Institutional power balance and decision-making in EU agency boards Nuria Font. Universitat Autònoma de Barcelona. Paper presented at the:

1 Institutional power balance and decision-making in EU agency boards Nuria Font Universitat Autònoma de Barcelona Paper presented at the: IPSA World Congress Madrid, 8-11 July 2012 PAPER IN PROGRESS. VERY PRELIMINARY DRAFT! Please do not quote without the permission of the author Comments are welcome: nuria.font@uab.es

2 Introduction The proliferation of EU specialized agencies over the last two decades has become an increasingly important phenomenon in the configuration of EU institutional order. The agencification process has raised numerous questions on how this phenomenon has reshaped the inter-institutional balances. Much of the specialized literature has addresses this broad topic by focusing on agency creation and design (Kelemen 2002, Wonka and Rittberger 2010; Koop 2011, Thatcher 2011) whereas only a few comprehensive studies have focused on the institutional dynamics by looking at attitudes and behaviors during the post-delegation phase (Egeberg and Trondal 2011, Wonka and Rittberger 2011, Busuioc 2012). Yet, comprehensive studies addressing the effects of institutional structures on actual behavior are scant. This study empirically explores the effects of institutional structures on the level of influence the three main political principals the Commission, the Member States and the European Parliamentexert on agencies major decision-making. When analyzing decision-making at agencies, most scholars have underestimated the role of agency boards by considering this is a secondary and non-dominant body (Egeberg and Trondal 2011, Wonka 2011). Yet, boards adopt highly relevant decisions for the agency. Recent studies have highlighted the importance of boards as they adopt key decisions affecting the functioning and the activity of agencies (Groenleer 2009, Busuioc 2012). Most management boards are responsible for adopting decisions on programs, budget, appointments, administrative procedures and internal organization, in addition to hierarchically supervising the activity of the director and the works of the agency (Busuioc 2012). They are predominantly intergovernmental bodies as are normally composed by national delegates of each Member State, one or two but in some cases up to six- representatives of the Commission, and in some of them two experts nominated by the European Parliament. Some boards have also stakeholders representation. This study examines which of the three political actors exerts higher levels of influence on major decision-making and sketches out which institutional structures affect the inter-institutional balance of powers during the post delegation phase. The specialized literature provides alternative views on the institutional dominance, with some scholars suggesting that the Commission is the central actor (Busuioc 2010; Schout and Pereyra 2011; Egeberg and Trondal 2011) while others suggesting that the Member States dominate agency governance (Christensen and Nielsen 2010). Beyond this, the literature has overlooked two important aspects. One, the role of the European Parliament has largely been neglected. And second, current research has hardly explored in a comprehensive way what are the main drivers of political actors influence on agencies decision-making. Based on such empirical and theoretical concerns, this paper addresses the following research questions: What is the actual distribution of powers among the three political principals? How can variations in the influence of such actors be accounted for? In order to address these questions, this study provides empirical data measuring the level of influence of the Commission, the Member States and the European Parliament in agencies major decisions. In addition, it has the purpose to identify the main conditions under which political actors exert varying levels of influence. Dataset includes an online survey providing primary empirical data and

3 documentary analysis of 22 agencies founding regulations. 1 The paper is structured as follows. The first section discusses the theoretical debates assessing the importance of institutional structures in the influence political principals may exert in decisionmaking during the post-delegation phase. This section also includes the main hypothesis on the effects of institutional structures on the influence of the Commission, the Member States and the European Parliament. The next section describes the dataset and the methods employed in this research, whereas the following one presents the operationalization of the dependent, independent and control variables. The next section presents the results of the study, while the last one presents some concluding remarks discussing the theoretical implications of this contribution. Theoretical debates The specialized literature on EU agencies has recently addressed the question of institutional dominance among the political principals (Groenleer 2009, Christensen and Nielsen 2010, Egeberg and Trondal 2011). Theoretical debates provide competing views about the actual distribution of decision-making powers between the Member States and the Commission. The underlying assumption of these contributions is that political principals intervening in the institutional design of agencies endow these bodies with decision-making powers while at the same time they expect to exert political dominance once they have been created (Groenleer 2009). Yet, the literature has overlooked some theoretically relevant aspects. Research examining political principals has generally focused on the Commission and the Member States, but has overlooked the European Parliament (Lord 2011). As the design of EU agencies has been the outcome of interactions between the Commission, the Council and the European Parliament (Kelemen 2002), the latter is also likely to exert influence not only in agency design but also during the post-delegation phase (Lord 2011). Further, studies offering an aggregate view of agencies identifying the relative influence on decisionmaking powers by the three main political principals on decision-making are still lacking. This paper shares the assumption that agencies operate under principals political constrains (Kelemen 2002, Dehousse 2008) and is concerned with the influence of each of the three main principals -the Commission, the Member States and the European Parliament- in major decision-making during the post-delegation phase. When addressing the question of institutional dominance, the theoretical debates provide different interpretations. Some scholars suggest that the Member States exert wider influence, thus reflecting the allocation of formal decision making powers 1 The agencies covered by this study include: European Centre for the Development of Vocational Training (CEDEFOP), European Police College (CEPOL), Community Fisheries Control Agency (CFCA), Community Plant Variety Office (CPVO), European Aviation Safety Agency (EASA), European Centre for Disease Prevention and Control (ECDC), European Chemicals Agency (ECHA), European Environment Agency (EEA), European Food Safety Authority (EFSA), European Institute for Gender Equality (EIGE), European Medicines Agency (EMA), European Monitoring Centre for Drugs and Drug Addiction (EMCDDA), European Maritime Safety Agency (EMSA), European Network and Information Security Agency (ENISA), European Railway Agency (ERA), European Training Foundation (ETF), European Agency for Safety and Health at Work (EU-OSHA), European Union s Judicial Cooperation Unit (EUROJUST), European Union Agency for Fundamental Rights (FRA), European Agency for the Management of Operational Cooperation at the External Borders (FRONTEX) and Office for Harmonisation in the Internal Market (OHIM).

4 between the Commission and the Member States (Christensen and Nielsen 2010). An alternative explanation suggests that the Commission is the most influential actor at agency governance (Schout and Pereyra 2011; Egeberg and Trondal 2011), with some scholars suggesting that agencification contributes to the emergence of a new European executive order (Egeberg and Trondal 2011). Somewhere in between, other interpretations argue that agencies are still highly dependent on both national governments and the Commission in the undertaking of implementation activities (Groenleer et al 2010). The lack of consensus on the distribution of decision-making powers among political principals may generally be attributable to the fact that these studies have employed different research approaches and methodologies. Some studies have examined the formal institutional design and more in particular agencies formal-institutional independence and institutional balances (Wonka and Rittberger 2010; Christensen and Nielsen 2010, Hanretty and Koop 2011). While they have generally offered an overarching perspective of agencies formal aspects, these studies imprecisely capture the dynamics and influences of political principals during the post-delegation phase. Other studies have readdressed the empirical focus to agencies actual practice, thus capturing actors behaviours and interactions. Many of such contributions have examined a sample of agencies (Trondal and Jeppesen 2008, Wonka 2011) or conducted single case studies, thus lacking an overall perspective. One particularity of many of these studies is their focus on agency staff for the purpose of capturing the behaviour of people involved in day-to-day and highly specialized policy-making activities (Egeberg and Trondal 2011) and identifying the influence of different type of actors on agencies work (Wonka 2011). While these studies tend not to expect boards to be dominant actors, the relevance of management boards has recently been highlighted by the literature (Groenleer 2009, Busuioc 2012). EU agency boards vary considerably as regards composition and functions. Beyond differences, most of them generally adopt crucial decisions for the agency, namely multiannual and annual plans containing strategic objectives and short and mid-term lines of action, appointments of executive directors, budget issues, administrative procedures and internal rules of organization, in addition to hierarchically supervising the executive director and the overall activity of the agency. Given the saliency of the decisions adopted by boards, the actual influences of political principals become a crucial aspect of agency governance since they may have a strong impact on agency activity (Busuioc 2012, Groenleer 2009). This study focuses on major decision-making in the sphere of boards. It firstly addresses the question of how influential are the Commission, the Member States and the European Parliament. In order to answer this question, this research provides primary data measuring such influences. The next step in the research consists of identifying the main institutional structures leading to varying levels of influence political principals. The research enhances the need to explore the relationship between agencies design and structure, on the one hand, and effective behaviour, on the other, as some scholars have already suggested (Groenleer 2009, Busuioc 2012). Based on such concerns, the study deals with the following research questions: Which aspects of agencies institutional structure affect the varying levels of influence of the three political principals? Are those effects similar across these three political actors? In sum, if political actors delegate powers to agencies but at the same time design them to secure certain degree of influence once they are created, what aspects of the institutional design affect their influence?

5 The specialized literature provides different accounts for actors behaviours and the inter-institutional dynamics of agencies once they are created. Some studies have developed theoretical explanations based on the core assumptions of the standard Principal Agent (PA) approach (Groenleer 2009), which emphasizes the ideas of informational asymmetries and incomplete contracting as regards principals oversight capacities to reduce agency losses (McCubbins and Schwartz 1984). Much of the literature on EU agencies have questioned the PA accounts for considering that it offers a disrupted view of the effective influence and control political principals are capable to exert. Interestingly, specific studies provide empirical evidence demonstrating that tight oversight is compatible with quasi independent agency action (Gehring and Krapohl 2007) and revealing that both the Commission and the Member States exert high levels of influence in agencies that were created as independent entities (Groenleer et al 2010). Grounded in such debates, this study examines what type of governance rules are more likely to affect the level of influence of the Commission, the Member States and the European Parliament. This approach relates to the idea that agencies institutional design has effects on the influences political principals may exert during the postdelegation phase (Kelemen and Tarrant 2011), but sketches out which specific aspects of the institutional structure affect the level of influence of each of the three political principals. The central aspect of this research explores what elements of the institutional design and structure are more likely to have effects on the influence political principals exert. The most important ones include governance rules, agency capacities and the type of task undertaken by agencies. Governance rules are institutionalized into agencies statues in different ways. Those taken under consideration in this study include fundamental rules for which one can find variation across agencies. They include the following ones: Commission representation and number of votes; Commission chairing the board; European Parliament nominating experts of boards, intervening in the nomination of executive directors and evaluating them; possibility of director dismissal; mandate of the director; existence of a bureau, a board of appeal, scientific committees and scientific networks; and board size. The study departs from the assumption that the effects of such rules on political actors level of influence varies considerably for each political actor. In particular, it suggests that certain governance rules, namely voting rights, the establishment of scientific committees and the nomination of experts by the European Parliament at boards are likely to affect the level of influence of political principals on decision-making in distinct ways and thus the actual distribution of institutional powers. Voting rights is the first of the governance rules examined. As agency boards are generally composed by one representative of each member state, the higher or lower presence of Commission representatives may be a measure of varying formal institutional balances between the two main political principals. Yet, not all Commission representatives have one vote each. I hypothesise that the influence exerted by the Commission is expected to be higher at boards with higher representation. Following the same reasoning, the influence exerted by the Member States and the European Parliament is expected to be lower at boards with higher number of Commission representatives. The hypothesis derives from the idea that formal allocation of voting powers between the Member States and the Commission, which varies across agencies mostly depending on the number of Commission representatives, has effects on actual behaviors and on political principals capacities to retain powers during the post-delegation phase (Kelemen 2002).

6 The second rule of governance under examination is the establishment of panels of experts. To this respect, the literature on delegation has already suggested that scientists and experts performing independently from policy-makers contribute to enhance the efficiency and credibility of agencies (Majone 1997). Following this rationale, agencies with scientific committees and panels, which are part of the agency bodies of authority, are likely to be less vulnerable to the influence of the political principals, and in particular, of the Member States. The related hypotheses looks in this direction by suggesting that the influence of the Commission, the Member States and the European Parliament in agencies with scientific panels is expected to be lower, since highly qualified experts may counterbalance boards politization (Everson et al 1999). Finally, the third main rule of governance prioritized in the analysis is the presence of experts nominated by the European Parliament at agency boards. While the European Parliament is not explicitly represented at any agency, it has managed to create instruments of accountability and transparence in order to gain a more prominent role in agency governance (Lord 2011). I hypothesize that the influence exerted by the European Parliament is expected to be higher at boards with experts, whereas the influence of the Commission and the Member States is expected to be lower. The underlying rationale is to assess whether the indirect representation of the European Parliament affects the inter-institutional politics. In addition to institutional design accounts, alternative explanations focused on agency autonomy, which is closely related to the issue of institutional dominance, demonstrate that agency task expansion is a result of organizational capacity building and subsequent de facto actor-level autonomy (Trondal 2010). Organizational capacities and resources may encompass staff size, budget size, self-funding capacities and technical expertise. While all these type of resources build up organizational capacities, budget size is not considered as an independent variable in this study since it is strongly associated with staff size and may also be affected by the source of agency funding. Given that staff size is a more stable parameter of organizational capacities than budget size, I have selected the former against the latter. The hypothesis prioritizes staff size as one of the indicators of the structural characteristics of agencies because it not only refers to the human volume of the bureaucracy, but also to its organizational complexity and capacities. Taking such considerations into account, I hypothesize that political principals capacities to exert influence on key decision-making are expected to be lower in large agencies. Other than that, one might expect that agencies with selffunding capacities are less exposed to the influence of political principals that those that are fully financed by EU budget. For this reason, fully EU funding is a control variable in this study. The third concern lies in the type of activity conducted by agencies. The literature has examined whether the politics of decision-making may be affected by the dimension regulation versus non-regulation (Egeberg and Trondal 2011, Wonka 2011) or by the policy field, namely financial, utility and social regulation (Gilardi 2008). In line with such arguments, this study departs from the assumption that the type of activity is likely to affect the effective distribution of decision-making influences among political principals. Yet, the extent to which an EU agency is regulatory or falls into one of the several policy categories is not always easy to determine (Levi-Faur 2011; Egeberg and Trondal 2011). For this reason, and following some other scholars (Christensen and Nielsen 2010, Busuioc 2012), I employ the Commission classification of types of agencies according to the center of gravity of their activity. It differentiates five categories of tasks: decision-making, assistance, operations, information / networking

7 and services (Commission 2008). Based n the idea that the type of activity may have effects on the inter-institutional dynamics, the third hypothesis suggests that the influence of the Commission, the Member States and the European Parliament is expected to vary according to the main type of task undertaken by the agency. Political principals are expected to exert lower influence in those boards of agencies providing technical and scientific assistance and / or perform inspection reports because of the highly specialized nature of the activities they conduct and the dependencies they might generate. Table 1 summarizes the main hypotheses. Table 1. Main Hypothesis Hypothesis 1 Hypothesis 1a Hypothesis 1b Hypothesis 1c Governance rules Commission representation Scientific committees Experts nominated by EP The influence exerted by the Commission, the Member States and the European Parliament in agencies major decisions adopted by agencies is unevenly affected by institutional rules The influence of the Commission in agencies major decisions is expected to be higher where this political actor has higher representation, whereas the influence of the Member States and the European Parliament is expected to be lower The influence exerted by the Commission, the Member States and the European Parliament in agencies major decisions is expected to be lower at agencies having scientific committees or panels within their bodies of authority The influence exerted by the European Parliament in agencies major decisions is expected to be higher at boards including experts nominated by the European Parliament, whereas the influence of the Commission and the Member States is expected to be lower Hypothesis 2 Size The influence exerted by the Commission, the Member States and the European Parliament in agencies major decisions is expected to be lower at large agencies Hypothesis 3 Type of activity The influence exerted by the Commission, the Member States and the European Parliament in agencies major decisions is expected to be lower at agencies undertaking assistance and decisionmaking (regulatory and quasi-regulatory) activities

8 Data and Methods The dataset includes documentary analysis and the results of an online survey. The documents analyzed comprise the founding regulations and further amendments of all 28 EU agencies operating at the beginning of this research (January 2010), information provided at the agencies websites and other official documents. Such information has served to create a database with 15 independent variables clustered into three groups: governance rules, structural capacities measured by staff size and type of task. The data provided by documentary analysis has served to operationalize the independent variables, whereas the data obtained in the online survey served to measure the three dependent variables, that is, the influence of the Commission, the Member States and the European Parliament. The survey was administered online between April and August 2011. The study initially targeted all 28 EU agencies created under former First, Second and Third Pillars that were operating at the time of initiating this research. Yet, four agencies of former Second and Third Pillars refused to answer (EUROPOL, ISS, EUSC and EDA). Other two agencies, CdT and GSA, were excluded as in both cases answers were incomplete. The final sample covered 22 agencies, mostly from former first pillar (see Table 2). Given the diverse composition of EU agency boards, the survey was addressed to national representatives and excluded other type of board members. The main reason is that boards are normally integrated by one representative of each member state (with the exception of EFSA and EIGE, with 14 and 18 national representatives each). Most boards also include between one and six representatives of the Commission (with the exception of CEPOL and EUROJUST, with none). In addition, some agencies have large representation of stakeholders (CEDEFOP, EUROFUND and EU-OSHA) and those recently created may have two experts nominated by the European Parliament (ECDC, ECHA, EEA, EMA, EMCDDA and ERA). Given the group targeted in the survey, the sample is not purely representative of boards composition and may incorporate a selection bias and. Yet, national delegates constitute the only large homogeneous sample of board members. In the interpretation of the data, two caveats are needed. First, responses reflect national delegates aggregate perceptions and not the perceptions on the management boards as a whole. And second, survey data provide individual perceptions of political actors influence that may not fully reflect actual influences. As perception errors are unknown and immeasurable, this caveat must be taken into account when interpreting the results. I obtained 210 responses representing a universe of 565 surveyed. The response rate represents 37.1 % of the national representatives surveyed, and 36.7% of the total national delegates of all 22 agencies under study (see Table 2). The questionnaire introduced a question asking for the member state the surveyed represented in order to know the distribution of answers among nationalities and acknowledge a possible national bias in favor of a certain nationality. 80% of the surveyed responded such a question. I received a minimal of 2 respondents per Member State and a maximum of 9, with responses being distributed among the 27 nationalities. The responses of Germany, France and the United Kingdom represented almost 10% of the responses. Next, the number of respondents per agency varied from 2 to 17 (mean = 9,5; median = 10). As the number of answers provided by each agency was relatively low, the statistical exploitation of the survey does not allow for an individual analysis but for an aggregated one. Given that the research focuses on agencies and not on national delegates, working with means was preferred to working with raw data.

9 Table 2. Survey response rate Respondents Surveyed Respondents / Surveyed (%) MB members Respondents / MB (%) CEDEFOP 11 27 41 27 41 CEPOL 17 27 63 27 63 CFCA 7 27 26 27 26 CPVO 16 27 59 27 59 EASA 10 27 37 27 37 ECDC 10 24 42 27 37 ECHA 9 26 35 27 33 EEA 9 27 33 27 33 EFSA 2 14 14 14 14 EIGE 4 18 22 18 22 EMA 10 27 37 27 37 EMCDDA 12 27 44 27 44 EMSA 16 27 59 27 59 ENISA 11 27 41 27 41 ERA 6 26 23 27 22 ETF 12 27 44 27 44 EU-OSHA 13 27 48 27 48 EUROFUND 10 27 37 27 37 EUROJUST 4 27 15 27 15 FRA 4 26 15 27 15 FRONTEX 4 26 15 27 15 OHIM 13 27 48 27 48 Total analysis 210 565 37.1 572 36.7 Source: Author s own 2011 survey; N = 210. Operationalization of the Variables The data from the survey served to measure the dependent variables, with three values for each agency corresponding to the mean perceptions on the influence of the Commission, the Member States and the European Parliament on major decisions. I adopt the mean value of the responses relative to the influence of the three main political principals. The three variables are measured using respondents answers to the question: How influential are each of the following actors in the major decisions taken by your agency?: the Commission, the Member States and the European Parliament. The answers provided on a five-point scale (1, uninfluential; 2, scarcely influential; 3 somehow influential; 4 very influential; 5: extremely influential). The analysis includes the following set of independent variables: governance rules, organizational capacities (staff size) and type of tasks. The three sets of independent variables refer to the accounts outlined in the theoretical framework and are displayed in Table 3: governance rules, agency size and type of tasks. The first set of independent variables capture the main governance rules as defined in agencies founding regulations. These include 13 main variables. Four of them (Commission representatives, Commission votes, director mandate and board size) are scale variables

10 and include its original value. I have created dummy variables for the remaining 9 independent variables. The values of the governance rules variables are the following ones: number of Commission representatives at the board (scale value: 0-6); number Commission votes at the board (scale value: 0-6); the Commission chairing the board (=1, else=0); the EP nomination of experts at the board (=1, else=0); the EP intervening in the nomination of the executive director (=1, else=0); the EP evaluating the executive director (=1, else=0); the possibility of executive director dismissal (=1, else=0); the maximum years of mandate of the executive director (scale value 5-15); the existence of a bureau or an executive board (=1, else=0); the existence of a board of appeal (=1, else=0); the existence of scientific panels or committees (=1, else=0); the existence of scientific networks (=1, else=0); and the board size (original value). The second group referring to organizational capacities includes a single independent variable: agency staff size (scale: original value). Finally, the third group of variables refers to the main tasks undertaken by agencies. For the sake of simplification, I have adopted the classification by the Commission (2008) reflecting the center of gravity of agencies activities. In this regard, I have created dummy variables for each of the four types of tasks under study: decision-making (=1, else=0), direct assistance (=1, else=0), operations (=1, else=0) and information / network (=1, else=0). The task consisting of providing services to other agencies and institutions is not included in the analysis given that the only agency falling under this category, the Translation Center for the Bodies of the European Union, was excluded from the analysis. Finally, two control variables have been incorporated in the analysis: fully EU funding against total or partial self-funding capacities, on the one hand, and the year of creation of the agency, on the other hand. First, the analysis assesses if the influence of the Commission, the Member States and the European Parliament decreases in agencies with self-funding capacities (Koop 2011). The initial expectation is that agencies with capacities to have external revenue are less subject to EU financial oversight, thus principals may be less likely to be influential. I have created a dummy variable for the source of agency founding (fully EU funding = 1, else = 0). And second, the analysis controls for the year of creation by distinguishing between old and new agencies (created before 2000 = 1, after 2000 = 0). The main purpose of including this control variable is to assess if the influence of the three political principals on decision-making is affected by the agency age.

11 Table 3. Operationalization of the dependent and independent variables Dependent Variables Value Influence of the Commission 1-5 scale Influence of the Member States 1-5 scale Influence of the European Parliament 1-5 scale Independent Variables IV1 Governance rules Commission representatives 0-6 scale Commission votes 0-6 scale Commission chairs board Yes = 1, else = 0 EP nominates experts of board Yes = 1, else = 0 EP intervenes in nomination executive director Yes = 1, else = 0 EP evaluates executive director Yes = 1, else = 0 Executive director dismissal Yes = 1, else = 0 Mandate executive director 5-15 scale Bureau, executive board Yes = 1, else = 0 Board of appeal Yes = 1, else = 0 Scientific panel / committee Yes = 1, else = 0 Scientific network Yes = 1, else = 0 Board size Scale: original value IV2 Organizational capacities Staff size Scale: original value IV3 Type of task Decision-making Yes = 1, else = 0 Assistance Yes = 1, else = 0 Operations Yes = 1, else = 0 Information / Networks Yes = 1, else = 0 Control variables Agency age (old / new) Before 2000 = 1, after 2000 = 0 Fully EU-funding Yes = 1, else = 0 Findings The dependent variables are the influence of the Commission, the Member States and the European Parliament in the major decisions adopted by agency boards. The findings reveal that the Commission is perceived to be the most influential actor, followed by the Member States, with the European Parliament exerting lower influence. On a scale of 1 (uninfluential) to 5 (extremely influential), the mean score of Commission influence is 3.88, the mean of the Member States influence is 3.40 and the mean of the European Parliament influence is 2.49 (see Table 4). The results suggest that, in spite of agency boards being intergovernmental in composition, the aggregate perceptions of the national delegates indicate that there is a hierarchy of political actors as regards their influence on decision-making. The data also reveal that while the European Parliament is perceived to be considerably much less influential than the other two political principals, it is by no means perceived as unimportant. Thus, while this political actor has no direct representation at management boards, it is perceived to find alternative

12 means to exert certain degree of influence. The results also suggest that there is more variation on the level of influence by the Member States across agencies, probably because some of them are highly tied to national governments while others are much less. Table 4. Descriptive statistics of the dependent variables Influence of the Commission Influence of the Member States Influence of the European Parliament N 22 22 22 Mean (1-5 scale) 3.88 3.40 2.49 St. Dev. 0.67 0.78 0.66 Min. 1.50 1 1 Max. 4.80 4.50 3.63 Source: Author s own 2011 survey; N = 210; Original question: How influential are each of the following actors [Commission / Member States / European Parliament] in the major decisions taken by your Agency? Rate on a 1 (uninfluential) to 5 (extremely influential) scale. Table 5 displays the bivariate correlations between the dependent and the independent variables in the sample of 22 agencies. While the small sample of cases does not lead to highly significant coefficients, the data suggest that different patterns associated to the influence of each of the three political actors can be identified. Only staff size has a similar, negative and significant coefficient for the three political principals. In the case of the Commission and the Member States, the coefficient is significant at 0.05%, whereas in the case of the European Parliament it is almost insignificant (at 0.15% level). The type of task undertaken by agencies also seems to have effects on the influence exerted by political principals. Task assistance achieves a high, negative and significant coefficient at 0.05% level as regards the influence of the Commission and the European Parliament. This may reveal that agencies undertaking assistance tasks by means of providing highly technical and scientific input are less exposed to the influence of the Commission and the European Parliament and thus are more likely to increase the credibility of commitments. Those agencies undertaking operational activities present a high, positive and a significant coefficient only at 0.15 level for the influence of the Member States. While weak, the correlation suggests that agencies conducting operations that are largely coordinated and in some cases funded by the Member States remain particularly tied to national governments. As regards the independent variables grouped as governance rules, none of them achieves a significant coefficient for more than one of the three political principals. This is the case of the establishment of scientific committees or panels and networks of experts, achieving a high, negative and significant coefficient for the influence of the Member States (with networks of experts achieving lower level of significance). This correlation suggests that the presence of independent experts at agencies authority bodies is likely to reduce the degree of politization at boards decision-making. Another case is the number of Commission representatives represented on agency boards, achieving a high, positive and scarcely significant coefficient (at 0.15% level). For the other independent variables considered in this study, no significant correlation has been found.

13 Table 5. Bivariate correlations between the independent variables and the influence of the Commission, the Member States and the European Parliament Commission MS EP N Pearsons r Pearsons r Pearsons r Governance rules Commission representatives.327 ŧ.226.092 22 Commission votes.323 ŧ.224.106 22 Commission chairs board.227.155 -.110 22 EP nominates experts of MB.186.052.063 22 EP inverv. nomination ED.139 -.007 -.043 22 EP evaluates ED.215.047 -.233 22 ED dismissal.168 -.146 -.244 22 Mandate ED -.018.166 -.188 22 Bureau, executive board.149.060.175 22 Board of appeal -.097 -.093 -.156 22 Scientific committee / panel -.167 -.512**.075 22 Scientific network -.061 -.359 ŧ.218 22 MB size.285.162.306 22 Organizational capacities Staff size -.531** -.468** -.329 ŧ 22 EU funding.118.188.052 22 Type of task / Activity Decision -.048 -.060 -,077 22 Assistance -.452** -.285 -.522** 22 Operations.155.356 ŧ -.014 22 Information / Network.321 ŧ.011.333 ŧ 22 **Significant at 0,05 level, *significant at 0,1 level, ŧ significant at 0,15 level (bilateral) Table 6 offers a multivariate analysis by means of four ordinary least squares (OLS) regression analysis using key independent variables for each of the three political principals. Since the study only covers 22 cases, each model combines two independent variables. The purpose is to identify and privilege the most significant factors affecting political actors influence. Model I combines the number of Commission representatives and task assistance. It can receive clear confirmation for the influence of the Commission, and a lower confirmation for the European Parliament. As the results suggest, the model partially confirms hypothesis 1a since the number of Commission representatives at boards clearly enhances the influence of this political actor, with positive and non-small coefficients. Yet, and perhaps most importantly, its effects on the influence of the Member States and the European Parliament cannot be confirmed. The other variable of this model, task assistance, clearly confirms hypothesis 3 since those agencies providing assistance to the Commission as a major activity have the effect of reducing the influence of both the Commission and the European Parliament. In both cases, task assistance has a clear effect, with negative and high coefficients, and in general clearly significant. Model II excludes governance rules but instead combines task assistance and staff size. This model is also confirmed for the influence of the Commission and the European Parliament. Instead, Models I and II fail to reach statistical significance for the influence of the Member States. For this principal, Model

14 III combining staff size as a structural indicator of organizational capacity with the existence of scientific committees receives confirmation. The findings indicate that there is a negative and significant relationship between the variables included in the model and the influence of the Member States. By contrast, they fail to reach statistical significance for the Commission and the European Parliament. As the results reveal, the effect of scientific committees and panels in combination with a large bureaucracy clearly reduce the capacity of the Member States to exert influence. The negative effects of scientific committees on the Member States influence on decision-making at agency boards is also confirmed in Model IV, which incorporates the number of Commission representatives. In this model, the effects of scientific committees are confirmed but the number of Commission representatives fails to reach statistical significance. Model IV also fails to reach statistical significance for the influence of the Commission and the European Parliament. Table 6. OLS estimations for the influence of the Commission, the Member States and the European Parliament Commission Member States European Parliament I II III IV I II III IV I II III IV B B B B B B B B B B N Com.21** (.87) Panel -.08 (.28) Staff size -.001* -.002** (.001) (.001) Assist -.87** -.44 (.22) (.32) Const. 3.599 4.262 (.22) (.19).15 (.10) -.21 (.30) 3.610 (.28).16 (.11) -.64 ŧ (.39) -.002 ŧ (.001) -.20 (.40) 3.783 (.22) -.71** (.29) -.002* (.001).11 (.10) -.82** (.31) 3,415 (.29).10 (.09) -.88*** (.30) 2.468 (.23).00 (.001) -.72** (.32) 2.754 (.18).21 (.30) -.001 ŧ (.001) Adj. R 2.33.27.21.03.08.14.33.23.24.21.03 -.08 N 22 22 22 22 22 22 22 22 22 22 22 22 Non-standardized OLS regression coefficients. Non-standard errors in brackets 2,672 (.21) ***significant at 0,01 level **significant at 0,05 level, *significant at 0,1 level; ŧ significant at 0,15 level (bilateral).04 (.10).11 (.31) 2.363 (.30) Support is found for hypothesis 1 on the uneven effects of governance rules on the influence of the Commission, the Member States and the European Parliament. As regards this group of independent variables, hypothesis 1a on Commission representation is only confirmed for the influence of the Commission, but there is no significant indication that the higher number of Commission representatives at agencies boards reduces the influence of the Member States. These findings might indicate that the higher number of Commission representatives are indeed associated with higher influence of this political actor, but there is no evidence that this goes to the detriment of the influence of the Member States. Moreover, the latter present a specific pattern as their influence is clearly reduced by the presence of scientific committees and panels. Yet, hypothesis 1b on the effects of scientific committees is only partially confirmed as the coefficient is statistically insignificant for the influence of the Commission and the European Parliament. Hypothesis 1c on experts nominated by the European Parliament fails to be confirmed. No evidence is found for the effect of the nomination of independent experts by the European Parliament on the influence of this

15 political actor in the decision-making processes. This finding suggests that this mechanism by which might expect to exert indirect influence through the nomination of experts is in practice of lower importance. By contrast, support for the hypothesis 2 on staff size is found. Accounts based on organizational capacities, which in this study are primarily measured by staff size, are confirmed. The data reveal that staff size is negatively and significantly related to the influence of the three political principals, but reaching different significance for each of them. Finally, hypothesis 3 on the effects of the type of activity is partially confirmed for agencies providing direct assistance for the Commission and the European Parliament, but fails to reach a significant coefficient for the Member States. Finally, when assessing the effects of the control variables, none of them year of creation, fully EU-funding- appear to significantly alter the coefficients obtained in all four models (the related data are not included in the table for the sake of clarity). Concluding remarks The findings are consistent with an important line of research on institutional dominance of EU agencies highlighting the relatively dominant role of the Commission vis-à-vis the Member States (Egeberg and Trondal 2011). The results also confirm the less well-known idea that the European Parliament also exerts some degree of influence in decision-making, although it is much lower than that exerted by the Commission and the Member States. These findings suggest that, at an aggregate level, the Commission is on top of the hierarchy of principals, followed by the Member States, with the European Parliament far behind. The findings also reveal that institutional structures are important determinants of the influence political principals exert on agencies decisionmaking. The results confirm the idea that agency design has effects on the decisionmaking powers political principals can exert on the post-delegation phase. These findings are consistent with the argument that political principals try to exert influence on agency design in order to keep their influences and a self-interested balance of powers during the post-delegation phase (Kelemen 2002). Previous studies have examined the varying levels of agencies formal autonomy and the varying influences of actors external to agencies on management staff. Yet, studies measuring the influence of the three main political principals and providing accounts for the varying levels of influence at the main body of agencies decision-making were lacking in the specialized literature. This is the main empirical and theoretical contribution of this study. The findings also identify which specific institutional structures have varying effects on the level of influence of each of the three political principals. In order to provide accounts for these variations, the study has addressed three types of institutional structures: governance rules, organizational capacities and type of tasks. The results reveal that certain structural features of agencies, as some scholars suggest (Christensen and Nielsen 2010), affect the inter-institutional balance and the distribution of decisionmaking powers but in varying ways across principals. This study reveals that staff size is clearly determinant of the influence of political principals. This finding may be explained by the fact that large agencies are normally complex organizations, sometimes even larger in size than a single DG, and undertake highly technical tasks sometimes in highly sensitive fields for the EU and the Member States. They are normally highly knowledge-based bureaucracies whose large staff is embedded in networks of experts (Craig 2006) and are better prepared to make the most of the Commission s informational dependencies (Everson et al 1999). As long as large

16 agencies are more likely to become niches of expertise and are capable of exploiting this resource in order to exert wider bottom-up influence (Egeberg and Trondal 2011, Wonka 2011), the expectation is that political principals are less capable of overcoming informational asymmetries and give guidance and exert influence. The study also reveals that those agencies undertaking assistance activities are less exposed to the influence of political principals, especially in the case of the Commission, than agencies undertaking other type of task. This finding may be explained by the fact that the Commission is highly dependent on the activity these agencies conduct, as they provide technical and scientific input for the Commission legislative program and some of them undertake quasi-regulatory activity. Interestingly enough, the Member States present a distinct pattern. The existence of scientific committees or panels in agencies authority bodies contributes to reduce their practical decision-making powers. This idea supports the theoretically relevant argument emphasizing that scientific control modulates politization (Everson et al 1999). Yet, no evidence is found for the effect of scientific committees and panels on the influence of the Commission and the European Parliament. Ins um, the study reveals that certain institutional structures as designed by political principals have effects on the decision-making powers they effectively have during the post-delegation phase, but the effects of such structures vary across principals. This research has measured the relative powers of the three main political principals of EU agencies and has revealed that certain institutional structures matter more than others in the configuration of the inter-institutional dynamics. When political principals create large bureaucratic structures endowed with regulatory or quasi-regulatory powers and enhance the role of experts in agencies authority bodies, the overall expectation is that political principals will be under-resourced to exert influence in major decisions during the post-delegation phase. In-depth qualitative analysis exploring the micro-mechanisms of such relationships and further investigation on the variations of the inter-institutional politics across specific agencies and across policy sectors constitute promising lines of research. Acknowledgements This research has been funded by the Spanish Ministry of Economy and Competition. References Busuioc, M. 2009. Accountability, Control and Independence: The Case of European Agencies, European Law Journal, 15, 5, 599-615. Busuioc, M. 2010. The Accountability of European Agencies, Delf: Eburon. Busuioc, M. 2012. European agencies and their boards: promises and pitfalls of accountability beyond design, Journal of European Public Policy, 19, 5, 719-736. Christensen, J.G. and V.L. Nielsen. 2010. Administrative capacity, structural choice and the creation of EU agencies, Journal of European Public Policy, 17, 2, 176-204.

17 Commission of the European Communities. 2008. Communication from the Commission to the European Parliament and the Council, COM (2008) 135 final. Craig, P. 2006. EU Administrative Law, Oxford University Press. Dehousse, R. 2008. Delegation of Powers in the European Union: The Need for a Multi-Principals Model. West European Politics, 31, 4, 789 805. Egeberg, M. and J. Trondal. 2011. EU agencies: New executive center formation or vehicles for national control?, Journal of European Public Policy, 18, 6, 868-887. Everson, Michelle/Majone, Giandomenico/Metcalfe, Les/Schout, Adriaan (1999): The Role of Specialised Agencies in Decentralising EU Governance. Report Presented to the Commission. http://europa.eu.int/comm/governance/areas/group6/contribution_en.pdf Gehring, T. and Krapohl, S. 2007. Supranational Regulatory Agencies between Independence and Control: The EMEA and the Authorization of Pharmaceuticals in the European Single Market. Journal of European Public Policy, 14, 2, 208 226. Gilardi, F. (2008). Delegation in the Regulatory State. Independent Regulatory Agencies in Western Europe, Cheltenham, Edeard Elgar. Groenleer, M. 2009. The Autonomy of European Union Agencies: A Comparative Study of Institutional Development, Delft: Eburon. Groenleer, M., M. Kaeding and E. Versluis. 2010. Regulatory governance through agencies of the European Union? The role of the European agencies for maritime and aviation safety in the implementation of European transport legislation. Journal of European Public Policy, 17, 8, 1212-1230. Hanretty, C. and C. Koop. 2011. Measuring the formal independence of regulatory agencies, Journal of European Public Policy, 19, 2, 198-216. Kelemen, D.R. 2002. The Politics of Eurocratic Structure and the New European agencies, West European Politics, 25, 93 118. Kelemen, D.R. and A. Tarrant. 2011. The Politics of the Eurocracy: Designing EU Regulatory Agencies and Networks, West European Politics, 34, 5, 922-947. Koop, C. 2011. Explaining the Accountability of Independent Agencies: The Importance of Political Salience, Journal of Public Policy, 31, 2, 209-234. Lord, C. 2011. The European Parliament and the legitimation of agencification, Journal of European Public Policy, 18, 6, 909-925. Majone, G. 1996. Regulating Europe. London: Routledge. Majone, G. 1997. The New European Agencies: Regulation by Information, Journal of European Public Policy, 4, 2, 262-75. McCubbins, M. D. and T. Schwartz. 1984. Congressional Oversight Overlooked: Police Patrols Versus Fire Alarms, American Journal of Political Science, 28, 165 79.