Participation of Implementing Agencies in European Administrative Networks

European administrative networks are an important part of the functioning of the EU, mainly because of their role in bridging the gap between policy implementation knowledge and the policy-making activities of the European Commission. However, the extent to which implementing agencies actually participate in these networks, and which conditions explain network participation, have not yet been investigated. This article fills this gap by studying the conditions under which implementing agencies are expected to make more use of European administrative networks. Therefore, we turn to the participation of domestic implementing agencies in an administrative network during the common agricultural policy’s post-2013 reform. Relying on a fuzzy set qualitative comparative analysis, this case indicates that longer EU membership, in combination with a large staff and domestic implementation issues, contributes to more active participation of implementing agencies in administrative networks.


Introduction
In the EU there is a considerable distance between the formulation of policies in Brussels and their implementation domestically (Young, 2010, p. 64). European administrative networks, in which domestic implementing agencies participate, play an important function in overcoming this distance and help to close the gap between the formulation and implementation of policies (Mastenbroek, 2018;Trondal and Peters, 2013, p. 295). Although their actual impact remains a point of discussion (Kelemen and Tarrant, 2011), European administrative networks are often seen as tools for both domestic implementing agencies and the European Commission to improve member states' compliance with EU policies, and to increase their effectiveness (Hobolth and Martinsen, 2013). Via the exchange of (best) practices and processes of learning, these networks affect domestic administrative practices (Thatcher and Coen, 2008;Yesilkagit, 2011). Another function of networks is to provide an important channel for domestic implementing agencies to access the Commission, through which they may influence the agenda-setting and policy formulation stages of the EU policy process (Börzel andHeard-Lauréote, 2009, p. 141, Peterson andBomberg, 1999). The mobilization of implementation experiences through administrative networks allows the Commission to draw important lessons about the functioning of its policies (Trondal and Peters, 2013, p. 303) and can contribute to the formulation of new policy proposals (Hofmann, 2008, pp. 671, 674).
An important prerequisite for all these functions is that domestic implementing agencies actually make use of such networks to share their experiential knowledge. This means that in order to close the gap between the implementation and formulation of EU policies, and to make an impact, domestic agencies will have to participate in these networks. However, we lack insight into the extent to which domestic implementing agencies do so, and what explains this. Therefore, this article aims to explain the extent to which domestic implementing agencies participate in European administrative networks, by asking the question 'which conditions affect the participation of domestic implementing agencies in European administrative networks that function as a channel to the Commission?' In order to answer this question, we turn to the case of the common agricultural policy (CAP), an important policy field that has seen major changes over the past decades, amongst others, with regard to the policy instrument for farmers' income support (Daugbjerg, 2009;Garzon, 2006;Swinnen, 2008). In the run up to the post-2013 reform of the CAP, various implementing agencies tasked with administering this income support, called paying agencies (PAs), established the Learning Network for paying agencies and coordinating bodies. This network provided national PAs a platform to discuss their positions on the CAP and share their expertise with representatives of the European Commission, with the aim of influencing the Commission's reform agenda (Learning Network, 2012a).
Despite the Commission's interest in the input from this network, there were strong differences across member states in the extent to which their PAs participated during this reform process. Based upon literature on the policy process and studies of EU interest representation, we expected that these differences in participation could be explained in part by the motivations and resources of implementing agencies in the various member states. Through qualitative comparative analysis (QCA) we show how the role these conditions play in the participation of domestic implementing agencies in administrative networks. The analysis highlights the importance of exposure to European integration as a condition and resource for more active participation, in particular, in combination with other conditions, such as implementation issues, large staff size and low autonomy.
In particular, the study shows to what extent these conditions explain why PAs from some member states participate more actively than others. This helps us to understand why or when particular practical experiences from implementing agencies are likely to be mobilized and feed back into the Commission's policy agenda, although the findings of this specific case require, of course, further testing for other administrative networks. We turn to this in the conclusion. Studying the CAP is also empirically valuable, because scholars have not yet investigated these types of networks in the CAP (Mastenbroek and Martinsen, 2018). This is somewhat surprising, as the CAP is the EU's largest policy domain and still takes up around 40 per cent of the EU's budget. Therefore, this article also contributes to the literature of European administrative networks by extending its scope to a new policy field and highlighting the agenda-setting role of these networks. The study also adds to the literature on networks by providing insight into the functioning of these networks. If we want to grasp how these networks function we should not only focus on their impact (Van Boetzelaer and Princen, 2012), but also on which implementing agencies are more likely to play an active role in network activities.
Because of the lack of understanding about the participation of agencies in European administrative networks, the following section elaborates on the relevance of understanding the participation in, and use of such networks. Then, employing a mixed method approach, the article combines existing literature with semi-structured interviews with key actors in both the Learning Network and DG Agriculture and Rural Development (AGRI), to formulate meaningful expectations about the participation of PAs in the Learning Network. Subsequently, the QCA, which is used for further analysis, is introduced.

Participation in Administrative Networks
Most studies on European administrative networks pay attention to the establishment and functioning of networks (Mastenbroek and Martinsen, 2018). In this tradition, networks are studied mainly with respect to their role in the coordination and harmonization of implementation practices, or in providing the Commission with expertise (Trondal and Peters, 2013). Such studies often use functionalist approaches, theorizing that networks are established to deal with the complex interdependencies created by European integration, and to fill the gap between the formulation and execution of European policies (Busuioc, 2016;Trondal and Peters, 2013). However, while these functionalist approaches help to explain the demand for these networks they are less useful for understanding the way in which they operate in practice. Blauberger and Rittberger (2015) have argued that this is mainly a political process in which strategic considerations play a central role. In this study we follow this strategic perspective to understand network participation: we assumed that domestic implementing agencies are strategic actors that try to realize their preferences (Beyers et al., 2015;Sager et al., 2014) and do so in a strategically selective context that favours some strategies over others (Hay, 2002, p. 127). Administrative networks, from this perspective, can be seen as specific access channels (Coen, 2007;Zito, 2001) and participation as a strategy through which implementing agencies attempt to realize their preferences.
In this article we link this strategic perspective to EU studies on interest representation and multi-level governance, and various strands of policy process literature. In these studies the recurring question has been how actors mobilize their resources to obtain access to the policy-making process, with a chance of influencing EU policies (Coen and Richardson, 2009;Hooghe, 1995). Various actors have been studied in this light, such as lobby organizations and interest groups (Eising, 2008), social movements (Marks and McAdam, 1996), regions  and other subnational authorities (Jeffrey, 1997). While implementing agencies have not been studied as such, insights from these studies may be used to formulate expectations about the conditions under which domestic implementing agencies participate in an administrative network, as an access channel to the European Commission. Thus, the literature points to two main factors for understanding the mobilization of domestic actors through an administrative network: motivation and the right resources.
In the next section, hypotheses are presented about motivations and resources for the participation of domestic implementing agencies of the CAP in an administrative network. In addition to reviewing existing literature on interest representation, multi-level governance and approaches to the EU policy process, we also draw on the insights of several exploratory interviews with key actors from both the Learning Network and DG AGRI to fine-tune our hypotheses more specifically to PAs as implementing agencies in the CAP. Because it is both methodologically and practically impossible to account for all the factors that contribute to the participation of domestic implementing agencies in European administrative networks, this strategy was used to highlight the most important conditions related to the motivation and resources of PAs. In the method section the use of these exploratory interviews will be further explained.

Motivation Implementation Issues
According to scholars of policy feedback, motivation for mobilization may come from discontent with existing policies (either in scope or content) or with the direction of expected policy developments (Pierson, 1993). Similarly, theories of policy uploading suggest that the greater the costs of a policy for a member state, the more likely it is to push for changes (Börzel, 2002). This mechanism is very likely the same for implementing agencies; the more difficulties with implementation or the higher the (expected) costs, the more likely agencies are to mobilize their experiences and reach out to the European Commission. This mechanism is to be expected specially during the policy formulation stage and in a network that has the explicit goal of raising awareness about implementation issues with the Commission, such as the Learning Network (Interview 1). Following the logic discussed above, implementation issues are expected to contribute to a positive outcome, although the condition by itself is not necessary or sufficient. This leads to the following hypothesis: H1a Implementation issues contribute to the active participation of a member state's implementing agencies in the Learning Network.
The other way around, it is reasonable to expect that a lack of implementation issues contributes to less participation by implementing agencies in such a network: H1b A relative lack of implementation issues contributes to the absence of active participation of a member state's implementing agencies in the Learning Network.

Domestic Constraints: Farmer Representation
Aside from specific issues with the implementation of a policy programme, the institutional and political context in a member state offers additional incentives to make more active use of a European administrative network. The priorities of the national government are therefore important to consider. Actors will have a stronger incentive to explore other channels for interest representation when their national government is not responsive to their issues; for example, when other interests have the upper hand (Baumgartner and Jones, 2005). More direct channels of access to the EU often offer a way to escape such domestic constraints (Princen and Kerremans, 2008).
Generally speaking, the interests of implementing agencies and member state governments are likely to overlap, as governments want a good implementation track record, without too many additional costs. However, with regard to simplification and administrative burdens PAs may have to compete with farmers' interests, as fewer burdens for farmers usually imply more control issues for PAs (Interview 2). Farmer (employer) organizations are often found actively seeking access to the national government in EUrelated issues (Henning, 2009). As a result, national governments sometimes have to make a trade-off between prioritizing simplification for farmers, or for PAs. This means that a strong representation of farmers' interests in the national government would lead to more competition for PAs to have their concerns uploaded through their national government. As a consequence, this provides them with a strong incentive to look for other channels to mobilize their expertise, which brings us to the following set of hypotheses: H2a Strong representation of farmers' interests will contribute to the active participation of a member state's implementing agencies in the Learning Network.
The absence of strong representation of farmers' interests should be seen as one less incentive for PAs to use a direct access channel, because they have to compete less for the attention of their own national government. This may make participation in a transnational network organization superfluous: H2b Weak representation of farmers' interests will contribute to the absence of active participation of a member state's implementing agencies in the Learning Network.

Resources
Resources are defined as the capacity of actors to mobilize (Tatham, 2010). As most approaches to the policy process point out; for actors to affect policy-making they need the right resources to make use of an access channel, or actively participate in a network (Baumgartner and Jones, 1993;Kingdon, 1984). Of course, the technical expertise and experiences with policy implementation are perhaps the most important resources of implementing agencies, as this offers crucial information to the Commission about how its policies work (Bouwen, 2004;Polman, 2018). However, all implementing agencies should be considered as having valuable experiences for the Commission (Henning, 2009); so these experiences are unlikely to explain differences in mobilization. Other important resources for making use of networks or other access channels to the EU that emerge from the literature are: staff size, socialization, and autonomy. On the basis of existing research, it is hard to point out whether one of these resources is a necessary condition. There appears to be no clear hierarchy of relevance. However, it is expected that the presence of these resources, either by itself, or in combination, is associated with more active participation within the Learning Network. Conversely, the absence of these resources is associated with a negative outcome.

Staff Size
Firstly, a large organizational size is thought to be an important condition for participation. Larger organizations will have more staff and resources to allocate to lobbying activities (Tatham, 2010). Larger organizations may even have the luxury of being able to use multiple channels at the same time. Likewise, it is assumed that a large and professional staff can lobby effectively (Marks et al., 2002, p. 9). Finally, another advantage of a large staff size is that it offers the possibility to allocate resources within the agency to collect and process the practical experiences gained from the day-to-day activities of implementing a policy: H3a A large staff will contribute to the active participation of implementing agencies in the Learning Network.
H3b A small staff will contribute to the absence of active participation of implementing agencies in the Learning Network.

EU Membership Experience
A second important resource relates to experience with the EU policy process. As a consequence of EU membership administrations in the member states become accustomed to dealing with European policy-making over time (Kassim, 2003;. Implementing agencies in member states that have been part of the EU for a long period have had more time to adjust to European modes of policy-making and develop strategies for dealing with the Commission. This can be seen as a (social or strategic) learning process, in which repeated interactions lead to an improved understanding of good practices for interest representation (see Dunlop and Radaelli, 2013;Hay, 2002). Therefore, implementing agencies in old member states have greater experience with the EU policy-making process, and this is thought to provide them with insights in how to influence this process. In contrast, the experiences of administrations in new member states have largely consisted of a top-down integration process in which they have had very little influence on the integration process itself (Goetz, 2004). Therefore the tendency with regard to implementation in new member states is more to follow than to push for adjustments (Interview 3). Finally, and this is especially the case in the CAP, there is a greater chance that employees of domestic implementing agencies from old member states than the new ones have had previous working experience within the European Commission and vice versa (interviews 4, 5). A survey on the influence of actors in the EU-related networks of various domestic agencies corroborates these expectations, finding that the largest influence in such networks is attributed to agencies from EU-15 member states (Egeberg et al., 2015, p. 343): H4a Operating in an old member state will contribute to active participation of implementing agencies in the Learning Network.
H4b Operating in a new member state will contribute to the absence of active participation of implementing agencies in the Learning Network.

Autonomy
Thirdly, studies on subnational authorities point towards the importance of the organizational autonomy of agencies vis-à-vis their government. Tatham (2010) comes to the conclusion that autonomy from the central government is an important predictor for using direct representation channels to the EU. The rationale is that the more formal powers these authorities have, the more likely they are to act independently from their domestic government and seek their own involvement at the European level, bypassing the government's gatekeeping function (Hocking, 1999). Moreover, agencies at arms-length from their ministry have been associated with acting directly as agents to the Commission, again, effectively bypassing their member state government (Egeberg et al., 2015, p. 332). The autonomy of PAs also varies across member states, and following this line of thought we expect that implementing agencies with more formal autonomy from their government than those with less autonomy are more likely to seek direct access to the Commission: H5a High autonomy of implementing agencies vis-à-vis government in a member state will contribute to its active participation in the Learning Network.
H5b Low autonomy of implementing agencies vis-à-vis government in a member state will contribute to a lack of active participation in the Learning Network.
In the remainder of this article these hypotheses are tested through QCA by using data on the participation of implementing agencies over the period surrounding the CAP post-2013 reform.

I. Methodology
The Case of the CAP Post-2013 Reform The post-2013 CAP reform was the first such reform after the establishment of the Learning Network, which was founded in response to the increasing issues with implementation and administrative burdens. As other channels of interaction with the Commission were not perceived to be useful for influencing the Commission's policy agenda, the Learning Network offered a new platform for PAs to share their implementation experiences, with the goal of affecting the policy process (Polman, 2018). As the Commission and DG AGRI expressed a clear interest in the input of the Learning Network, and took the time to attend and host meetings with this network (European Commission, 2012;Learning Network, 2012a), it serves as a direct channel to the Commission, offering PAs a route for uploading their experiences and concerns.
Furthermore, the post-2013 reform gave PAs strong incentives to participate in the newly introduced Learning Network. One of the main elements of this reform was the introduction of new greening measures, which were additional requirements for farmers to make the system of direct payments more environment-friendly (European Commission, 2016). These new instruments, however, also demanded impactful adjustments on the PAs, who expected difficulties in implementing the greening measures in a timely and consistent manner. There were simultaneously some unresolved difficulties with the correct application of policy instruments from the previous reform (Learning Network, 2012b). These were strong reasons for PAs to participate in the Learning Network, with the aim of raising these issues with the Commission during the new round of policy formulation.
This combination of the successful establishment of a direct access channel to the Commission and the anticipation of the new reform made it a most likely scenario for PAs from all member states (N = 27) to participate actively in this network. 1 However, in practice there were great differences across member states in the extent to which they participated, and some did not even participate at all. Therefore the CAP reform post-2013 offers a good case for examining the extent to which a strategic perspective considering the motivation and resources of agencies is able to explain network participation.

II. Qualitative Comparative Analysis
The hypotheses was tested using QCA, a method based on Boolean logic. This methodology is suitable for gaining insights into broad patterns of representation, or in this case, participation (Spierings and Jacobs, 2018). Theoretically, different combinations of conditions may help to explain the active participation of PAs in different member states; an assumption often referred to as complex causality. Thus, many non-exclusive explanations of an outcome can coexist. QCA offers the most systematic way of analysing these types of relations (Legewie, 2013;Schneider and Wagemann, 2007). Moreover, the negative outcome; non-participation, may be explained by different paths that are not necessarily the negation of those associated with the positive outcome. Therefore it is common practice in QCA to run separate analyses for both the positive and negative outcome. Moreover, QCA has been frequently described as a suitable approach for medium-sized research groups or samples, ideally for 10-50 cases (Rihoux and Ragin, 2009;Schneider and Wagemann, 2012).
The specific fuzzy set approach to QCA employed in this article treats the different conditions as sets of which cases have different degrees of membership. Qualitative anchors are set to determine whether the different cases are a full member of a set (≥ 0.95), completely out of it (≤ 0.05), indifferent to it (0.5), or somewhere in between. For this purpose, the direct method of calibration, as described by Ragin (2008) was used.
The QCA produces complex combinations of conditions, or configurations, each of which can be seen as a sufficient solution for explaining the outcome (written in uppercase), or its negation (lower-case). The explanatory power and empirical relevance of a sufficient solution are evaluated through measures of consistency (to what extent the solution confirms the statements of necessity and sufficiency), and coverage (to what extent the presented solution explains the actual empirical cases). For the analysis I have used the software packages QCA and SetMethods in R (Duşa, 2007;Medzihorsky et al., 2018). In the next section the measurement of the outcome and conditions are briefly discussed. A more detailed explanation of the measurement and calibration process, and the raw and calibrated data can be found in the Supporting Information online (1-6).

Operationalization and Calibration
Below, different conditions are operationalized. These conditions are based on the literature, with additional support from five exploratory interviews. The interviewees were selected primarily on basis of their position in DG AGRI (a deputy director in frequent contact with Learning Network, and a representative of the unit responsible for the new greening instruments) and their role in the Learning Network (a former chair and former secretary). In addition, an interview with a senior adviser of the Dutch paying agency was held.

Participation in Learning Network (P)
The outcome variable is measured by looking at the participation in Learning Network general meetings by member state PA representatives between October 2011 and March 2014. The starting point corresponds with the release of the Commission's first proposal for the reform. From this point forward, PAs had a better understanding of which implementation issues had to be addressed during the reform process. Also, at this time, the Learning Network had already established an informal channel to the Commission, as the Commission had accepted their invitation to participate in regular informal meetings to discuss the agenda for reform. The end date was set because at that point the Commission presented the implementing and delegated acts related to the reform, effectively finalizing the CAP reform post-2013. Over this period, the Learning Network organized nine general meetings in which all member state PAs could participate and provide their input. Moreover, PAs had the option to join working groups on specific issues, which are a more systematic and specialized form of participation in the network. In order to measure active participation we used a combination of presence at meetings, input during these meetings and participation in working groups.

Implementation Issues (I)
The degree of implementation issues with the CAP is measured by looking at the CAP expenditure made by PAs in member states that was reclaimed by the Commission over the period between and 2013 (2014 was not included, as this procedure takes place only twice or thrice a year, and not before March 2014). These reclaimed payments provide an indication for the difficulties that member states' PAs have with the existing requirements of policy instruments of the CAP.

Competition from Farmer Interests (C)
Competition from farmer interests is measured by looking at the density of farmer employers' organizations in the year closest to 2011, based on the institutional characteristics of trade unions, wage setting, state intervention and social pacts database. Although it is hard to find a good measure for the influence of employers in a national government, density is regarded as a reliable proxy (Brandl and Lehr, 2016).

Large Staff (S)
Staff size is a measure often used for measuring organizational resources (Tatham, 2010). For most member states staff size was readily available through annual reports or activity reports, or even reported on the website of the PA. Data closest to 2011 were used (see table S1 in the Supporting Information for a complete overview).

EU Membership Experience (M)
Different scores were attributed on basis of the accession dates of the various member states.

Autonomy (A)
There is a vast body of literature on the autonomy of government agencies. What stands out is that there is not one clear way of looking at agency autonomy (Van Thiel et al., 2012). For this analysis the concept of formal autonomy will be used. For measuring formal autonomy, the various PAs are studied in terms of having a managing board that is separate from the parent ministry (Yesilkagit and Christensen, 2009). In these cases agencies are likely to have a level of discretion that allows them to mobilize themselves directly through administrative networks.

III. Results
So which conditions and configurations contribute to active participation in the Learning Network? As a first step of QCA it is common to run analyses for necessity, in order to check whether any individual necessary condition is present in the sets (Thomann, 2015). None of the conditions scores above, or around, the 0.9 consistency threshold that is commonly used to identify single necessary conditions (Legewie, 2013;Skaaning, 2011). This means that we found no evidence for a single condition to be necessary for implementing agencies to seek direct access to the European Commission. In addition, a sufficiency test was executed, in which no single condition was found to be sufficient by itself, for explaining active participation in the Learning Network (analyses of necessity and sufficiency can be found in Supporting Information S7 and S8). These findings fit our general expectation that the presence of the conditions in the analyses contributed to a positive outcome, although a single condition was not expected to be necessary or sufficient by itself.
As a second step, we looked at the different configurations that are sufficient for observing outcome P (based on all truth table rows with a consistency > 0.8: for further motivation, see the robustness checks in Supporting Information S14 S15 S16). 2 These configurations are presented in the sufficient solution in Table 1. Here we see three configurations associated with the outcome set. Overall, the results show one sufficient configuration for the active use of the Learning Network in which both motivation and resources play a role (I*M). In the other configurations only resource conditions were included. The findings suggest that to explain the active use of administrative networks by implementing agencies, resources may be more important than specific motivations.
When zooming in on the different configurations, the first path indicates that implementation issues in combination with a long membership of the EU are sufficient for observing the use of the Learning Network in eight cases. This configuration confirms the effects expected in H1a and H4a. Long EU membership proves to be an important element of all configurations in the parsimonious solution. The second configuration adds staff size into the mix. A large staff size was also found to contribute to the outcome, in line with the third hypothesis (H3a). However, there was one unexpected finding, which is that non-autonomy contributes to the outcome (a*M). This means that PAs in old member states without a board are associated with active participation in the Learning Network, while the hypothesized relation was that a lack of autonomy contributed to their non-use. This surprising result can be interpreted in two ways. Firstly, it may suggest that governments may push agencies to make active use of these networks, or even view administrative networks as additional channels for policy uploading, which disproves the fifth hypothesis. Secondly, this specific configuration may indicate that in old member states agencies may use participation in administrative networks to increase their autonomy and legitimacy towards their own government (see Verhoest, 2018, p. 336). The cases uniquely covered by this path are Belgium, Finland, and Luxembourg. Furthermore, a strong farmer representation did not show up in any of the configurations, thus there is no clear evidence supporting H2a.
The coverage of the solution formula indicates that 77.1 per cent of the outcome set is explained by these configurations. The solution consistency shows that 93.0 per cent of the outcome set is correctly explained the sufficient configurations.
In Table 2 the sufficient solution for non-use of the Learning Network is presented. Here two configurations come forward. Firstly, new member states with a strong farm lobby (m*C) are associated with non-use. The contribution of a strong farm lobby to a path associated with the negative outcome is also in contradiction to the expectations raised in H2a and H2b. This means that in new member states a strong farm lobby actually hampers the active participation of PAs. This finding may imply that a strong farm lobby directly affects the behaviour of PAs. The cases uniquely covered by this path are Malta, Slovenia and Romania. Secondly, autonomy is now associated with non-use, indicating that agencies that are more autonomous are less likely to make use of administrative networks when combined with few implementation issues and a small staff size (i*A*s). This finding is consistent with the interpretation of autonomy; namely, that governments actually play a role in pushing agencies to participate in administrative networks, and that autonomous agencies have to figure this out for themselves. Furthermore, the other conditions in the path (few implementation issues and small staff size) are in line with the hypothesized effects in H1b and H3b.
The coverage of the solution of 0.647, however, is relatively low, which means that only 64.7 per cent of the outcome set is explained by these solutions, which makes the explanatory power of the solution formula relatively weak.
It must also be noted that in the truth table (Supporting Information S9) there are many contradictory rows with so-called deviant cases consistency or true logical contradictions. Because these rows all score low on consistency (< 0.8) they have not been included in the minimization process leading to the solutions presented above. However this does mean that four cases that are part of the positive outcome set (Estonia, Latvia, Lithuania and Hungary) are not explained. For the negative outcome set, three cases are not explained because they are in rows with low consistency scores and contradictory cases (Slovakia, the Czech Republic and Poland).
The most surprising case here is Estonia, which scores < 0.5 on all conditions, but is seen as an active participant, unlike the other country in this row, Slovakia. The other contradictory cases are all in a row that has only membership in the large staff set. This part of the findings indicates that there must be other relevant conditions that have not yet been included in the analysis.

Conclusion and Discussion
In this article we have taken a close look at the participation of implementing agencies in an administrative network in the CAP, with the particular aim of explaining why implementing agencies from some member states participate more actively than others. Therefore, this article has brought together theories of European administrative networks and interest representation. Based on a strategic perspective, the expectation was that a combination of specific motivations and resources would contribute to a more participation in these networks. The case of the Learning Network has allowed us to identify three paths associated with active participation in this network. In line with the expectations, long membership of the EU, in combination with either a high level of implementation issues or a large staff size, are associated with active network participation. These conditions are neither necessary nor sufficient by themselves, but in combination they contribute to active participation.
These results indicate that the strategic perspective presented in the article helps to shed light on the conditions affecting the participation of domestic implementing agencies in European administrative networks. The third path for active participation, however, deviated from our expectations. Autonomy was found to have a reverse effect: the absence of formal autonomy contributes to participation, while the presence of autonomy is part of a path leading to non-participation. We also found a reverse contribution of strong farmers' interest representation, which was associated with non-participation, while we expected this to contribute to more active participation. With regard to the unexpected findings, uniquely covered cases may serve as input for studies on the effect of the (farm) lobby and autonomy on network participation.
From a strategic perspective, these findings may indicate that, at least in some cases, active network participation is not born out of a lack of domestic opportunities but should be seen as a strategy that overlaps the implementing agency and the national government. In order to understand this mechanism it will be interesting to see whether other, informal types of autonomy also show a similar contribution to the network participation of implementing agencies.
The most consistent finding is that long EU membership is part of all the paths explaining a positive outcome, in combination with other conditions. This emphasizes the importance of practical experience with the implementation of EU policies and dealing with Brussels, for understanding network participation. However, the importance of this factor for explaining participation also brings in a normative concern: it means that in the experiences shared through the Learning Network there may be a strong bias toward old member states and an underexposure of the experiences and issues faced by new member states (see Mastenbroek and Martinsen, 2018, p. 429). Additional research may also focus on this selectivity in policy feedback.
The cases uncovered by the solution paths need to be examined further. In particular, the motivations for PAs in new member states that have participated actively in the Learning Network deserve further examination. Moreover, looking at the coverage and consistency scores, the conditions in the analysis appear to be better at explaining active participation than non-participation. Therefore, as a next step, we need to look at conditions aimed more specifically at explaining non-participation.
To do so our strategy might be complemented by other perspectives. Sociological institutionalist explanations, which have so far been little used in the literature on administrative networks, may add additional insight (Mastenbroek and Martinsen, 2018). Further research should focus on whether the findings from this article hold for other administrative networks. Moreover, the article has shown that administrative networks need not focus only on coordination, harmonization or enforcement: they can also provide the Commission with policy feedback and adopt an agenda-setting role. The Learning Network is a good case to study which conditions contribute to participation within such networks, but other studies are needed to find out whether the findings hold for similar (agenda-setting) networks, and the extent to which our findings apply to more traditional administrative networks. Finally, future research may explore new directions related to network participation, for example, whether participation affects the impact of administrative networks in the EU or how participation varies over time. For the latter, other methods, such as network analysis, should be considered.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.  Table S6. Calibrated Data Matrix S7 Analysisof Necessity of Single Conditions S8 Analysis of Sufficiency of Single Conditions S9 TruthTables S10 QCA Solution Types S11Simplifying Assumptions S12 Conservative and Intermediate Solutions S13 Plots of Parsimonious Solutions S14 Robustness Check I: Alternative Calibrations S15 Robustness Check II: Alternative Consistency Levels S16 Robustness Check III: Use of Alternative Variables S17 Additional Literature for Supporting Information