Beyond the myth of nationality: Analysing networks within the European commission

Abstract

Literature on the European Commission refers to the influence of nationality in the functioning of the Commission and, in particular, to the reliance on networks based on nationality, but it fails to provide much evidence apart from anecdotes. This empirical study takes a systematic approach by applying concepts from organisational network analysis to examine the networking patterns of Commission officials and to explore the effect of nationality therein. The data clearly show that nationality is not a significant factor in shaping officials' task-related informal networks. While variables related to nationality and socialisation fail to explain the variation, the size of the member state in terms of the number of officials it has and whether the contacts occur within the Directorate-General determine whether an official relies on compatriots for information and advice. The organisational structure of the Commission renders nationality irrelevant for its daily work.

By the nature of their composition, nationality is a salient factor within international organisations. On the one hand, member states demand to be represented within these organisations through their citizens; on the other hand, the same organisations are expected to be independent, i.e. to have an international outlook free of national influences. The need for representation is linked both to the question of legitimacy and to making successful policies that take national circumstances into account. The need for independence in turn stems from the desirability of assuring an overarching interest for all member states. International organisations need to deal with this dilemma1 by starting with the recruitment of their officials. The United Nations was the first to formally acknowledge this dilemma by embedding the principles of ‘recruiting staff on as wide a geographical basis as possible’ and ‘not seeking or receiving instructions from any government or from any other authority’ into its Charter (Claude 1971: 193). The European Commission followed suit by endorsing and applying both of these principles to its College of Commissioners (through the treaties) and to its officials (through its staff regulations).

Career aspects, however, are just one facet of the nationality issue within an international bureaucracy. Personnel management in a multinational context has challenges of its own, as the diverse cultural backgrounds of officials might be seen as a barrier to creating a common working culture (Claude 1971: 192; Mazey and Richardson 1996: 419; McLaren 1997: 57–59; Page 1997: 97) or because the working culture might be ‘cross-cut and fractured by particularistic attachments and cultures’ (Nugent 2000: 297), as has been argued in the case of the Commission.

The European Commission is not only the largest of the EU institutions; it is unique for its exclusive formal competence in initiating and drafting EU legislation (Hooghe 2005: 863). Additionally, the Commission is an interesting institution to study from an organisational perspective because its formal organisational structure makes it less likely that nationality and national attachments will play a role. Comparable to national ministries, the Commission is divided into Directorates-General (DGs) which reflect sectors or functions instead of territories or national components (Egeberg 2004).2 Furthermore, the geographical balance principle is applied from the bottom up, starting with the multinational composition of units to avoid national enclaves (ibid.: 212).

Setting up a formal structure based on the multinationality ideal, however, is only one side of the story. Ethnographic studies have shown that most bureaucratic organisations tend to develop disorganised yet successful informal sectors that follow very different cultural codes and principles beneath the external appearance of order and formality (Shore 2000: 207). It is this informal culture of the Commission that has been at the origin of the reference to national influences, clubs and networks. It has been argued that these national networks provide instrumental functions such as the exchange of information, contacts with more influential compatriots within and outside the Commission, and political opinionating (Hooghe 1999: 415).

Indeed, the idea of compatriots seeking and supporting compatriots seems plausible. As organisational network theory asserts, people interact more with their own kind, that is, with people similar to themselves, and this basic human tendency structures network ties of every type, from marriage and friendship to work, advice, support and information transfer (McPherson et al. 2001). But what about the networks of Commission officials? Are national networks a myth or everyday reality?

This article assesses whether European Commission officials are homophilic in terms of their networks. Put differently, do Commission officials have national networks and how can this be explained? In answering this question, this study moves beyond the current literature which confines itself merely to stating that networks are key to the functioning of the Commission (Hooghe 1999; Shore 2000; Stevens and Stevens 2001) and that they are partly shaped by nationality (Egeberg 1996: 731; Hooghe 1999: 405; Laffan 2004: 90; Shore 2000: 199). The concept of networks is defined and deconstructed, and a differentiation is made between task-related (formal and informal), career, and leisure networks. The task-related, informal networks are, in turn, singled out in order to analyse whether and how nationality plays a role in shaping the networking behaviour of Commission officials. The research is based on original empirical data consisting of 82 interviews with Commission policy officials of four Commission DGs between April and June 2005. The analysis is derived from both the quantitative and qualitative survey data.

The effect of nationality in perspective

The role of national origins and culture within the European Union is a popular theme in the European integration literature. Michelle Cini (1996: 125) has argued that national affiliation remains a fundamental characteristic of internal Commission affairs, in spite of the non-nationalistic vision of the EU set out by the founders of the European Commission. It has been observed that the convergence that results from a common professional experience does reduce national identification to a certain extent and brings distance to relations with the native state (Bellier 1995: 60), but that individuals are notably still conscious of their differences and sensitivities (Abélès and Bellier 1996: 435). The assumption is, ‘However open-minded, flexible and adaptable an official may be, … background matters because he or she will bring into work at least some of the values, presuppositions and habits that have been acquired in early life’ (Stevens and Stevens 2001: 116).

The core issue here is not that officials have different backgrounds, which is a given, but that the differences are perceived to influence the behaviour of officials from different backgrounds and that these perceptions ‘can feel empirically true’ (Abélès et al. 1993: 42; McDonald 2000: 66). To manage these differences, officials tend to make assumptions about each other that emphasise such diverging characteristics (Page 1997: 87). Such assumptions involve national stereotyping by a significant proportion of officials, which especially comes to the fore during times of stress (Michelmann 1978: 494). One blatant example of this was when the Jacques Santer Commission (1995–99) had to resign due to allegations of fraud and nepotism. In particular, Edith Cresson (then the French Commissioner responsible for Science, Research, and Development) was accused of favouritism involving the appointment of several associates to well-paid positions in the Commission.

[M]any Commission officials agreed that Cresson had been victim of an ‘Anglo-Saxon political crusade’ and deplored the way the ‘Germans had joined the [N]ortherners in a Protestant crusade against the [S]outhern culture of state administration’. As Cresson declared, much to the embarrassment of her colleagues, she was ‘guilty of no behaviour that is not standard in the French administrative culture’. (Shore 2000: 202)

As this example demonstrates, even though the EU bureaucracy is relatively culturally homogenous compared to, for instance, the UN Secretariat, the differences between the North and South have been often cited as a dividing line (Abélès et al. 1993; Beyers and Dierickx 1997, 1998; Egeberg 1996; Hofstede 1994; McDonald 2000), where Austria, Britain, Denmark, Finland, Germany, Ireland, Luxembourg, Netherlands, and Sweden constitute the Northern group and Belgium, France, Greece, Italy, Portugal, and Spain comprise the Southern group.3

Arguably, the North–South division influences how officials behave in the Commission. Culturally, Northerners are taken to be well adapted to the rational, impersonal Weberian forms of bureaucracy, whereas Southerners are assumed to link loyalties and virtues to personal or patron–client type obligations (Egeberg 1996: 727; McDonald 2000: 67–68). To take the example of the issue of information flow, the Northern officials have been observed to complain of the lack of readily shared information in the Commission, whereas Southerners are said to find it simple to obtain information by making friends (McDonald 2000: 67).

The foregoing examples are not insignificant observations about life in the multinational administration of the Commission. Yet, for these observations to become practically significant for scholars of public administration and policy studies, the question has to be answered as to how these differences influence the behaviour of administrators when it comes to their substantive work, so that it is possible to analyse whether this eventually has any consequences for policy-making. This is why nationality and national differences have to be put into perspective by looking into the organisational context of the Commission.

Such an organisational approach has been brought into the European integration literature primarily by Morten Egeberg (2001, 2004, 2006a, 2006b). While acknowledging the fact that Commission officials are pre-packed with national experiences, norms and values, he argues, ‘Although these personal attitudes may be seen as some sort of paradigm, belief system or conceptual lens that might somehow make a difference in a given decision situation, they are, nevertheless, of a relatively general nature. To become relevant in a given decision context, they have to be operationalised and pass several potential organisational filters' (Egeberg 2004: 212).

As mentioned above, the organisational structure of the Commission closely reflects that of a national administration in which DGs are the ministries, each with its own policy area or function. The DG attachment is particularly important for understanding the actual decision-making behaviour of Commission officials (Cram 1994; Cini 1996, 2000; Egeberg 1996, 2004, 2006b). The fact that the officials' posts are organised according to purpose and function makes it less likely that officials will focus on territorial (national) concerns (Egeberg 2004: 212–213). Moreover, in order to avoid national clusters, units are multinationally composed and staff immediately above and below a given senior post are of a different nationality.

Still, ‘Formal relations coordinate roles or specialised activities, not persons’ (Selznick 1957: 8). This is why nationality can still be expected to play a role in contrast to or simultaneously with the formal role expectations, especially in areas where formal tasks and obligations leave leeway for individual fulfilment. It is when individuals use informal channels to conduct their formal tasks that the informal becomes relevant for the performance of an employee and in turn for the functioning of the organisation.

Networks in the commission and organisational network theory

Having access to the right information at the right time is vital, especially for conducting knowledge-intensive work such as professional services which require employees to solve complex problems within short time horizons (Cross and Borgatti 2004: 137). For the most part, such issue-specific and problem-centred information is in individuals' heads since ‘specialisation (among other factors) ensures that each individual maintains different bundles of knowledge’ (Cross et al. 2001: 216).

When interactions with people are the only way to get information that matters (Cross and Borgatti 2004: 137), connections to other colleagues become very valuable. This is the underlying idea behind networks as social capital (Borgatti and Foster 2003: 993). The social capital of individuals consists of their personal networks and their chances of accessing whatever is circulating there, e.g. information (Cross and Parker 2004: 11). Accordingly, ‘whom you know has a significant impact on what you come to know, because relationships are critical for obtaining information, solving problems, and learning how to do your work’ (ibid.). As such, personal networks are an important factor in an employee's performance within an organisation (ibid.). This is why it is of particular interest to look into the networking behaviour of individuals in organisations.

Stevens and Stevens (2001: 177) underscore the value of information within the policy-making processes of the Commission and argue that information is a key resource in daily relationships within this organisation. Information flow constitutes an important informal aspect of power relationships in the Commission (Abélès et al. 1993: 6). Consequently, it becomes a tool for both incorporating and excluding colleagues. ‘The Commission is riven with internal divisions and inclined to habits which prevent the free flow of information and ideas’ (Stevens and Stevens 2001: 243). In such a system, key personal contacts become vital for building support and alliances within the fixed deadlines (ibid.: 178).

Before continuing with the discussion, it is important to offer a clear definition of the concept of networks. Networks are taken here as the ‘personal contacts within the context of organisational contacts’ (Jönsson and Elgström 2005: 3). As such, they are complementary to formal hierarchical channels that are set by the hierarchical division of labour and formal rules and procedures in an organisation. Whereas hierarchical channels are used for task-oriented formal communication, networks come into play to fill in the gaps or to make up for the inefficiencies of the hierarchical channels. James March and Herbert Simon even argue that ‘formal hierarchical channels tend to become general-purpose channels to be used whenever no special-purpose channel or informal channel exists or is known to the communicator’ (March and Simon 1958: 167–168). Networks involve a relationship of mutual exchange and dependency and usually remain highly informal and to a degree invisible (Morgan 1986: 174). Networking may occur over the telephone, through old-boy networks and other friendship groups, through informal meetings, or through chance contacts (ibid.). ‘The informality of networks rests on the personal relationships that develop as a result of frequent interaction’ (Jönsson and Elgström 2005: 3).

Networks are central to understanding the way the Commission works in practice (Shore 2000: 200). Liesbet Hooghe (1999: 405) refers to some nationalities in the Commission as having a strong reputation of clubness, which she defines as ‘a set of formal and informal networks within which members tend to act in concert’. Hooghe argues further that officials with weak national networks are at a disadvantage, as successful policy-making in the EU often depends on the quality of intelligence (ibid.: 415). ‘Access to information can create a form of power parallel to the official hierarchies and sometimes much more efficient than them’ (Abélès et al. 1993: 57–58).

In their study of the communication networks of national officials within European Council working groups, Jan Beyers and Guido Dierickx (1997: 436) found nationality to be a major factor when negotiators have to select partners to forge informal communication links with. Moreover, they tend to view their partners in terms of larger ‘regions’ (ibid.: 465). Their study showed that Northern negotiators communicated more with other Northern Europeans and Southern Europeans were more attached to each other than to Northern Europeans (ibid.: 437). The influence of nationality within the Council setting is perhaps not so puzzling since the officials are expected to represent their nation-states there. It is more intriguing to look into the networks of Commission officials, where nationality is not supposed to play a role due to the Commission's sectoral organisational structure. In turn, one would expect officials to have supranational intra-organisational4 networks, maintaining contacts with all member state officials. Is this the case or do officials rely on their national intra-organisational networks, predominantly contacting officials of their own nationality?

In order to be able to answer this question, the concept ‘network’ needs to be specified further. Daniel Brass (1984: 519) refers to the existence of three types of social networks within organisational structures as the workflow, communication, and friendship networks. I apply this tripartite division to the Commission as follows:

1.	Task-related formal network: formal contacts dictated by the official's task description and obligations.
2.	Task-related informal network: informal contacts used to conduct one's tasks.
3.	Leisure network: contacts during social activities and gatherings which fall outside working hours and obligations.

In the case of the Commission, the fieldwork leads me to propose adding a fourth type:

1.	Career network–contacts maintained for one's own career advancement.

The theoretical expectations and the pilot interviews conducted with Commission officials to fine-tune the questionnaire utilised for this study led to making the task-related informal networks the primary focus for studying how nationality shapes the intra-organisational networks of Commission officials. The reasoning behind this is as follows. First, the task-related formal networks of Commission officials are hierarchically defined. There is no choice element involved here, so the organisational structure and formal responsibilities are the factors that determine these networks. Accordingly, nationality is by default not an explanatory factor here since these contact persons are pre-defined. Second, there is no direct link between the pure leisure and career networks and policy-making processes within the Commission. These networks are a case apart. This does not exclude the possibility of using these contacts for work purposes as well. Whether these contacts are activated for substantive work-related issues, however, depends on the relevance of these contact persons to the official's work. The task-related informal networks are where the formal and the informal merge. Moreover, the information and advice obtained through these networks have a direct effect on an official's performance and simultaneously on the functioning of the policy-making process in the Commission. I therefore argue that these are the most crucial networks for testing the effect of nationality.

I refer here to information that is not directly accessible through the formal channels, such as the intranet and official mailing lists. This is the kind of information one can usually only obtain through personal contacts, with the aim of receiving complementary information such as background information on an issue, analysis/interpretation of a problem, and advice on how to proceed with a given situation. When the issue at hand involves another DG or unit, this information might also merely be figuring out what is going on in the other DG since this information is not readily available until a draft proposal reaches a certain stage of maturity. These are mainly the types of information and advice which enable an official to build on what is formally available.

On the whole, informal practices require a measure of trust and a faith in commonality (Middlemas 1995: 680). ‘Similarity breeds connection’ (McPherson et al. 2001: 415). This principle, formulated by Aristotle simply as ‘[people] love those who are like themselves’ (quoted in ibid.: 416), is called ‘homophily’ in social network research. Homophily is a simple principle which asserts that contact between similar people occurs at a higher rate than among dissimilar people (ibid.: 416). This principle rests on experimental literature in social psychology which establishes that similarities in attitudes, beliefs, and values lead to attraction and interaction (ibid.: 428). Interacting with similar others is efficient as it facilitates transmission of tacit knowledge, simplifies coordination and avoids potential conflicts (Borgatti and Foster 2003: 999). Crude as it may be, network research has shown that the homophily principle structures network ties of every type (McPherson et al. 2001: 415).

An interesting result of homophily research is that the strongest effect of homophily occurs with regard to race and ethnicity on a wide range of relationships from the most intimate bonds of marriage to work relations, networks of discussion about a particular topic, and ‘knowing about’ someone else (ibid.: 420). Furthermore, many facets of ethnicity (e.g., mother tongue, national origins, ethnic group, and region of birth) also display this characteristic (ibid.). Research on multi-national companies is a case in point: ‘People in different countries preferred to interact with others of the same nationality’ (Cross and Parker 2004: 16). Moreover, nationality is an easy characteristic to assess in comparison to personality and abilities (Pratt 2001: 25).

In this sense, the homophily theory forms a basic rationale to test the argument that Commission officials will generally try to deal, at least in the first instance, with people from the same country or from the same general geographic area (Page 1997: 136; Stevens and Stevens 2001: 180). The first step in finding out whether Commission officials rely on similar others for information or advice is to answer the question, ‘Who do Commission officials contact?’

	Hypothesis 1: Officials rely relatively more on officials with the same nationality for information or advice.
	Hypothesis 2: a. North European officials have relatively more contacts from the North and South European officials have more contacts from the South. b. South European officials have relatively more same region contacts than North European officials. c. South European officials have relatively more same nationality contacts than North European officials.
	Hypothesis 3: Officials have more contacts who speak their native language for obtaining information or advice.

To test these hypotheses, the contacts with similar others in terms of nationality-related variables have been counted to see what their sheer numbers reveal in terms of the networking patterns of Commission officials (see the Appendix for details of research design and methodology). This analysis is followed by an explanation of these results.

National networks in a multinational organisation?

When openly asked whether nationality or culture has an effect on the networks of Commission officials, 57.3 per cent of the interviewed officials said nationality mattered and 61 per cent of them said culture mattered. Commission officials acknowledge that it is easier to establish contacts with officials from their own nationality (Officials #2, #22). The contacts with compatriots are more spontaneous (Official #76) and smooth (Official #82). This is mainly due to the simple fact that they ‘speak to each other in their own language and share the same cultural references' (Official #69). The communication is therefore faster and easier since they understand each other more quickly (Officials #75, #76, #119). You know how to approach someone from your own country (Official #87) and you are more open to helping a compatriot when approached (Official #17). You also have higher expectations of obtaining an answer or a favour when approaching a compatriot (Official #22). These aspects of same nationality contacts do indeed point to the advantages of speaking with someone of your own kind. Consequently, Commission officials might well have been homophilic if cultural commonalities had determined their communication patterns.

The quantitative network evidence, however, strongly refutes this argument: Only 17.8 per cent of the contact persons are of the same nationality. The average of same nationality contacts is 0.53 out of the maximum possibility of three. At the opposite end, 49 of the 81 officials have no same nationality contacts, i.e. 60.5 per cent of the officials have a purely multinational network which does not include any official of their own nationality. Officials with none or only one same nationality contact add up to 87.7 per cent of the sample. These indicators strongly suggest that task-related informal contacts are overwhelmingly supranational in the Commission and that national networks are a myth when it comes to asking for information and advice on task-related issues in the Commission.

Still, it is often argued that some nationalities have a stronger reputation of national clubness (Hooghe 1999, 2001) or, in the officials' words, a much higher tendency to stick to each other even within this multinational environment (Officials #3, #7, #17, #26, #29, #69, #72, #116, #117, #120). The next question that has to be answered is: Are there really differences5 among nationalities when it comes to contacting one's own nationals? Table 1 presents the network patterns per nationality. The results should be looked at with caution, however, due to the small size of officials per nationality in the sample.

TABLE 1 AGGREGATED DISTRIBUTION OF SAME NATIONALITY CONTACTS PER NATIONALITY

Nationality	Ratio contacts with same nationality: total contacts	Percentage of same nationality contacts
Italy	7/20	35
France	10/30	33.3
United Kingdom	3/9	33.3
Belgium	6/20	30
Germany	10/39	25.6
Greece	4/21	19
Czech Republic	1/6	16.7
Ireland	1/9	11.1
Spain	1/15	6.7
Finland	0/15	0
Austria	0/12	0
Denmark	0/12	0
Sweden	0/9	0
Netherlands	0/6	0
Portugal	0/3	0
Other	0/15	0
Total	43/241	17.8

Table 1 indicates that contacts between compatriots, if any, are a large member state phenomenon, with officials from Italy, France, United Kingdom, Belgium,6 and Germany having the most same nationality contacts. Officials see this, however, as a mere effect and automatic consequence of size of these member state contingents (Officials #30, #31, #76). This observation is also reflected in the quantitative analysis. While the mode is ‘0’ for both small and large member states,7 the median is ‘0’ for small member states and ‘1’ for large member states. Even though large member state officials have relatively more same nationality contacts, their networks are still predominantly supranational.

The accounts of officials are supportive of these results: ‘There is not more contact with one nationality more than others … Normally, the most part of colleagues who are here don't think in national terms’ (Official #10). ‘Nationality is not a dominating factor in networks. Work-related networks are multinational’ (Official #52). Nationality does not matter for work (Official #117). A lot of it is perception (Official #105).

Yet five of the six nationalities which have no same nationality contacts are North European. Is the difference between the individualistic Northerners and the collectivistic Southerners reflected in the data? Do the Southern officials stick more than the Northern officials to their own region and nationality? Or are both groups homophilic (contacting more people from their own region) such as the Council of Ministers working group participants of Beyers and Dierickx (1997, 1998)?

The North–South division: Another myth?

The successive enlargements of the EU have resulted in an increasing number of different nationalities. Arguably, as nationalities increased and spread throughout the Commission, regional identities have become more relevant reference points for officials. In terms of cultural differences, the reference point ‘North–South’ which had already been present at the time of the study of Abélès et al. (1993) became even more established with the Northern enlargement in 1995.

The relevance of the regional affiliation categories of North and South was demonstrated in the interviews. Namely, 52.4 per cent of the officials referred to the existence of a North–South division, the so-called ‘wine-belt vs. beer-belt division’ (Official #50), especially when asked whether culture has any influence on the shape of networks within the Commission. Officials from the North and the South say they understand and communicate better with people from their own cultural region because they have the same mentality (Officials #9, #75, #78, #108). Northerners acknowledge that they would call other Northerners if they wanted to get to the heart of the matter and receive a direct answer to the question they had in mind (Officials #9, #13, #42, #114). Southerners, on the other hand, feel that they can count on Northerners, obtain direct and reliable information from them (Official #53, #76) and they appreciate the fact that Northerners are easy-going, open, and not hierarchical (Officials #53, #55, #59, #95).

The perception of cultural differences with regard to the networking behaviour of North and South European officials is clearly present. Northerners have the impression that the Southerners are better at networking than Northerners due to the negative connation of networking/lobbying in the North, but that they are catching up (Officials #38, #67, #111, #113, #114). In contrast, Southerners have the impression that the concept of networking is itself Northern, but that Southerners, if by chance they find themselves in a network, maintain it better (Officials #22, #39, #58). The differences are also related to attitudes, i.e. introverted Northerners vs. extraverted Southerners (Official #35), and to preferences with regard to working methods, i.e. Northerners opting for autonomous work, written communication (e-mail) vs. Southerners' willingness to set up meetings and coffee breaks to enable face-to-face talking (Officials #13, #55). Given the ease of contact due to these cultural common denominators (Official #52), some Northerners or Southerners admit that they sub-consciously have respectively more Northerners or Southerners in their network as a result (Officials #12, #47, #52, #92, #106). Some even go so far as to argue that the Northern and Southern networks tend to be separate (Officials #75, #92). Are these perceptions reflected in the task-related informal networks of Commission officials or is this just another myth? Table 2 displays the distribution of contacts within and across Northern and Southern officials.

TABLE 2 DISTRIBUTION OF REGIONAL CONTACTS PER REGION (NORTH–SOUTH)¹

Region	North	South	TOTAL
Same region contacts	53.2%	48.8%	118
Different region contacts	46.8%	51.2%	114
Total	111	121	232
^1The East European officials (only two of them in the sample) and the officials with double nationality (a combination of North-South European) were excluded from the calculations.

Comparing North and South Europeans, the difference8 is very small between the two groups in terms of contacting other Northerners and Southerners. The results are contradictory to all expectations: There is neither an overall regional homophily (North with the North, South with the South), nor is homophily a tendency of the ‘collectivistic’ Southerners. The contacts are almost evenly distributed between same and different region contacts, and if there is a group that prefers people of their own region to those of the others, it is the Northerners. The difference between the proportion of same region contacts of Northerners and Southerners, though, is a mere 4.4 per cent.

In terms of contacting officials of the same nationality, however, the patterns of Northerners and Southerners are slightly different, as Table 3 shows. This time it is the Southern officials who contact their own nationality more than the Northerners. The Southerners have a relatively stronger tendency to rely on compatriots: Southerners have twice the number of same nationality contacts as Northerners (28 vs. 14). In sum, the results of the bivariate analysis of the difference between North and South are mixed and call for a recheck in a multivariate analysis.

TABLE 3 DISTRIBUTION OF SAME NATIONALITY CONTACTS PER REGION (NORTH–SOUTH)

Region	North	South	TOTAL
Same nationality contacts	12.6%	23.1%	42
Different nationality contacts	87.4%	76.9%	190
Total	111	121	232

Some officials also tend to dismiss this North–South division, calling it a prejudice (Official #82) that is often exaggerated (Official #73). It is a misconception, and the differences are rather in modality (the way Northerners and Southerners are perceived to network) and not in approaches (Official #105). As another official emphasises, ‘Of course, people may be different if they come from the North or from the South …, but I don't think that influences our work. No, I don't see anything of that’ (Official #71).

The language issue in a multilingual environment

When asked whether nationality matters in terms of shaping networks, 46.3 per cent of the officials claimed that it was language that mattered more. As the explanations of officials in the previous two sections also suggest, speaking the same native language is seen as an important door-opener (Officials #9, #80). After all, one can only express and understand the nuances in one's native language (Officials #80, #81). This ease in communication also creates trust (Official #69), which at times becomes vital: ‘Sometimes if you need sensitive background information, then it's easier if you speak the same language’ (Official #81).

It comes as no surprise in a union with 23 official languages that officials tell anecdotes on misunderstandings due to language (Officials #46, #90, #95, #97) and emphasise the extent to which it is important to make yourself understood in such a multilingual environment (Official #97). Yet, at the same time, Commission officials are champions of speaking foreign languages. The interviewed officials speak on average 3.63 languages, the maximum being six.

The real question is to what extent the multiplicity of languages comes to life in the daily work of Commission officials. The Commission has three official working languages: English, French, and German. In practice, however, English and French are the languages used on a daily basis. The dominant language, however, has clearly shifted from French to English, a trend which has become even stronger with the Eastern enlargement of the EU. Most written documents are also increasingly produced first in English (Official #2). The diversity of languages is thus not reflected in daily practice. As one official explained, as the EU has become larger and more official languages have been added, the number of languages actually being used has decreased (Official #82).

The predominance of English and the decline in the use of French is also echoed in the empirical data. Respondents were asked to indicate in which language they communicate with the three officials they chose as their most important contacts. The aggregated responses showed that 52.3 per cent of the exchanges were in English, 25.7 per cent in French, 6.6 per cent in German, and 5.4 per cent in a mixture of English and French.9 Only in 32.8 per cent of the cases do officials speak in their native language with contacts in their network.

Another issue that has been identified by the officials is whether one has English or French speaking networks (Officials #29, #73, #74, #90, #113, #117). However, this is not a significant divide either since Commission officials are very much accustomed to constantly shifting from English to French in their daily work (Official #92). More importantly, as one official also explained, when people are working on the same specific issue, there is no language barrier (Official #48).

Explaining networks in the commission

The foregoing empirical evidence demonstrates that the homophily variables of nationality, region, and language do not shape the task-related informal networks of Commission officials. How can this be explained? Plausibly, the effect of homophily may be weakened or annulled by other factors. Being forced to interact with people different from oneself may be one such factor (Cross and Parker 2004: 83–84). In the specific case of the Commission, officials might contact officials of different nationalities due to their socialisation (prior to or during their time in the Commission), or they might be constrained in their willingness to contact their own nationals by organisational structures that counteract territorial clustering.

These socialisation and organisational variables that come to the fore in the EU literature have been tested in the following model to explain the choice of contact persons:

•	Control variable: Size of member state to account for the lower/higher probability of contacts from small/large-member states due to the number of officials they have in the Commission.
•	Independent variables: Testing for the homophily, socialisation, and organisational variables. North–South dummy–Retesting Hypothesis 2c in a multivariate equation. Experience in national administration–Testing prior national socialisation: officials who have previously worked at their home administration (who are socialised into defending national interests) are more likely to contact compatriots than officials who have no previous experience in national administration. Number of languages spoken–Testing prior transnational socialisation: the more languages an official speaks (the more affiliated an official is with other cultures), the less likely the official is to have same nationality contacts. Tenure–Testing Commission socialisation: the longer officials work for the Commission (the more an official is used to working in a multinational environment), the less likely they will have same nationality contacts. Number of contacts within the DG–Testing organisational structure: contacts within a DG are shaped primarily by functional requirements and are thus less likely to be shaped by nationality.
•	Dependent variable: Explaining the occurrence of same nationality contacts.

The dependent variable ‘same nationality contacts’ was dichotomised into the values 0 and 110 in order to run logistic regression. The choice for logistic regression results from the distribution of the responses: there is no normal distribution, and as mentioned above, 49 of the officials have zero same nationality contacts. The combination of these factors makes a linear regression analysis an unsuitable choice. Logistic regression, however, accounts for a high number of zero responses when the dependent variable is dichotomous with the values 0 and 1. As such, logistic regression predicts the likelihood of the occurrence or non-occurrence of events in terms of odds (Pampel 2000: 11): ‘Odds express the likelihood of an occurrence relative to the likelihood of a non-occurrence.’

The dependent variable ‘number of same nationality contacts’ has not been measured as a dichotomous variable, which means that dichotomising it leads to some loss of information. However, there were 9 and 1 respondents with 2 and 3 same nationality contacts, respectively, which meant that there was already not enough variance to be detected for a variable with four categories. Even though individuals with 1 and 3 nationality contacts may differ, this trade-off had to be made in order to be able to conduct a methodologically sound test. The SPSS output of the logistic regression including these variables is presented in Table 4.11

TABLE 4 EXPLAINING NETWORKS WITH LOGISTIC REGRESSION

Independent variables^1	B	S.E.	Sig.	Exp(B)
Size of member state**	2.205	.741	.003	9.069
North–South dummy	.366	.613	.550	1.442
Tenure	.031	.040	.439	1.031
Experience in national administration	.343	.609	.574	1.409
Number of languages spoken	−.473	.321	.140	.623
Number of contacts within DG*	−.677	.307	.027	.508
Constant	.581	1.717	.735	1.788
Constant	.581	1.717	.735	1.788
**p<0.01, *p<0.05N=75	Overall percentage correct: 73.3%Nagelkerke R Square: .389
^1The variables have been coded as follows:
- Size of member state: Small member state=0 (Austria, Denmark, Finland, Greece, Ireland, Netherlands, Portugal, Sweden), Large member state=1 (Belgium, France, Germany, Italy, Spain, UK);
- North–South dummy: North=0 (Austria, Denmark, Finland, Germany, Ireland, Netherlands, Sweden, UK), South=1 (Belgium, France, Greece, Italy, Portugal, Spain);
- Tenure: Number of years working for the Commission;
- Experience in national administration: No=1, Yes=1;
- Number of languages spoken: minimum 2, maximum 6;
- Number of contacts within DG: minimum 0, maximum 3;

The model performs quite well with correctly predicted cells at 73.3 per cent and with a Nagelkerke R² of .389.12 Only two variables pass the significance test, namely those of size of member state and the number of same DG contacts. In logistic regression, an odds ratio, i.e. exponentiated coefficient–Exp(B), higher than 1 increases the odds of an event occurring and a coefficient smaller than 1 decreases the odds (Pampel 2000: 22). The control variable size of member state has the highest coefficient, and it is positive. This means that an official belonging to a large nationality group is far more likely to contact an official of the same nationality group. The coefficient of DG is relatively smaller, but it is smaller than 1 (negative) and significant: an official who has more contacts within a DG is less likely to turn to compatriots. This implies that contacts within a DG are less influenced by nationality compared to contacts outside the DG, which are more influenced by nationality.

These results find support in the accounts of the interviewed officials. Size of a national group is indeed an important factor that determines whether an official can build a network with compatriots at all. The underlying reason for this is that Commission officials are specialised in specific files which in some cases require very specific technical or scientific expertise. Especially when you are an official from a small member state, the chances are low that there is someone else from your nationality who is working in the same field of expertise (Officials #25, #100, #110, #116). Therefore, in practice nationality has limited effect because your nationality is not represented everywhere (Official #119). Furthermore, it would not be sufficient to have a solely national network as a small member state official (Official #67). In this sense, nationality potentially matters more for large member state officials. It is not as if there is a choice between various persons in your field who all have the answer to a particular question. If that were the case, some officials admitted, they might choose the person from their own nationality over someone who is not (Official #95). But the chances of this hypothetical case are very low, and the exceptions are large member states which have their officials in almost every unit (Official #106).

The fact that the DG is significant for the task-related informal networks of Commission officials is a reflection of the centrality of one's field of expertise. When it comes to performing one's tasks, the official's specific file (dossier) is the most important consideration. As one official put it, ‘[P]eople have their area of work that's specified, so you really have to talk to them. Sometimes you can talk to one of their colleagues, but normally in the Commission people have quite specific responsibilities’ (Official #13). Within this organisational context, expertise determines a contact person's competence and utility. As a result, the policy issue one is working on ‘creates things that are in common’ and ‘transcends the differences that might exist because of the main cultural differences’ (Official #24). In the words of a French official:

I don't see nationality playing a role because what comes first is the function, the technical aspect. I am not going to ask a French colleague, simply because he is French, for information on a file managed by a Swedish colleague who knows the subject … It is the technical knowledge, the function, the fact of being responsible for a subject that comes first. (Official #91) [Author's translation]

The identification with one's file (dossier) and expertise is so strong that there is little room left for general questions (Abélès and Bellier 1996: 437). The only kind of useful general information one can get from a fellow national tends to be on ‘how things work in another DG’ (Officials #81, #113). Seeking out a compatriot also comes in handy for finding out who is responsible (Officials #80, #81): ‘It's just a point of first contact, but they can tell you to whom you have to turn. And then nationality doesn't play a role anymore’ (Official #81). Especially in such circumstances, it is easier to contact people in other DGs in your own language (Official #3). However, there are also officials who do not appreciate this ‘door-opener’ function since it is not related to their expertise: ‘I don't like someone who's phoning, “I was just checking someone who's Spanish on the list and wanted to ask you.” To be Spanish, this is not a professional mark’ (Official #46).

The North–South division had given mixed results in the bivariate analysis, but in the multivariate analysis this variable fails to pass the significance test. The socialisation variables also fail, which is in line with previous quantitative research on socialisation in the Commission (Hooghe 1999, 2001, 2005). Amongst the socialisation variables, however, the prior transnational socialisation variable ‘number of spoken languages’ performs the best.

Discussion and conclusion

The foregoing analysis of the empirical data on the networks of Commission officials demonstrates that national networks are indeed a myth when it comes to the substance of their work. This is not only supported by the sheer numbers which demonstrate that same nationality contacts are the exception rather than the rule, but also through the accounts of Commission officials who stress that ‘nationality is not a category that counts’ in the Commission (Official #25) and that there is actually no need for advanced national networks (Official #31). Clearly, officials do not contact fellow nationals just because they come from the same country. By virtue of their supranational networks, they fulfil their role of independent international civil servants as designated in the treaties.

The analysis has also shown once more the significance of working with clearer concepts and the relevance of borrowing definitions and insights from other bodies of literature, in this case from organisational and network theory for investigating ‘networks’. In particular, the deconstruction of the different types of networks has proven essential for analysing networks in the Commission. This study was only the first step in mapping out and explaining the networking behaviour of Commission officials in a systematic manner. The career and leisure networks still require a systematic study.

These results are in line with the results of a Commission survey in 1974, cited in Hans Michelmann's article (1978). Michelmann reports that the survey found no statistically significant relationship between nationality and interaction (ibid.: 492), and furthermore that the quality of interaction was also independent of nationality (ibid.: 493). This led Michelmann to conclude that ‘under normal circumstances officials react to fellow civil servants as individuals and not as members of national contingents’ (ibid.). These conclusions hold for Commission officials interviewed three decades later.

The results also provide support for the organisational/institutional perspective on the study of EU institutions on a more generic level (Egeberg 2001, 2004, 2006a, 2006b; Trondal 2006a, 2006b). The extent to which the networking patterns of Commission officials are shaped (or constrained) by the organisational structure of the Commission demonstrated that the organisational affiliation of officials influences their behaviour. The contrast between the networking behaviour of Council working group participants and Commission officials is a case in point: the Council fora remain ‘after all a negotiation process among nations’ (Beyers and Dierickx 1998: 313), whereas the Commission is where the European/supranational element visibly comes to life.

There are also interesting parallels to be drawn between the results presented and previous research conducted on the College of Commissioners and their Cabinets. Egeberg's research (2006a) had shown that the portfolio role was the most important factor in shaping decision-making behaviour in the College. As for the effect of size of member states, Joana and Smith (2004: 39) have previously argued that ‘Commissioners from large countries are most often in a position of comparative advantage because they and their cabinets are able to call upon networks of national actors when preparing arguments and objections on non-portfolio issues’ (emphasis added).

As for the Commissioners and their Cabinets, the portfolio/dossier is the most important factor shaping the daily work of Commission officials. The extra information and advice they need to obtain sometimes falls on the periphery or outside this portfolio, in which case they might turn to a compatriot. In such cases, coming from a large member state might become advantageous since there are by default more contact points for these officials to turn to.

The lessons to be drawn with regard to the effect of socialisation are more ambiguous. Even though the results confirm those of Hooghe (1999, 2001, 2005) in terms of dismissing the role of socialisation, the accounts of the Commission officials interviewed for this study, as well as earlier research focusing on the identities of EU officials (Risse 2004), point in a different direction: Commission officials have internalised the values of the institution, which is reflected in their daily behaviour at work. The failure of socialisation variables in quantitative analyses might therefore be related to the fact that the indicators used for tapping into socialisation are not powerful enough to capture this concept. In fact, the Commission attracts individuals with a cosmopolitan background, as the number of languages they speak also demonstrates. In this sense, I argue that self-selection plays a larger role than previously argued (Hooghe 2005). The Commission now recruits younger officials who have grown up in a globalised world, whereas Hooghe's research only focused on the earlier generation of senior Commission officials.

This study has shown the extent to which the Commission's organisational structure shapes the networks of officials. Even though the empirical data covered four social regulation DGs, the organisational structure is valid for the whole organisation. In sum, national networks in the Commission are a myth when it comes to officials performing their tasks. The ideal of civil servants ‘whose nationality [is] supranationality’ (quoted in Spence and Stevens 2006: 151) seems to be manifested in the everyday reality of the Commission.

Acknowledgements

I would like to thank Mark Bovens, Deirdre Curtin, Morten Egeberg, Martijn Groenleer, Frank Häge, Liesbet Hooghe, Michael Kaeding, Sebastian Princen, the participants of the CONNEX Research Group 1 Workshop in Vienna (May 2006) and the ECPR Conference in Istanbul (September 2006), and the anonymous reviewer for their comments and suggestions on earlier versions of this article.

Notes

1. This dilemma has previously been termed ‘consociationalism vs. Weberianism’ by Liesbet Hooghe (1999) and ‘territorialisation vs. autonomisation’ by Morten Egeberg (2006b).

2. Territorial subdivisions do, however, exist in some units of a few DGs.

3. It remains to be seen how the Eastern enlargement has added an ‘East’ category to this classification (Cyprus and Malta being the two exceptions which plausibly belong to the Southern group) or whether the new countries will be included and/or integrated in the Northern or Southern groups in the long term.

4. The networks of Commission officials with external actors on the EU or member state level fall outside the scope of this study.

5. Unfortunately, the differences cannot be tested for statistical significance since the data violates the basic assumptions of ANOVA due to the skewed distribution of the sample, the unequal distribution of the cells, and the violation of the assumption of independence of observations in the aggregated data (see among others Field 2005: 324).

6. Due to the location of the European Commission in Brussels, Belgium is also a large member state in terms of the number of officials.

7. Belgium, France, Germany, Spain, and United Kingdom were coded as large member states based on the number of their A-level officials in the Commission.

8. Similar to the case of the nationalities, the differences cannot be tested for statistical significance since the data violates the basic assumptions of ANOVA due to the skewed distribution of the sample and the violation of the assumption of independence of observations in the aggregated data (see among others Field 2005: 324).

9. The results should not suffer from a language bias since the respondents were given a choice between English and French for the language of the interview.

10. Responses 1, 2, and 3 have been aggregated and recoded as 1.

11. Collinearity diagnostics have shown that the variables do not suffer from a multicollinearity problem. Residual analyses resulted in only two cases with standardised residual values around 3.

12. It should be noted that the Nagelkerke R² tends to be lower than the corresponding linear regression R². See, for example, http://www2.chass.ncsu.edu/garson/PA765/logistic.htm.

13. The new EU staff regulation adopted in 2004 has changed the names of these officials to Administrator (AD). Yet, since Commission officials themselves still refer to the term ‘A-level’, this term will be employed in the rest of the paper. Note that the sample also includes permanent and temporary officials (seconded national experts and temporary agents) with policy-making functions.

14. I rely on the classification Hooghe (2001) uses to group the DGs of the Commission into six policy areas: Administration, External Affairs, Market-Oriented, Social Regulation, Supply Side and Provision.

15. Two out of the 120 sampled officials were later excluded and not re-contacted after the first letter asking for an interview when it became clear that they were not A-level officials.

16. Social network scholars conducting research on ego-centred networks have asked their respondents to name three respondents (Cross et al. 2001; Cross and Borgatti 2004). This is based on the finding that an average adult has 20 regular interlocutors, only three of which are confidants (Degenne and Forse 1999: 20–21).

Appendix: Research design and methodology

The research questions have been answered by means of interviews with policy-making, that is, Administrator level (A-level)13 European Commission officials. The interviews contained both structured and open questions derived from the literature and exploratory pilot interviews. To get the most interesting results with this number of informants, the study was limited to one policy area. Four DGs belonging to the ‘Social Regulation’14 family have been included in the sample: Employment, Social Affairs and Equal Opportunities; Environment; Health and Consumer Protection; Justice, Freedom and Security. Methodologically, obtaining a high response rate is crucial, and reducing the respondents' burden is one of the main factors that influence the response rate (Lohr 1999: 261). At the time of sampling, this group of DGs employed a total of 1100 A-level officials, more or less equally distributed across the DGs. This was an important factor in terms of distributing the respondent burden across DGs. Reducing the respondents' burden is one of the crucial factors that influences the response rate (Lohr 1999: 261).

Since the DG attachment is the central variable for testing the effect of organisational structure, it is vital to ensure an adequate representation of all the DGs to allow comparison between the DGs. To obtain a real ‘miniature of the population’ that reflects the population with respect to the size of the DGs, the sampling technique that has been used is a type of stratified random sampling called proportional allocation (Lohr 1999: 104). Proportionate stratified sampling typically results in more precise survey estimates by reducing sampling error (Edwards et al. 1997: 58–59).

The online Commission directory was utilised to compose the list of the population of officials from which a random sample of 120 officials has been drawn. A proportionally equal number of officials were thus randomly selected within each DG. The interviews were conducted within the span of a total of five weeks, during the two periods between 5–22 April 2005 and 23 May–3 June 2005. The resulting overall response rate (out of 118 officials15) was 69 per cent, i.e. 82 officials, which is considered to be a very good response rate (Babbie 1992: 267). Item non-response rate was very low: Only one of the 82 officials only responded to half of the survey questions and was excluded from most of the analyses.

First, some background information on the officials was gathered to account for potential explanatory factors that determine the shape of an official's network, based on the studies and questionnaires of Beyers and Dierickx (1997, 1998), Hooghe (2001), and Trondal (2001). These elements included multinational family background and experience, command of foreign languages, past work experience (in national or international organisations) and length of service.

In social network analysis, a network is a set of actors connected by a set of ties. A single focal actor is called an ‘ego’ and the set of actors that ego has ties with are called ‘alters’. The ensemble of ego, his alters and all ties among these (including those to ego) is called an ego-network (Borgatti and Foster 2003: 992). Since ego-networks can be collected for unrelated egos, ego-network studies combine a network-theoretical perspective with the conventional, individual-oriented methods of collecting and processing data (ibid.).

To obtain the network of each official, the respondents were asked to reflect upon the officials they regularly turn to for information or advice, specifying that this does not (necessarily) involve the officials they have to contact due to their task description and obligations. The respondents were then asked to select the three officials 16 they considered to be the most important for conducting their policy-making work. The interviews focused subsequently on the attributes of and relationships with these three contact persons.

The structured interview questions were followed up by open questions concerning networks in the Commission, such as how Commission officials build their networks and more importantly to what extent, how, and why their networks affect policy-making processes. This qualitative data has been recorded, and the recorded material and the notes of the interviews have been transcribed. The results of the quantitative part of the survey were analysed using the statistical software package SPSS 12.0.1.