This is the accepted version of the journal article: Font, Nuria; Pérez Durán, Ixchel. «The information phase of accountability : The role of management boards in European Union agencies». International Review of Administrative Sciences, Vol. 88 Núm. 3 (September 2022), p. 882-898. DOI 10.1177/0020852320946813 This version is available at https://ddd.uab.cat/record/269090 under the terms of the license The information phase of accountability: The role of management boards in European Union agencies Nuria Font Universitat Autònoma de Barcelona, Spain ORCID 0000-0002-2629-6829 Ixchel Pérez-Durán Universitat Autònoma de Barcelona, Spain ORCID 0000-0003-3665-3380 Published at: International Review of Administrative Sciences, 88(3), 882–898. https://doi.org/10.1177/0020852320946813 Abstract The creation of European Union agencies has prompted normative and theoretical discussions about their accountability and democratic legitimacy. This study examines the extent to which different features concerning the institutional design of agencies incentivize information provision on their performance. To do so, it measures four aspects for which their governing bodies are expected to provide information: governance decisions, policy performance, financial issues, and interactions with stakeholders. Using a new data set of informative statements included in 199 annual activity reports of 29 agencies during the period 2010–2016, this study demonstrates that formal accountability positively affects information provision on governance and policy issues, and that the establishment of scientific bodies within agencies improves information provision on governance aspects, policy performance and their interactions with stakeholders. Finally, this article finds that agencies that perform more direct regulation are subject to greater transparency demands about their governance decisions and stakeholder interactions. 1. Introduction The spread of the regulatory state has been one of the main transformations in the public sector in recent decades and has resulted in the creation of different types of independent agencies (Jordana et al, 2011). At the European Union (EU) level, there are currently 33 decentralized agencies covering a wide range of policy issues, including food safety, financial markets and asylum processes. The growth of independent bodies that are not subject to majoritarian democratic procedures has increased concerns about their legitimacy and accountability (Olsen, 2013). In the context of (democratic and) contemporary modes of regulatory governance it is crucial to examine whether accountability mechanisms of agencies are able to strengthen input and output legitimacy of regulatory regimes. Prior research on European Union agencies (EAs) has advanced our understanding of these agencies’ political independence (Wonka & Rittberger, 2010), their relationships with different political principals (Egeberg & Trondal, 2011; Font, 2015b; Kelemen, 2002) and with stakeholders (Borrás et al, 2007; Pérez-Durán, 2019; Wood, 2018), their parliamentary control (Lord, 2011), their reputation (Busuioc & Rimkuté, 2019; Rimkuté, 2019), and their accountability (Buess, 2015a, 2015b; Busuioc, 2013; Font, 2015a). However, the question of whether information provision in agencies —an essential condition for agencies to be held accountable— is affected by their institutional design remains an underexplored area of research. The present study fills this literature gap by addressing the following research question: to what extent do different features of the institutional design of EAs incentivize their governing bodies to provide information about their performance? We focus on a core phase of accountability: the provision of information (Brandsma & Schilemmans, 2013). While we acknowledge that transparency is not a sufficient condition for accountability (Bovens, 2017), this phase represents a sine qua non for the subsequent phases of the accountability cycle, namely justification and sanction (Meijer, 2013). By addressing the question “Accountable to whom?” (Bovens, 2007), we examine the role of agency management boards (MB) —understood as account-givers (actors)— in the provision of information that can be accessible either for their political principals and external audiences —understood as account-holders (forums). In addition, by addressing the question “Accountable for what?” (Bovens, 2007), this article assesses the provision of information on four aspects upon which MBs are expected to provide information: governance issues, policy performance, financial aspects and their interactions with stakeholders. Hence, we empirically examine four dependent variables, which correspond to the provision of information about the decisions and actions of management boards on governance issues, policy performance, financial aspects, and their interactions with stakeholders. For each of these operational aspects, the study tests theoretically grounded hypotheses on the effect of agency institutional design on information provision. In particular, we argue that crucial aspects of institutional arrangements in agencies, such as formal provisions of accountability, participation of scientists, representation of stakeholders and agency tasks affect the provision of information that is carried out in practice. The study is based on an original dataset including 921 informative statements about decisions and actions of management boards —on the four aspects examined— included in 199 Annual Activity Reports (AARs) of 29 EAs during the period 2010-2016. The study examines AARs, as these allow agencies to inform a wide variety of forums (e.g. political principals, stakeholders and the public in general) about their decisions and actions. As current studies on EU agencies confirm (Busuioc & Rimkuté, 2019; Rimkuté, 2019), AARs are suitable sources of information because their mandatory nature allows us to examine and compare the same type of empirical evidence across agencies. More importantly, as neither the amount nor the type of information included in AARs is formally established in the basic regulations of agencies, examining the conditions under which agencies provide different amount of information on each of the four aspects under study is especially pertinent. This study contributes to the literature on agencification and EU governance in three ways. Firstly, the study directly engages with debates on transparency and accountability of the European regulatory state and its non-majoritarian institutions (Brandsma et al, 2016; Wille, 2016). In particular, this article provides a systematic assessment of information provision performed by agencies’ MBs — which represent their highest level of authority—, demonstrating that agencies perform reporting activity differently. Secondly, our study also engages with debates on the tension between Eurocracy and democracy (Büttner et al, 2015; Georgakakis, 2017; Georgakakis & Rowell 2013). On the basis of Majone’s (1997) argument that EU agencies were created to provide expert knowledge to decision- making and enhance credible commitment, some scholars have suggested that although agencies cannot achieve input-oriented (democratic) legitimacy, they can increase output-oriented (technocratic) legitimacy (Wonka & Rittberger, 2010). In response to this argument, other scholars have stressed that both type of legitimacy tools are necessary in a democratic context that demands accountability to a wide range of stakeholders, societal actors, public institutions and political institutions that operate within the multilevel EU governance (Olsen, 2013). Our study provides evidence about the interplay between both types of legitimacy tools, by showing to what extent and under which conditions the provision of information covers different types of aspects, such as those dealing with agencies’ performance and those dealing with their governance decisions. Finally, from a normative perspective, the analysis of the scope of information provision on regulatory institutions is particularly important to assess whether it could serve as a tool to enhance trust in European Union democracies and their institutions. 2. ‘Accountability to whom’ and ‘Accountability for what’ in EAs Theoretical approaches of accountability This article adopts the “social mechanism approach” of accountability (Dubnick, 2005: 381), which is widely used in public administration studies. Specifically, accountability is understood as a social relationship between at least two parties, an actor and a forum, in which the former is obliged to inform and justify his/her conduct, while the latter has the mechanisms to evaluate and sanction such conduct (Bovens, 2007 & 2010; Mulgan, 2000). According to Brandsma and Schilemmans (2013), there are three core phases of accountability: information, discussion and consequences. The first consists of the provision of information about an actor’s decisions and actions in front of a forum. The second is based on the questions placed by the forum and the reasons provided by the actor regarding her/his decisions. For its part, the consequences phase is focused on the evaluation mechanisms carried out by the forum and the sanctions derived —either positive or negative— from the actions performed by the actor. Given that accountability implies a systematic set of mechanisms to inform, justify and sanction the conduct of the actors (Lindberg, 2013), some studies have examined such mechanisms from a formal perspective by analysing the legal framework that regulates the procedures for informing about and sanctioning this behaviour (e.g. Koop, 2011). Other studies focus on de facto accountability by examining how formal accountability provisions are implemented in practice (e.g. Busuioc, 2013). All in all, both formal and de facto approaches on accountability typically deal with two questions: “Accountability to whom” and “Accountability for what” (Dubnick, 2005; Bovens, 2007). Regarding the first question, public administration scholars have emphasised at length that, in the public sector, there are many accountability relationships in which “the forums are not principals of the actors” (Bovens, 2010: 951) or else in which “the principal is normally just one of multiple account-holders” (Schillemans & Busuioc, 2015). For example, one can generally differentiate between upwards accountability —based on vertical/hierarchical processes between subordinates and superiors—, horizontal accountability —carried out by state agencies—, diagonal accountability —characterised by intermediary forums of accountability— (Bovens, 2007), and downwards accountability —which refers to accountability to the people or community at large (Hood, 2010). Concerning “Accountability for what,” relationships can be differentiated depending on the issue subject to accountability. From this perspective, accountability can be disentangled into different issues that fall under the responsibility of a given actor. As Behn (2001) emphasises, this approach allows for the examination of, for instance, whether the policies and activities of the actor are producing the expected results. In a similar line, and by focusing on the substantive contents of accountability, Bovens (2007) distinguishes between financial accountability —which refers to budgetary scrutiny—, procedural accountability —mainly focused on the examination of the decision-making process—, and product accountability —which refers to the scrutiny of the produced outcome. Informative role of MBs in EAs: informing about governance, policy performance, budget and stakeholder interactions This study focuses on assessing one core phase of accountability: the provision of information by those actors who are required to be accountable (Brandsma & Schilemmans, 2013). A more precise conceptualization when referring to this component is transparency, understood as “the availability of information about an actor that allows other actors to monitor the workings or performance of the first actor” (Meijer, 2013: 430). We acknowledge that transparency is a necessary but not sufficient condition for accountability. As Bovens (2007: 453) argues, “transparency as such is not enough to qualify as a genuine form of accountability, (…) because transparency does not necessarily involve scrutiny by a specific forum.” Although we recognise the limitations of focusing on the information phase, this study is based on the assumption that this phase represents a crucial step in the accountability process since only through the availability of information is it possible to carry out evaluations and sanctions (Meijer, 2013; Grimmelikhuijsen, 2012). When it comes to assessing “Accountability to whom” in EAs, scholars of agencification have generally emphasised that, rather than representing a clear principal-agent relationship, these bodies respond to a wide array of political principals such as the European Commission, the European Parliament, and the Member States (Egeberg & Trondal, 2017; Kelemen, 2002; Kelemen & Tarrant, 2011). In addition, for reasons of legitimacy, agencies can also be accountable to external forums/audiences, such as stakeholders and the general public (Wood, 2018). This study investigates the provision of information provided by the highest level of authority within EAs: their MBs. In particular, we examine the extent to which MBs provide information about their performance that can be accessible to their political principals and to external audiences. The analysis of decisions and actions of MBs allows the identification of agencies’ core activities and priorities. While some scholars have criticized the importance attributed to MBs by arguing that these bodies perform multiple and complex tasks (e.g. Schout & Pereyra, 2011), others acknowledge the significance of their role (e.g. Buess, 2015a, 2015b). Although the type of tasks varies across agencies, in general MBs are responsible for adopting crucial accountability mechanisms of agencies, such as their annual work programs, AARs, and annual budget accounts. The composition of MBs varies across agencies, but the majority of them include representatives of the Commission and the Member States, and in some cases representatives from the European Parliament, stakeholders and other EU agencies, as well as non-EU countries. We assume that coexistence of both intergovernmental and supranational logics (Jordana & Triviño, 2020) also emphasizes the importance of investigating these bodies. In terms of specific objects (“Accountability for what”), the work by Bovens (2007) is particularly relevant for identifying the issues upon which MBs are expected to provide information. Against this backdrop, we examine the provision of information on governance decisions, policy performance and financial aspects. In addition, as current studies also emphasise the importance of providing information about interest group participation in decision-making (e.g. Wood, 2018), we also assess information on the interaction between agencies and stakeholders. The transparency of this aspect is crucial for examining to what extent and for what purpose agencies interact with stakeholders. That is, to investigate whether such interactions seek to enhance the exchange of knowledge and information, to generate credibility in front of societal actors, and/or to favor their involvement in policy decisions of agencies (Borrás et al, 2007). In sum, we distinguish four types of issue where the provision of information is involved: a) Information on governance, which deals with MB decisions and actions in relation to the governance structure of agencies. For example, decisions on the rules of decision-making, the code of conduct, declarations of conflicts of interest, transparency/public access to internal documents and anti-fraud strategies of agencies. b) Information on policy performance, which covers decisions and actions related to the policy outcomes of the agency in question. For example, the development of indexes on policy coverage and performance, the insurance of certifications and the deployment of joint Member State operations. c) Information on financial aspects, regarding MB decisions and actions on budget management. For example, decisions on the approval of annual accounts, budget implementation, and budgetary procedures. d) Information on interactions between agencies and stakeholders, which deals with MB decisions and actions in relation to the involvement of stakeholders (firms, NGOs, trade unions, business associations and consumer groups) in the decision-making of agencies. An example of this is the information provided through public consultations and meetings. 3. Institutional design and the informative dimension of accountability In order to assess the factors that help explain the provision of information across agencies, this study builds on the assumption that the way in which EAs are designed at the delegation stage provides constraints and incentives in the day-to-day operations of these bodies during the post-delegation phase. This argument is widely sustained in recent research on agencification in the EU assuming that the actions of agencies could be conditioned by their institutional structures (Egeberg et al. 2012). This study investigates the effect of key features of agency design. In particular, we focus on broad classifications of agencies according to their obligations of accountability, scientific profile, inclusion of societal actors in their organizational structures, and formal tasks. Formal accountability One of the main concerns of pioneer studies on agencification was determining the extent to which EAs are accountable and subject to the control of their political principals (Busuioc, 2013; Wonka & Rittberger, 2010). With a focus on formal rules, some scholars (Font, 2015a) have identified and assessed the core provisions of accountability included in the founding documents of agencies. Since such formal provisions are mainly the responsibility of MBs, we expect those agencies whose regulations make more explicit references to formal accountability to be stronger performers when it comes to the de facto provision of information. This expectation is based on the idea that the provision of information can be a consequence of formal requirements set out in the institutional design of agencies. In the EU context, while all EAs are required to publish AARs, we expect that variations in the amount and type of information provided might be influenced by their formal requirements. In particular, we expect that MBs of agencies whose founding regulations include more accountability provisions would be more inclined than those with lower accountability requirements to supply more information about their performance, governance issues, financial aspects, and interactions with stakeholders. Hypothesis 1: MBs of agencies with higher degrees of formal accountability provide more information on governance issues (H1a); policy performance (H1b); financial aspects (H1c); and their interactions with stakeholders (H1d) Formal participation of scientists Previous studies show that the inclusion of scientific experts in EU institutions as key information providers is not only meant to remedy information deficits, but also to use scientific knowledge as a way to gain legitimacy (e.g. Ossege, 2015). While there are several types of experts who can participate in the policy-making of EAs, we expect that the involvement of scientists would improve the provision of information about agencies performance. This expectation is built on two main arguments. Firstly, MBs and scientific committees normally interact regularly with each other since scientific experts are appointed by the boards (e.g. FRA, EMCDDA and EEA) —or else, in consultation with them (e.g. EMA). This close relationship between both bodies enhances the fluency in the provision of information on decisions dealing with the policy issue and with more technical aspects of agency tasks. Secondly, we know from previous studies that scientific knowledge itself also needs to achieve legitimacy and credibility through a set of requirements, such as the absence of conflict of interest of such experts (Jasanoff 2005, 211; Krick, 2018). Since some studies have highlighted an increasing demand to democratise expert advice in EU institutions (Ossege, 2016), we could also expect that the institutionalised involvement of scientists would positively affect the provision of information on other aspects, such as governance decisions of agencies and stakeholder interactions. Hypothesis 2: MBs of agencies with a permanent participation of scientists provide more information on governance issues (H2a); policy performance (H2b); financial aspects (H2c); and their interactions with stakeholders (H2d) Formal participation of stakeholders Previous studies have examined the decision of agency designers to ensure formal representation of stakeholders in crucial decision-making bodies of EAs, namely their MBs, advisory/expert forums and stakeholder groups (Pérez-Durán, 2018). Some scholars suggest that stakeholders (e.g. representatives of industry, trade unions and civil society organizations) should be involved in the decision-making of agencies as a way to generate credibility and legitimacy at the post-delegation stage (Borrás et al., 2007). By contrast, others argue that such agencies need to be independent not only from politicians, but also from private interests in order to avoid regulatory capture (Stigler, 1971). Faced with such concerns, one would expect that MBs of agencies with formal representation of stakeholders, as compared to those without, would have more incentives to improve their provision of information on their decisions and actions. This strategy would be aimed at reducing potential criticism concerning agency capture and avoiding reputational losses in front of a wide range of societal actors (Busuioc & Lodge, 2017; Carpenter, 2010). We expect formal representation of stakeholders to increase the provision of information along the four issues covered in this study. Hypothesis 3: MBs of agencies with a formal representation of stakeholders provide more information on governance issues (H3a); policy performance (H3b); financial aspects (H3c); and their interactions with stakeholders (H3d) However, it would also be plausible to expect the opposite effect: that MBs of agencies whose formal design ensured the representation of stakeholders might consider that there is less need to inform external forums about their performance, under the assumption that they already have the consent of the most important parties for the policy issue in question. Hence, an institutional feature that was initially provided to enhance the inclusiveness of the agency might end up being a closed mechanism to favour accountability practices towards a reduced number of interested parties. This process of biased accountability might occur when agencies "stack the deck" (McCubbins, Noll & Weingast, 1987) in favour of specific types of constituents (Guidi, 2015). Hypothesis 4: MBs of agencies with a formal representation of stakeholders provide less information on their governance decision (H4a); policy performance (H3b); financial aspects (H4c); and their interactions with stakeholders (H4d) Agency task: agencies performing a direct way of regulation Scholars on agencification (Kelemen, 2002) have commonly classified EAs according to the nature of their ‘primary tasks’. These types of tasks, which are defined in their institutional design, can generally be classified as: agencies performing a direct way of regulation (Majone, 1997), related to agencies that assist the Commission in providing technical and/or scientific advice on developing or amending EU legislation (Busuioc, 2013); agencies performing an indirect way of regulation (Majone, 1997), related to informative agencies in charge of gathering and providing information that the Commission and/or the Member States use for policy-making purposes; and operational agencies, referring to those in charge of coordinating operational activities between Member States. Since agencies performing direct regulation rely on legally binding standards and are more politically independent than both operational and informative agencies (Wonka & Rittberger, 2010), some scholars (e.g. Borrás et al., 2007) have noted that holding such a predominant place has led to a growing need for greater credibility. Against this backdrop, we expect MBs of agencies performing a direct mode of regulation to be subject to higher credibility pressures and, as a result, to provide more information compared with their counterparts. Hypothesis 5: MBs of agencies performing direct modes of regulation provide —compared to their informative and operational counterparts— more information on governance issues (H5a); policy performance (H5b); financial aspects (H5c); and their interactions with stakeholders (H5d) 4. Data and method This study is based on a novel database that includes statements related to MB decisions and actions on the four operational aspects examined here. These statements have been extracted from the annual activity reports published by 29 EAs during the period 2010-2016. The analysis covers all agencies in charge of quasi-regulatory tasks (BEREC, EMA, EFSA, ACER, EASA, ECHA, ESMA, ERA, EUIPO, CPVO, EBA and EIOPA), the totality of informative agencies (EEA, EIGE, EMCDDA, EU- OSHA, EUROFOUND, Cedefop, FRA, and ENISA), and nine out of the ten operational agencies (Frontex, ECDC, EFCA, EMSA, EASO, SatCen, EUROPOL, EUROJUST and EDA) (see Table A1 in Appendix). The analysis does not cover one operational agency, the EUISS, due to our lack of access to information on this body. Agencies dealing with management tasks (ETF, GSA, CdT, CEPOL and EU-LISA) have not been included either, as these are responsible for the management of particular programs or for the provision of specific services —e.g. the CdT provides language services. The study covers all published AARs for 29 EAs during a seven-year period (199 documents). A shorter period has been considered for ACER (2011-2016), EASO (2011-2016) and EBA (2012- 2016), as these more recently created agencies released reports for the first time in 2011 and 2012, respectively. By covering a seven-year period, we provide more comprehensive empirical evidence and avoid the potential bias that might occur by just observing a single year. Annual reports are a suitable source of information for our study (see also Busuioc & Rimkuté, 2019). Firstly, the publication of annual activity reports is compulsory for all agencies. This mandatory nature implies that the same type of empirical evidence can be systematically analysed and compared across agencies. As annual reports clearly reflect the obligation of agencies to inform about their activity, this source of information is a key instrument from the perspective of accountability. Secondly, while publishing annual activity reports is mandatory, the amount and type of information included in annual reports is the result of a voluntary decision by agencies. This allows for the identification of variations across agencies regarding their core activities and priorities. Third, as Busuioc & Rimkuté (2019) highlight annual reports are sources for mapping how agencies present themselves to a broad range of audiences, such as their political principals, societal actors and citizens in general.While EAs also produce other documents that could be useful for assessing their provision of information, for instance press releases, such sources of information may focus on the most topical issues and may not necessarily cover in a systematic fashion the four operational aspects deemed to be crucial in our study. We acknowledge that by focusing on annual reports there is a risk that certain aspects related to the information phase might be missed, for instance, information on activities conducted by agencies but not reported in annual reports. Nevertheless, annual reports allow us to compare the type of decisions and activities on which agency boards report and, therefore, are suitable for examining agencies’ informative choices. Although one might argue that the contents of annual reports are usually prepared by agency officials, it is worth noting that the institutional responsibility for such documents formally falls to MBs. The study has four dependent variables: the provision of information contained in each annual activity report about MB decisions and actions on governance issues (i), policy performance (ii), financial aspects (iii), and their interactions with stakeholders (iv). In order to operationalise the dependent variables, we proceeded as follows. In an initial step, we mapped the extent to which MBs were mentioned in each AAR. Mentions that did not specifically refer to MB decisions and actions were excluded. Then, we proceeded with the codification stage by identifying all statements reflecting MB decisions and actions. As a coding rule, an informative statement —our unit of analysis— is comprised of a sentence or groups of sentences that provide information on a given decision or action made or implemented by MBs. Once we defined our coding criteria, a research assistant was instructed on how to identify and codify the statements on the informative stage of accountability (see Table 1 for examples). The coder was supervised through the process, and any particular doubts were identified and resolved in a first sample of statements (20 percent). The coding of the totality of the identified statements was rechecked by the authors. Minor disagreements on a few statements were resolved and agreed on by the authors. As the value of the dependent variables might be influenced by how agencies present information in their annual reports, for instance, by splitting a certain activity into several actions —compared to describing it as a single action—, we checked that each mention was coded only once. While statements might not reproduce the entire content of a given decision or action, we are confident that they unambiguously capture the essence of such activity. It is also worth noting that our measurements focus on the quantity of statements, which leaves for further research the need to investigate qualitatively the content of the information examined here. The first dependent variable —provision of information on governance issues— is operationalised as the number of statements in each annual activity report about MB decisions and actions regarding institutional rules, codes of conduct, conflicts of interest, transparency and public access to documents, anti-fraud strategies, as well as internal control systems or audits. The second dependent variable —provision of information on policy performance— is operationalised as the number of statements about MB decisions and actions on the performance of agencies in their specific policy area. The third dependent variable —provision of information on financial issues— is operationalised as the number of statements in each annual activity report about MB decisions and actions on their budgetary activity. Finally, provision of information on stakeholder accountability includes statements about MB decisions and actions involving their relationships with interest groups (e.g. firms, NGOs, trade unions, business associations and consumer groups), such as those made in meetings and consultations (see Table 1 for examples). We identified 921 informative statements about MB decisions and actions for the seven-year period under study (see Table A2 in Appendix). Table 1. Examples of types of informative statements Governance information “(ECHA) Welcomes the transparency approach adopted by the Management Board (MB/61/2014), hereby also responding to the European Ombudsman’s request” (AAR ECHA 2014: 116). Policy performance information “[T]he Institute’s Management Board (...) agreed on a number of issues, in particular, on the procedure of developing the reports for the Presidency of the Council and approach to the concept of the Gender Equality Index” (AAR EIGE 2011: 5). Financial information “The Agency’s Financial Regulation was adopted by the Administrative Board in September 2011” (AAR ACER 2011: 35). Information about stakeholder interactions “The Board adopted a framework for interaction between EMA and industry associations” (AAR EMA 2015: 70). We operationalised the independent variables as follows. We used the index of formal accountability constructed by Font (2015a). The measure is based on the accountability provisions included in the founding regulations of EAs, which are coded according to whether they refer to one of the three phases of accountability: information, explanation and sanction. For example, the information component includes provisions establishing the obligation on the agency to inform the Commission, the Council and the European Parliament about its activities. The measure equals the mean value of the three components, and ranges from 0 (no formal accountability) to 1 (highest degree of formal accountability). Concerning the participation of scientists, a binary variable takes the value 1 when a scientific panel/committee or an advisory/expert forum was included in the organisational structure of an agency at the time it was created, and 0 otherwise. Only seven EAs —EMA, EEA, EFSA, ECDC, EIGE, EMCDDA, and FRA— include such a provision. To capture stakeholder representation, another binary variable takes the value 1 when the founding documents of an agency set out stakeholder representation in the decision-making bodies of agencies, namely management boards, advisory/expert forums and stakeholder groups. The variable takes the value 0 otherwise. Nine agencies ensure stakeholder representation in their MBs (Cedefop, EASA, ECHA, EFSA, EMA, EMSA, ERA, EUROFOUND and EU-OSHA), whereas two agencies do so in their expert forums (ECDC and EIGE); four agencies have permanent stakeholder groups (EBA, EIOPA, ENISA and ESMA). While stakeholders can participate through various formal and informal channels, the focus on formal arrangements is the most appropriate one, as these facilitate the direct and permanent involvement of stakeholders in agency decision-making. To operationalise agency task, we follow Majone (1997) and Busuioc (2013) and create a categorical variable that indicates whether the agency performs direct regulation (quasi-regulatory agencies), indirect regulation (informative agencies), and operational tasks (operational agencies). In a recent study, Rimkuté (2019) define regulatory tasks as being both quasi-regulatory and informative. However, we opt for an operationalisation distinguishing between agencies that merely provide information from those that perform, for instance, quasi-certification, sanctioning or inspecting activities. As we are interested in examining the role of agencies that have direct regulation, quasi- regulatory agencies are our reference category. As the reporting activity of agencies may be influenced by their size, we include agency budget for each year as a control variable. Budget size is controlled for to decrease the risk of unobserved heterogeneity, that is, to control for differences among agencies which are not included in the analyses. The budget for each year was taken from the Court of Auditors’ annual reports on EU agencies as well as other documents published in the websites of agencies. Table A3 in the Appendix summarises the descriptive statistics on the dependent, independent and control variables. 5. Analysis and results As our dependent variables are count variables and there was evidence of over-dispersion in our data, we perform negative binomial regression models. For each dependent variable, we ran a model including all our independent variables. A second model controlled for the annual budget and included fixed effects for years to capture variation over time in our dependent variable. Given that our data include observations per agency, standard errors were clustered at the level of agencies. We discard problems of collinearity as VIF coefficients for all variables are under 1.8 in all models (see Table 2). As expected, the results indicate that formal accountability mechanisms have positive and significant effects on the provision of information about governance issues (H1a) and policy performance (H1b), the effect of the latter being substantial. Contrary to our expectations, no association was found between formal accountability and the provision of information about financial issues (H1c) and their interactions with stakeholders (H1d). Based on models 1 and 3, we calculated the predicted changes of governance and policy mentions at different levels of formal accountability, respectively, keeping other variables constant. The predicted count of governance mentions moves from 0.5 to 1.3 at the minimum and maximum levels of formal accountability, respectively. Changes in the predicted count of policy performance mentions move from 0.06 at the minimum degree of formal accountability to 2.5 at the maximum one. By demonstrating that formal accountability only seems to influence policy performance and governance-related information, our findings show that its effects vary across aspects upon which agencies governing bodies are expected to provide information. This reinforces the idea that the relationship between formal and de facto accountability is not always straightforward (Busuioc, 2013). One potential explanation of this result is that, since agencies are non-elected bodies, those that were designed to be highly accountable may be particularly incentivized to be transparent on how the agency governs itself in order to improve their reputation and legitimacy among their political principals and audiences in general. Our results would also indicate that agencies subject to higher accountability demands may be highly incentivized to provide information on policy-related issues as a way to demonstrate optimal policy performance and dissipate concerns among certain audiences on the added value of agencies in the EU regulatory regime. Table 2. The results also indicate that agencies that have a more scientific profile —that is, those agencies that have scientific committees in their organizational structures— are more likely to provide higher number of informative statements about governance issues (H2a), policy performance (H2b) and their interactions with stakeholders (H2d). No significant effect of scientific bodies on the provision of financial information was found. Based on models 1, 3 and 7, respectively, we calculated the expected count of statements for governance, policy and stakeholder issues for agencies with scientific panels, holding continuous variables constant at their means and dichotomous variables at their modes. The expected count of governance mentions is 0.8 in the absence of scientific panels and moves to 1.3 when scientists have institutionalised representation in agency structures. The predicted count increases from 0.3 to 4.1 for policy performance and from 0.2 to 1.3 when it comes to interactions with stakeholders. The results are consistent with the argument that the institutionalised presence of scientific experts in the organizational structures of agencies could improve the circulation of information within the leading structures of agencies, and more specifically, the provision of information on agencies’ policy performance. In addition, since previous studies have also highlighted that scientific knowledge itself needs to achieve credibility (Jasanoff 2005, 211; Krick, 2018; Ossege, 2016), our results also provide evidence that the institutionalised involvement of scientific bodies could enhance the transparency of other aspects of agencies, such as their governance decisions and interactions with stakeholders. The results also indicate that agencies performing direct regulation (quasi-regulatory agencies) — compared with informative ones— provide more information both on governance and stakeholder issues, therefore providing partial support to hypotheses H5a and H5d. Based on model 1, the predicted count of governance mentions is 0.8 for agencies performing direct quasi-regulatory tasks and 0.4 for informative ones, holding continuous variables constant at their means and dichotomous variables at their modes. Based on model 5, the predicted count concerning mentions of interactions with stakeholders in quasi-regulatory agencies is 0.15, compared with 0.03 in the case of informative agencies. These results are consistent with the argument that, since agencies performing direct quasi- regulatory tasks are generally more challenged to enhance their credibility, their MBs have more incentives to provide information on governance rules and practices. These agencies also have a higher motivation to report on their interactions with stakeholders as a way of avoiding criticism about their lack of transparency or regulatory capture. Because of the far-reaching type of activity that regulatory agencies perform and because they typically make decisions that affect large industrial interests —e.g. pharmaceuticals— directly, they may be more exposed to higher pressures by their political principals and civil society actors, as opposed to informative agencies. If this holds, agencies performing quasi-regulatory tasks might have more incentives to report on critical governance and stakeholder issues —for instance those dealing with conflicts of interest— or on the type of relationships that the agency establishes with private interests. Contrary to our expectations, hypotheses on the effect of formal representation of stakeholders on any of the four dependent variables have not been validated. Finally, models controlling for agency budget show that budget size slightly increases the count of mentions on policy performance and barely decreases the count on financial aspects. While debatable, one possible interpretation of the latter counterintuitive result could be that small agencies provide more information on budget allocation and expenditure, as they have greater concerns with justifying their added value and ultimately their raison d'être vis-à-vis large and well-established agencies. 6. Conclusions This article contributes to the literature on accountability of EAs by providing an empirical analysis regarding information provision by their boards. To do this, we focus on four aspects upon which boards are expected to provide information: governance decisions, policy performance, financial aspects and interactions with stakeholders. Current research on agencification has emphasized that the intergovernmental and supranational logics in which EAs operate complicate their accountability relationships (Egeberg & Trondal, 2017: 9; Jordana & Triviño, 2019). Given that EAs have a wide array of accountability forums, some scholars have focused on agencies’ interactions with specific political principals. For example, on the basis of the existence of close ties between the European Commission and agencies, Egeberg and Trondal (2017) have investigated agency relations with this executive and administrative body of the EU. Other studies, such as Buess (2015a, 2015b) have provided important insights on accountability interactions between agencies and Member States. Our study contributes to previous studies by providing an assessment of the provision of information on MBs decisions and actions through the analysis of annual reports, a source that can be used to inform to a wide variety of audiences or accountability forums, such as stakeholders and political institutions. In this article we show that the provision of information is not only focused on technical aspects (e.g. financial procedures) —mainly targeted at specific forums—, but also, on issues dealing with other aspects that could attract the interest of a wide range of forums, such as their governance decisions. One interpretation is that MBs inform about different aspects of their activities to respond to the multiplicity of agencies’ goals and the variety of pressures of multiple masters (Olsen, 2013). This is also done to cultivate their reputation vis-à-vis (relevant) audiences (Busuioc & Rimkuté, 2019: 2, based on Carpenter, 2010). The most important contribution of this study is that it demonstrates that certain properties of agencies’ institutional design could influence their information provision. More specifically, we show that these institutional design properties matter in the provision of information unevenly depending on the type of aspect one which agencies inform. Our study shows that boards of agencies subject to more stringent accountability requirements are more likely to provide information on governance and policy operational aspects. Hence, by disentangling information provision into four operational aspects, our findings contribute to revealing the ways in which formal accountability design matters for an agency’s provision of information in post-delegation. This study also finds that agency boards whose founding regulations ensured a more scientific profile are more likely to deliver information about governance issues, policy performance and stakeholder interactions. Therefore, as institutionalised scientific bodies reinforce expert knowledge —which is assumed to improve the technical competences and performative capacities of agencies— EAs with permanent scientific bodies are more likely to be open about their policy performance. This finding also supports the interpretation that the formal inclusion of scientific expertise is correlated with the transparency of agencies’ governance decisions and interactions with stakeholders. The results also confirm the idea that EAs perform differently depending on the type of task that they undertake. Generally, the output of agencies performing quasi-regulatory tasks concerning information provision on governance and stakeholder aspects is higher than that of informative agencies. This supports the interpretation that the agencies that perform direct regulation are exposed to higher transparency demands. Our research also raises new questions with important normative implications. More specifically, it engages with the normative debate around whether enhanced transparency and accountability practices by non-majoritarian bodies would enhance the input- and output legitimacy deficit of the EU —and their institutions (Olsen, 2013). Since information practices cannot qualify as accountability practices, further analysis would benefit from the examination of the extent to which the performance of such non-elected bodies is subject to evaluation and sanctioning mechanisms. As this study focuses on information included in annual reports, analyses on other types of informative channels used to report about agency performance would further enrich the picture. Subsequent research should also address whether reporting practices indeed help improve the effectiveness and credibility of agencies towards stakeholders and citizens at large. References Behn, R.D. (2001). Rethinking Democratic Accountability. Washington, DC: Brookings Institution Press. Büttner, S.M., Leopold, L., Mau, S & Posvic, M. (2015) Professionalization in EU Policy-Making? The typology of the transnational field of EU affairs, European Societies, 17:4, 569-592 Borrás S., Koutalakis, C., & Wendler, F. (2007). European Agencies and Input Legitimacy: EFSA, EMeA and EPO in the Post-delegation Phase’, Journal of European Integration 29, 583–600. Bovens, M. (2007). Analysing and assessing accountability. A conceptual framework, European Law Journal 13(4): 447–468. Bovens, M. (2010). Two concepts of accountability: accountability as a virtue and as a mechanism. West European Politics 33(5): 946–967. Brandsma, G.J. & Schillemans, T. (2013). The Accountability Cube: Measuring Accountability, Journal of Public Administration Research and Theory, 23, 4, 953–975. Brandsma, G.J, Heidbreder, E., & Mastenbroek, E. (2016). Accountability in the post-Lisbon European Union. International Review of Administrative Sciences, 82(4), 621–637. Buess, M. (2015a). European Union agencies and their management boards: an assessment of accountability and demoi-cratic legitimacy, Journal of European Public Policy, 22:1, 94-111. Buess, M. (2015b). Accountable and Under Control? Explaining Governments' Selection of Management Board Representatives. JCMS: Journal of Common Market Studies, 53: 493– 508. Busuioc, M. (2013). European Agencies: Law and Practices of Accountability, Oxford: Oxford University Press. Busuioc, E.M., & Lodge, M. (2017). Reputation and Accountability Relationships: Managing Accountability Expectations through Reputation. Public Administration Review, 77, 91-100. Busuioc, M. & Rimkutė, D. (2019). Meeting expectations in the EU regulatory state? Regulatory communications amid conflicting institutional demands, Journal of European Public Policy, DOI: 10.1080/13501763.2019.1603248 Carpenter, D. (2010). Reputation and power: Organizational image and pharmaceutical regulation at the FDA, Princeton, NJ: Princeton University Press. Dubnick, M. (2005). Accountability and the promise of performance. In search of mechanisms. Public Performance & Management Review 28(3): 376-417. Egeberg, M., M. Martens & J. Trondal (2012). Building executive power at the European level: on the role of EU-level agencies. In: Busuioc, M., Groenleer, M., & Trondal, J. (2012). The agency phenomenon in the European Union: Emergence, institutionalisation and everyday decision-making. Manchester: Manchester University Press. Egeberg M. & Trondal, J. (2011). EU-level Agencies: New Executive Centre Formation or Vehicles for National Control? Journal of European Public Policy 18, 868–887. Egeberg, M., & Trondal, J. (2017). Researching European Union Agencies: What Have We Learnt (and Where Do We Go from Here)? JCMS: Journal of Common Market Studies, 55: 675–690. Font, N. (2015a). Designing Accountability Regimes at the European Union Level. In: Bianculli A, Jordana J, Fernández-Marín X (eds) Accountability and Regulatory Governance. Audiences, Controls and Responsibilities in the Politics of Regulation, pp. 123–142. Palgrave, London. Font, N. (2015b). Policy properties and political influence in post-delegation: the case of EU agencies. International Review of Administrative Sciences 81: 773–792. Georgakakis, D. & J. Rowell, (dir.) (2013). The Field of Eurocracy, Mapping EU Actors and Professionals, UK: Palgrave Macmillan. Georgakakis, D. (2017). European Civil Service in (Times of) Crisis A Political Sociology of the Changing Power of Eurocrats, London: Palgrave Macmillan. Grimmelikhuijsen, S. (2012). Linking transparency, knowledge and citizen trust in government: an experiment. International Review of Administrative Sciences, 78(1):50–73. Guidi, M. (2015). Modelling the Relationship Between Independence and Accountability of Regulatory Agencies. In: Bianculli A, Jordana J, Fernández-Marín X (eds). Accountability and Regulatory Governance. Audiences, Controls and Responsibilities in the Politics of Regulation, pp. 105–122. Palgrave, London Hood, C. (2010). Accountability and Transparency: Siamese Twins, Matching Parts, Awkward Couple? West European Politics, 33:5, 989-1009. Jasanoff, S. 2005. “Judgment under Siege: The Three-Body Problem of Expert Legitimacy.” In Democratization of Expertise? Exploring Novel Forms of Scientific Advice in Political Decision- Making, edited by S. Maasen and P. Weingart, 209–224. Dordrecht: Springer. Jasanoff, S. 2011. “Quality Control and Peer Review in Advisory Science.” In The Politics of Scientific Advice: Institutional Design for Quality Assurance, edited by J. Lentsch and P. Weingart, 19–35. Cambridge: Cambridge University Press. Jordana, J., Levi-Faur, D. & Fernández-i-Marín, X. (2011). The global diffusion of regulatory agencies: channels of transfer and stages of diffusion. Comparative Political Studies, 44: 1343– 69. Jordana, J, Triviño‐Salazar, JC. (2020). EU agencies' involvement in transboundary crisis response: Supporting efforts or leading coordination? Public Administration. 1– 15. https://doi.org/10.1111/padm.12652 Kelemen, D.R. (2002). The Politics of ‘Eurocratic’ Structure and the New European Agencies. West European Politics 25 (4), 93–118. Kelemen, D.R. & Tarrant A.D. (2011). The Political Foundations of the Eurocracy. West European Politics 34, 922–947. Koop, C. (2011). Explaining the accountability of independent agencies: The importance of political salience. Journal of Public Policy 31(2): 209–234. Krick, E. & Holst, C. (2019). The Socio-political Ties of Expert Bodies. How to Reconcile the Independence Requirement of Reliable Expertise and the Responsiveness Requirement of Democratic Governance. European Politics and Society 20 (1): 117–131 Lindberg, S. I. (2013). Mapping accountability: core concept and subtypes. International Review of Administrative Sciences, 79(2), 202–226. Lord, C. (2011). The European Parliament and the legitimation of agencification. Journal of European Public Policy 18(6):909–925. Majone, G. (1997). The New European Agencies: Regulation by Information. Journal of European Public Policy, 4, 262-275. McCubbins, M.D., Noll, R. & Weingast, B. (1987). Administrative procedures as instruments of political control. Journal of Law, Economics and Organization 3: 242 – 79. Meijer, A. (2013), Understanding the Complex Dynamics of Transparency. Public Administration Review, 73: 429-439. Mulgan, R (2000). “Accountability”, an ever-expanding concept?, Public Administration 78(3): 555- 574. Olsen, J.P. (2013). The Institutional Basis of Democratic Accountability. West European Politics, 36, 447-473. Ossege, C. (2015). Driven by Expertise and Insulation?: The Autonomy of European Regulatory Agencies. Politics and Governance, 3(1): 101–113. Pérez-Durán, I. (2019). Political and stakeholder’s ties in European Union agencies. Journal of European Public Policy. 26 (1): 1–22. Pérez-Durán, I. (2018). Interest group representation in the formal design of EU agencies. Regulation & Governance. 12, 238–262. Rimkuté, D. (2019). Building organizational reputation in the European regulatory state: An analysis of EU agencies' communications. Governance. 1– 22. https://doi.org/10.1111/gove.12438 Schillemans, T. & Busuioc, M. (2015). Predicting Public Sector Accountability: From Agency Drift to Forum Drift, Journal of Public Administration Research and Theory, 25(1): 191–215. Schout, A. & Pereyra, F. (2011). The Institutionalization of EU Agencies: Agencies as ‘Mini Commisssions’. Public Administration, 89(2): 418-432. Stigler, G.J. (1971) The Theory of Economic Regulation. The Bell Journal of Economics and Management Science 2, 3–21. Strøm, K. (2000). Delegation and Accountability in Parliamentary Democracies. European Journal of Political Research 37: 261–289. Wille, A. (2016). The evolving EU accountability landscape: moving to an ever denser union. International Review of Administrative Sciences, 82(4), 694–717. Wonka A, & Rittberger, B. (2010). Credibility, Complexity and Uncertainty: Explaining the Institutional Independence of 29 EU Agencies. West European Politics, 33: 730–752. Wood, M. (2018). Mapping EU agencies as political entrepreneurs. European Journal of Political Research, 57(2): 404-426.