Differentiation from Below: Sub-national Authority Networks as a Form of Differentiated Cooperation Pier Domenico Tortola and Stefan Couperus

Transnational networks of local authorities are an established and growing phenomenon in Europe, where they perform a number of (soft) governance functions for their membership, often in direct connection with EU institions. This paper examines networks from the angle of institutional differentiation – an inherent trait of these organisations – in order to expand the analysis of this concept beyond its state-centric confines. The paper also adds to the study of (differentiated) integration more traditionally defined, both by analogy and insofar as networks are part and parcel of the Union’s system of multi-level governance. Building on original empirical data, we identify three dimensions of differentiation generated by networks – insider-outsider, compound and multi-level differentiation – and discuss their implications for the efficiency, effectiveness and legitimacy of networks’ actions. Based on our analysis, we formulate three broad recommendations for policy-makers involved in networks: first, strive for inclusion, especially of smaller and less administratively capable sub-national governments; second, improve available information on networks in order to make their landscape more efficient; third, exploit the paradiplomatic advantages of networks both outside and inside the European Union.


Executive summary
Transnational networks of local and regional authorities are an established and growing phenomenon in Europe.Dozens of networks of various size currently exist within EU territory, working in a wide range of policy areas -with environmental policy being particularly prominent.
Networks perform a number of functions for their members, such as promoting exchanges of best practices and policy learning; developing projects; providing information and technical support; and representing sub-national interests.In doing so, networks often interact directly with the EU, which in turn plays a role in encouraging and supporting them.
Networks are inherently differentiated organisations.Studying networks from the angle of differentiation helps us move the notion of differentiation beyond its state-centric confines.At the same time, it can enrich our knowledge of state-led (differentiated) integration, either by analogy or insofar as networks are part of the EU's system of multilevel governance.
Building on original empirical data, we identify three main dimensions of differentiation engendered by networks: • Insider-outsider differentiation separates the participants in a network from those who remain outside.This type of differentiation affords networks institutional flexibility and coherence.On the other hand, it may lead to free-riding problems, undue exclusion from club goods, diminished diversity and barriers to policy innovation.
• Compound differentiation is generated by the coexistence of several networks in the same policy area.This increases linkage opportunities for sub-national governments, and may stimulate healthy competition and synergies among organisations.At the same time, it may lead to duplication and fragmentation of efforts, and the loss of economies of scale.Membership overlaps, and the ensuing creation of "super-insiders" may facilitate cooperation between networks, but also reinforce existing hierarchies among sub-national governments.
• Multi-level differentiation occurs in interactions between the EU and networks, whenever the countries represented in the latter mismatch with EU membership.The absence of certain member states in a network's coverage may affect the network's representativeness, and hamper Europeanisation at the sub-national level.The presence of non-EU-based sub-national governments in a network, on the other hand, may present accountability issues, but also provide EU policy-making with fresh input, and facilitate the spread of EU ideas and practices beyond its borders.
Based on our analysis, we formulate three broad recommendations for policy-makers at both the sub-national and EU level: • Increase the inclusiveness of networks, focusing especially on smaller sub-national governments, which often lack the capacities to effectively participate in transnational organisations.The EU, together with existing networks as well as big cities and regions,

Introduction
In the context of European governance, differentiation -defined as the creation of flexible and non-homogeneous forms of cooperation among territorial authoritiesis a topic most commonly conceived and studied in state-centric terms, namely by seeing EU member states as the primary instigators, subjects and implementers of differentiated integration (DI) arrangements.While this view is understandable -DI is indeed the foremost embodiment of differentiation -focusing only on nation-states risks obscuring other arenas in which differentiation plays an important role, and which can contribute to our understanding of this concept and its implications for the politics and governance of Europe.Such is the case for transnational networks of sub-national authorities. 1Sub-national authority networks are an established and growing phenomenon in Europe, some with historical precedents dating back to the early 20th century (Couperus 2011, Payre 2007).They span a wide number of policy areas, and provide their members (primarily cities and regions 2 ) with a range of benefits and tools of (soft) governance.Networks can be seen as both a consequence and a manifestation of a number of epochal transformations in Europe and elsewhere, such as globalisation, transnationalism and the rescaling of governance functions (Brenner 2004, Leitner 2004, Sassen 2001, Taylor 2005).More concretely, they are part and parcel of Europe's system of multi-level governance, insofar as they participate in the EU's policy-making cycle at various stages and in various ways (Hooghe 1995, Perkmann 2007, Tortola 2013and 2017).
1 Sub-national government organisations do not, of course, exhaust the set of transnational networks in Europe: the latter also include a wide range of organisations gathering non-territorial entities, e.g., professional, corporatist or NGO networks.In this paper we focus exclusively on networks of sub-national governments not only because of their strong connections to the EUwhich supports and collaborates with many of them -but also because their territorial nature makes them conceptually closer to the sphere of traditional (differentiated) integration.This, in turn, makes sub-national authority networks more amenable to being studied through some of EU IDEA's key analytical categories, such as governance effectiveness, legitimacy or Europeanisation. 2 Throughout this paper we will use the term "region" generically, to indicate any sub-national unit of government larger than a town or a city -hence not only regions properly named, but also provinces, Lander, etc. EU IDEA Policy Papers No. 8 Sub-national authority networks are loci of institutional flexibility and differentiation par excellence, due to their very nature and characteristics, above all the fact that they gather specific sub-sets of local and regional governments in a non-compulsory fashion.Differentiation is also engendered by the ways networks interlock with one another and with other levels of government in the continent, primarily the supranational one.Studying networks, therefore, can enrich our overall understanding of differentiation in Europe by expanding this notion beyond its predominant statecentric focus, in line with the goals of the EU IDEA project.At the same time, examining sub-national authority networks can potentially offer new perspectives and lessons for more traditional forms of differentiation, either indirectly by analogical reasoning, or directly insofar as network governance connects to state-led (differentiated) integration.This paper presents an analysis of sub-national authority networks in Europe through the lenses of differentiation, supported by an original dataset and a number of in-depth interviews with network practitioners.We proceed as follows: the next section introduces sub-national authority networks by briefly discussing their history, functions and main empirical traits.We then examine, in the following three sections, three distinct dimensions of differentiation generated by sub-national authority networks dubbed, respectively, insider-outsider, compound, and multilevel differentiation.The fifth and final section concludes and presents some policy recommendations aimed at increasing the effectiveness, efficiency and legitimacy of networks.

Sub-national authority networks in the European Union
For the purposes of this paper, we define sub-national authority networks as horizontal, voluntary and independent organisations connecting local and/or regional authorities across state boundaries in a stable manner, with the aim of achieving some common goal and/or producing mutually beneficial services.This broad definition includes a wide range of networks in terms of type of members, size, geographic span, organisational density and thematic focus.It leaves out, however, all-inclusive organisations of sub-national authorities formally embedded in a wider institutional structure -such as the EU's Committee of the Regions, or the Council of Europe's Congress of Local and Regional Authorities -and collaborations that are episodic or mere emanations of specific projects (for instance in the context of the EU's regional policy).
The activity of transnational networking among sub-national authorities can be traced back to the late 19th and early 20th centuries during which municipalities, taking up ever more public tasks and responsibilities, shared experiences, best practices and (administrative) technologies across borders in the realm of public utilities, public transport, municipal bureaucracy and social services (Saunier and Ewen 2008, Hietala 1987).After the Second World War, the institutional make-up Differentiated Cooperation of transnational local and regional networks gradually thickened, for instance with the creation of a number of experiments involving cities as well as regions, such as town twinning (Couperus and Vrhoci 2019), Euroregions (e.g., the Dutch-German one, established in 1958) and larger organisations like the Council of European Municipalities, founded in 1951.However, it is really in the past three decades or so that networking has gathered pace, stimulated by a number of converging factors such as improvements in cross-border communications, power and competence gains on the part of local and regional authorities, encouragement (and in some cases initiation) by inter-and supra-national institutions (like the United Nations, the Council of Europe and above all the European Community/Union) and, finally, the emergence of a number of new and inherently transnational challenges, in the first place environmental ones (Ewen and Hebbert 2007, Kuznetsov 2015, Le Galès 2002, Murphy 1993, Tavares 2016).Taken together, these trends have contributed to the establishment of dozens of networks of various kinds, in particular in Europe, which stands today as one of the most densely networked areas in the world (Acuto 2013, Acuto andLeffel 2020).
Comprehensive and reliable data on sub-national authority networks in Europe (and elsewhere) is notoriously hard to come by. 3 For this paper we therefore collected and coded an original dataset of 96 networks operating, fully or partially, in the EU's territory, comprising 30 city networks, 30 regional networks and 36 mixed membership networks (see appendix for more details on our data collection procedures). 4While this dataset is likely to underestimate the total number of sub-national authority organisations in the continent, we believe that it provides us with a reliable and broadly representative sample of Europe's current network landscape.Figure 1 shows the distribution of networks in our sample by decade of establishment.
The picture emerging from the graph is consistent with the above account: over two thirds of the networks in our sample were established in the 1990s and 2000s.Environmental and climate change activism on the part of many sub-national governments, in conjunction with the United Nations, played an important role in accelerating networking in this period, by spurring the creation of a number of important networks -such as the International Council for Local Environmental Initiatives, the Climate Alliance, Energy Cities and later the C40 network -and more generally promoting the notion of sub-national participation and leadership in actions against global warming, above all through the UN's Agenda 21 (Acuto and Leffel 2020, Bansard et al. 2017, Bulkeley et al. 2003, Keiner and Kim 2007).
3 To the authors' knowledge, the only existing comprehensive and publicly accessible database of sub-national authority networks is included in the Yearbook of International Organization maintained by the Union of International Associations (https://uia.org/ybio),which collects information on over 70,000 inter-and transnational organisations.However, the absence of a specific classifier on subnational authority networks in the database hinders the extraction of data for our purposes.4 For the purposes of our dataset, which was collected before the official Brexit date, the United Kingdom is considered part of the EU.The prominence of environmental themes among sub-national authority networks is confirmed by figure 2, which shows the static breakdown of our network dataset Differentiated Cooperation by policy area.Slightly over a quarter of the organisations surveyed work exclusively or predominantly on environmental policy or in the adjacent area of energy.At the same time, taken together, sub-national authority networks cover a wide array of policy sectors, oftentimes combining more than one, as shown by the considerable portion of "generalist" networks, which work in two or more fields without any stated hierarchy among them.In this latter category one finds a number of broad organisations -such as Eurocities, the Assembly of European Regions or the Association of European Border Regions -as well as regionally specific ones, which tackle a range of topics within a well-delimited geographic area, like the Alps-Adriatic Alliance, the Alpine network Arge-Alp or the French-German PAMINA Eurodistrict.These findings are in line with previous surveys of the field (e.g., Davidson et al. 2019, Keating 1999, Niederhafner 2013, Rapoport et al. 2019).
Regardless of the field(s) in which they work, the functions of sub-national authority networks tend to be similar.As voluntary, non-coercive and often not particularly rich organisations, the main resources on which networks base their activities (and authority) are knowledge and information, in their various forms.Among the activities to which networks devote most of their times one finds policy learning and innovation, exchanges of know-how and best practices, technical support to local and regional governments, interest representation and lobbying, project formulation, development and implementation, and further networking and matchmaking among members (e.g., for project consortia) (Keating 1999, Kern and Bulkeley 2009, Leitner 2004, Niederhafner 2013).
With the growing importance of sub-national authority networks in Europe's political and institutional landscape has come increasing scholarly interest in these organisations.Over the past couple of decades, research across several disciplines -notably political geography, urban studies, regional studies, international relations and more recently European studies -has produced important findings on aspects such as the drivers of networks (e.g., Huggins 2018a, Keating 1999, De Sousa 2013), their main types (e.g., Callanan andTatham 2014, Murphy 1993), their organisational dynamics (e.g., Bulkeley et al. 2003, Mocca 2018) and their relationships with other governance levels, above all the EU (e.g., Heinelt andNiederhafner 2008, Hooghe 1995).What still remains largely unexplored in this literature is the angle of differentiation.The latter is an inherent feature of networks, above all because they gather together specific sets of sub-national governments in a voluntary and flexible way.This, in turn, can have significant implications for the nature, effectiveness and even legitimacy of their actions.The remainder of the paper presents a first stab at filling this gap, by describing and reflecting on three main dimensions of institutional differentiation engendered by sub-national authority networks: insider-outsider differentiation, compound differentiation and multi-level differentiation.EU IDEA Policy Papers No. 8

Insider-outsider differentiation
The first type of differentiation created by sub-national authority networking is, very simply, that separating members of each network from non-members.This "insideroutsider" differentiation is the most straightforward kind, and the most similar to traditional state-based differentiation.Unlike the latter, however, insider-outsider differentiation accompanies networks almost by definition: given these institutions' voluntary nature and often specialised focus, it would be hard to imagine a network comprising each and every potentially eligible member.Our data reveals tremendous variability in territorial membership size: ranges for city, region and mixed networks are 2 to 9,741; 4 to 151; and 2 to 1,709 respectively.5But even the largest city and regional organisations -the Covenant of Mayors and the Conference of Peripheral Maritime Regions -do not come close to being all-inclusive.Sheer numbers do not tell the whole story, however.Broadly speaking, insideroutsider differentiation in networks can be broken down into two dimensions: the first concerns limits established by the network's mission to who can participate, based on certain attributes, which can be geographic, demographic, institutional or of other kinds.Cross-border cooperation networks are a typical case of this "statutory differentiation".Other examples in our dataset include the metropolitan areas network METREX, the Association of European Border Regions, the Association Internationale des Maires Francophones, the Unión de Ciudades Capitales Iberoamericanas and the European Straits Initiative, just to mention a few.The second dimension concerns actual participation in the network compared to the latter's potential coverage as statutorily defined.6All the networks in our dataset entail at least one of these two dimensions of insider-outsider differentiation.
Differentiation has several advantages for sub-national authority networks.The most obvious is that, in principle, it allows sub-sets of cities and/or regions sharing similar features, needs and problems to get together in a flexible fashion and define the perimeter of the organisation to best fit their common goals.This, in turn, should increase the network's internal coherence in terms of the framing and definition of problems, governance outlook and policy agendas, ultimately improving the effectiveness of joint activities (interviews 3 and 4).Consider, for instance, these remarks made by one of our interviewees: I know that […] other regions would like to participate [in our network], but at this point, the political level […] decided not to go too quick too broad, but rather to deepen work within the network.[…] And we stick to our definition, [whereby a member must be] a region with constitutional responsibilities Differentiated Cooperation (and not just an autonomous region within a nation-state […] with some legislative competencies), and economically strong -strong enough to [be able to] share in with the other [member] regions.(interview 4) For organisations that are more open-ended in terms of (potential) membership, differentiation also means greater flexibility for groups of pioneers to begin joint endeavours to which more sub-national governments will adhere later on.From an organisational standpoint, and especially for more exclusive organisations, having smaller groups of insiders may also facilitate the management of the network and its internal flow of information and communication, and increase the density and depth of cooperation among members (Leitner 2004, interview 4).Finally, insideroutsider differentiation may also be regarded as advantageous to the outsiders of an organisation, insofar as it involves the freedom of sub-national authorities that are not willing or ready to join a network to remain unencumbered by it, organisationally and financially.
On the other hand, insider-outsider differentiation also presents a number of drawbacks, of which we identify four main ones.The first is that differentiation may lead to free-riding on the part of outsiders, whenever they are able to benefit from public goods produced by the network, not only in the area of "external governance", such as lobbying and interest representation, but also in "internal governance" -that is, the generation of know-how, best practices, standards, etc. (Hooghe 1995, Kern andBulkeley 2009).While this is, admittedly, not the biggest problem for networkssome of which have, in fact, the ambition to provide leadership beyond their confines -free-riding may nonetheless lead to the suboptimal production of some of the network's services.
A second, and bigger drawback of differentiation concerns those services that take the form of club goods, and from which outsiders can therefore be excluded, such as technical support, training or the access to privileged information (Capello 2000).Clearly, the disadvantage here is particularly marked for those outsiders which are not so by virtue of a deliberate choice, but rather because they are unable to join, e.g., for financial or administrative reasons.For smaller sub-national governments, exclusion might even end up feeding a vicious circle whenever the capacities that would facilitate participation in a network -which can be as simple as having an officer in charge of international and EU relations -are exactly those that would be boosted by joining the network in the first place (Kern and Löffelsend 2008, Tortola 2012and 2016, interview 1).
The third drawback is the mirror image of the coherence argument presented above.Reducing diversity within a network might become a disadvantage in all those areas in which variety -of various kinds: institutional, cultural, political, etc. -is a plus, for instance in the areas of policy learning and innovation.To the extent that these are seen as priorities, a network might be better served by an inclusive strategy, which tries to increase the number, or at least the types, of participants.
Finally, and connected to the foregoing, insider-outsider differentiation may diminish the representativeness of networks in their interactions with other actors (most EU IDEA Policy Papers No. 8 notably EU institutions, but also other inter-and transnational organisations), and possibly impinge on the legitimacy of their actions in these contexts (Hooghe 1995, Pirozzi et al. 2017).This problem might be particularly pronounced for networks characterised by a geographically skewed membership.Kern and Bulkeley (2009) list a few examples of such unequal representation in the crowded environmental field, noting how most cities participating in the prominent Climate Alliance network are located in Germany, Austria and the Netherlands, while Cities for Climate Protection and Energie-Cités have their strongholds in, respectively, Finland and the UK, and France.To the extent that interactions with the EU are concerned, the unequal distribution of network participants might also produce a more specific mismatch between the countries represented in the network and EU member states, which will be discussed in greater detail in section four.

Compound differentiation
The differentiation picture becomes more complicated as we move from considering networks individually to looking at them collectively within EU boundaries.Unlike what happens at the state level, where we usually find one main cooperation or integration arrangement per policy area, at the sub-national level several -and in some cases many -networks usually coexist within each policy area, often with (partly) overlapping missions and memberships.7This multiplies the points of differentiation created by networks, generating what we call here compound differentiation: a situation in which sub-national governments can have various degrees and types of "insiderness" and "outsiderness", depending on how many (if any) and which networks they take part in.Needless to say, compound differentiation is especially important in network-dense fields, such as environmental policy.
The presence of different networks in the same field may be taken as a mitigating factor for some of the drawbacks of insider-outsider differentiation.Coexisting networks may, for instance, multiply opportunities for cooperation among subnational governments, while preserving a certain degree of coherence within each organisation (interviews 1, 2 and 3).Consider, just to make an example, the case of C40, a network working on climate change with a specific focus on large cities, whose interests and preferences might be overly diluted in the context of a more generalist organisation.Multiple networks can also be a stimulus to healthy competition among organisations, especially when it comes to policy innovation and the elaboration of projects (Keiner andKim 2007, Mocca 2018, interview 2).Finally, the existence of different organisations within the same policy areas may incentivise a virtuous division of labour in the network's modus operandi and specific mix of services, which can in turn facilitate ad hoc cooperative arrangements that take advantage of the synergies among different organisations (Keiner and Kim 2007, Differentiated Cooperation Kern and Bulkeley 2009, interviews 2 and 3).One of our interviewees expresses this idea rather clearly: […] I think what we'll see more and more is that all these networks of cities and regions -because actually there are many of them -will start to get more and more specialised.(interview 2) At the same time, the coexistence of multiple networks may increase the risk of duplicating efforts and even wasting resources, both on the side of networks themselves, and on that of financing institutions -especially, but not only, the EU -which might be tempted to spread their support broadly across organisations rather than picking winners and losers (interviews 1 and 2).Additionally, excessive fragmentation among networks can lead (some) sub-national authorities to lose some of the advantages and economies of scale that come with unity and size, such as political weight or the ability to establish an effective administrative and policy infrastructure at the centre of the network (Barber 1997, Capello 2000).This is exacerbated by the fact that a sub-national government's adherence to one network rather than another is sometimes driven not so much by functional reasoning with respect to the size, shape and mission of the organisation, but by more contingent factors such as personal connections, pre-existing links among sub-national authorities or even effective public relations on the part of the network (Barber 1997, Tavares 2016, interviews 3, 4 and 6).
As noted above, membership overlaps among networks further increase the range of differentiation among sub-national governments, creating what one could dub "super-insiders", namely local or regional governments that participate in several networks per policy area.Previous research has found, unsurprisingly, that the greatest number of transnational connections tend to be found among the biggest and most resourceful sub-national governments: for instance, major cities such as Paris, Barcelona, Brussels, Berlin or Rome (Acuto andLeffel 2020, Rapoport et al. 2019).The number and size of networks in our dataset does not allow us to draw a comprehensive and granular picture of network membership overlap.As a rough proxy, however, one can look at network membership by EU member state, shown in figure 3.
As one could expect, the figure suggests a general relation between country size and membership coverage: cities and regions in bigger countries tend to participate in a greater number of networks.Interestingly, however, there seems to be also a broad connection between age of EU membership and number of networks, whereby older member states are generally those whose sub-national governments are more connected across borders.This is consistent with we have noted above about the central role of the EU in encouraging, supporting and sometimes even creating transnational networks of sub-national governments in its territory.EU IDEA Policy Papers No. 8 Figure 3 | Sub-national authority network participation by country Membership overlaps between networks may help redress some of the problems generated by the existence of multiple networks in the same policy area, as cities and regions participating in several organisations can play an important role in internetwork communication, mediation and collaboration (Keiner andKim 2007, Kern andBulkeley 2009, interview 5).On the other hand, membership overlaps could also produce some undesirable effects for the super-insiders, who might at times find themselves having to juggle conflicting demands from different networks.
The greatest potential drawback of membership overlaps, however, is that it may further magnify the hierarchies among sub-national governments (or clusters thereof) that already exist within, and are reinforced by, each individual network (Acuto 2013, Kern and Bulkeley 2009, Mocca 2018).Being in the middle of different networks allows super-insiders to benefit from the club goods produced by each of them, provides them with different arenas to which they can turn according to their needs, and affords them with better connections and access to other levels of government, above all the EU (Keiner and Kim 2007).

Multi-level differentiation
With multi-level differentiation we indicate the sort of differentiation originating in the intersection of sub-national authority networks and state-based (differentiated) integration in the European Union.To a significant degree, networks work on policy areas in which the EU is also involved, in a more or less extended fashion.It is, therefore, Differentiated Cooperation very common for networks to be in contact with the EU and its work, either directly (for instance by taking part in EU-supported projects, or by playing a consultative role in EU decision-making), or indirectly by operating in areas that are exposed to processes of Europeanisation (interviews 2 and 4).This raises the issue of possible mismatches between the membership of the Union (or some of its differentiated integration arrangements) on the one hand, and the countries represented in the various sub-national authority organisations on the other.
Simplifying, we identify here two kinds of multi-level differentiation.The first occurs whenever sub-national authorities from one or more member states are excluded from a network interacting with the EU.This is the modal case in our dataset, where the Covenant of Mayors -a network initiated by the European Commission -is the only organisation including sub-national governments from each and every EU member state.More generally, networks in the dataset occupy the whole gamut between one and 28 member states, with an overall average a bit lower than 10.
In several cases, partial national participation is a direct result of the exclusive design of the network, which we have discussed above: we would not expect, for instance, a cross-border cooperation organisation to involve countries far from the border in question, or landlocked countries to be represented in a network of coastal regions.In other cases, however, the skewed representation of countries in the network is less obviously justifiable, and potentially problematic.In the first place, it may add a distinctly national dimension to the issues of representativeness and legitimacy mentioned above, whenever networks participate, more or less directly, in the EU's decision-making process.This would be an example of what Lavenex and Križić (2019) refer to as "organisational differentiation" -the differential participation of sub-national entities in European processes of governance -which is especially tricky because the involvement of sub-national organisations in supranational policymaking is often regarded as an important channel to alleviate the democratic deficit of EU institutions (Davidson et al. 2019, Heinelt and Niederhafner 2008, Hooghe 1995).The absence of certain countries in the representation of networks might also pose a problem of policy effectiveness insofar as it may skew the policy input provided by these organisations by removing nation-specific needs, considerations, administrative experience and so on.
Partial national representation in networks may also create a number of top-down issues in the relationship between supra-and sub-national governments.First and most tangibly, it could limit the ability of sub-national authorities from certain countries to take advantage of EU projects and funds.More importantly, it could truncate the intra-network flow of information, know-how, best practices, etc. along national lines.This is more so if we consider that members of a transnational network may, in some cases, also act as nodes that relay domestically (e.g., through national networks) what they have learnt from the EU as well as their peers abroad.To the extent that intra-network learning processes are connected to EU policies, not being represented in one or more networks may also affect a country's capacities and opportunities in terms of sub-national level Europeanisation (Huggins 2018b, Kern and Bulkeley 2009, Tortola 2016).
The second form of multi-level differentiation occurs when sub-national authority networks operating on EU territory include members from non-EU countries.Figure 4   The questions raised by this second type of multi-level differentiation are, in part, a mirror image of those just described.Insofar as networks function as interfaces between their members and EU institutions (in representing local and regional interests, participating in policy-making, evaluating policies, etc.) one could see here not only issues of representativeness, but also a problematic mismatch of accountability, given that a portion of the constituency of the networks in question lies outside the borders of the EU.At the same time, the presence of non-EU-based members in networks can also be a source of fresh ideas and out-of-the-box policy thinking, both for EU-based network members and for EU institutions, notably the Commission.
Networks that reach beyond the EU can, finally, be channels for the diffusion of certain ideas and practices promoted by the Union beyond its borders -thereby extending the scope of integration outcomes beyond their immediate targets, to use Lavenex and Križić's (2019) terminology -either directly, via interactions between Differentiated Cooperation EU institutions and non-EU sub-national governments, or indirectly to the extent that networks are carriers of sub-national level Europeanisation.This paradiplomatic function is all the more valuable when it provides a backdoor through which to counter the limits or deterioration of traditional diplomatic relations between the EU and outside actors (Duchacek 1984, Tavares 2016).
We argue that there are three instances in which transnational networking beyond EU borders can be especially helpful.First, networks can (indirectly) help prospective EU members prepare for accession.This was, for instance, the role played by some Euroregions bridging the EU and post-communist Europe in the 1990s (Pasi 2007).Kern and Löffelsend (2008), for one example, have shown how a number of transnational structures of governance, among which the Union of the Baltic Cities, played a role in promoting the environmental Europeanisation of their postcommunist members, and strengthening their links with Brussels, well before the 2004 enlargement.
Second, networks can help establish and maintain productive transnational relations with those countries -above all in the EU's neighbourhood -in which the practice of liberal democracy is currently wanting.Obydenkova (2005), for instance, has documented how transnational cooperation between European and Russian regions has solidified processes of sub-national democratisation in the latter.Pintsch (2020) has explored the increasing connections between the EU and newly decentralised Ukraine via community twinnings and transnational municipal networks as vehicles for the societal Europeanisation of Ukraine and, more generally, the increasing legitimacy of the country's European agenda.In this respect, sub-national authority organisations can reproduce, on a larger scale, one of the historical functions of town twinning: to promote linkages and engagement beyond international rivalries.
Finally, networks have a role to play in post-Brexit Europe, as a tool to foster transnational communications between the EU and the United Kingdom -whose sub-national governments participate in 48 of the networks surveyed here -and perhaps help mitigate some of the political and policy rifts that might emerge between the two sides in the years to come.

Conclusions and policy recommendations
This paper has presented a systematic analysis of sub-national authority networks in the European Union from the hitherto unexplored angle of institutional differentiation.Building on an original dataset and a number of interviews with network practitioners, we have identified and discussed three main dimensions of differentiation generated by networks: insider-outsider, compound and multi-level differentiation.For each, we have highlighted some of the main advantages and drawbacks, primarily in terms of the effectiveness, efficiency and legitimacy of networks' actions.
This study has presented a first overview of a number of important aspects and issues of sub-national authority networking, which should be of interest to scholars as well EU IDEA Policy Papers No. 8 as practitioners in this area.Further research should look at each of these in greater depth, both analytically and empirically, focusing primarily (though not exclusively) on the areas in which networking among European sub-national governments so far has been denser and more extensive.That said, we can already identify, based on our survey, a few areas for improvement in the landscape and work of networks.These will be of interest to policy-makers at different levels: primarily members and practitioners of networks, but also EU institutions, which play an important role in promoting, supporting and therefore influencing sub-national authority networks.In the remainder of this section, we elaborate on three main such recommendations, broadly connected to the three dimensions of differentiation explored above: 1) strive for inclusion in network membership; 2) improve information on networks; and 3) take advantage of networks' paradiplomatic potential.
First, strive for inclusion.As shown in this study, the set of EU-based networks is highly diverse, comprising organisations of various types, size and focus.For many of them, and especially the more generalist ones, there is a constant danger of underrepresenting smaller local and regional governments -in particular among the EU's most recent member states.Networks are often the main, if not only way for these towns, cities and regions to establish links and fruitful exchanges beyond national borders.At the same time, joining and participating in transnational organisations already requires capacities -in terms of personnel, skills, information, know-how, etc. -that are often in scarce supply in smaller administrative contexts.This makes it important for actors in a stronger structural position to proactively engage and promote the participation of smaller sub-national governments in transnational networking.The overall goal here should be to improve on the representativeness and internal diversity of (some) networks, while retaining the advantages of insideroutsider differentiation especially in terms of flexibility and coherence of purpose among network members.The EU already has a number of initiatives in this area (for instance the URBACT programme in the area of urban development), which should be continued and expanded, and complemented by greater outreach efforts on the part of existing networks, as well as bigger cities and regions -in the first place those that have been dubbed super-insiders in this paper.
Second, improve information on networks to make their landscape more efficient.Sub-national authority networking is a crowded field, the traits and boundaries of which remain partly concealed.This amplifies the risks of duplication, inefficiencies and missed opportunities for cooperation, simply because full and clear information on "who does what" in the world of networks is hard to come by.While cooperation and joint projects are common among the most established networks, there is a widespread feeling among scholars and practitioners that much remains under the radar, especially when it comes to younger and smaller organisations.Generating more comprehensive data and information -for instance through the creation of a comprehensive EU-wide network registry -is a crucial step for shedding light on the entirety of the sub-national authority network landscape, and stimulating a more effective division of labour, as well as the establishment of synergies among networks, thereby minimising the drawbacks of the compound differentiation of networks described above.Differentiated Cooperation Third, explore and exploit the paradiplomatic advantages of networks.Networks can provide effective additional, or even alternative channels for transnational communication and engagement that can make up, at least partially, for the deterioration or disruption of traditional diplomatic relations.As discussed above, this is particularly true with respect to the EU's external relations (especially, but not exclusively with its neighbourhood).However, this logic may also apply within the Union itself.Increasing, for instance, the network participation of sub-national governments in the post-communist member states may provide a further waymore pragmatic and less confrontational -to support the ideas and practice of Europeanisation in those countries in which these values currently are in retreat.Subnational governments (and particularly cities) can provide more friendly contexts for the thriving of progressive elites, and above all the conduct of more problemoriented and less ideological politics and governance (Barber 2013).This in turn can facilitate the establishment (or continuance) of cooperative transnational links where international relations are suffering.
Finally, multi-level differentiation can also help vis-à-vis more traditional and formal dimensions of differentiated integration in Europe.The EU IDEA project is premised on the idea that DI -and, at least as regards Brexit, differentiated disintegration -will be a new normal for the EU in the years to come (Pirozzi andTortola 2016, Schimmelfennig 2018).As noted time and again, DI entails a number of advantages for the participants in new integration arrangements, but also some political and legitimacy challenges for the countries remaining outside, which can be especially acute if domestic preferences on integration are polarised (Lavenex and Križić 2019, Rensmann 2020).Network linkages among sub-national actors may provide a way to mitigate these problems, by preserving a concrete sub-national level connection between participants and non-participants in formal DI arrangements (in the form of joint projects as well as exchanges of practices, ideas and policy innovations), and more generally by multiplying societal channels of communication between countries on the two sides of differentiated integration.This is an additional reason for the EU to keep encouraging the development of strong, inclusive and independent sub-national authority networks throughout the continent.EU IDEA Policy Papers No. 8

Appendix: Data collection procedures
Our dataset of sub-national authority networks was constructed in two phases.In the initial phase, we established the perimeter of our sample by combining data from two sources.The first is the EU Transparency Registry (https://ec.europa.eu/transparencyregister), in which many sub-national authority networks operating in Europe are registered under categories III (non-governmental organisations) or VI (sub-national organisations).We downloaded the entire lists of organisations under these two categories as of the end of September 2019, and subsequently proceeded to manually select, with the help of a research assistant, independent and transnational organisations of sub-national authorities with at least one EU-based local or regional government among their members.We then combined the resulting set of organisations with those listed in Tavares (2016).This is, to our knowledge, the only recent and publicly available systematic list of existing sub-national authority networks.After purging Tavares's list of duplicates, and organisations not involving EU-based sub-national governments, we ended up with a combined set of 99 organisations.
In the second phase we manually coded the networks in our dataset for a number of basic variables, using the organisation's websites and official documents as sources (with the only exception of the variable "Year founded", for which the UIA yearbook was occasionally consulted: https://uia.org/ybio).At the coding stage we dropped three organisations from our dataset for lack of reliable data, which brought our final sample number to 96.The table below presents a summary version of our codebook.Differentiated Cooperation

Year founded
Year in which the organisation was founded.

Policy area
Main policy areas (up to three, ranked by importance) in which the organisation works.
Categorical.Its value range corresponds to the 21 categories of the Comparative Agendas project's codebook (https://www.comparativeagendas.net),plus an additional category for generalist organisations.

Total members
Total number of members of the organisation, of any kind.
Numerical; unbounded.(Members that are themselves sub-national authority networks are counted as one.) Non-territorial members Indicates whether the organisation includes members that are not territorial units of government.

Territorial unit type
Type(s) of territorial units in the organisation.
Categorical.Its values are: City: the organisation is composed exclusively of urban actors (regardless of size); Region: the organisation is composed of larger sub-national units, such as regions, provinces, and the like.Mixed: the organisation includes both cities and regions.

Total territorial members
Total number of territorial members in the organisation.

No. of EU members
Total number of territorial units located in EU member states.
No. of non-EU members Total number of territorial units located in non-EU member states.

Total countries
Total number of countries in which the territorial members of the organisation are based.

No. of EU countries involved
Total number of EU member states in which the territorial members of the organisation are based.
Numerical; range 1 to 28 (includes the United Kingdom).

EU countries names
Names of EU member states in which the territorial members of the organisation are based.
Categorical; the range corresponds to the list of EU28 member states (includes the United Kingdom).

No. of non-EU countries involved
Total number of non-EU countries in which the territorial members of the organisation are based.
Differentiation has become the new normal in the European Union (EU) and one of the most crucial matters in defining its future.A certain degree of differentiation has always been part of the European integration project since its early days.The Eurozone and the Schengen area have further consolidated this trend into long-term projects of differentiated integration among EU Member States.
A number of unprecedented internal and external challenges to the EU, however, including the financial and economic crisis, the migration phenomenon, renewed geopolitical tensions and Brexit, have reinforced today the belief that more flexibility is needed within the complex EU machinery.A Permanent Structured Cooperation, for example, has been launched in the field of defence, enabling groups of willing and able Member States to join forces through new, flexible arrangements.Differentiation could offer a way forward also in many other key policy fields within the Union, where uniformity is undesirable or unattainable, as well as in the design of EU external action within an increasingly unstable global environment, offering manifold models of cooperation between the EU and candidate countries, potential accession countries and associated third countries.
EU IDEA's key goal is to address whether, how much and what form of differentiation is not only compatible with, but is also conducive to a more effective, cohesive and democratic EU.The basic claim of the project is that differentiation is not only necessary to address current challenges more effectively, by making the Union more resilient and responsive to citizens.Differentiation is also desirable as long as such flexibility is compatible with the core principles of the EU's constitutionalism and identity, sustainable in terms of governance, and acceptable to EU citizens, Member States and affected third partners.

Figure 1 |
Figure 1 | European sub-national authority networks by decade of establishment [W]e are even thinking to […] start working with [organisation names omitted] based on their membership, because we are becoming … specialised in infrastructure, […] in finance, in […] delivering those projects.[…] [W]e feel more and more that other actors are getting better than us at working with the regions on capacity building, [and being] the voice of the sub-national [level].
shows the incidence of this type of differentiation in our dataset, by ordering networks according to the percentage of EU-based versus non-EU-based members.As the figure shows, of the 96 networks in our sample, only 35 have an all-EU-based membership.The remaining 61 include one or more members based outside the EU, and 25 of these have a majority of non-EU-based members.

Figure 4 |
Figure 4 | EU vs. non-EU membership of sub-national authority networks

Interview 1 :
Sub-national authority network officer (online interview), 6 July 2020 Interview 2: Sub-national authority network officer (online interview), 6 July 2020 Interview 3: Sub-national authority network officer (online interview), 8 July 2020 Interview 4: Three sub-national authority network officers (joint online interview), 14 July 2020 Interview 5: Sub-national authority network president (online interview), 31 August 2020 Interview 6: Sub-national authority network officer (online interview), 11 September 2020 /company/euidea spreaker.com/show/euideaThis project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 822622