Steering or Networking: The Impact of Europe 2020 on Regional Governance Structures

This article probes into how regions organize themselves to deal effectively with the Europe 2020 reform program. More specifically, it maps governance structures of regional policy-making and implementation of Europe 2020 and explains variation in these structures between policy domains and policy stages. The empirical focus is Flanders as this Belgian region possesses substantial legislative and executive autonomy and is therefore highly affected by the Europe 2020 program. The article distinguishes between policy-making (upload) and implementation stages (download) in education, energy and poverty policies. It is hypothesized that the varying impact of Europe 2020 can be attributed to the varying adaptational pressure of EU programs and a set of domestic intervening factors. Findings indicate variation between policy domains and policy stages on a continuum from lead-organization governed networks to shared participant governance networks. Overall, the extent to which Flanders is competent seems to be crucial. In addition, a substantial administrative capacity is needed to firmly steer and coordinate the governance structures that manage Europe 2020 policies. The level of integration further increases the extent to which Flemish Europe 2020 policies are steered.


Introduction
Regions, Member States and the European Union (EU) itself are all confronted with major social, economic and budgetary challenges.The sovereign debt crisis, the euro-zone upheaval and the stagnation of the national and regional economies have been dominating the European and domestic agendas since 2008.The EU has implemented several policies to cope with these challenges.One of these tools is the large-scale Europe 2020 reform program whereby the EU aims for high levels of employment, productivity and social cohesion through smart, sustainable and inclusive growth.Europe 2020 is not implemented by legislation but by the coordination of national policies, inspired by the Open Method of Coordination (OMC) of the earlier Lisbon Strategy.The OMC is "in line with the principle of subsidiarity in which the union, the Member States, the regional and local levels, as well as social partners and civil society, will be actively involved, using variable forms of partnership" (European Council, 2000, para. 38).Its architecture is based on soft law and its main EU goal is to disseminate best practices in order to achieve greater convergence among member states and regions (Tucker, 2003).
Unlike its predecessor, the Europe 2020 program is expected to generate substantial effects as it is incorporated in the framework of the European Semester.Through this European Semester, the Commission is empowered to monitor the economic and budgetary policies of Member States and to take action when agreed targets are not reached.While Europe 2020 policies are predominantly decided by European level institutions, Member States play a major role during the policy preparation and policy implementation stages.In federal and highly decentralized member states, both national and regional authorities are challenged by Europe 2020 policies.Via intra-state as well as extra-state channels (Jeffery, 2000) regions are involved in the policy formulation and implementation of Europe 2020 policy measures whenever the latter touch upon their competences.This paper probes into how the regional level organizes itself to deal effectively with Europe 2020 policy-making and implementation.As regions can be involved in several stages and policy domains we expect variation in the way they deal with Europe 2020.
Based on findings from the Europeanization literature (Bursens, 2012), we expect a differential impact of Europe 2020 due to domestic intervening variables.Our core research question is therefore how we can explain the variation in regional governance structures established in response to the Europe 2020 program.More specifically, this paper aims to map the governance structures of regional policy-making and implementation of Europe 2020 and to explain variation in these governance structures between policy domains and policy stages.We distinguish between the policymaking stage (upload) and implementation stage (download) in education, energy and poverty policies.Our empirical focus is Flanders as this Belgian region possesses substantial legislative and executive autonomy in the selected policy fields and is therefore highly affected by the Europe 2020 program.
The paper is structured as follows.Firstly, it discusses the relevant literature on Europeanization and governance in order to conceptualize the way regional governance structures are steered and to identify the variables that translate the adaptational pressure put on the regional governance structures by the Europe 2020 program.Next, the cases are presented.In the third part we operationalize and measure the variables.The final part maps the Flemish governance structures in response to the Europe 2020 program and accounts for the variation between the selected policy domains and policy stages.

Europeanization of Regional Governance Structures
Research on the adaptation of the regional level to European integration is relatively recent and has not delivered univocal conclusions (Graziano & Vink, 2008).In a literature review, Bursens concluded that "all empirical findings reveal some impact of European integration on the regional level, but there is no agreement on the intensity or the direction of the impact."(Bursens, 2012, pp. 400-401).The literature suggests that regional authorities implement EU policies in diverse ways contingent to the varying national contexts (Borghetto & Franchino, 2010;Sturm & Dieringer, 2005).Furthermore, the focus has mainly been on explaining variation between EU policy coordination mechanisms (Kassim, Peters, & Wright, 2000;Wessels, Maurer, & Mittag, 2003;Zeff & Pirro, 2006), the integration of the OMC in domestic policymaking arenas (Macphail, 2010;Weishaupt, 2009;Zeitlin, 2009) and regional actors' preferences and strategies (Bache, 2008;Dyson & Goetz, 2003).This paper builds upon the Europeanization literature but shifts the focus from comparing regions towards comparing policy domains and policy stages.In addition, by looking at regional governance structures in response to Europe 2020, it adds to the analyses of regional implementation of EU legislation.
The extent to which the regional level responds to European integration depends on the adaptational pressure, or how well European and regional policies fit together (Börzel & Risse, 2000).The introduction of the OMC by the Lisbon Strategy aimed at bringing European integration in line with the principle of subsidiarity.After the mid-term review, the focus of the Lisbon Strategy shifted towards the goals of competitiveness, growth and jobs (Zeitlin, 2009) since the EU was not able to urge member states to participate more intensely in the OMC procedures for social inclusion and sustainable development.As the Europe 2020 program also functions along the lines of the open method of coordination, the extent to which regional authorities are confronted with adaptational pressure is reflected by the EU's ability to push forward on the Europe 2020 objectives.A high adaptational pressure constitutes a trigger for regions to adapt governance structures and policies.To explain how regional authorities respond to the Europe 2020 program, the Europeanization literature points to a range of intervening factors that facilitate or constrain the adaptation process.In this paper we borrow these intervening factors from rational and sociological institutionalism.

Mapping Governance Structures
A necessary step to address our research question is to describe how regions deal with Europe 2020 in different policy domains and policy stages, by mapping how political and administrative actors as well as societal organizations are embedded in governance structures.Newman (2001) describes governance as a mechanism for solving common problems on various levels.More specifically, governance structures are about coordination mechanisms that settle decision-making and im-plementation between actors by allocating tasks and resources among those actors (Carver, 2000;Kooiman, 1993;Lowndes & Skelcher, 1998).According to Pierre (2000), "governance refers to sustaining coordination and coherence among a wide variety of actors with different purposes and objectives".The performance and effectiveness of governance structures mainly depends on how they are steered (Provan & Milward, 1995).Over time coordination mechanisms have been gradually transformed from traditional governmental steering, characterized by hierarchical, direct top-down control towards more autonomy and self-responsibility for administrative actors and a stronger involvement of societal organizations (Kickert, 2005;Nelissen, 2002).This implies a tendency to manage or steer actors and processes rather than controlling them in a top-down fashion (Borgason & Musso, 2006).Such more horizontal and mixed public/private policy networks are argued to increase the problem-solving capacity of governmental action as they combine the expertise and means of both state and non-state actors (Wolf, 2001).
However, steering instead of controlling actors may still take place under the shadow of hierarchy (Börzel & Risse, 2005), since some organizations may have advantages over others to assign tasks and goals, sometimes assisted by (financial) means.There is, in other words, a wide variety in policy networks.One useful classification is provided by Provan and Kenis (2008).Based on the involvement of actors and the decisionmaking procedures within mixed networks, Kenis and Provan identify two ideal types of networks with shared participant governance networks and leadorganization governed networks at the far ends.Whereas shared participant governance networks are characterized by collective decision-making procedures based on unanimity and a high degree of involvement of all actors thereby resulting in high density networks, lead-organization governed networks function in the opposite way by allocating decision-making powers to one or a few actors who thereby obtain a central position in the network.Although both ideal types of networks increase the problem-solving capacity of governmental regulation, empirical research (Brower & Choi, 2006;Creech, Huppé, & Knoblauch, 2012;Kenis & Provan, 2006;Provan & Milward, 1995) suggests that networks that are steered by just one or a few actors are more effective in reaching their goals.This paper doesn't aim to explain the effectiveness of governance structures, but seeks to understand why specific governance structures are installed, the latter being considered as one of the variables that can explain effectiveness.We use social network analysis to operationalize the network types suggested by Kenis and Provan (2006) in the context of Europe 2020 policies.By calculating the network's density and the actors' degree of centralization in the network while accounting for their competences to initiate and coordinate policy initiatives, we define the governance structures as either shared participant governance networks or lead-organization governed networks.

Explaining Governance Structures
We now turn to the factors that can account for variation between governance structures when adapting to Europe 2020 policies.First, adaptational pressure is exerted via the OMC procedures of Europe 2020.A high adaptational pressure triggers regions to adapt governance structures and policies in order to comply with Europe 2020.This adaptational pressure is the independent variable in our model.Next, we expect that variation in Europe 2020 governance can be explained by how domestic intervening variables tap into this pressure.We look at the domestic division of competences and the regional administrative capacity to deal with Europe 2020 (derived from rational choice institutionalist theories) and the political and administrative support for Europe 2020 policies (taken from sociological institutionalism).
As EU competences have broadened and deepened over time, the EU's influence on national politics and policies increased (Birk, Gos, Haas, & Tadini, 2010).So we argue that the extent to which adaptational pressure is exerted via Europe 2020 mainly depends on the level of EU competences and thus differs across policy domains.According to Pfetsch (2004) a higher degree of integration is likely for policy domains in which the EU has acquired significant regulatory powers.We expect that the extent to which the EU is able to put pressure on regions to act on the European growth strategy is contingent on the degree of legislative powers and the amount of regulatory measures.The absence of such powers undermines EU undertakings to reach the Europe 2020 targets.The European Commission is aware of its powers and has drawn lessons from the failed Lisbon Strategy (Schoukens, 2014).Attempting to overcome its lack of competences in certain policy areas, the Commission increased the entanglement of the Europe 2020 objectives and strategies.However, despite the Commission's efforts to fuse policy issues from different policy areas (such as education and employment), its capacity to guide or even urge regions to act on Europe 2020 continues to depend on its legislative and regulatory powers.In case of only modest competences the EU will find it hard to stimulate regions to establish strong governance structures aimed at coordinating Europe 2020 policies.On the other hand, if the EU is able to play out its competences, regions and particularly regional governments will be more inclined to act on Europe 2020 by coordinating policy initiatives and actors.Hence our first hypothesis: the more integrated the policy domain in the EU sphere, the more the regional governance structure will be organized according to the lead-organization governed network.
Although the adaptational pressure triggers regions to adapt governance structures, domestic intervening factors are expected to differentiate the impact.In federal systems, legislative as well as executive competences are allocated at various levels (De Vicq, Van Hecke & Buyst, 2014).This constitutional setting functions as an opportunity structure in which domestic actors' behavior is shaped.The division of competences between governmental levels includes the allocation of means and the authority to initiate policies (Provan & Kenis, 2007) and therefore also the responsibility to assure a performant governance structure.When a regional government has competences in a certain policy domain, it has the capacity to coordinate the governance structure and to determine the policy content.According to Saunders (2006), decentralization indeed strengthens the capacity of the sub-national constituent units.This is even moreso the case when regions are granted the competence to conduct foreign relations with respect to their competences.Being competent for a wide range of policies also affects the quantity and complexity of policy issues.A high number and high variety of policy measures and involved actors requires high levels of coordination.We expect strong coordination in those policy domains for regions which are strongly competent (Provan & Kenis, 2008).Hence our second hypothesis based on the extent to which a region is competent: The stronger the regional competences in a policy domain, the more the regional governance structures will be organized according to a lead-organization governed network.
Besides the institutional environment of governance structures, we also look at the actual efforts actors put into influencing and implementing Europe 2020 policies.The EU is a multilevel system requiring the establishment and management of coordination capacities.This administrative capacity of domestic actors is equally a part of the opportunity structure in which domestic actors try to maximize their preferences: it is considered as a facilitating factor for the successful implementation of Europe 2020 policies.We expect that the ability of regions to deal with European policies also depends on the administrative capacity to upload and download EU policies (Börzel, 2002).Cadri (2014) defines administrative capacity as "the process through which individuals, organizations and societies obtain, strengthen and maintain the capabilities to set and achieve their development objectives over time".More specifically related to Europeanization, Börzel and Risse (2000) and Deforche and Bursens (2012) provide a more specific approach considering only the administrative capacity that deals with European issues.Administrative capacity is highly relevant in this context as research has pointed out that most instances of non-compliance with international agreements are due to a lack of capacity (Jacobson & Weiss, 1998;Perkins & Neumayer, 2007).Being able to mobilize ca-pacity provides regions with steering capabilities in particular governance structures (Milio, 2007).We expect that the specific role of the minister's office or the department, which includes high-level policy-making and planning tasks, is strengthened by a high degree of administrative capacity.Our third hypothesis therefore puts that the more administrative capacity devoted to Europe 2020, the more the governance structure will resemble a lead-organization governed network.
Finally, we consider the actors operating in the structures.How do administrative and political actors react to the Europe 2020 program?Do they support all Europe 2020 policies to the same extent?We expect that the level of support affects the efforts regional actors invest in uploading and downloading Europe 2020 policies.The intervening variable support thus departs from the logic of appropriateness: domestic actors deal with European 2020 as they see fit with their position towards the program.According to Sorensen and Torfing (2005) networks establish a frame of mutual interest for consensus building among various stakeholders.They argue that when actors are strongly involved, they are more likely to be supportive.Due to that mutual interest, supportive actors have a similar focus on what is to be done, but more importantly also on how things should be done.Furthermore, widespread support facilitates cooperation and exchange of information within a governance structure.In shared participant governed networks, the performance and effectiveness depends to a great extent on the consensus between the actors.The more a network lacks consensus, or the less supportive actors are, the higher the need for a lead-organization in the network in order to assure the performance and effectiveness (Van Oorschot, 2015).Therefore we expect weak steering mechanisms in cases of high support.Hypothesis 4 stipulates that the more supportive actors are, the more the governance structure will be structured according to a shared participant governance network.
To conclude, previous research has pointed to the advantages of lead-organization governed networks over shared participant governance networks in terms of performance and effectiveness.Hence, the way regional governance structures are steered is crucial for achieving Europe 2020 policy objectives.We expect that regions install varying structures in different domains and suggest four hypotheses to account for this variation.In the next sections we present and motivate our empirical cases before turning to the operationalization and analysis.

Case-Selection
Europe 2020 focuses on smart, sustainable and inclusive growth.These priorities break down into 10 integrated guidelines which serve as themes in the national reform programs and trigger specific policy initiatives.
We selected one integrated guideline for each priority, taking into account the Belgian division of competences.For smart growth we selected reducing the rates of early school leaving below 10%, from the policy field of education which is a quasi-exclusive regional competence; in the area of sustainable growth we opted for 20% of energy from renewables, energy policy being a mixed federal/regional competence; regarding inclusive growth 20 million (on EU level) fewer people in or at risk of poverty and social exclusion was selected as poverty policies predominantly belong to the federal level.
From the perspective of regions, one can distinguish between four different policy stages in the context of Europe 2020: participation in the Open Method of Coordination (OMC) processes of Europe 2020, the draft of the Regional (RNP) and National Reform Programs (NRP), the implementation of the NRP and the follow-up and feedback on the national level.In order to reduce complexity, we simplified these to two stages: (1) the European semester, i.e. participation in the OMC and the drafting of the Reform Programs, which is a process mainly oriented at the European level and (2) the national semester, i.e. the implementation and the follow-up of the agreed reform program which takes place at the (sub)national level.This leaves us with in total six cases, as listed in Table 1.

Operationalization and Measurement of the Dependent and Independent Variables
To map the governance structures as a whole and to position the actors within those structures we used Social Network Analysis (SNA).SNA captures the complexity of social relationships by analyzing numerical data and visualizing the set of actors involved as well as the relations between those actors (Hawea & Ghali, 2008).A social network can be defined as "a specific set of linkages among a defined set of persons with the additional property that the characteristics of these linkages as a whole may be used to interpret the social behavior of the persons involved."(Mitchell, 1969).We conducted 31 interviews with political and administrative actors as well as societal organizations from September 2013 till January 2014 (appendix-list of respondents).We presented the respondents a list of actors that may have been active with respect to one of the selected Europe 2020 integrated guidelines.Respondents were asked to indicate the frequency and direction of the contacts and to qualify the relation as information transfer or substantial cooperation.Moreover, respondents were given the opportunity to add actors to the list which were then included in the following interviews.The extent to which data is missing is crucial for SNA as it has an impact on the outcome of SNA-indicators.The impact of missing data differs among SNA-indicators, meaning that some indicators are more robust than others.The indicators centrality degree and density are very robust measures in SNA (Wangh, Shi, Mcfarland, & Leskovec, 2012).The robustness of degree centrality and density holds when 80% of the data is taken up (Costenbader & Valente, 2003), meaning that the output on those indicators will not change significantly by adding more data.Although not all involved actors could be interviewed, the percentage of missing data is sufficiently low to allow for a valid interpretation of the network data.(69%, 55% and 69% response rate and 8.3%, 19% and 8,3% missing data in the education, energy and poverty case respectively).The robustness of our network data is further strengthened as we have data on all the spillactors of the governance structures.Furthermore, the network data have been double-checked with the core members of the network.
In order to determine to what extent a governance structure is organized according to a shared participant governance network rather than a lead-organization governed network, we look at the degree of density and the position of authoritative actors (Creech, Huppé, & Knoblauch, 2012;Kenis & Provan, 2006).The boundaries of those governance structures are fixed by the cases, i.e. by the regional competences for the policy domains energy, poverty and education.Hence, the description of regional governance structures is based on the relations between actors within those regional boundaries.Actors from the federal or the local level are only relevant if they have a functional role for the regional level, i.e. when conceived as relevant by the actors that are by definition part of the regional governance structure.Both density and degree of centrality are used to define the type of steering.While the density indicator is a measure on the level of the governance structure, the degree of centrality allows us to measure for each actor the extent to which it has acquired a central position in the governance structure.We use the normalized indicators as this allows for comparisons between cases of different sizes.
First, we measure the networks' degree of density.The more dense a network, the more it reflects a shared participant governed network, as the latter's performance depends on collective action decisions by unanimity.Next we look at whether the networks possess authoritative actors and-if so-how many.This indicates to what extent governance structure is steered by a lead-organization (one or two authoritative actors).More specifically, when an authoritative actor has the most or is close to the most central position, it can be assumed to be highly capable to steer the governance structure which points to a leadorganization governed network.On the contrary, when such actors are not the most or not even close to the most central position and in case of highly dense networks, the governance structure will be labeled as a shared participant governance network.
Next, we turn to the operationalization of the independent and intervening variables.The adaptational pressure is indicated by the level of EU integration.We use the Treaty on the Functioning of the EU (TFEU) to determine the extent to which the EU exerts powers in different policy fields.In addition, we use the amount of regulations and directives for which data are derived from EUR-Lex.Variation between the networks can further be accounted for by domestic intervening factors.Concerning the division of competences, the Belgian constitution clearly addresses the allocation to the different governmental levels.Secondly, the administrative capacity is measured by the amount of personnel that directly deals with Europe 2020.Although human resources only cover a partial aspect of administrative capacity, it is considered as a strong indicator for financial resources too (Beyers & Kerremans, 2007).Hence, respondents were asked to provide an adequate evaluation of the number of full-time equivalents (FTE) directly working on Europe 2020.Finally, the extent to which actors support the way Europe 2020 is dealt with is extracted from the questionnaire: we asked respondents to express their opinion with respect to the regional and European policy initiatives based on the following questions "Compared to the other Integrated Guidelines, how do you perceive the importance of this (case-specific) Integrated Guideline in the Europe 2020 strategy?"and "Compared to the other Integrated Guidelines, how do you perceive the importance of this (case-specific) Integrated Guideline in the (sub)national reform program?".The questionnaire data were aggregated as the independent variables are situated on the level of the governance structures and not on the level of individual actors.

Governance Structures and Intervening Factors in Six Europe 2020 Networks
Considering the density of the networks, we found little variation between cases with exception of the governance structure for energy during the European Semester (see Table 2).The network density of the latter (15%) is very low in contrast to the other cases (between 37% and 46%).This implies that the case of energy during the European Semester cannot be categorized as a shared participant governance network.For the other cases, the absence or presence of centrally positioned authoritative actors will further define the type of network.We found clear variation between the governance structures, not only with respect to the type of network, but also with regard to the involvement of different types of actors.The policy domain of education strongly resembles a lead-organization governed network for both the European and the national semester (respectively Figure 1 and 2; Table 3 provides the legend for interpreting all visualized governance structures, whereas Table 4 list actors in the networks of education).In both stagesand even more outspoken during the national semester, the minister of education and his personal staff (MIN EDUCATION) are positioned in the center, followed in second place by the department for education and training (DEP EDUCATION).Furthermore, taking a look at the visualized governance structure, also the VLOR, the advisory council on educational policies, seems to have acquired a rather central position.Remarkably, the VLOR's position is far less central during the national semester, which was confirmed by its respondent who suggested that the VLOR is specifically crucial in conveying European policies and actions to educational actors in society.
Governance structures in the energy domain (Table 5 lists actors in the networks of energy) differ substantially across the stages, as the variation in density of relations has already made clear.During the European semester (see Figure 3) the authoritative actors (the minister of energy and the department of environ-ment, nature and energy) lack a central position, pointing at the absence of a clear lead-organization.The low number of relations among the actors suggests that Europe 2020 is not very salient among energy-related actors.This was confirmed by several respondents who stated that Europe 2020 is perceived as being of merely secondary importance.The EU takes a lot of legislative initiatives in the energy domain which is intensively monitored and discussed at the Flemish level at the expense of the Europe 2020 energy targets.During the national semester (see Figure 4), on the contrary, the minister of energy is more intensively involved.Looking at the degree of centrality, he is not the foremost central actor, but the relatively high score still pictures him as being able to lead the governance structure.Remarkably, however, the executive Agency for Energy (VEA) and the Flemish Chamber of Commerce and Industry (VOKA) are considered to be stronger involved than the minister, suggesting a more shared participant governance approach that is neither centrally nor collectively steered.Hence, we classify the governance structure for energy during the national semester as a mixed lead-organization governed/shared participant governance network.The governance structures in the policy domain of poverty most clearly resemble shared participant governed networks (see Figure 5 and 6).During the national semester the most central positions are obtained by societal organizations or administrative actors (Table 6 lists actors in the networks of poverty), suggesting a lack of steering by the competent department or the responsible minister and his staff.Although the department of welfare, public health and family is the second most central actor in the governance structure during the European semester, the corresponding score on the degree of centrality is rather weak (6.5%).
On the contrary, and especially during the national semester, societal organizations are far more involved.But even their score on the degree of centrality indicator remains quite modest.From this, we conclude that poverty governance structures are organized in accordance to a shared participant governance network.
In the remaining part of this section we map the independent and intervening variables that have been identified as potential explanations for the variation in governance structures (see table 7 for an overview).First, the degree of EU competences is expected to affect regional governance structures as this variable indicates the amount of adaptational pressure.Article 4 of the TFEU explicitly stipulates that the EU has a substantial degree of authority regarding energy issues: member states can act only in so far the EU has not acted.The EU has no such powers for education or poverty.Education is mentioned in article 6: the EU is allowed to coordinate, support or supplement policies developed by member states.Concerning poverty, the EU may only act conform article 5, i.e. provide arrangements such as the European platform against poverty and social exclusion in which member states participate to coordinate their policies on social inclusion.Having substantial powers is one indicator, the extent to which those powers are exercised is yet another.EUR-Lex is a useful tool to determine the level of actual policy practice.A search for the keywords energy, education, poverty and social inclusion, delivered 450, 98, 1 and 6 hits respectively.Both indicators make clear that EU exercises considerably more regulatory powers in the policy domain energy than in the fields of education and poverty.
Next, the Belgian constitutional set-up reveals the competences allocated to the regional level.Education policy is a quasi-exclusive regional competence as it is allocated to the Belgian subnational Communities.In the domain of energy policy, the division of competences is more nuanced.The regional level enjoys legislative and executive powers with respect to energy renewables, isolation of houses and buildings, traffic (including public transport) and road infrastructure, whereas the federal level acts on matters of energy infrastructure (transmission grid), nuclear energy and energy prices.In other words, energy is a mixed competence, making both levels responsible to comply with a series of Europe 2020 energy objectives.In the policy domain of poverty too, both the federal and the regional level are equipped with competences to combat poverty and to increase social inclusion.The balance heads over to the federal level, though, as the latter runs the social security system, including unemployment benefits and income support.The regional level has only supplementary powers, meaning that it is allowed to take measures in the area of social inclusion within the range of other policy competences (such as education).

Shared participant
During the interviews respondents indicated to what extent they support the way the EU and the regional level deal with the Europe 2020 integrated guidelines.Support was scaled from no support at all (1/5) to highly supportive (5/5).The data reveal clear differences: while actors in the policy domain of education are highly supportive (on average 4,4/5), those in the domain of energy are moderately supportive (on average 3,6/5) and those in the field of poverty are only weakly in favor (on average 2,7/5).
Finally, regarding administrative capacity we asked respondents to determine the number of people working explicitly on the Europe 2020 targets.The policy domain of education is clearly the best equipped to deal with Europe 2020, having on average 1,6 FTE's per actor in the network, compared to 1,2 FTE's in the field of energy and 0,7 in the field of poverty.More importantly, the authoritative actors in the policy domain of education, the ministerial staff and the department of education, employ respectively 2,5 and 2 FTE's working on Europe 2020, rendering them comparatively well equipped to lead the governance structure.With respect to the policy domain of energy both the department and the ministerial staff reported only 1 FTE dealing with Europe 2020.A significantly higher capacity (3 FTE) is reported for the Flemish Energy Agency (VEA).Concerning the policy domain of poverty, the average amount of FTE's is only 0,7, which is the lowest of all policy domains.The ministerial staff only slightly exceeds the average administrative capacity with a total of 1 FTE.The highest number of staff is recorded for the Flemish Representation within the Belgian Permanent Representation (VL-PV), employing 2 FTE's to deal with Europe 2020.

Explaining the Variation in Governance Structures' Type of Steering
The three policy domains of education, energy and poverty are subjected to a varying degree of adaptational pressure induced by the Europe 2020 program.Moreover, domestic intervening variables further differentiate the impact of the adaptational pressure.Across all three policy domains, we found that regional governance structures activated to deal with Europe 2020 are managed by varying steering mechanisms.In this section we seek to explain this variation by analyzing how the identified domestic variables nuance the adaptational pressure on governance structures.
Education and poverty clearly have opposite steering mechanisms in both policy stages while the degree of EU integration and thus the adaptational pressure however only slightly differs.In the case of poverty, EU competences are weak, and although the EU has stronger somewhat competences in the case of education, national law remains dominant.Still, we found that the governance structures in the field of education are characterized by lead-organization governed networks, whereas the governance structures in the field of poverty are organized by shared participant governance networks.The answer to this puzzle is found by looking at domestic intervening variables which vary substantially.The support for EU 2020 among administrative and political actors, administrative capacity and the extent to which the regional level is competent are all high in the domain of education, whereas the domain of poverty shows lower scores on all these variables.A first comparative assessment thus suggests that strong regional competences coincide with high administrative capacity.As theorized above, these features enhance authoritative actors' capacity to steer and coordinate other actors resulting in a lead organization type of network.The field of education is very outspoken in this respect.Hence, domestic intervening variables seem to strongly affect the Flemish governance structures.Furthermore, the strongest support for Europe 2020 is also found in the governance structures with the strongest steering.These findings so far are valid for both stages in the fields of education and poverty.The lead organization approach in the field of education can be attributed to the quasi-exclusive regional competences, the high administrative capacity and the strong adherence to the Europe 2020 policies, while the shared organization approach coincides with regional supplementary powers, low administrative capacity and weak administrative and political support for the Europe 2020 objectives.
The policy field of energy delivers more puzzling results.The EU has strong competences is this field, yet the governance structures differ in the two policy stages: a weak governance network during the European semester and a more mixed governance network during the national semester.Given the relatively high capacity and the status of shared competence, the weak governance network during the European stage is rather surprising.The explanation may be that precisely because of the strong EU integration of the policy domain of energy, the Europe 2020 targets are rendered less salient compared to the high amount of legislative proposals.Some respondents pointed out that the energy policies of Europe 2020 are far less frequently discussed than legislative proposals, making the presence of lead organizations during the European semester less necessary.This is different during the download stage as both legislation and Europe 2020 policies have to be addressed through transposition and reform programs respectively.From this perspective, the European stage of the energy case can be considered as an isolated case.Secondly, the governance structure of the energy case during the national semester nuances our conclusions so far.The mixed network emerges in a context of strong support and a high level of EU integration, which is even stronger than in the education cases and which therefore induces a lead organization type of network.However, what makes the energy case different from education is the central position of VEA and VOKA in the network, VEA's strong administrative capacity and the federal competences in energy policy.These features seem to decrease the emergence of strong authoritative actors (such as the ministerial staff or the competent department) as lead organizations.Clearly, the capacity and involvement of other regional stakeholders join the regional authoritative actors as central players in the network, making the governance structure less hierarchical.
From this we conclude that the type of governance structures established to deal with the Europe 2020 program is determined less by the extent to which a policy domain in integrated in the EU or by the degree of support among the actors involved, but rather by the domestic division of competences and the level of administrative capacity of authoritative actors.The more competences, the more capacity, the more authoritative executive actors are able to put themselves in the center of the network and therefore in charge of uploading and downloading Europe 2020 policy.Only when other regional stakeholders manage to mobilize capacity, the authoritative actors have to share their central position, resulting in a more mixed type of steering.

Conclusion
Europe 2020 is the European growth agenda that covers a wide area of policy domains.Yet the impact of Europe 2020 plays out differently across policy domains and policy stages.This paper has assessed the differentiated Europeanization at the regional level in three policy domains: education, energy and poverty.Furthermore we have considered two policy stages, providing us with six cases situated in the Belgian region of Flanders.Based on the interactions between political, administrative and societal actors, the governance structures were mapped and the type and strength of their steering assessed.The underlying relevance is that the type of steering affects the effectiveness of the governance structures, whereby the literature suggests that strongly steered networks perform better than weakly steered networks or topdown controlled structures.
Our findings indicate variation between policy domains and policy stages.The governance structures in the field of education were defined as lead organization networks, whereas those in the domain of poverty were considered as shared participant networks.In the field of energy, variation was even found between policy stages: the European semester governance structure is considered a weak governance network, whereas the the national semester network features a more mixed network.
Overall, the extent to which Flanders is competent seems to be crucial.Due to extensive regional powers lead-organization governed networks are likely to be put in place.In order to exercise these powers, a strong administrative capacity is needed to steer and coordinate the governance structures.EU integration may further increase the extent to which Flemish governance structures are steered.However, strong EU integration can also entail a high degree of EU legislation which may overshadow Europe 2020 policies, as was the case in the upload stage in the energy case.In other words, competence and capacity matters for the way the Europe 2020 policies are governed.Support among the actors for Europe 2020, however, was not found to affect the strength of steering.
Clearly, our conclusions are bound to the policy domains and the Flemish region examined in this article.Future research may broaden the scope to other strong legislative regions and other policy domains in order to gain a better understanding of the impact of Europe 2020 on regional governance structures and of the intervening role of domestic variables.Furthermore, increasing the number of cases may open the door for more systematic (small N) comparative analysis enhancing the scope of generalization.

Table 2 .
Degree of centrality and type of network in six cases.

Table 7 .
Overview of variables in six cases.