European Journal of Political Research ••: ••–••, 2008 1

doi: 10.1111/j.1475-6765.2008.00825.x

European Union transgovernmental networks: The emergence of

a new political space beyond the nation-state?

PAUL W. THURNER1 & MARTIN BINDER2

1
Empirical Political Research and Policy Analysis, University of Munich, Germany;
2
Mannheim Center for European Social Research (MZES), Mannheim University, Germany

Abstract. Does the European Union (EU) represent a new political order replacing the old
nation-states? The assessment of the real character of political orders requires the identifi-
cation of political key actors and of the specific structure of their interactions. Transgovern-
mental networks have been considered to be one of the most important features of EU
integration. Unfortunately, the network structures, processes and the impact of these infor-
mal horizontal inter-organisational relations between nation-states are mostly unknown.The
main objective of this article is to measure and explain the selective pattern of informal
bilateral relations of high officials of the EU Member States’ ministerial bureaucracies on
the occasion of an EU Intergovernmental Conference. The quantitative data used rely on
standardised interviews with 140 top-level bureaucrats. The statistical estimation of network
choices is based on recent developments of exponential random graph models.

Emerging new political spaces: The fusion of European administrations

Does the European Union (EU) represent a new political order replacing the
old nation-states? The suggestions of the grand theories of EU integration
are well known: for the liberal intergovernmentalist approach of Moravcsik
(1998), national governments are and continue to be the ‘masters of the treaty’.
Therefore, the author focuses on the making of constitutional rules as repre-
senting the most fundamental rules of this intergovernmental regime. He
understands the dynamics of integration to be driven mainly by the economic
interests of states with defined territories and boundaries. To the contrary,
neofunctionalism accentuates the role of civil society, reorienting its demands
and support to supranational actors who are supposed to be better able to
manage social change (Haas 1961). In this view, functional needs drive the
integration process and shape its form. Cross-sectional spillovers expand
the political competencies of supranational actors and, eventually, replace the
authority of the nation-state.
Are these grand theories incommensurable? The assessment of the real
character of a political order requires the identification of key political actors
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
Published by Blackwell Publishing Ltd., 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main Street, Malden,
MA 02148, USA
2 paul w. thurner & martin binder

and the specific structure of their interactions. In their volume The Institution-
alization of Europe, Stone Sweet, Sandholtz and Fligstein (2001) offer a per-
spective that has the potential to overcome the clash in EU grand theorising.
They define ‘political spaces’ as ‘social spaces wherein actors meet to make,
apply, interpret and enforce rules; they are thus sites of collective governance’
(Stone Sweet et al. 2001: 13). So-called ‘skilled actors’ within the existing
political boundaries of governance develop new forms of governance. This
perspective is open for any type of actor in the making and the application of
any type of rules of a political order. Accordingly, we suggest neglecting (for
the moment) competing claims with regard to the preponderance of different
actors (national governments versus actors below or above the governmental
executive). Rather, applied research should specify the respective action arena
(constitutional treaty-making versus legislative policy making and/or imple-
mentation in different policy areas) and focus on the ‘skilled actors’ involved
in order to identify the specific structure of their interactions.
For the following analysis we have chosen a constitutional conference as
our research case.1 However, contrary to Moravscik (1998), we will not focus
on formal interstate bargaining, but on the underlying, embedding informal
transgovernmental interactions of the national ministerial bureaucracies.
According to several theorists of EU integration, government officials are the
most important national ‘skilled actors’ for the preparation of EU inter-
governmental conferences (Hayes-Renshaw & Wallace 1997). Following
Slaughter (2004) and Wessels (1997), transgovernmental relations constitute
the most important process of integration. Regular formal meetings of gov-
ernments and administrations at the EU level institutionalise manifold forms
of mutual participation and facilitate a common perception of problems.
Wessels (1997) diagnoses a complex fusion of national governmental admin-
istrations with supranational bureaucracies. He argues that the ongoing trans-
border bureaucratisation has already produced a special type of a mixed
multilevel administration. Focusing on the intergovernmental policy-making
styles of state officials,2 we will provide a completely new perspective on the
political order of the EU-15. Making transparent the informal communication
networks of EU-15 government officials on the occasion of an intergovern-
mental conference, we want to delineate inter-organisational relations – that is,
activities at the boundaries of semi-open political systems.
In the next section, we will outline the approach of transgovernmental
networks and provide definitions. Next, we introduce the case: the EU Inter-
governmental Conference 1996. Then we will derive hypotheses with regard to
the micro-incentives to form transgovernmental networks. A descriptive over-
view of the different ministerial networks is followed by a statistical explana-
tion of the individual network choices. Knowing the micro-incentives to form
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 3

such networks contributes to our understanding of the emergence of new

political spaces and new authority structures (Stone Sweet et al. 2001; Grande
& Pauly 2005). More specifically, we investigate whether informal bureaucratic
networks have been a challenge to the national governments: Have incentives
for ‘bypassing’ their own government been effective? Our data are based on
standardised interviews with 140 top-level bureaucrats in the governments of
the EU-15 who were involved in the preparation of the negotiations preceding
the Amsterdam Treaty. For the statistical estimation we apply recently devel-
oped exponential random graph models based on Markov Chain Monte Carlo
maximum likelihood estimation (Snijders et al. 2005; Robins et al. 2006).

Transgovernmental networks

According to Slaughter (2004), transgovernmental networks are the most

important feature of the new world order. They replace the state, formerly
conceived as a hierarchical unitary actor. In her view, government officials
form horizontal networks in order to prepare and enforce global policies
informally. Central governmental authorities may completely lose their func-
tion in favour of issue-specific border-crossing networks; or, the state may
evolve into a strategic manager of a ‘networked polity’ (Ansell 2000). The
discussion of transgovernmental relations was mainly initiated by the impor-
tant contribution of Keohane and Nye (1974), where they provided what has
become a classic definition:

We define transgovernmental relations as sets of direct interactions

among sub-units of different governments that are not controlled or
closely guided by the policies of the cabinets or chief executives of those
governments. Thus we take the policies of top leaders as our benchmarks
of official government policy. (Keohane & Nye 1974: 43)

The authors take the interests of the top leaders as a reference point
representing the highest governmental authority. They distinguish two types of
transgovernmental relations. As long as transgovernmental relations of gov-
ernmental officials are fully consistent with the targets and intentions of
top leaders, these relations are called ‘transgovernmental coordination’. To
the contrary, where the central executive is weak and the officials ‘perceive a
greater common interest with another government, or sub-units of another
government’, the resulting relations are called ‘transgovernmental coalitions’
(Keohane & Nye 1974: 48).3 It may seem surprising that the authors contrast
transgovernmental coalitions with transgovernmental relations, because the
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
4 paul w. thurner & martin binder

latter are already characterised as ‘not controlled or closely guided’. However,

this simply mirrors the modern conception of bureaucracies, and of delegation
in general (Hammond & Knott 1996; Lupia 2003). By definition, every agency
relation is characterised by an asymmetry of the distribution of information
and the risk of moral hazard – otherwise, delegation, specialisation and self-
organisation would be useless. A transgovernmental coalition à la Keohane &
Nye distinguishes itself by the fact that the bureaucratic agent actually works
against its own principal. Note that in their comparative analysis of top officials
in Western European states, Page and Wright (1999) come to the conclusion
that the question of political control of the bureaucracies has become the most
crucial one.
The formal structure of governmental organisations lays down the chain
of delegation and the hierarchy of positions. Competencies of decision mak-
ing and of control are formally assigned and allocated to positions. However,
every organisation is also built on informal structures of administrative
self-organisation (Breton 1998: 187). Whereas the existence of informal trans-
governmental bureaucratic networks is not new, their extent and the issue
areas covered have reached completely new dimensions in the international
sphere. Despite her rather optimistic expectations with regard to the diverse
functions and effectiveness of such networks, Slaughter acknowledges their
potential problems: their lack of accountability due to their informality, the risk
of uncontrolled technocracy and the consequences of depoliticisation. Her
suggestion for countering these risks is to secure transparency – that is, to make
the networks visible (Slaughter 2004: 235). Unfortunately, quantitative empiri-
cal evidence with regard to the structure, processes and impact of horizontal
inter-organisational networks between nation-states’ bureaucracies does not
exist.4 As a consequence, questions concerning how these networks come into
being, how they are structured and how they further develop remain open.

The case: The EU Intergovernmental Conference 1996

The Intergovernmental Conference (IGC) 1996 and the resulting Amsterdam

Treaty constituted a step – like Maastricht, Nice or Lisbon – in an ongoing
reform process that contributed to the formal and explicit constitutionalisation
of the EU. The IGC 1996 took place from April 1996 to June 1997. It had the
purpose of completing Political Union, of (re-)balancing the division of power
and especially of preparing the institutional setting for an EU enlargement. So
far, EU constitution-building has happened incrementally, with Member States
consenting gradually to voluntarily incomplete contracts. Like national con-
stitutions, intergovernmental treaties contain global goals as well as provisions
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 5

for institutional arrangements. However, the public impression of one-shot

intergovernmental conference negotiations during summits of heads of states
is invalid. Negotiations span months of formal meetings and informal discus-
sion between Member States. Therefore, an appreciation of negotiation out-
comes has to take into account the specific form of the underlying processes.
More specifically, we argue that the specific, selective pattern of transgovern-
mental networks of the responsible top officials are an indicator of already
existing enhanced informal cooperation – among at least some of the national
ministerial bureaucracies. The existence of subsets of Member States already
exhibiting a deeper degree of an administrative integration is expected to
constitute a valid measure of the implicit integration of the EU-15 at that time.
Actually, the possibility of allowing subsets of Member States to form
sub-union(s) on the basis of concrete policies (‘Europe à la Carte’) or on the
basis of more general criteria (‘Core Europe’, ‘Europe of Concentric Circles’)
was one of the most discussed issues in the negotiations leading to the Amster-
dam as well as the Nice treaties, and it continues to determine the public and
scientific debate. Finally, the Amsterdam Treaty inserted the opportunity for
majorities to engage in a closer cooperation of subgroups with the possibility
of using the existing institutions and procedures of the EU.5
Hitherto, our knowledge about the embeddedness of EU intergovernmen-
tal negotiations in informal transgovernmental networks of officials in the
relevant ministries has been at best incomplete. In order to identify and char-
acterise the informal bureaucratic interactions, we follow graph-theoretical
approaches defining ‘networks’ simply as relations between nodes. Nodes in
our case are top officials of the ministerial bureaucracies. They indicate their
communication partners in equivalent departments of the other Member
States. These relational data allow us to reconstruct the emerging networks
from local bilateral transgovernmental communication channels of the minis-
tries involved and the respective responsible officials during the IGC 1996.
In the next section we discuss the micro-incentives to initiate and maintain
such informal transgovernmental networks in order to formulate testable
hypotheses.

Hypotheses on the network formation of the ministerial bureaucracy

Why should agents of the ministerial bureaucracies initiate and/or maintain

transgovernmental relations before and during ongoing international negotia-
tions? Slaughter (2004: 208) argues that transgovernmental networks encour-
age ‘multilateral discussion prior to all decisions’ and are therefore ‘likely to
produce more creative, more reasoned, and more legitimate solutions to many
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
6 paul w. thurner & martin binder

of the problems that members face’. Fearon (1998) enumerates several reasons
to discuss issues before making a collective decision: revealing private infor-
mation, information aggregation, justification of demands, promoting a
consensus and so on. However, these advantages of debate and preplay com-
munication are not unconditional (Austen-Smith 1990a, 1990b). For example,
the attraction of revealing private information depends on a perception of
common, or at least non-opposite, interests. Hence, for theoretical reasons it is
necessary to identify preconditions for different incentives of informal trans-
governmental preplay communication in varying contexts.
In the following we leave it open whether transgovernmental communica-
tions efforts are intended to influence the addressees and whether resources
like support are offered or demanded. In principle, in every relation there is
a flow of information that can be of considerable value. At the same time,
initiating and maintaining relations entails opportunity costs: the time budget
of politicians and top officials is extremely constrained. Therefore, we assume
that informal bilateral transgovernmental networks are formed on the basis
of a cost-benefit analysis: connections are valuable, but their initiation as well
as their continuation requires the allocation of resources (e.g., attention) that
may be more useful for other tasks or relations (Bueno de Mesquita 2003: 85).
As a consequence, actors limit their network efforts and we expect selective
and directed networks. ‘Selective’ means that not every bureaucratic agent
takes every possible opportunity to entertain relations due to opportunity cost
considerations. ‘Directed’ means that network efforts originate from one agent
and are addressed to another agent in an equivalent ministerial jurisdiction.
The revealed selectivity and directedness of network choices can be inter-
preted as an indicator of the relative valuation of a link at a certain point in
time. As a result, the pattern of the networks mirrors a ‘spatial cost topology’
(Jackson & Wolinsky 1996) – that is, the respective spatial distribution of costs
and benefits. ‘Spatial’ differences are not necessarily of a geographic nature –
they may represent any social similarity or conflict. The assumption of a spatial
cost topology supposes that the resulting similarity structure indicates the
more or less constant attributes of the nodes/actors leading to intensified
interactions. Starting from this general opportunity cost perspective, we now
have to derive concrete micro-incentives for officials to bear the cost of
informal communicative efforts – additional to the already formalised regular
meetings in the sectoral Councils.
Applying the delegation perspective of bureaucracies (Bendor et al. 2001)
to the Keohane/Nye conjecture on the incentives for building transgovern-
mental coalitions to ‘bypass’ their own government, we distinguish between
the principals (premiers, ministers) and the respective bureaucratic agents
(responsible ministerial officials). We propose to separate four situations
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 7

Table 1. Formal authority and transgovernmental networking

Agent is in line Agent is not in line

Principal is against No networking Networking (hidden action,

informal network transgovernmental
formation coalitions = bypassing)
Principal is in favour of Networking No networking (shirking)
informal network (transgovernmental
formation coordination, screening
and signaling efforts)

resulting from the combination of the following two dimensions: the principal
is in favour of or against the creation of informal networks of self-organisation;
the agent self-organises transgovernmentally in line with or against the explicit
or implicit goals of the principal. Table 1 summarises this combination.
The table nicely illustrates that it is not straightforward to derive the
underlying incentives for network formation from simple observation of infor-
mal communication. The absence of informal networking may be in harmony
with the principal’s explicit and implicit command or it may be simply due to
a shirking bureaucracy. Analogously, the principal may be in favour of decen-
tralised informal networking – for example, in order to screen another
governments’ bureaucracy, to signal negotiation positions, to signal domestic
resolve or to prearrange coalitions. Or the agent may actually hide his or her
networking efforts – for example, in order to bypass the principal. It should be
noted, however, that the bypassing conditions as formulated by Keohane and
Nye are sufficient and necessary separating conditions for detecting moral
hazard! Formulating their definition of ‘transgovernmental coalition’ as a
hypothesis, we expect officials to bypass their principals under the following
conditions. Hypothesis 1a formulates the spatial distance component of the
Keohane/Nye bypassing conjecture:

H1a: Bypassing Incentive – Issue Distances

The more often a bureaucratic agent perceives a larger distance to the
negotiation position of his or her government as compared to the dis-
tance to the position of a bureaucratic agent of an equivalent ministry
of another Member State, the higher the probability of forming a trans-
governmental coalition with this agent.

With regard to EU summits of the heads of governments, such incentives have

been highlighted by Hayes-Renshaw and Wallace (1997), who argue that many
ministers as well as officials have a rather ambivalent relation toward these
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
8 paul w. thurner & martin binder

summits. They fear that their jurisdictional interests are put aside in these
highly political meetings. Assuming that the issues on the table can be mea-
sured on a quantitative scale, it is possible to compare the issue distances
between the bureaucratic agent and the official bargaining position of the
executive and the one between the bureaucratic agent and his or her colleague
in an equivalent ministry in another Member State.6
According to Keohane and Nye (1974: 48), transgovernmental coalitions
are expected to be more frequent in governments with a low degree of hier-
archy or, conversely, with a high degree of ministerial discretion or autonomy.7
Hypothesis 1b formulates this part of the conjecture as follows:

H1b: Bypassing Incentive – Low Hierarchy

Ministries in governments that are characterised by a low degree of
hierarchy in their organisation8 are expected to exhibit a higher propen-
sity to engage in transgovernmental networks.

Bypassing may be especially strong in such an environment, or even exist only

there (Keohane & Nye 1974: 48, see Note 3 below):

H1c: Bypassing Incentive and Low Hierarchy9

The incentive for bypassing their own government should be effective
more frequently or exclusively in low-hierarchy environments.

Our bypassing hypothesis relies on one of several distinct processes of network

formation as identified by the network literature – that is, on social selection
due to attribute similarity or so-called ‘homophily’. The probabilities of selec-
tive network choices often depend on ‘types’ (i.e., on the attributes of actors).
This process is called ‘assortative mixing’ (Newman 2003: 191). It reflects the
tendency ‘to associate preferentially with people who are similar to ourselves
in some way’ (Newman 2003: 191). In the following we propose additional
hypotheses based on ‘assortative mixing’, with regard to the effect of (dis)simi-
larities of policy preferences, of being a founding member and of bilateral
economic interdependencies.
It is often argued that forming transgovernmental networks with equiva-
lent ministerial jurisdictions in other states is caused or facilitated by similar
policy interests. Conceiving policy decision making as being located in a mea-
surable political space (Enelow & Hinich 1984), we argue that similar locations
of the state officials with regard to the negotiation issues of the conference
should induce transgovernmental interactions. For example, Grofman (1982)
accentuates the process of proto-coalition building of actors with similar inter-
ests. Scharpf (1997) argued that in the process of ‘negative coordination’, small
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 9

subsets of actors bilaterally engage in avoiding negative externalities. Based on

these arguments and on theoretical results of signalling theory (Austen-Smith
1990a, 1990b), we expect state officials with similar policy preferences to form
transgovernmental links. On the other hand, if we assume complete informa-
tion on political positions and high opportunity costs, we would not expect
rational actors to bear the cost of additional communicative efforts directed to
officials with similar or even identical preferences. If it is rather the second
incentive that is effective in transgovernmental network formation, agents
should invest more in networks with distant agents – for example, to screen or
influence other officials:

H2: Political Distance10

The larger the issue distances between the bureaucratic agents of differ-
ent Member States, the lower/higher the propensity of officials to enter-
tain informal transgovernmental networks.

According to concepts of a ‘Core Europe’ or a ‘Europe of Concentring

Circles’ – that is, according to the idea that only a subset of Member States
should deepen their political integration (Stubb 2002) – one should expect that
especially the ministerial bureaucracies of the founding members (Germany-
France-BeNeLux-Italy) would be engaged in a relatively closer network as
compared to agents of late entry states. For the officials of these Member
States, established long-term relations may imply lower transactions costs of
transgovernmental coordination. Contrary to this hypothesis, and along the
arguments of adherents of the thesis that ‘deepening is not in contrast with
widening the EU’, we should expect that transgovernmental network forma-
tion would spread evenly across all Member States:

H3: Informal ‘Core Europe’

The ministerial bureaucracies of the founding Member States exhibit a
higher degree of informal transgovernmental interactions.

The central goal of EU integration as addressed in the guidelines of the

European Community (Article 2 of the founding treaty) is economic coopera-
tion and a free trade area. According to geopolitical approaches, international
economic cooperation follows security concerns and already existing alli-
ances.11 Neofunctionalism and liberalism in international relations assume
the causal impact to be reverse: regional integration is driven by economic
demands. More recent approaches propose a more complex causal logic of
regional integration. For Mattli (1999: 190ff), the initial demand for regional
integration comes from market actors. However, the political supply side has
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
10 paul w. thurner & martin binder

to show a ‘willingness . . . to accommodate demands for functional integra-

tion’. Fligstein and co-authors (Fligstein & Mara-Drita 1996; Fligstein & Stone
Sweet 2002) put forward the view that the dynamics of political and economic
integration mutually interact. More importantly, markets and political gover-
nance systems are socially embedded and constructed from networks. Regard-
less of the postulated direction or process of causality, these perspectives lead
us to expect those states with higher bilateral economic interdependence or
connectedness to have closer underlying transgovernmental relations. Further-
more, strong economic interdependence requires a politico-administrative
embedding, regardless of which ministerial jurisdictions are considered. If, to
the contrary, transgovernmental relations reflect special jurisdiction-specific
interdependencies, we would not expect a generalised relation between trans-
governmental networks and economic flows across ministerial jurisdictions:

H4: Bilateral Economic Interdependencies

The higher the bilateral economic interdependence between two Mem-
ber States, the higher, ceteris paribus, the propensity to form trans-
governmental ties.

If Fligstein’s idea of the social embedding of markets and politics is valid,

we should also observe typical patterns of social interaction. The literature on
social networks has long experience with micro-incentives of network forma-
tion. First, social processes give rise to structural reciprocity and transitivity.
The investigation of preferences for reciprocated relations in an observed
network implies the question: ‘How strong is the tendency for one actor to
“choose” another, if the second actor chooses the first?’ (Wasserman & Faust
1994: 507). Referring to our example, if the bureaucratic agent A has been
trying to achieve a pre-arrangement with an equivalent ministry B in another
country, then how likely is it that the bureaucrat B has been trying to do the
same? Another prominent preference of actors in social relations is to be party
to closed and therefore trustful relations. This incentive of social network
formation is represented, for example, by transitivity structures. In many social
networks we observe that if A is connected to B, and B to C, then there is an
increased probability that A will also be connected to C. Transitivity is a
formalisation of the proposition that ‘the friend of my friend is my friend’. The
existence of such configurations would be an indicator for informal self-
organising teams and even clique-like structures – below highly formalised
conferences. Measuring transitivity, therefore, allows us to assess the stability
and the degree of the institutionalisation12 of transgovernmental relations.
Alternative social processes of network formation are the asymmetric
attraction of ties (popularity) as well as asymmetric efforts to initiate links
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 11

(expansiveness). Both types lead to core-periphery network structures and

indicate the emergence of informal transgovernmental hierarchies. Non-
hierarchical structures are reflected by so-called ‘cycling structures’ – that is,
we observe a tie from A to B, from B to C, and from C to A. Communication
flows moving in a circle indicate decentralised deliberation. To the contrary,
acyclic configurations are interpreted as indicators of hierarchical networks
(De Nooy et al. 2005: 213 ff).
Last but not least, there is a market for political brokers of information.
These actors invest in bridging ‘structural holes’ (Burt 1992) and serve as
mediators. There are two possible consequences: either these bridges institu-
tionalise as social positions without leading to more direct interactions, or the
probability of a direct connection between two officials increases if there is
at least one indirect relation between them. Whether these local processes of
reciprocity, transitivity, asymmetric popularity and expansiveness, deliberation
and indirect relations are present in transgovernmental networks has never
been assessed empirically. In the next section we provide formal definitions
and assumptions in order to make intersubjectively transparent the opera-
tional prescriptions for the set up of the empirical model.

Definitions and assumptions

We consider the set N consisting of the 15 EU national ‘governments’

(i = 1, . . . , 15). Each government is regarded as a system with specific chains of
delegations (Strøm et al. 2003) and a policy-specific assignment of competen-
cies to partial ministerial jurisdictions. The government can therefore be dis-
aggregated into a team with varying numbers of actors (j = 1, . . . , 9). We call
them for brevity ‘premiers’ and ‘ministers’. Denote aij as a bureaucratic agent
in the premier’s office or in jurisdiction j of Member State i. Each of the agents
involved is endowed with varying formal policy-specific decision-making com-
petencies. However, this simplified formal structure may be embedded in
informal, potentially deviating, authority systems due to or based on informal
interaction patterns. Let F be the set of jurisdiction-specific transgovernmental
networks with F = {PO, MFA, MI, MJ, MF, MEco, MLab, MSoc, MEnv}.13 Thus,
we conceive transgovernmental networks as consisting of multiple function-
specific relations between different equivalent actors in each government of
the Member States.
The assumption of directed relations requires a differentiation between
relations of agent A to agent B and agent B to agent A. For a relation of agent
A to agent B, we consider agent A as the initiator of a communication/
coordination effort whereas agent B is considered as the addressee. Therefore,
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
12 paul w. thurner & martin binder

every network consists of a set of ordered pairs indicating whether the agents
have relations one to another. Each of the jurisdiction-specific networks can be
represented by g ¥ g sociomatrices XPO, XMFA, . . . XMEnv, where g is the number
of actors involved in that network. In case of the occurrence of a directed
relation between agent A and agent B, the cell entry xa,b takes a value of 1, and
takes the value of 0 otherwise.
Each bureaucratic agent aij has a location zij in the multidimensional politi-
cal space Z with dimensions supposed to be separable. A subspace of this
overall political space consists of the set M of negotiation issues (m = 1, . . . ,
46), over which the governments tried to reach an agreement in the inter-
governmental conference. Each negotiation issue can be mapped onto the unit
interval. The official bargaining position of a government on an issue is called
its national bargaining position (NBP). Bureaucratic agents are assumed to
optimise an objective function (Snijders et al. 2005: 38) with respect to a local
network configuration when forming transgovernmental ties. Analogously to
random utility models where non-deterministic discrete choices are assumed
due to uncertainty, random graph models may capture the agents’ uncertainty
when choosing network partners: uncertainty about the attributes of the alteri
and their organisational environment, about the quality of a bilateral relation,
and about the structure of the overall network. Alternatively, random graph
models may reflect a lack of information on the part of the observing scientist
– that is, they include measurement error (McFadden 1974; Manski 1977). Let
us therefore assume that bureaucratic agents are not completely certain about
the exact location of the alteri in the political space. However, each official
is supposed to have the same subjective expectations. In the following it will
be argued that these expectations are represented by probabilistic choice
functions.

Exponential random graph models: A new method for the

explanation of network choices

The main objective of this study is a statistical analysis of each of the different
transgovernmental networks of the bureaucratic agents of the premiers’
offices and of the ministerial jurisdictions. However, in additional to a quanti-
tative description of aggregate network patterns we want to test whether the
observed network choices are purely random or whether their selectivity
exhibits regularities that reflect the hypothesised incentives and processes. As
is well known from the extensive literature on social networks, real-world
networks are far from being completely random. There are interdependencies
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 13

due to reciprocal attraction, transitive closures and so on. These complex

patterns of interdependencies between the units of observation make statisti-
cal estimation a nontrivial task.
Random graph models try to capture these interdependencies by construct-
ing so-called ‘dependence graphs’ where the presence and relevance of specific
local configurations are hypothesised.14 According to the Hammersley-Clifford
theorem (Besag 1974), each hypothesis on the presence of specific dependen-
cies in an observed network requires a particular specification of ‘sufficient
statistics’.The most simple, but also completely unrealistic, model is the assump-
tion of the independence of all ties (Bernoulli Graph). This implies uniform
probabilities for all edges of a given network. A more complex model assumes
dyadic independent tie formation. In this so-called ‘p1 model’, the sufficient
structural parameters control for the number of ties and the number of mutual
ties. Newer approaches, the so-called ‘Markov random graphs’ or ‘p* models’,
construct even more complex dependence graphs. p* models assume that
network ties are conditionally dependent as soon as common actors are shared
– thereby taking into account triadic and even higher order configurations.
The functional form of these models is as follows. Suppose our g ¥ g
sociomatrices XPO, XMFA, . . . XMEnv to be random matrices with diagonal ele-
ments of 0. xa,b indicates whether there is a tie from a to b. Assuming a type 1
extreme value distribution for the stochastic component and a given vector of
sufficient statistics y(x) of a behavioural model, the following exponential
probability function results (Wasserman & Robins 2005: 152–153):

exp {θ ′y ( x )}
Pr ( X = x ) = ,
κ (θ )

where q represents a vector of model parameters, and k(q) is a normalising

constant guaranteeing that estimated probabilities sum to 1. The estimated
coefficients can be interpreted as in conditional logit regression models – that
is, the change in a network statistic, in a actor or in a dyadic covariate, refers to
the change of the log odds of observing a tie or not. The estimated parameters
reflect the relative values of the incentives incorporated in our hypotheses on
(dis-)similarities of the actors, on the quality of their relations (dyadic covari-
ates), as well as on the discussed network processes of reciprocity, transitivity,
attractions and so on. Applying standard maximum likelihood techniques
is now considered inappropriate for statistical testing. Therefore we rely on
a new estimation technique as implemented by Snijders in the StOCNET
module SIENA. Snijders uses a Monte Carlo Markov Chain maximum likeli-
hood estimation technique based on Metropolis-Hastings Sampling (Snijders
et al. 2005).
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
14 paul w. thurner & martin binder

Data and descriptive results

In the standardised interviews with 140 bureaucratic agents involved15 in the

preparation of the IGC 1996 the following question was asked:
Sometimes, it proves to be useful for a ministry – before taking the final
national official position – to come to an agreement with an equivalent
ministry of another Member State. (Interviewer: Please show list F16).
Could you indicate the Member States where you have practiced such an
agreement building?
This network generator is neutral with regard to the analytical distinction
between transgovernmental coordination and transgovernmental coalition-
building as put forward by Keohane and Nye (1974). It does not carry
meanings such as ‘influencing others’ or ‘demanding information’. We only
investigated a communicative effort for ‘distilling and disseminating credible
information’ (Slaughter 2004: 178).
For illustrative reasons, we present the emerging transgovernmental
network of the officials of the ministries of environment in Figure 1. The
visualisation is based on the concept of PageRank (Brin & Page 1998).17
PageRank-centrality operationalises a concept of centrality that is constructed
not only by simply counting the number of received choices (in degrees) of an
actor, but also by taking into account the number of sent and received choices of
the respective initiator. Therefore, this measure also accounts for indirect ties,
and the importance of the initiator.
The most central bureaucratic agents, the officials of Germany and Sweden,
are located in the middle of the circle. All other actors are depicted on
concentric circles around the centre according to their decreasing centrality
values. Actors beyond the outer dashed circle have a value of zero. They have
not been involved in the domestic process. The directedness of choices is
represented by arrows. A unilateral offering of or demand for a relation is
represented by a single arrow. Non-directed edges indicate symmetrical trans-
governmental relations. Circles representing the respective Member State are
used as long as the number of indegrees and targeted agents (outdegrees) are
identical. Ellipses indicate the asymmetry between indegrees and outdegrees.
The more indegrees as compared to outdegrees, the flatter the ellipse of an
actor in the vertical direction (e.g., Denmark, DK) and the more outdegrees
as compared to indegrees, the flatter the ellipse of an actor in the horizontal
direction (e.g., Austria, AUT). The size of the ellipse/circle grows with the
overall number of relations.
Figure 1 shows that only the officials of the ministries of the environment
of the Scandinavian countries, Austria, the Netherlands, Germany and Great
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 15

Figure 1. Informal transgovernmental communication channels: Environmental ministries

(EU-15) (PageRank).

Britain were engaged in transgovernmental networking. Most of these are well

known as ‘green countries’. This transgovernmental network seems to include
proactive policy pioneers rather than coalitions of negative coordination. As
we can see in Table 1, this network is especially peculiar due to its extremely
high degree of reciprocatedness. Visual inspection already makes clear the
nearly complete cliqueness of this network.18
In Table 2 we present several network descriptive measures for those
networks large enough to be treated statistically. Other ministries, like the
ministries of defence or agriculture, have not been, or have only sporadically
been, involved in the preparation of the intergovernmental conference. For a
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 16

© 2008 The Author(s)

Table 2. Synopsis of relevant network descriptives (calculated with StOCNET; Boer et al. 2006, Snijders et al. 2005)

PO MFA MI MJ MF MEco MLab MSoc MEnv

N of involved actors 13 15 11 14 14 13 12 11 11
Maximum N of directed ties 156 210 110 182 182 156 132 110 110
N of observed ties 29 55 29 44 35 31 31 24 32
Density 0.186 0.262 0.188 0.224 0.179 0.170 0.185 0.156 0.208
N of possible dyads 78 105 55 91 91 78 66 55 55
Mutual (M) 8 8 7 6 11 7 8 5 15
Asymmetric (A) 10 39 10 30 11 15 11 11 2

Journal compilation © 2008 (European Consortium for Political Research)

Null (N) 60 58 38 55 69 56 47 39 38
Index of reciprocity 0.615 0.291 0.583 0.286 0.667 0.483 0.593 0.476 0.938
Maximum N of triads 286 455 165 364 364 286 220 165 165
N of observed triads 81 181 76 107 123 93 90 64 154
paul w. thurner & martin binder

N of transitive triads 36 84 23 54 40 33 36 21 96
Index of transitivity 0.444 0.464 0.303 0.505 0.325 0.355 0.400 0.328 0.623
 european union transgovernmental networks 17

complete network with bureaucratic agents from all 15 Member States there
are

⎛ N⎞
⎜⎝ 2 ⎟⎠ = ( N − N) 2 = (15 − 15) 2 = 105
2 2

possible unordered pairs and 210 ordered pairs. Comparing the densities,
corrected for missing nodes, we get an impression of the relative network
activity of the officials of ministries. As expected, the highest degree of com-
munication activities is reported by the officials of the ministries of foreign
affairs. Traditionally, they have a formal monopoly over the external represen-
tation of the nation-state. As a rule, they managed the coordination units
responsible for preparing the negotiation positions. Note also the relatively
high network activity of the officials of the ministries of justice, and the min-
istries of the environment showing by far the highest degrees of network
activity. The expertise on European Law of the officials of the ministries of
justice is highly appreciated during such conferences. They have to assess the
compatibility of negotiation positions with existing European and national
laws and are therefore expected to communicate with the legal experts of
other Member States. The officials of the ministries of the environment not
only stand out as especially communicative, but the configuration of this trans-
governmental network is, as already shown in Figure 1, extremely reciprocal.
Obviously, these officials had a preference for reciprocating relations. On the
other hand, the network of the officials of the ministries of foreign affairs is
highly asymmetric. This indicates the existence of informal authorities or at
least dependencies. The relatively low density of the network of the officials
of the premiers’ offices reflects the usually applied ex-post control, leaving
transgovernmental precoordination efforts to the officials of the partial
jurisdictions.
According to the network densities, all networks under investigation have
been highly selective, and therefore we conclude already at this stage that the
officials followed a cost-benefit analysis when establishing and maintaining
informal relations with equivalent ministries in other EU Member States.
We point again to the asymmetric nature of the dyads – that is, we distinguish
between addressing and being addressed. Asymmetric dyads are often con-
ceived of as ‘intermediate states of relationships that are striving for a more
stable equilibrium of reciprocity, or complete nullity (devoid of either arc)’
(Wasserman & Faust 1994: 510–511). Further interpretations argue that asym-
metric dyads reveal an unequal distribution of resources. Another important
attribute of (sub-)graphs is the degree of transitivity given the distribution of
triads and triplets. Triads consist of relations connecting triplets (three actors)
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
18 paul w. thurner & martin binder

and are therefore structurally embedding dyads. Transitive relations constitute

a social equilibrium state allowing for reciprocal exchange and its control.
They secure the control of compliance and policy implementation. Again, the
network of the officials of the ministries of environment stands out as espe-
cially transitive. Do these descriptive measures indicate a significant deviation
from pure randomness given the overall network configuration? And is the
selectivity of these network choices driven by the hypothesised incentives?
This will be tested with the following statistical analysis.

Test of hypotheses

Table 3 shows the estimated parameters for each of the effects discussed.
We begin the discussion with the block of structural parameters. Most of the
jurisdictional networks are characterised by non-random reciprocity and tran-
sitivity effects. This indicates that reciprocated ties and network closure have
been valued positively. Officials have been more inclined to communicate
transgovernmentally with other officials if such relations were reciprocated
and if there were also indirect ties connecting them. Both results corroborate
the thesis of an institutionalisation of transborder interactions among national
bureaucracies. Remember, however, that this process of institutionalisation is
obviously very selective. Both non-random reciprocal and transitivity effects
are absent in the transgovernmental network of the officials of the ministries
of economics. The networks of the officials of the ministries of labour and the
ministries of the environment exhibit no additional preference for transitive
closure, whereas the network of the officials of the ministries of foreign affairs
network is characterised by a lack of reciprocated relations. However, this
network is characterised by a tendency towards transitive structures. We inter-
pret this as an indicator that foreign offices pay attention to communication
efficiency and non-redundancy.
The network of the ministries of the environment is remarkable insofar
as the highly reciprocal nature of the relations absorbs all other effects. Addi-
tionally including our hypothesised micro-incentives based on actor attributes
or on dyadic attributes does not improve the models in a way that is required
statistically (Snijders et al. 2005: 24). Therefore we present only the reduced
form model. Interestingly, the negative effects of ‘three-cycles’ in the networks
of the premiers offices, the ministries of the interior, the ministries of finance
and the ministries of social affairs point to a statistically significant low occur-
rence of this configuration. We conclude, therefore, that in four out of nine
transgovernmental networks, we observe tendencies towards informal hierar-
chies whereas ‘deliberative’ tendencies have been completely absent.
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 Table 3. ERGM: Sufficient statistics and dyadic covariates

PO MFA MI MJ MF Meco MLab Msoc Menv

Reciprocity 4.209* 0.098 3.495* 1.123 4.044 1.477 1.815* 2.748* 6.322*
(1.303) (0.707) (1.168) (0.815) (0.946) (0.857) (0.901) (1.210) (2.185)
Expansiveness 0.569* -0.051 0.251 0.114 0.099 0.350* 0.124 0.030 -2.174
(out-2-stars) (0.160) (0.140) (0.308) (0.115) (0.219) (0.161) (0.293) (0.251) (1.617)
Popularity (in-2-stars) 0.070 -0.116 0.187 -0.205 -0.224 0.122 -0.229 -0.830 -1.973
(0.348) (0.143) (0.261) (0.190) (0.309) (0.255) (0.367) (0.542) (1.441)
Indirect Relations -0.074 -0.147 -0.003 -0.319* 0.189 0.017 0.042 0.588 2.486
(2-paths) (0.251) (0.136) (0.250) (0.147) (0.240) (0.172) (0.295) (0.340) (1.566)
Transitivity 0.616* 0.533* 0.616* 0.645* 0.753* 0.256 0.417 0.944* 0.924
(0.201) (0.149) (0.259) (0.181) (0.187) (0.246) (0.324) (0.312) (0.855)
Deliberation (3-cycles) -1.533* -0.527 -2.098* -0.493 -1.988* -0.143 0.066 -2.796* -2.223
(0.682) (0.418) (0.888) (0.534) (0.667) (0.687) (0.909) (1.133) (2.587)
Bypassing Incentive 0.001 -0.028 -0.004 0.002 -0.008 -0.018 -0.001 -0.116
(0.042) (0.017) (0.016) (0.013) (0.017) (0.026) (0.026) (0.133)
Low Hierarchy -1.432* -0.056 0.938 0.431 0.205 -0.054 0.473 0.354
(0.700) (0.372) (0.513) (0.383) (0.410) (0.431) (0.420) (0.719)
Bypassing Incentive x -0.034 -0.010 0.009 -0.005 0.011 0.021 -0.006 0.102
Low Hierarchy (0.031) (0.011) (0.017) (0.011) (0.017) (0.027) (0.024) (0.132)
Issue Distance -0.030 -0.078* 0.002 -0.016 0.003 -0.021 -0.032* -0.028
(0.033) (0.026) (0.021) (0.012) (0.014) (0.014) (0.014) (0.016)
Informal ‘Core Europe’ 1.124 0.274 -0.079 -0.222 0.334 0.243 0.893* 0.425
european union transgovernmental networks

(0.691) (0.367) (0.520) (0.409) (0.356) (0.433) (0.437) (0.531)

Bilateral Economic 0.117* 0.066* 0.089* 0.089* 0.060* 0.055* 0.134* 0.121*
Interdependencies (0.038) (0.024) (0.040) (0.030) (0.026) (0.026) (0.041) (0.047)

Notes: * p < 0.05. Standard errors are in parentheses. Premiers Offices = PO; Ministries of Foreign Affairs = MFA; Ministries of the
Interior = MI; Ministries of Justice = MJ; Ministries of Finance = MF; Ministries of Economy = MEco; Ministries of Labour = MLab; Ministries
19

© 2008 The Author(s)

Journal compilation © 2008 (European Consortium for Political Research)
of Social Affairs = MSoc; Ministries of the Environment = MEnv.
20 paul w. thurner & martin binder

The second block of estimation results pertains to our hypotheses on

micro-incentives, which relate to actor attributes or dyadic attributes. The
hypothesis that issue proximity (i.e., the similarity of policy preferences)
induces informal transgovernmental relations is partly corroborated. In the
case of the networks of the officials of the ministries of foreign affairs and
the ministries of labour affairs, the functional relationship between policy
distance and network choice is negative: the larger the policy distances
between the respective ministries, the smaller the chance that they entertain
relations. We interpret this as corroboration of the expectation that trans-
governmental efforts have been made to build or rely on coalitions with
like-minded state officials. One of the most astonishing results can be seen in
the coefficients on bilateral trade interdependencies. For the first time, results
from a complex relational analysis show that informal transgovernmental
interactions have an economic background: the higher the relative export
shares from Member State A to Member State B, the higher the probability
of a transgovernmental relation, regardless which jurisdictional network is
considered.
Finally, Keohane and Nye’s bypassing conjecture is refuted for all the
networks. Neither the main effects of the existence of the simple bypassing
incentive and strong ministerial autonomy, nor their interaction effect are
statistically significant. Governments characterised by a low-hierarchy do not
induce more transgovernmental relations, nor do they increase the impact of
the bypassing incentive.There is one exception: officials of the premiers’ offices
have a significantly reduced tendency to form transgovernmental network
in governments where the ministries have high autonomy. This is plausible
insofar as we expect that transgovernmental self-organisation in such contexts
is exclusively carried out by specialised ministries endowed with high discre-
tionary decision-making competencies.
What could be the reason for this quasi non-existence of moral hazard in
EU transgovernmental networks? Do we observe political control over the
national bureaucracies (Page & Wright 1999)? At first glance, it seems that the
national governmental executives effectively exercised command and control,
that they even mastered the political control of transgovernmental messages.
Probing deeper into necessary conditions (Braumoeller & Goertz 2003) of the
Keohane/Nye conjecture, we formulated two further hypotheses: in order to
be effective the bypassing incentive (a) must be simultaneously present for
both agents (‘mutual bypassing incentive’), and (b) occur only in trustful
bilateral situations. The latter condition has been operationalised as an inter-
action effect of the bypassing incentive and the situation of a reciprocated
relation. Testing these additional hypotheses we get the result that a ‘mutual
bypassing incentive’ does not contribute to the explanation of the occurrence
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 21

of transgovernmental relations. Bypassing in our transgovernmental networks

is only effective conditional on the existence of a reciprocal communication
situation – namely in the case of the networks of the ministries of the interior,
ministries of finance, ministries of economy and ministries of social affairs.
Here the interaction effect of bypassing and reciprocity proves to be statisti-
cally significant.19 We interpret this as an indication that the risk of bypassing
is taken only in mutually trustful relations.

Conclusion

For the first time we have reconstructed the informal transgovernmental

organisation of an important EU intergovernmental conference. We would
like to emphasise that it is already difficult to trace out such behaviour in
private organisations and at lower levels of public and private organisations.
Naturally, the existence of transgovernmental relations has been acknowl-
edged in the scientific literature for decades. The pattern of the emerging
overall structure – often beyond individual and bilateral will and perception –
and its selectivity were so far unknown. Our objective in this article was
explanatory: we wanted to contribute to the question of ‘why’ informal trans-
governmental communities are formed.
The main results of our statistical analyses are: bureaucratic agents carry
out a cost-benefit analysis when entertaining transgovernmental networks;
these networks are highly selective and they exhibit low densities. Given the
statistically significant effects of reciprocity, transitivity and acyclical relations,
there are strong indications of a high degree of institutionalisation of these
interactions and of the existence of mutual trust. Third, the existence of acy-
clical relations corroborates the expectation of the emergence of informal
administrative hierarchies. Structures of informal deliberation are absent.
The existence of these structural effects confirms the hypothesis that politics
and markets are socially constructed and embedded (Fligstein & Stone
Sweet 2002). Fourth, the often repeated conjecture that preferential similarity
induces and facilitates transgovernmental coordination is at least partly cor-
roborated by our data; affinities with regard to political issues facilitate com-
munication between the officials of different states. Fifth, bilateral economic
interdependencies are related to the formation of transgovernmental rela-
tions, regardless of the type of ministry considered – that is, regional political
integration and economic exchange patterns mutually interact. Our research
design and our results are in line with a research agenda suggested recently by
Fligstein (2005: 195): ‘The expectation is that one will observe global gover-
nance where markets exist and a push for new governance, where markets
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
22 paul w. thurner & martin binder

openings have occurred.’ This hypothesis has never been investigated in the
way that we have proposed it in this article.
Last but not least, we put forward a measure for the effectiveness of state
boundaries amid processes of regional integration. The proposed bypassing
incentive seems to have been at work – but only in several actor networks
(networks of the ministries of the interior, ministries of finance, ministries of
economy and ministries of social affairs) and conditional only on mutually
confirmed relations. In order to bear the risk of bypassing, agents had to
consider the relationship as stable and trustworthy. We emphasise that we not
only put forward an operational hypothesis of the Keohane/Nye conjecture,
but we also offered a theoretical extension insofar as we proposed and
corroborated an alternative necessary condition for the effectiveness of this
behavioural mechanism. We conclude that moral hazard is existent and
exploited in transgovernmental relations, but the breakup of nations by trans-
governmental relations is far from being complete.
Despite acknowledging the transient character of the observed trans-
governmental structure of European regional integration, we claim that we
have been able to derive insights of a wider spatial-temporal process of the
development of European integration. As a rule, those high officials having
been or being in charge of the preparation of negotiation issues have been and
continue to be those who are responsible and competent in the every-day
policy making. State officials are not nominated completely anew, nor are
responsible teams composed from scratch to prepare an Intergovernmental
Conference. These state officials are experts and have experience with
Brussels’ comitology – and they rely on their established networks. Naturally,
we do not argue that the observed network structure is time-independent.
Especially, EU enlargement should have changed the overall structure. Nev-
ertheless, due to the human tendency to build stable coalitions, the network
links between the ‘old members’ are expected to have strong inertia and are
path-structuring. As classical studies of the dynamics of bureaucracy (Blau
1955) have shown, successful and legitimate forms of informal administrative
self-organisation potentially institutionalise to new forms of accepted gover-
nance. The institutionalisation of informal hierarchies potentially leads to new
taken-for-granted forms of political governance. Our analyses of informal
inter-organisational relations clearly indicate the emergence of informal trans-
governmental hierarchies of informal administrative leaders and followers.
What are then the broader theoretical implications of this study? First, the
development of Eurpopean integration is driven mainly by the Member States
– just as liberal intergovernmentalism claimed. However, states and govern-
ments are, and have always been, ‘legal fictions’. Consequently, we disaggre-
gated the state into those parts mostly responsible for policy making in
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 23

parliamentary systems: the ministries. State officials of the ministries prepare

the sectoral councils of ministries, they control the comitology and they
prepare Intergovernmental Conferences. The transgovernmental cooperation
of administrations is a strategic process – including coalition-building and
strategic communication. There is no smooth global functional administrative
‘fusion’, but rather a competitive process of politicians and bureaucrats with
vested career concerns. A by-product is the emergence of informal authority
structures. Focusing on the transgovernmental nerves of governance it is pos-
sible to identify specific patterns of community formation below the processes
of formal intergovernmental conferences. The understanding of the inter-
organisational networks of these state officials is a prerequisite for the under-
standing of the process of EU governance – and its management.

Appendix. Operationalisation of hypotheses

Bypassing

46

∑ BY a ,b
iid =1
Bypassing a ,b = 46

∑ VT
iid =1

⎧1 if Positiona − Position b < Positiona − NBPA

where BYa ,b = ⎨
⎩ 0 otherwise

NBPA is the publicly declared bargaining position of Member State A

⎧ 1 if Positiona , Position b and NBPA are not missing

VT = ⎨
⎩0 otherwise

Low Hierarchy20

⎧1 if Member State A has a high degree

⎪
⎪ of ministerial autonomy
Low Hierarchy A = ⎨
⎪ 0 if Member State A has a low degree
⎩⎪ of ministerial autonomy
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
24 paul w. thurner & martin binder

Issue Distance

46

∑ Positiona − Position b
iid = 1
Distancea , b = 46

∑ VD
iid = 1

⎧ 1 if Positiona and Position b are not missing

where VD = ⎨
⎩0 otherwise

Informal Core Europe

⎧ if Member State A is a founding member of

⎪1
Informal Core EuropeA = ⎨ the EU
⎪0
⎩ otherwise

Export21

Export from Member State A to B

Export A, B =
∑ all Exports of A

Notes

1. For empirical studies on legislative decision making in the EU, see Bueno de Mesquita
and Stokman (1994); Thomson et al. (2006).
2. ‘State officials develop policy-making styles . . . that are provoked by, and are used to
respond to, perceived dysfunctionalities’ (Fligstein & Stone Sweet 2002: 477).
3. ‘For a transgovernmental coalition to take place, a sub-unit of one government must
perceive a greater common interest with another government, or sub-units of another
government, than with at least one pertinent agency in its own country; and central
executive control must be loose enough to permit this perception to be translated into
direct contacts with the foreign governments or agencies in question’ (Keohane & Nye
1974: 48).
4. For recent case studies, see Newman (2008); Eberlein and Newman (2008).
5. For more details on these flexibility provisions, see Title VII of the Treaty of the EU. An
overview of the debate is provided in Stubb (2002).
6. We calculate the proportion of issues where an agent has a larger distance with regard
to the official national position than to the position of the agent in the other Member
State. We provide formal operational definitions for all hypotheses in the Appendix to
this article.

© 2008 The Author(s)

Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 25

7. For an operationalisation of ministerial autonomy, see Laver and Hunt (1992: 125).
8. More concretely, low governmental hierarchy can be thought of as a high degree of
ministerial autonomy (see Laver & Hunt 1992: 125).
9. This effect should become visible by specifying an interaction effect of both variables.
10. Operationally, we add up the issue distances between two agents over 46 issues. For
reasons of inter-individual comparability, we normalise with regard to the number of
positions taken simultaneously by both agents.
11. For the context of the EU, see Moravscik (1998: 27–35, 476–478).
12. Note that we suggest here a network-oriented micro-foundation of institutionalisation
processes. ‘Institutionalisation’ has recently been defined as the ‘self-enforcing feedback
dynamics of heightened legitimacy and enhanced-taken-for-grantedness’ (Colyvas &
Powell 2006: 306).
13. Premiers Offices = PO; Ministries of Foreign Affairs = MFA; Ministries of the
Interior = MI; Ministries of Justice = MJ; Ministries of Finance = MF; Ministries of
Economy = MEco; Ministries of Labour = MLab; Ministries of Social Affairs = MSoc;
Ministries of the Environment = MEnv.
14. For recent overviews on random graph models, see Robins et al. (2006); Wasserman and
Robins (2005); Snijders (2002); Snijders et al. (2005).
15. ‘Involved’ means that bureaucratic agents have written proposals for the national co-
ordination unit. These persons were indicated by the delegation leaders in a first wave
of data collection. This form of snowball sampling implies that all relevant actors were
surveyed.
16. List F showed the EU-15 Member States in an alphabetical order.
17. Figure 1 was produced with the software package visone.
18. Cliques are characterised by maximal connected subgraphs (i.e., all member of clique
are directly connected) (Wasserman & Faust 1994: 254). These are the cliques: AUT-
GER-NL-SWE; FIN-GER-NL-SWE; DEN-GER-NL-SWE; DEN-GER-GB-SWE.
19. Estimation of these additional models are available from the authors on request.
20. Laver and Hunt (1992: 125): ‘High autonomy’ is assumed to take values of 1–5 of their
index.
21. Data are available online at: http://comtrade.un.org/db/

References

Ansell, C. (2000). The networked polity: Regional development in Western Europe. Gover-
nance 13(3): 303–333.
Austen-Smith, D. (1990a). Credible debate equilibria. Social Choice and Welfare 7(1): 75–93.
Austen-Smith, D. (1990b). Information transmission in debate. American Journal of Political
Science 34(1): 124–152.
Bendor, J., Glazer A. & Hammond, T.H. (2001). Theories of delegation. Annual Review of
Political Science 4: 235–269.
Besag, J. (1974). Spatial interaction and the statistical analysis of lattice systems. Journal of
the Royal Statistical Society B 36: 192–236.
Blau, P. M. (1955). The dynamics of bureaucracy: A study of interpersonal relations in two
government agencies. Chicago, IL: University of Chicago Press.
Boer, P. et al. (2006). StOCNET User’s Manual. Groningen: ICS.

© 2008 The Author(s)

Journal compilation © 2008 (European Consortium for Political Research)
26 paul w. thurner & martin binder

Braumoeller, B.F. & Goertz, G. (2003). The statistical methodology of necessary conditions.
In G. Goertz & H. Starr (eds), Necessary conditions: Theory, methodology and applica-
tions. Lanham, MD: Rowman & Littlefield.
Breton, A. (1998). Bureaucracy. In P. Newman (ed.), The new Palgrave dictionary of eco-
nomics and the law, vol. 3. New York: Stockton Press.
Brin, S. & Page, L. (1998). The anatomy of a large-scale hypertextual web search engine.
Computer Networks and ISDN Systems 30(1–7): 107–117.
Bueno de Mesquita, B. (2003). Principles of international politics: People’s power, preferences
and perceptions. Washington, DC: Congressional Quarterly Press.
Bueno de Mesquita, B. & Stokman, F.N. (eds) (1994). European Community decision making
models, applications and comparisons. New Haven, CT: Yale University Press.
Burt, R.S. (1992). Structural holes: The social structure of competition. Cambridge, MA:
Harvard University Press.
Colyvas, J. & Powell, W.W. (2006). Roads to institutionalization: The remaking of boundaries
between public and private science. Research in Organizational Behavior 21: 305–353.
De Nooy, W., Mrvar, A. & Batagelj, V. (2005). Exploratory social network analysis with
Pajek. Cambridge: Cambridge University Press.
Eberlein, B. & Newman, A. (2008). Escaping the international governance dilemma? Incor-
porated transgovernmental networks in the European Union. Governance 21(1): 25–52.
Enelow, J.M. & Hinich, M.J. (1984). The spatial theory of voting:An introduction. Cambridge:
Cambridge University Press.
Fearon, J.D. (1998). Deliberation as discussion. In J. Elster (ed.), Deliberative democracy.
Cambridge: Cambridge University Press.
Fligstein, N. (2005). The political and economic sociology of international economic arrange-
ments. In N. Smelser & R. Swedberg (eds), Handbook of economic sociology. Princeton,
NJ: Princeton University Press.
Fligstein, N. & Mara-Drita, I. (1996). How to make a market: Reflections on the European
Union’s single market program. American Journal of Sociology 102: 1–33.
Fligstein, N. & Stone Sweet, A. (2002). Constructing politics and markets: An institutionalist
account of European integration. American Journal of Sociology 107: 476–512.
Grande, E. & Pauly, L.W., eds. (2005). Complex sovereignty: Reconstituting political authority
in the 21st century. Toronto: University of Toronto Press.
Grofman, B. (1982). A dynamic model of protocoalition formation in ideological N-space.
Behavioral Science 27: 77–90.
Haas, E.B. (1961). International integration: The European and the universal process. Inter-
national Organization 15: 366–392.
Hammond, T.H. & Knott, J.H. (1996). Who controls the bureaucracy? Presidential power,
Congressional dominance, legal constraints and bureaucratic autonomy in a model of
multi-institutional policymaking. Journal of Law, Economics and Organization 12(1):
121–168.
Hayes-Renshaw, F. & Wallace, H. (1997). The Council of Ministers. New York: St Martin’s
Press.
Jackson, M.O. & Wolinsky, A. (1996). A strategic model of social and economic networks.
Journal of Economic Theory 71: 44–74.
Keohane, R.O. & Nye, J.S. (1974). Transgovernmental relations and international organiza-
tions. World Politics 27: 39–62.
Laver, M. & Hunt, W.B. (1992). Policy and party competition. New York: Routledge.

© 2008 The Author(s)

Journal compilation © 2008 (European Consortium for Political Research)
 european union transgovernmental networks 27

Lupia, A. (2003). Delegation and its perils. In T. Bergman, W.C. Müller & K. Strøm (eds),
Delegation and accountability in West European parliamentary democracies. Oxford:
Oxford University Press.
Manski, C.F. (1977). The structure of random utility models. Theory and Decision 8: 229–254.
Mattli, W. (1999). The logic of regional integration: Europe and beyond. Cambridge:
Cambridge University Press.
McFadden, D. (1974). Conditional logit analysis of qualitative choice behaviour. In
P. Zarembka (ed.), Frontiers in econometrics. New York: New York Academic Press.
Moravcsik, A. (1998). The choice for Europe: Social purpose and state power from Messina
to Maastricht. Ithaca, NY: Cornell University Press.
Newman, A. (2008). Building transnational civil liberties: Transgovernmental entrepreneurs
and the European Data Privacy Directive. International Organization 62: 103–130.
Newman, M.E.J. (2003). The structure and function of complex networks. SIAM Review
45(2): 167–256.
Page, E.C. & Wright, V. (eds) (1999). Bureaucratic elites in Western European states: A
comparative analysis of top officials. Oxford: Oxford University Press.
Robins, G. et al. (2006). Recent Developments in Exponential Random Graph (p*) Models
for Social Science Networks. Unpublished manuscript, University of Melbourne.
Scharpf, F.W. (1997). Games real actors play: Actor-centered institutionalism in policy
research. Boulder, CO: Westview Press.
Slaughter, A. (2004). A new world order. Princeton, NJ: Princeton University Press.
Snijders, T.A.B. (2002). Markov Chain Monte Carlo estimation of Exponential Random
Graph Models. Journal of Social Structure 3: 1–40.
Snijders, T.A.B. et al. (2005). Manual for SIENA, version 2.1. Groningen: ICS, Department of
Sociology.
Stone Sweet, A., Sandholtz, W. & Fligstein, N. (eds) (2001). The institutionalization of
Europe. Oxford: Oxford University Press.
Strøm, K., Müller, W.C. & Bergman, T. (2003). Delegation and accountability in parliamen-
tary democracies. Oxford: Oxford University Press.
Stubb, A. (2002). Negotiating flexibility in the European Union:Amsterdam, Nice and beyond.
Basingstoke: Palgrave.
Thomson, R. et al. (eds) (2006). The European Union decides. Cambridge: Cambridge
University Press.
Wasserman, S. & Faust, K. (1994). Social network analysis: Methods and applications.
Cambridge: Cambridge University Press.
Wasserman, S. & Robins, G. (2005). An introduction to random graphs, dependence graphs
and p*. In P.J. Carrington, J. Scott & S. Wasserman (eds), Models and methods in social
network analysis. Cambridge: Cambridge University Press.
Wessels, W. (1997). An ever closer fusion: A dynamic macro-political view of integration
processes. Journal of Common Market Studies 35: 267–299.

Address for correspondence: Paul W. Thurner, Empirical Political Research and Policy
Analysis, University of Munich, Geschwister-Scholl-Institut für Politische Wissenschaft,
Oettingerstrasse 67, D-80538 München, Germany. Tel.: +49 (0)89 2180 9080; Fax: +49
(0)89 2180 9092; E-mail: paul.thurner@mzes.uni-mannheim.de; Website: www.gsi.uni-
muenchen.de\lehreinheiten\le_policy_analysis\index.html

© 2008 The Author(s)

Journal compilation © 2008 (European Consortium for Political Research)