Perspectives on EU governance: an empirical assessment of the political attitudes of EU agency professionals

European Union (EU)-level agencies have emerged as important actors on the EU's policy-making scene. To date, we know relatively little about the personnel working in EU agencies: what attitudes do EU agency staff members hold on issue-dimensions relevant for the EU integration process? How do they perceive of their role in EU policy-making? Moreover, we know little about the cohesiveness of attitudes of agency staff within and between different EU agencies. The aim of this contribution is conceptually and empirically descriptive. It draws on original data from an online survey of professionals working in EU agencies to gain insights into the attitudes held by EU agency staff on three substantive attitudinal dimensions: conceptions relating to legitimate and accountable EU governance, conceptions about the preferred level of centralization of political authority in the EU, as well as views on economic governance in the EU. While the conceptual focus of this paper is on attitudes and not on behaviour, the attitudes held by EU agency staff and their relative homo- or heterogeneity is likely to affect perceptions and evaluations of the political environment and interpretations of the challenges agency staff members face in their substantive area of work. The findings of the survey will enable us to draw broader conclusions about the type and quality of accountability relationships as well as of the EU's democratic legitimacy. Moreover, the data will permit to inform arguments about the actor quality of EU agencies, which are often conceived as efficient institutional solutions to overcome credibility problems.


INTRODUCTION 1
In an article published in 1973, Robert Putnam posed a rhetorical question when asking whether there could be much doubt as to 'who governs our complex modern societies' (Putnam 1973: 17). His argument that nonelected bureaucratic élites and their monopoly of policy-relevant information and expertise had gained a position of dominance in the policy-making process in modern societies is all but faintly echoed by contemporary scholarship which sees 'the rise of the unelected' (Vibert 2007) as a defining attribute of our times. There is broad agreement among scholars that 'non-majoritarian institutions' (Gilardi 2008;Majone 2001;Thatcher and Stone Sweet 2002), such as regulatory agencies, central banks or constitutional courts, have been empowered to deliver better informed, more credible or even 'fairer' policy solutions to various public policy problems. Over the past decades, non-majoritarian institutions have thus come to play an ever more important role in affecting the political agendas of modern government, both at the domestic as well as on the international level.
In this contribution, we take a closer look at European Union (EU)-level agencies, which have emerged as important actors on the EU's policy-making scene. Given the variable but undisputed impact of EU agencies on EU policy-making, we know surprisingly little about the personnel working in EU agencies. Who are the experts working in EU agencies, called upon to deliver seemingly 'technical' opinions, which may nevertheless carry profound political implications? How do they perceive of their role in the world of politics? Do professionals working in EU agencies see themselves primarily as 'technocrats' answerable merely to the scientific standards defined by their peers? Or do they consider their profession as one, which is inherently 'political'? Questions like these have occupied political scientists and élite theorists for a long time: 'Some have hailed technocracy as the wise and disinterested role of philosopher-kings, whereas others fulminated against technocrats as despots of a new and peculiarly inhuman sort' (Putnam 1977: 383; see also Centeno 1993).
The aim of this contribution is conceptually and empirically descriptive. It draws on original data from an online survey of professionals working in EU agencies to gain insights into the attitudes held by EU agency staff on three substantive attitudinal dimensions: conceptions relating to legitimate and accountable EU governance, conceptions about the preferred level of centralization of political authority in the EU, as well as views on economic governance in the EU. We thus approach the study of EU agencies by turning our attention to the professionals working in EU agencies, and explore the content and relative cohesiveness of their attitudes on the above-mentioned issues of European integration. The attitudes of agency professionals can be thought to have an effect on the behaviour of individual agency staff members or agencies as collective actors (Aberbach et al. 1981: 30 -3;Hooghe 2001: 11;Olsen 2008: 199), especially in situations where no standard routines apply, for instance, in situations of high or unprecedented technical complexity and, more generally, in situations of uncertainty (Denzau and North 2000: 33 -7;North 1990: 23). In contrast, in situations where professional assessments are uncontested, where administrative rules offer clear behavioural guidelines in specific situations and where administrative and political leadership formulates a clear mandate and engages in active supervision, the attitudes of agency professionals should carry less weight in the activities of EU agencies. This paper, however, does not investigate the impact of attitudes on agency professionals' behaviour in policy-making. This paper is organized as follows: in Section 2, we introduce three substantive dimensions of ideology in more detail: conceptions relating to legitimate and accountable EU governance; conceptions about the preferred level of centralization of political authority in the EU; views on economic governance in the EU. We follow Hooghe (2001) in deriving different analytically relevant categories to study the attitudes of agency staff on each dimension. Section 3 provides an overview of the online survey and information on our sample. In Section 4, we present our empirical results by assessing the substantive content and the relative cohesion of the attitudes of agency personnel along the different political authority-relevant dimensions. We conclude by discussing the implications of our findings for the study and analysis of 'executive governance' in the EU and other international organizations.

ASSESSING THE CONTENT AND COHESION OF IDEOLOGIES OF EU AGENCIES' STAFF
Our study will focus conceptually and empirically on different conflict dimensions that are of crucial importance in structuring political contestation in EU politics, namely on how competencies and authority should be distributed between the member state and the EU level and the extent to which political authority should intervene in the economy (Marks and Steenbergen 2004). Moreover, we will investigate the sources underpinning legitimacy and accountability of EU agency governance by exploring the attitudes agencies' staff holds when it comes to legitimizing their work and to holding EU agencies accountable. We will treat each of these dimensions in turn.

Legitimacy and accountability
Since agencies form part and parcel of the regulatory policy-making process in the EU by providing policy-relevant information and expert opinions, agencies can be said to exercise political authority, which they share with a multitude of other actors involved in the policy process. We distinguish between two basic principles to legitimize political rule in modern democracies: the democratic and the technocratic principle (see Centeno 1993;Hooghe 2001). The democratic principle is reflected by the chain of representation, which places a premium on representative institutions (parliaments) and citizens' involvement via regular and competitive elections. While representative democracy marks one way to define the democratic principle, democratic theory provides alternative conceptualizations of the democratic principle. Participatory democracy emphasizes that the democratic process ought to offer venues to provide access for all interests potentially affected by particular decisions, i.e. citizens as well as organized groups. The rapid process of 'agencification' in the EU has provided fuel to the debate about the role of expertise and a technocratic logic to legitimize public policy more generally and EU policy-making in particular (see Boswell 2008;Majone 2005;Radaelli 1999aRadaelli , 1999bSchrefler 2010;Shapiro 1997;Sosay 2006;Vibert 2007). The technocratic principle rests on the assumption that decision-making should ensue from the 'use of value-free, objective criteria' (Centeno 1993: 311) and promises to deliver socially and politically neutral decisions generating 'Pareto-optimal' solutions. Unlike the democratic principle, the technocratic principle is rooted in an 'ideology of method', that is 'a belief in the ability to arrive at the optimal answer to any discussion through the application of particular practices' (Centeno 1993: 312).
Moreover, the belief that in a democratic polity various actors have a legitimate claim to exercise authority depends 'on whether or not the actor is accepted as having appropriate accountability relationships with others' (Black 2008: 150). Below, we will discuss two types of accountability relationships, professional accountability (which is derived from the technocratic legitimation principle) and the democratically inspired concept of social accountability. The thrust of the term accountability posits a social relationship between an actor and a forum, in which the actor 'accounts' for her behaviour by explaining her conduct and providing information. The forum, in turn, renders assessment and judgement of the conduct and decides on whether or not to sanction the accounting actor (Bovens 2007;Curtin and Egeberg 2008). What does this definition imply for the accountability of EU agencies? Some argue that the nature of EU governance in regulatory policy-making, characterized by governance through networks, defies the strict logic embodied in the notion of a chain of delegation from clearly defined principals to agents (Curtin and Egeberg 2008: 654). Sabel and Zeitlin (2010: 12) claim that 'principal-agent accountability gives way to peer review through fora, networked agencies, councils of regulators, and open methods for coordination: the full repertoire of processes by which EU decisionmakers learn from and correct each other even as they set goals and performance standards for the Union'. These peer-review processes are said to contribute to agencies' professional accountability (Bovens 2007). Professional accountability works via explaining and justifying actions to an accountability forum in which actors possess the knowledge and information to evaluate and understand the explanations provided by the agent, which are unlikely to be ordinary citizens but rather professional peers (Bovens 2007: 456;Sabel and Zeitlin 2010: 12) or disciplinary 'epistemic communities' (Haas 1992). From a functionalist perspective on delegation, the focus on professional norms of accountability whereby 'scientists are seen to be totally independent from policy-makers' might contribute to agencies' reputation to efficiently and credibly support EU regulatory policymaking (Majone 2000: 300). If these professional norms of accountability are of relevance for EU agency staff, they should be reflected in their attitudes towards accountability vis-à-vis their professional peers.
Social accountability rests on the assumption that the relevant accountability forum is composed of public and private stakeholders. It is a derivative of calls for more participatory democracy and contradicts the conception of professional accountability in that social accountability suggests that agencies should 'feel obliged to account for their performance to the public at large' or, at least, to affected stakeholders, such as 'civil interest groups, charities and associations of clients' (Bovens 2007: 457). The fine line between social accountability of agency personnel vis-à-vis organized groups or the broader public and the collusion of agency personnel and interest groups in policy subsystems whose members 'share preferences, or at least share understandings concerning the basic dimension of conflict' (Baumgartner and Jones 2009: 19) is empirically difficult to draw.
Agency personnel might value interactions with private groups not because they feel obliged to give account to these, but for instrumental reasons, i.e. to pursue their individual and common interests. The extent to which 'accountability' or 'collusion' dominates the relationship will depend to a large degree on whether preferences in a specific decision are indeed shared and groups therefore do not see the need to raise public pressure on an agency. Importantly, however, it will also depend on the extent to which agency personnel feels obliged to give account of their work also to the public at large instead of just relying on interactions with like-minded actors (Bovens 2007: 457).

Level of centralization
This dimension enquires whether the members of staff working in EU agencies favour a more centralized or decentralized organization of authority in the EU, i.e. whether they favour a more supranational or a more intergovernmental allocation of competencies and authority. Why and how does this matter for how EU agencies fulfil their tasks? If we follow the logic of the 'Europeanization thesis' (Eberlein and Grande 2005: 92), regulatory functions are increasingly delegated to the EU level to reduce transaction costs, since EU agencies provide independent expertise and information, thereby increasing the efficiency and the transparency of EU regulatory policy-making. Moreover, EU agencies are also said to improve the implementation of regulatory decisions given their enmeshment in regulatory networks with national administrations well acquainted with the situation in their respective country (see Groenleer 2009: 100-1). Delegating tasks to EU agencies can thus be regarded as a functional solution to the problems arising from uncoordinated or domestic-level activities alone (Majone 1997(Majone , 2005. If agency personnel share this assessment, we would expect to observe attitudes favouring supranational centralization. An alternative logic underpinning support for supranational centralization is not rooted in functional reasoning but in the bureaucratic self-interest of agency personnel. Kelemen (2002) claims that EU agencies should be considered part of an 'Eurocratic structure' in which the Commission tries to expand its regulatory powers by pushing for the establishment of agencies. As governments are increasingly hesitant to delegate further powers to a Commission, which is perceived to be too 'entrepreneurial', the agency solution seemed to be the best available alternative for the Commission (Kelemen 2002: 95, 98). This argument is in line with a set of institutional theories of bureaucracies, which posit that the designers of bureaucratic organizations are able to structurally determine 'political preferences' of these organizations (Moe 1990). It is also in sync with theories of bureaucracies claiming that bureaucrats have an inherent interest in empowerment and aggrandizement of their organization (Majone 1996;Niskanen 1971). If these theories of bureaucratic empowerment provide a valid interpretation for EU agencies' political motivation, their staff should hold attitudes that are strongly in favour of European integration and centralized, supranational political solutions and in particular of a strong and powerful Commission.
In contrast, the 'nationalization thesis' (Eberlein and Grande 2005: 92-3) posits that the functionally best solution to regulate domestic markets is not found by abdicating sovereignty to the EU level, given the relative economic, administrative and political heterogeneity of the EU-27 and the continuing dominance of member state politics in the regulation of markets. Followers of this thesis will see the regulatory capacity of the EU-level much more critically and may be more inclined to express support for decentralized forms of political regulation. One way to avoid regulatory fragmentation in the common market is to acknowledge the heterogeneous situation in the EU-27 and respond with a regulatory strategy that operates according to the principle of subsidiarity. The EU's regulatory networks comprising national and EU regulatory institutions can be seen as an institutional realization of this principle (see Coen and Thatcher 2008). Given the EU's heterogeneity, decentralized regulatory solutions might be better able to find efficient responses to the heterogeneous institutional and economic situations in member states, enhancing trust among regulators in these networks and, as a consequence, acceptability and effectiveness of EU regulation (Majone 2000: 295-8).

Economic governance
EU agency professionals might also differ attitudinally in the extent to which they endorse an active role of politics in economic governance. Economic governance can be broadly understood as the relationship between employers and employees, the redistribution of material benefits between different strata of society, as well as regulatory activities such as rule or standard setting in matters relating to workers' health and safety, consumer protection or environmental standards, etc. Two positions can be contrasted along this dimension: while economic liberals and conservatives emphasize a minimal role of the state in economic governance, which finds its expression in less regulation and an emphasis on market liberalization, proponents of welfare state interventionism conceive of the state as an active player in economic governance, emphasizing the positive role of state interventionism and regulation to correct for market excesses and failures (Hooghe et al. 2002;Kriesi et al. 2008). In academic and public parlance, we also refer to these two positions as 'right' and 'left'.
As far as legislation in the area of economic governance is concerned, EU agencies have no formal powers of agenda-setting or co-decision. While some are responsible for the administrative implementation of particular programs and EU regulations, others have a more informal role in providing national and EU political actors with policy-relevant expertise. Depending on their ideological convictions of the role of the state in economic governance, they might come up with different solutions to address regulatory policy problems. While free-market 'liberals' might ask for non-binding 'voluntary agreements' and forums in which relevant actors share their experiences, 'interventionists' might prefer and push for 'command and control' instruments, such as binding regulations. If EU agency professionals share a common outlook on questions of economic governance, this is likely to be reflected in a uniform view on the problems they identify, the information they provide to political decision-makers and the possible solutions they suggest. A widely shared particular ideological outlook on economic governance might also lead to greater attention and responsiveness of an agency towards representatives of political parties, interest groups and firms sharing the same ideological outlook. Such an ideological bias would, however, compromise agency independence and its ability to provide untainted, neutral 'expertise'.

THE SURVEY
We conducted an online survey including a total of 720 agency staff members from nine different agencies to assess the content and cohesion of EU agency staff attitudes (see Table 1). 2 To secure broad comparability, we drew on items from Hooghe's élite study of Commission officials (Hooghe 2001), Putnam's work on bureaucratic élites (Putnam 1977) and Eurobarometer items on economic governance and European integration (Brinegar et al. 2004). Where necessary, we slightly adapted them to our object of investigation and theoretical interests. In order to grasp potentially systematic differences in agency staff attitudes, we selected agencies according to the following criteria (see Table 1): first, we selected agencies displaying different levels of formal institutional independence, following the agency independence index developed by Wonka and Rittberger (2010). Second, we made sure that our agency sample contained agencies performing regulatory and non-regulatory tasks, i.e. agencies predominantly occupied with informational or executive functions. These two selection criteria reflect the argument that agencies with different levels of formal institutional independence and different tasks potentially display differences with regard to how agency employees perceive of their role.
Our survey targeted the multinational staff of these agencies working 'on the ground', i.e. those involved in taking decisions, preparing recommendations and providing policy-relevant expertise. We did not survey members of the different agencies' management boards, which are formally responsible to oversee the work of an agency. We have also excluded support staff and clerical staff from our sample. To obtain information on agency staff members, we consulted the home pages of the respective agencies and the staff lists included therein. If information provided on the home page was missing or incomplete, we consulted the agency administration to obtain information on, for example, email addresses of the staff. The survey was carried out in the period between April and June 2010. Among the 720 staff surveyed, we received between 178 and 186 responses for individual attitude items, which bring us to an overall response rate for individual items between 25 and 26 per cent (Tables  2 -4). Response rates vary, however, considerably for individual agencies (see information in Table 1). 3 Given the moderate response rate and the fact that we do not cover all agencies, we cannot generalize our results to the overall universe of EU agencies. The fact that responses of agencies with high and low response rates are very similar, however, indicates that self-selection among respondents did not lead to responses of only a very particular type of agency employee, thus increasing our confidence in the reliability of our data.

ANALYSIS
The goal of our conceptual discussion in Section 3 was to introduce three different dimensions -principles of legitimate governance, political centralization and economic governance -to capture the attitudes of agency professionals on key aspects of political authority. Some of the items relate directly to the experiences of agency professionals in their respective agency (see questions 4, 5, 7 and 8 in Table 2), while other items demand of our respondents to abstract from their immediate working environment in the agency. These items hence intend to capture attitudes on the EU and the nature of democratic politics more generally (see questions 1, 2, 3, 6 and 9 in Table 2). The items employed to tap into different aspects of the ideology of EU agency staff are hence not meant to capture one single underlying concept of legitimacy, centralization or economic governance but are employed as indicators to explore how EU agency professionals think about these different issues at different levels of abstraction. 4 Turning to the data, we find that the attitudes expressed by agency staff on the different items hardly differ between the nine agencies in our sample. 5 Significant differences could only be found for the question on whether agencies should be concerned with the public approval of their work (x 2 (with ties): 33.7, df: 8, p: 0.0001) (question 5, Table 2) and the question whether citizens' influence on political decisions should be expanded (x 2 (with ties): 16.0, df: 8, p: 0.04) (question 6, Table 2). Individuals working in different agencies thus hold fairly similar attitudes on issues of legitimacy, EU centralization and economic governance. The (institutional) working environment thus does not seem to shape the attitudes of agency professionals. In the ensuing discussion, we will therefore only look at the overall distributions and not at individual agencies.
We begin our analysis with the legitimacy and accountability dimension for which we devised nine questions to tap agency staff attitudes (see Table 2). We analytically subdivided these nine questions to capture attitudes on the merits of majoritarian political institutions and politics (questions 1-4), attitudes on the direct involvement of citizens (questions 5 -6) and stakeholders (question 7), as well as attitudes towards professional and expertise-based (as opposed to political) forms of accountability (questions 8 and 9). While questions 4, 5, 7 and 8 demand of the respondents to relate their answer to their work and experience in the agency, the other questions tap their general attitudes towards majoritarian, participatory and expertise-based legitimation principles.
Several findings are eye-catching. For one, as regards the role of 'professionalism' and expertise, we find that more than 86 per cent of our respondents agree or strongly agree with the notion that their own accountability should flow from upholding professional standards, while 61 per cent consider expertise to be more important than political considerations in policy-making more generally (questions 8 and 9, Table 2). This is not overly surprising, given that agency staff members are not political 'generalists' but are trained as specialists in particular professions. Yet, it is not only professional accountability that is broadly supported as accountability standard among our respondents, social accountability has an equally central status in the minds of EU agency staff: not only is direct participation and influence of citizens in policy-making seen positively, agency professionals also tend to place a high value on their work being publicly approved (questions 5 and 6, Table 2). Taken together, these findings indicate that agency staff members consider a purely professional or expertise-based form of accountability to be insufficient to legitimize their work. On the level of individual agency staff members, however, agency professionals do not seem to directly relate public approval of their work and professional accountability (r(t B ): 0.06, p(z): 0.34; questions 5 and 8, Table 2]. 6 Yet, those who approve of professional accountability also tend to be generally in favour of more direct citizen influence in policymaking (r(t B ): 0.14, p(z): 0.04; questions 6 and 8, Table 2). Interestingly, agency staff members draw a distinction between citizen influence and public approval of their work on the one hand, and the role attributed to 'stakeholders', i.e. those organized interests directly affected by an agency's work  on the other. There is strong disagreement among our respondents on the role and the value of information provided by stakeholders. While 35 per cent of respondents think that those directly affected provide the 'best' information, almost the same proportion of respondents disagrees with this statement (see question 7, Table 2). Turning to agency staff members' attitudes on the role of majoritarian politics, the findings are more ambivalent. This holds especially with regard to the role ideology should play in the work of agency staff as well as with regard to the role of political ideology and politics more generally (questions 1 -4, Table 2). While around 30 per cent of the respondents disagree with the statement that political considerations should drive contestation among agency staff, 41 per cent do approve of 'politicization' inside the agency (see Table 2, question 4). Thus, while agency professional have a 'technocratic' inclination as indicated by the broad consensus on 'professionalism', this does not strictly rule out approval for majoritarian politics and 'politicized' decision-making inside the agency. 7 Positive attitudes towards professional and social accountability as well as split attitudes on the desirability of majoritarian politics imply that EU agency staff members entertain a more nuanced set of attitudes as might be expected by those who argue that the modern regulatory state entertains a clear division of labour between 'technocrats' and 'politicians' (see Vibert 2007). While demanding professional standards for the evaluation of their own work, agency staff members seem to approve of 'normal' politics in general and also, to some degree, in their work as experts inside EU agencies. This should open up the potential for conflicts among 'experts' within agencies that may go beyond mere 'scientific' debates on the nature of evidence and the application of appropriate methods. This view is underpinned by the observation shared by a substantial share of our respondents who report that it 'normally takes quite a while' to arrive at a common position and that 'regularly' decisions are taken even consensus could not have been established. 8 What are the implications of these findings for our discussion about the sources of legitimacy of governance by and through agencies? As shown, the 'technocratic' and 'participatory' principles to legitimize policies loom large in the responses of agency staff members to our survey as reflected in the broad support for more citizen influence (participatory principle) and the role accorded to expertise vis-à-vis political considerations (technocratic principle). In turn, the democratic principle is more contested. The views our respondents hold on political representatives and their 'methods' to provide efficient policy solutions is quite divided: while approximately 30 per cent of the respondents display 'tolerance to politics', a slightly larger share of about 37 per cent consider politicians and their methods a discomforting state of affairs (see Table 2, question 3).
To put our findings into a broader perspective and to allow for a comparison between EU agencies and, given its regulatory 'mission' and non-majoritarian quality, another 'relevant' EU organization -the Commission -we adopted and adapted a number of survey items employed by Liesbet Hooghe in her study of high-ranking Commission officials (2001) (questions 2, 3, 4 and 7,  Table 2). 9 It turns out that Commission officials have a more 'technocratic' inclination than professionals working in EU agencies: 54 per cent of the surveyed Commission officials agree that ideological principles never provided solutions to the problems of European citizens (question 2, Table 2) and even 72 per cent, compared to 37 per cent of EU agencies' staff, object to politicians and their methods (question 3, Table 2). Moreover, Commission officials attribute even less value to the information provided by 'stakeholders' than agency officials: 57 per cent of the surveyed Commission officials disagree with the statement that directly affected stakeholders usually provide the best advice on a proposed policy, compared to 32 per cent of the surveyed agency officials (question 7, Table 2). Yet, with respect to their work inside the Commission, Commission officials show greater sympathy for ideological convictions: 59 per cent, compared to 41 per cent of agencies' staff, do think that they should express their ideological convictions in internal Commission discussions, even if they risk conflict with their colleagues (question 4, Table 2). Turning to the second substantive dimension -the preferred level of centralization in EU policy-making -our respondents hold diverging views on whether or not they support a more 'intergovernmental' or a more 'supranational' Europe (see Table 3, questions 1 -4). While 24 per cent of our respondents want to see the European Commission as the true government of Europe, more than 40 per cent oppose such a view (see Table 3, question 3). At the same time, while a considerable share of agency staff (approximately 45 per cent) do see member states' pursuit of their respective national interest as a threat to the survival of the EU, a significant minority does not consider this to be problematic. Interestingly, almost 50 per cent of agencies' employees disagree with the claim that the EU level is the 'natural' place for addressing European political problems, while less than one-fifth do have a strong and unconditional preference for EU policy solutions (see Table 3, question 4). Hence, a 'functional' perspective, according to which problems should be addressed at the level where they can be tackled most efficiently, seems to be of considerable relevance in the views of the respondents and might support the claims of the proponents of multi-level network governance that efficient and legitimate policy solutions can be best found by linking EU-level experts with experts possessing 'local knowledge'. As Majone argues 'preferences vary locally and local conditions often affect both the costs and benefits of regulation, decentralized rulemaking and enforcement can provide a better match between local public goods and citizen preference [. . .] subsidiarity is an important source of regulatory legitimacy' (Majone 2000: 295). Finally, there is broad agreement among our survey respondents that conflicts driven by national interests and misunderstandings based on professional working cultures across EU countries do not substantively impact on their everyday work (see question 2, Table 3).
Comparing these findings with the results of Hooghe's study of Commission officials shows that more than 60 per cent of Commission officials share with agencies' staff their positive inclination towards subsidiarity (question 4, Table 3). However, with almost 70 per cent, the share of Commission officials who perceive member states' pursuit of their respective national interests as a threat to EU integration is considerably larger than among agencies' staff (question 1, Table 3). Moreover, Commission officials are almost perfectly divided over the question whether the Commission should become the government of the EU (question 3, Table 3). Finally and almost perfectly in line with EU agencies' staff, there is broad agreement among Commission officials that problems resulting from colleagues' different national backgrounds do not play a significant role in their day-to-day work (question 4, Table 3). Thus, while being in favour of policy solutions respecting the principle of subsidiary and while holding differentiated views on its own role in the political system of the EU, overall Commission officials are more wary about the merits of the intergovernmental mode of EU policy-making than agency staff seems to be. In sum, while the majority of agency professionals favours European integration and political cooperation among EU member states, there seems to be a less clear-cut notion on how this should be realized institutionally. There is no common outlook on the preferred allocation of authority across different territorial levels. The picture which emerges from the questions tapping the attitudes of agency staff members on the state of European integration and the EU's future is therefore less straightforward as might have been expected by those scholars arguing that EU agencies are part of an 'Eurocratic' structure aiming at the expansion of their competencies together with the European Commission (Kelemen 2002) or by structuralist bureaucratic theories that directly infer the interests of an organization from its functional and territorial institutional location.
The last analytical dimension we are discussing in this contribution deals with the attitudes of EU agency staff on the role the state should play in economic governance in general and the role of the EU in economic governance in particular (questions 1 -3, Table 4). Around 60 per cent of our respondents see a positive role for welfare states in promoting a 'fairer' society (Table 4, question 2), while only around one-tenth of our respondents oppose such an active role of the state. There is hence a strong inclination towards support for state intervention among those working in EU agencies, which is also reflected in attitudes on the role of the EU level and its role in political regulation. There is almost unanimous agreement that the EU should engage in 'positive' regulation and not reduce itself to 'merely' putting a common market in place (see question 3, Table 4). Among Commission officials, there is even stronger agreement on this issue, with the vast majority of respondents in agreement that Europe should be more than a common market (Hooghe 2001). This relationship also holds at the individual level: agency employees who have a positive attitude towards state intervention in general also tend to see such an interventionist role for the EU [r(t B ): 0.26, p(z): 0.0001]. On regulatory matters, agency staff members' attitudes on state intervention might thus result in a strong inclination to suggest more direct and hierarchical instead of non-binding forms of regulation.
In their self-placement along an eleven-point left (1) -right (11) scale, agency staff shows a rather centrist political outlook (mean value: 5.1; SD: 1.8), slightly bending to the (very moderate) left. This rather centrist political (self-) positioning is in line with the findings of other studies on members of national administration (Aberbach et al. 1981). More generally, one might claim on the basis of the data presented here that members of EU agencies do see a positive role for political intervention, but have no very strong and polarized ideological convictions on the form and content this intervention should take, as could be expected from ideologically dedicated (redistributive) 'leftist' and (liberal) 'rightists'.

CONCLUSION
This contribution has explored the attitudes of EU agency staff members. Our descriptive analysis has shown that the ideologies of agency professionals are by no means uniform and cohesive across the different dimensions under scrutiny. In this regard, our findings are largely in sync with Hooghe's (2001) analysis of the attitudes of Commission top officials. We find strong cohesion among agency staff with regard to the perceptions they hold on legitimacy and accountability of their work. According to our respondents, legitimacy and accountability of EU agencies builds on expertise and should be based on professional standards. Yet, legitimacy does not flow from professionalism and expertise alone; the results clearly indicate that agency staff members also seek public approval of their work. Moreover, the professionals working in EU agencies do also attribute an important role to the classic elements of majoritarian politics. There thus seems to be a selfunderstanding among EU agency staff members rooted in a strong sense of professionalism with, at the same time, an acute awareness of the political character (and impact) of their tasks. EU agency staff members might consider the exclusive reliance on professional norms insufficient to legitimize their work and are therefore also attentive to the political preferences and sensitivities of the broader public as well as of their political principles. Such attentiveness might exist for a number of principled, functional and instrumental reasons: first, agency professionals may not belief in the possibility to always separate 'objective facts' from 'value judgements'. They might therefore see a need for extra-professional norms underpinning accountability and legitimation of their work (see Brown 2009;Shapiro 1997). Second, agency staff might see a (functionally driven) need for public and political approval of their work to see their suggestions and proposals being taken up and implemented by political decision-makers (Scharpf 2009). From such a perspective, professionally derived norms of efficiency and democratic responsiveness are mutually reinforcing rather than exclusive (Majone 2010). Finally, agencies might consider public and political approval of their work instrumental in order to secure their organizational survival in the medium to long run.
Our interpretation of the data presented here does not lend itself to a perspective on EU agencies emphasizing a cohesive ideology or regarding EU agencies as a very special 'breed' or 'class' of like-minded experts, all sharing the same goals and outlooks, while -at the same time -being disconnected from the realm of politics. 10 While most agency staff members might not consider themselves as key players in the political game, they appear to be sensitive and responsive to the political discussions going on around them. Yet, whether agency professionals approve of democratic legitimation for intrinsic or functional/instrumental reasons cannot be answered on the basis of our data. A number of our empirical findings lend support to the interpretation that the general ideological outlook of EU agency professionals is conducive to their institutionally and politically prescribed roles to improve quality and effectiveness of EU regulation. The approval of agency staff of 'governance by subsidiary' in the EU and their relatively strong scepticism of a 'federalist' EU with the European Commission working as the EU's government does not lend support to the thesis that their primary (and unconditional) goal is an extension of the EU's regulatory and bureaucratic powers. Moreover, their attentiveness towards public approval and, at least in parts, acceptance of democratic politics and majoritarian institutions should have a positive impact on their responsiveness and their ability to recognize and react to the culturally, economically and politically heterogeneous situation in the EU. Whether and how these factors will play out in agencies' behaviour in individual policy decisions will have to be the object of policy studies.

NOTES
1 This paper is part of the research project 'Agency governance and its challenges to the EU's system of representation', jointly led by the two authors and based at the Mannheim Centre for European Social Research (MZES). The project is affiliated to the FP7 Integrated Project 'Reconstituting Democracy in Europe' (RECON) and financed by ARENA (Centre for European Studies) at the University of Oslo. 2 We sent our survey to 734 individuals, seven of whom let us know that they are only involved in clerical tasks and therefore not part of the population, which is of interest to us in this study. Another seven persons were not working for the respective agency anymore. 3 Respondents in our sample have an average age of 45 and are fairly equally distributed as regards their sex (42 per cent women vs. 58 per cent men). Almost all earned a university degree (96 per cent; 38 per cent hold a doctoral degree), with natural scientists/mathematicians constituting the largest group (33 per cent), followed by social scientists (18 per cent) and business administrators/economists (13 per cent). Moreover, our sample covers nationals from 20 EU member states and five non-EU member states, with Italians (18 per cent) and Germans (15 per cent) being the most frequent respondents. Given the relative heterogeneity of our respondents along relevant criteria, the results should not be driven by respondents with particular characteristics. Since, however, we have no information on the demographics in the overall target population, let alone in all EU agencies, we are not able to assess the representativeness of our sample. 4 Consequently, correlation coefficients among different factors (items) are moderate at best, reflecting the fact that agency personnel can, for example, have a very positive attitude towards majoritarian politics in general (questions 2 and 3 in Table 2) but at the same time hold the opinion that their own work should be evaluated strictly in terms of professional standards (questions 5 and 8 in Table 2). 5 Since variances for some items differ between agencies and sample sizes for individual agencies vary considerably, we chose a non-parametric test of variance (Kruskal-Wallis) to test for differences between agencies. 6 Correlation coefficients are non-parametric, rank-based, Kendall tau-b values. A list of pairwise correlations of all items, ranging between 20.25 and 0.30, is available from the authors upon request. 7 While not ruling out ideological conflicts, staff members do not seem to see it as a complement to professionally based accountability, as the two items are not strongly positively related (r(t B ): 0.10, p(z): 0.14). 8 We included a number of questions on the 'mode' of decision-making inside agencies in our survey: 'Normally there is agreement among the employees of the agency on how to proceed, and me and my colleagues do not really differ on the positions we upon different issues' (54 per cent: agree somewhat, 26 per cent: agree strongly; N ¼ 189); 'When taking decisions, it normally takes quite a while until we find a common position within the agency and are able to take a decision in consensus' (38 per cent: agree somewhat; 14 per cent: agree strongly); 'We regularly take decisions, even if consensus among the relevant members of the agency cannot be established' (38 per cent: agree somewhat; 14 per cent: agree strongly). 9 Hooghe originally used a four-point scale for her answer (disagree with reservation, disagree without reservation, agree without reservation, agree with reservation). She subsequently coded those not answering as 'undecided' and introduced this middle category (Hooghe 2001: 69). In our discussion, we treat Hooghe's scale as equivalent to ours. The number of responses to Hooghe's survey for the items discussed here varies between 103 and 105. 10 The findings of this paper support the conclusions drawn by Trondal (2010) and Trondal and Jeppesen (2008).