interest group (homophily)				0.03			0.02
				(0.03)			(0.06)
Industry interest group (incoming)				1.29*			0.94*
				(0.08)			(0.17)
Industry interest group (outgoing)				−0.85*			−0.62*
				(0.13)			(0.20)
Industry interest group (homophily)				0.16*			0.09*
				(0.04)			(0.04)
Party group (homophily)					0.76*		1.65*
					(0.03)		(0.06)
Environmental position (homophily)					0.03*		−0.06*
					(0.01)		(0.01)
Environmental salience (incoming)						0.15*	0.01*
						(0.00)	(0.00)
Environmental salience (outgoing)						−0.03*	0.11*
						(0.00)	(0.00)
Environmental salience (homophily)						−0.03*	−0.08*
						(0.01)	(0.01)
Weak experience (incoming)						−0.85*	−0.06
						(0.13)	(0.13)
Strong experience (incoming)						−0.75*	−1.26*
						(0.06)	(0.07)
Weak experience (outgoing)						0.45*	−0.34*
						(0.12)	(0.12)
Strong experience (outgoing)						1.50*	2.65*
						(0.04)	(0.06)
Experience (homophily)						0.03	0.04
						(0.05)	(0.07)
AIC	1230.74	939.83	821.92	888.26	903.83	846.94	700.91
BIC	1236.77	963.96	900.34	984.78	958.13	937.43	893.95

*p < .05

Table 4 summarizes the robust, significant and substantive effects from the full model. First, we discuss the evidence with a view on the effects of institutional pre-structuring and, second, with a view on the distributional, informational and partisan perspectives. Regarding the effects of prior institutional choices, ENVI membership enhances the popularity of MEPs. Yet, it depresses their activity. Holding an additional institutional position has positive effects on popularity and activity. Thus, H2 is fully supported. However, the evidence does not support the second element of H1, expecting enhanced activity of committee members. It appears that outsiders are more active than committee members. In other words, one assumption in the study of committee seat allocation does not seem to hold.

Table 4.

Substantive effects of robust and significant parameters: the table shows the odds associated with a unit change in the independent variable. In the third column the percentage change in the odds is shown

Variable	Odds	Percentage change in the odds
ENVI (incoming)	2.72	171.83
ENVI (outgoing)	0.04	−95.63
Institutional position (incoming)	8.08	708.49
Institutional position (outgoing)	2.36	136.32
Outlier (outgoing)	1.68	68.20
Industry (incoming)	2.56	155.10
Industry (outgoing)	0.42	−57.00
Party group (homophily)	5.21	420.70
Strong expertise (incoming)	0.28	−71.63
Strong expertise (outgoing)	14.15	1315.40

Turning to the remaining three perspectives on specialization, there is some evidence for the behavioural dynamics expected in the second element of H4. Although there is no over-representation of preference outliers, they are more active in the negotiations. Note that there are no industry outliers so that this effect captures the behaviour of pro-environment MEPs. Turning to interest group ties, there are no consistent effects of affiliation to green groups. Industry affiliates tend to be more popular, yet they are less active in terms of approaching other actors. One possible interpretation would be that activity stems from actors with green preferences and is directed at industry affiliates. Overall, H4 was based on the assumption that outlying preferences would matter on one side of the policy space only. Here we observe a two-sided effect. On the basis of our data, it cannot be said that either green or industry preferences dominate. Hence, while outlying preferences or interest group affiliation matter for understanding parliamentarians' behaviour, the main idea of the distributional perspective captured in H4 is not supported. At least the relational structure of negotiations does not indicate distributional implications of specialization.

Turning to the informational perspective, we find support for the second element of H6. MEPs with strong experience, that is with previous EP and ENVI experience, are more active than others. One also finds that such MEPs are less popular, possibly because actors who seek to promote their policy positions turn to other actors. There is no support for the second element of H5 because environmental salience does not have consistent effects. Overall, given that there is no one-sided distributional effect and at least some informational effects, the evidence is more in line with the informational than the distributional account. As Krehbiel (1991) argues in his study of the US Congress, the absence of distributional effects is indeed partly evidence in favour of the informational account of legislative politics.

Regarding the partisan perspective, the descriptive evidence is further strengthened. Any two actors from the same EP party group are likely to cooperate. We do not find an effect, though, of the proximity of environmental positions of national parties. Thus, the intra-party coordination, we have discussed, seems to take place mainly within EP groups and not between national party delegations. An explanation could be that the former process makes the latter process superfluous. When the EP group develops a common position, this is necessarily also the common position of similar national party delegations because those delegations tend to join the same EP group (McElroy and Benoit 2010). H7 is thus supported but not H8. Although the latter finding is unexpected, it can arguably be explained. Therefore, the results can be considered compatible with the partisan perspective.

Summing up, modelling the network yields the following results. First, the assumptions of the literature on office allocation are valid with the exception of the activity of committee members. Second, the central idea of the distributional perspective is not supported. Preferences matter but not in the way suggested by this perspective. Third, the informational account is partly supported because experts behave as expected and because the distributional account does not find support. However, the expected effects of salience do not hold. Finally, the evidence is most compatible with the partisan hypotheses 7 and 9.

Discussion

The aim of this article has been to take a first step towards understanding legislative negotiations. This ambition is situated in a wider debate on the implications of legislative specialization in the EP. Specialization effectively means that the parliament has to grant added influence to a sub-set of the legislature on every given decision. What are the implications of this choice? The literature has debated this question primarily by studying processes of organizational specialization such as the allocation of committee seats and rapporteurship positions. Yet, in doing so, one has to assume the importance of these offices for subsequent negotiation processes. An aim of this article has been to test the validity of such assumptions. Additionally, by focusing on office allocation one cannot study the factors driving variation in actual activity, popularity or cooperation. For instance, it seems worth analysing whether party leaders exert control over committee seat allocation although the party members subsequently do not cooperate; or whether preference outliers are over-represented in a committee while being inactive in practice. The other aim of this article has been to offer a first take on such questions. Although ours is a single case study, the results provide several insights.

First, there is reason for confidence in the studies of office allocation. The assumptions made about office-holders' behaviour in negotiations are largely supported. Committee members constitute the majority in negotiations and (shadow) rapporteurs are central actors. Furthermore, outsiders participate in negotiations and are even more active than committee members without additional positions. At the same time, not all committee insiders are necessarily involved. Thus, committee membership seems to be a useful yet not overly precise proxy for actual participation in decision-making. Second, it should be stressed that the question of how one could explain more or less overlap between the set of committee members and the set of negotiators has not been resolved here. It was suggested that the pursuit of outlying policy goals, expertise or salience motivates parliamentarians to participate in negotiations. However, the set of participants does not differ notably from the EP as a whole. It remains an open question what besides prior institutional choices explains participation and non-participation in legislative negotiations.

Third, the most intriguing question has been whether the formally unconstrained choice of MEPs to participate and act in negotiations creates bias or informational gains or serves EP party groups. The evidence casts doubt on the first view. There is no over-representation of outliers in the first place, and when it comes to behaviour we observe effects of outlying preferences both on the green and industry sides of the policy space. There is some limited evidence for the informational view. Experts or MEPs who care about the issue in question are not over-represented. But, experts who are in the negotiations are more active.

Overall, though, the evidence is strongest for a partisan perspective. Central negotiators are also central party group actors and any two MEPs from the same group are likely to cooperate. As we pointed out in the theory section, the focus on the relational structure of negotiations does not allow us to study whether party members actually negotiated a common position, and whether this position actually formed the basis of inter-party exchange via lead negotiators. Yet, since EP groups do not have issue-specific positions on all legislative dossiers, they have to form a group line at some point. It is hard to see how this would happen if the party group members active in the negotiations did not interact. It is also hard to see how a group line could be maintained in inter-party exchange absent of lead negotiators. The patterns of cooperation we find indicate both intra-party coordination and inter-party exchange. Although further research is required to substantiate this conclusion, we would suggest that, rather than serving outliers or creating informational gains, negotiations mainly serve EP party groups to form and exchange issue-specific positions.

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Footnotes

Notes

References

Aldrich

(1995) Why Parties? The Origin and Transformation of Political Parties in America. Chicago: University of Chicago Press.

Benedetto

(2005) Rapporteurs as legislative entrepreneurs: The dynamics of the codecision procedure in Europe's parliament. Journal of European Public Policy 12(1): 67–88.

Benoit

Laver

(2006) Party Policy in Modern Democracies. Routledge: London.

Borgatti

Everett

(1999) Models of core/periphery structures. Social Networks 21(4): 375–395.

Bowler

Farrell

(1995) The organizing of the European Parliament: Committees, specialization and co-ordination. British Journal of Political Science 25(2): 219–219.

Corbett

Jacobs

Shackleton

(2007) The European Parliament. John Harper Publishing.

Coman

(2009) Reassessing the influence of party groups on individual members of the European Parliament. West European Politics 32(6): 1099–1117.

Cox

(2006) The Oxford Handbook of political economy. In: Cox

Weingast

Wittman

(eds) The Organization of Democratic Legislatures. Oxford University Press, 141–161.

Cox

McCubbins

(1993) Legislative Leviathan: Party Government in the House. Berkeley: University of California Press.

10.

Doreian

Woodard

(1992) Fixed list versus snowball selection of social networks. Social Science Research 21(2): 216–233.

11.

Everett

Borgatti

(1999) Peripheries of cohesive subsets. Social Networks 21(4): 397–407.

12.

Faust

(2005) Using correspondance analysis for joint displays of affiliation networks. In: Carrington

Scott

Wasserman

(eds) Models and Methods in Social Network Analysis. Cambridge University Press, 117–147.

13.

Frank

(1977) Survey sampling in graphs. Journal of Statistical Planning and Inference 1(3): 235–264.

14.

Freeman

(1979) Centrality in social networks conceptual clarification. Social Networks 1(3): 215–239.

15.

Fruchterman

TMJ

Reingold

(1991) Graph Drawing by Force-directed Placement. Software - Practice and Experience 21(1): 1129–1164.

16.

Golub

(1999) In the shadow of the vote ? Decision making in the European Community. International Organization 53(4): 733–764.

17.

Goodman

(1961) Snowball sampling. The Annals of Mathematical Statistics 32(1): 148–170.

18.

Goodreau

Kitts

Morris

(2009) Birds of a feather, or friend of a friend? Using exponential random graph models to investigate adolescent social networks. Demography 46(1): 103–125.

19.

Hausemer

(2006) Participation and political competition in committee report allocation: Under what conditions do MEPs represent their constituents? European Union Politics 7(4): 505–530.

20.

Heckathorn

(1997) Respondent-driven sampling: A new approach to the study of hidden populations. Social Problems 44(2): 174–199.

21.

Hix

Noury

Roland

(2007) Democratic Politics in the European Parliament. Cambridge: Cambridge University Press.

22.

Hoyland

(2006) Allocation of codecision reports in the fifth european parliament. European Union Politics 7(1): 30–50.

23.

Hunter

(2007) Curved exponential family models for social networks. Social networks 29(2): 216–230.

24.

Judge

Earnshaw

(2011) ‘Relais actors’ and co-decision first reading agreements in the European Parliament: The case of the advanced therapies regulation. Journal of European Public Policy 18(1): 53–71.

25.

Kaeding

(2004) Rapporteurship allocation in the European Parliament: information or distribution? European Union Politics 5(3): 353–371.

26.

Krehbiel

(1991) Information and Legislative Organisation. Ann Arbor, MI: University of Michigan Press.

27.

Krehbiel

(2004) Legislative organization. Journal of Economic Perspectives 18(1): 113–128.

28.

Mamadouh

Raunio

(2003) The committee system: Powers, appointments and report allocation. Journal of Common Market Studies 41(2): 333–351.

29.

McElroy

(2006) Committee representation in the European Parliament. European Union Politics 7(1): 5–29.

30.

McElroy

Benoit

(2010) Party policy and group affiliation in the European Parliament. British Journal of Political Science 40(02): 1–21.

31.

Müller

(2000) Political parties in parliamentary democracies: Making delegation and accountability work. European Journal of Political Research 37(3): 309–333.

32.

Neuhold

(2001) The ‘legislative backbone’ keeping the institution upright? The role of European Parliament Committees in the EU policy-making process. European Integration Online Papers 5(10): 1–27.

33.

Robins

Pattison

Kalish

Lusher

(2007) An introduction to exponential random graph (p) models for social networks. Social Networks 29(2): 173–191.

34.

Salganik

Heckathorn

(2004) Sampling and estimation in hidden populations using respondent-driven sampling. Sociological Methodology 34(1): 193–240.

35.

Schultz

König

(2000) Institutional reform and decision-making efficiency in the European Union. American Journal of Political Science 44(4): 653–666.

36.

Settembri

Neuhold

(2009) Achieving consensus through committees: does the European Parliament manage? Journal of Common Market Studies 47(1): 127–151.

37.

Shepsle

Weingast

(1987) The institutional foundations of committee power. The American Political Science Review 81(1): 85–104.

38.

Shepsle

Weingast

(1994) Positive theories of congressional institutions. Legislative Studies Quarterly 19(2): 149–179.

39.

Snijders

TAB

(1992) Estimation on the basis of snowball samples: How to weight? Bulletin de Méthodologie Sociologique 36(1): 59–70.

40.

Snijders

TAB

(2002) Markov chain Monte Carlo estimation of exponential random graph models. Journal of Social Structure 3(2): 1–40.

41.

Thurner

Binder

(2009) European Union transgovernmental networks: The emergence of a new political space beyond the nation-state? European Journal of Political Research 48(1): 80–106.

42.

Votewatch.eu (2010) Voting Behavior in the New European Parliament: The First Year. URL http://www.votewatch.eu/blog/wp-content/uploads/2009/09/votewatch_report_30_june_2010.pdf.

43.

Wasserman

Faust

(1994) Social Network Analysis: Methods and Applications. Cambridge University Press.

44.

Wasserman

Robins

(2005) An introduction to random graphs, dependence graphs and P*. In: Carrington

Scott

Wasserman

(eds) Models and Methods in Social Network Analysis. Cambridge: Cambridge University Press, 148–161.

45.

Weingast

Marshall

(1988) The industrial organization of congress: Or, why legislatures, like firms, are not organized as markets. The Journal of Political Economy 96(1): 132–163.

46.

Whitaker

(2005) National parties in the European Parliament: An influence in the committee system? European Union Politics 6(1): 5–28.

47.

Yordanova

(2009) The rationale behind committee assignment in the European Parliament: Distributive, informational and partisan perspectives. European Union Politics 10(2): 253–280.

48.

Yordanova

(2011) The effect of inter-institutional rules on the division of power in the European Parliament. West European Politics 34(1): 97–121.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.71 MB

0.00 MB