Taking stock of the multilevel governance research programme: a systematic literature review

ABSTRACT

Multilevel governance (MLG) is one of the most successful ideas in contemporary political scholarship. Coined in the early 1990s by Gary Marks to examine developments in the European Union’s regional policy, MLG has since been adopted and further elaborated in numerous publications, which have expanded the concept well beyond its initial context, and afforded it wide currency in academic and policy-making circles. As the term turns thirty, this article maps MLG scholarship by presenting the first systematic review of this literature, based on a dataset of 590 publications. In particular, the article identifies key conceptual and empirical developments in MLG studies, assesses strengths and weaknesses of existing research, and informs future work by locating gaps and suggesting directions. By examining MLG research comprehensively, the article also provides a solid basis on which future scholarship may, possibly, reassess some of the longstanding ‘existential’ questions on multilevel governance as a research programme.

KEYWORDS:

• Multilevel governance
• systematic literature review
• European Union
• regional policy
• environmental policy

Introduction

Multilevel governance (MLG) is without a doubt one of the most successful ideas in contemporary political scholarship. Coined in the early 1990s by Gary Marks (1993) to make sense of recent developments in the regional policy of the European Union (EU), MLG has since been adopted, and further elaborated – in the first place by Marks himself, together with Liesbet Hooghe (e.g. Hooghe and Marks 2001, 2003) – in numerous publications, which have over time expanded the concept well beyond its original empirical context, and afforded MLG wide currency in academic as well as policy-making circles (European Commission 2001). To put it like Michael Zürn (2020, 784–785) MLG is ‘one of the most important political science research programmes of the last decades’, and ‘a thriving theme and approach with contributions from all the subfields of political science, from both sides of the Atlantic and with applications to various different issue areas’.

The growth of research on MLG has elicited, over the years, a number of critical reflections aimed at assessing the boundaries of the concept, its empirical applications, the analytical value added of the MLG framework, as well as its limits and blind spots.¹ In this article we contribute to this stock-taking endeavour by presenting the first systematic review of the MLG literature, based on an original dataset of 590 publications. A well-established method in hard science fields such as health and clinical studies, systematic literature reviews are becoming widespread within the social and political sciences – including in such areas as public administration (e.g. most recently Bartocci et al. 2022), policy and governance (Frisch-Aviram, Beeri, and Cohen 2020; Gjaltema, Biesbroek, and Termeer 2020) and, increasingly, European Union studies (Boykin 2023; Bunea and Baumgartner 2014; Höing and Kunstein 2019; Tortola 2014; van der Veer and Haverland 2019). These exercises are aimed at providing comprehensive assessments of the contours, trends, and findings of specific fields of research. The value added of this type of analysis has been summarized effectively by Dacombe (2018, 149–150):

Everyone working in the discipline will, at some point, have carried out a literature review. Indeed, the importance of presenting one’s work in the light of previous research is an academic convention […]. However, systematic reviews are potentially important to political science because they question the underlying approach to these activities. By conforming to the standards usually expected of primary research, systematic reviews are able to demonstrate significant advantages by reducing bias […] increasing transparency […] and recording results and synthesis in a manner which is transparent enough to allow replication.

Clearly, the comprehensiveness of a systematic literature review comes with limits in terms of analytical depth. Systematic reviews afford accurate tracing of ‘the development of research on an issue within and across different research traditions’ (Gough, Thomas, and Oliver 2012, 4), and can inform future work by assessing strengths and weaknesses, locating gaps and suggesting directions. However, the reach of a systematic review in terms of analytical detail is necessarily shorter than that of narrower and more qualitative examinations. The goal of our analysis, therefore, is not to supplant but rather complement previous analyses of the MLG scholarship, by bringing to light aspects and trends of this scholarship that may not be as evident in less comprehensive studies, and possibly by providing future evaluative work with a more accurate backdrop on which to build.

A feature of MLG scholarship is the extent to which the expansion of this research has been accompanied by ‘existential’ (self-)reflections on the ontological status of MLG as a research programme. To a large extent, these reflections stem from the uncertain boundaries and ambiguities of MLG as a concept (which have, ironically, also contributed to the growth of MLG scholarship by providing the concept with a remarkable degree of flexibility). Quite regularly, proponents as well as critics of MLG raise questions about, among others, the applicability of MLG outside of the EU’s sui generis polity; its distinctiveness vis-à-vis pre-existing subjects such as federalism; its normative implications; finally, whether there is a proper ‘theory of MLG’ to speak of (Benz, Broschek, and Lederer 2021b; Jeffery and Peterson 2020b; Piattoni 2010; Stephenson 2013; Tortola 2017; Trein 2022). These are crucial questions which, however, are usually tackled on the basis of unsystematic (and often small) samples of the MLG literature. By examining existing MLG work systematically and comprehensively and offering ‘a forensic account of existing knowledge’ (Dacombe 2018, 151) our study provides a sounder empirical footing on which to base, examine, and possibly revise at least some of these longstanding ‘existential’ questions on multilevel governance in future analyses.²

To summarize our main findings, our systematic review shows that MLG as a concept has travelled in many respects outside its original territorial, disciplinary and policy confines: from the EU as a polity and region of interest to a variety of settings characterized by different types of ‘multilevelness’ outside and beyond the EU (including centralized countries and transnational governance systems); from European studies and political science as academic disciplines to environmental, regional and urban studies among others, with the corresponding shifts in terms of empirical anchorage. However, this exceptional expansion also entails limits regarding coverage, and the closer scrutiny of conceptual material suggests that the expansion has not clarified, and perhaps has even amplified original definitional ambiguities, and that it has not prevented some blind spots from persisting.

The rest of the article proceeds as follows: the next section describes the methodology of our systematic literature review. The third section reflects on the development of the MLG research programme by looking at its expansion over time and across disciplines. Section four isolates the empirical portion of our sample and examines it through the lens of ‘normalization’ (defined in the article as the extent to which MLG has departed from its initial and sui generis empirical context). Section five concentrates on the conceptual portion of our sample – zooming in on articles that are particularly ‘information-rich’ in that respect (van der Heijden 2021, S127) – to assess the advancement, or lack thereof, of the theoretical debate on MLG. In the concluding section we recap our findings and offer some reflections for future research, aimed at remedying knowledge fragmentation, avoiding the pitfalls of conceptual stretching, and developing MLG as a coherent research programme.

Methodology

In line with standard practice on systematic literature reviews, our data collection proceeded in two phases: first, setting the perimeter of our sample; second, defining our variables of interest and corresponding coding. In this section we go over the main steps of both phases.

Sample collection

As mentioned in the introduction, the birth of multilevel governance as a distinctive concept can be located quite clearly in the early 1990s, and more precisely in Marks’s 1993 piece ‘Structural policy and multilevel governance in the EC’ (Marks 1993),³ where the term was first used. Accordingly, we set the starting point of our sample collection at 1993, extending up to and including 2018, thereby covering the first 26 years of MLG research. Within this temporal range we focused on articles published in English, conforming to standard practice on systematic literature reviews.

We are aware of the limitations posed by excluding work published in formats other than articles – primarily monographs and edited volumes, which still make up an important part of publishing in the social sciences⁴ – as well as sources in languages other than English. The latter, in particular, may lead our sample to underrepresent research on non-anglophone countries (a case in point is the German speaking scholarship, in which the use of MLG has a long tradition).⁵ However, we believe that limiting our analysis to a single publishing format presents three clear advantages, which more than make up for its shortcomings: first, it increases our confidence in the relevance of our sources, given the more uniform peer review process underpinning academic journals. Second, it allows us to avoid selection biases due to absence of comprehensive and recognized indices for books and non-English scholarly publications (quite simply, if we expanded to these source types, we would have to arbitrarily decide what publishers and/or languages to include in the study). Finally, it presents us with a single text format on which to code across the entire sample, therefore increasing intercoder reliability, as shown below.

We retrieved journal articles from two of the most comprehensive and widely used scholarly databases, namely Scopus and Web of Science (WoS).⁶ We ran searches for ‘multilevel governance’ OR ‘multi-level governance’ restricting to the title, abstracts, and keywords fields in Scopus, and to the ‘topic’ field in WoS, which includes title, abstract, keywords and ‘keywords plus’ (keywords generated by an algorithm using text references). This search yielded 2,775 items, of which 1,381 from Scopus and 1,394 from WoS.⁷ From this initial sample we removed 998 duplicates, one article retracted for plagiarism, and 25 additional entries, which a closer inspection revealed to be items other than research articles (books, book chapters, book reviews, and a conference report). We further purged the remaining 1,751 items of 103 articles written in languages other than English (albeit published in journals with an English title). This left us with 1,648 articles, of which we could access 1,632 in full text. We had to treat the 16 inaccessible items (less than 1% of our sample) as missing data. These 1,632 articles have been published in 568 journals.⁸

The popularity of MLG means that this term is sometimes used just as a buzzword, without real interest in and engagement with this research agenda. A quick but efficient way to remove this sort of articles from the rest of the sample, is to look at the frequency with which MLG is mentioned in the article, in its extended or abbreviated form. This is consistent with van der Heijden’s observation that ‘a low use of keywords may indicate that the publication is not explicitly dealing with the topic of the meta-research’ (van der Heijden 2021, S127). More precisely, we set a minimum of ten occurrences as a threshold to separate relevant from irrelevant articles.⁹ We are, of course, aware that any such threshold includes a degree of arbitrariness. However, we believe that ten mentions is a high enough number to ensure that the resulting final sample is composed entirely of articles engaging meaningfully with the idea of MLG – thereby safeguarding sample homogeneity and relevance to our analytical goals.¹⁰ This procedure took our final sample down to 590 articles, published in 251 journals. Figure 1 summarizes our sample selection steps by means of a PRISMA diagram (Page et al. 2021) for transparency and replicability.

Figure 1. PRISMA flow chart.

Coding

Due to the nature of our analysis and the variables in which we are interested, we decided to code the 590 articles in our sample manually. As a first step, we coded articles for the variable Academic discipline, using the disciplinary affiliation of the journals in which they were published as a proxy. To do so, the three authors allocated the 251 journals to 13 categories (12 disciplines plus a residual category), first coding separately, and subsequently solving disagreements via discussion and consensus. We then proceeded to transfer the resulting codes to the articles in our sample.

For the remaining variables, we coded our articles based on the text of their abstracts, with the only exception of the variable Countries, for which we expanded to introduction and conclusion whenever the abstract did not contain enough information for the coding.¹¹ We used a codebook developed by the authors through an iterative process of categories definition, trial and refinement to achieve sufficient reliability. After drafting a first version of the codebook, the authors coded a random sub-sample of 60 articles independently – one author coded the whole subsample, while the other two coded 30 articles each, so that each article was coded twice – and later discussed discrepancies, amending the codebook where needed. The revised codebook was subsequently tested on a different subsample of 60 articles. This process was repeated once more for the Policy area variable, which had not reached sufficient reliability after the first codebook revision. Table 1 presents a summary of our final codebook, and intercoder reliability scores as measured through the Cohen’s kappa statistic. A full version of the codebook is attached as Appendix A to this article.

Table 1. Summary codebook.

Variable	Values	Cohen’s k
Analytical orientation	Conceptual / Empirical	0.942
If empirical (n = 480):
Method	Single case / Comparative	0.907
Unit of comparison (if comparative: n = 165)	Territorial / Policy / Both	0.769
Policy area	External relations / Macroeconomic policy / Market regulation / Regional and cohesion policy / Infrastructural and industrial policy / Environmental policy / Urban policy and housing / Research and education / Social and health policy / Agriculture / Institutional and administrative matters / Law and order.	0.910
Highest governance level	National / Supranational (EU) / International	0.940
Countries	All countries, plus EU	n/a

The development of multilevel governance scholarship

Figure 2 shows the distribution of the articles in our sample by year of publication. The first thing to be noted is a rather steady growth of MLG scholarship over time, which in turn confirms more impressionistic assessments of this as a research agenda with ever increasing appeal. Interestingly, our sample includes no articles prior to 1995, namely two years after the introduction of MLG as a distinctive term. This might be, in part, the effect of lag times between research and publication. However, it could also indicate some initial hesitancy among scholars to embrace and expand on this new concept.

Figure 2. MLG publications over time.

The expansion of MLG research seems largely unaffected by the vicissitudes of European integration over the past three decades, such as the Eurozone crisis, the migrant crisis, or Brexit. This is notable given the traditionally close connection between European integration scholarship and politico-institutional developments in the EU (Hooghe and Marks 2019; Tortola 2015). Then again, the foregoing is probably less surprising if one considers two aspects: first, compared with other EU studies paradigms, MLG scholarship has traditionally been more focused on everyday policy-making and less on the high politics and ‘history-making’ junctures of integration. This may make it less susceptible to the Union’s political and institutional problems du jour. The second aspect is the extent to which MLG research has transcended its original disciplinary confines over time. Figure 3 tracks the split of MLG research across disciplines, grouping the latter under three concentric headings: (a) EU studies; (b) political science (inclusive of all its subfields but European integration); (c) other disciplines. Approximately until 2010 MLG work was conducted predominantly within EU studies and the broader field of political science. This changed visibly in the last decade or so under examination, with a number of other disciplines taking over research on multilevel governance, and publications in major EU journals (such as the Journal of Common Market Studies, the Journal of European Integration, and the Journal of European Public Policy) becoming a small minority.

Figure 3. MLG disciplinary breakdown over time.

Figure 4 zooms in on the interdisciplinary nature of MLG scholarship by displaying a static breakdown of our publications sample across individual disciplines. As the chart shows, work in EU studies represents, at this point, only one tenth of published MLG research. The share goes up to 43% if we expand to include all publications in the field of political science (generalist journals, plus journals in the subfields of domestic/comparative politics, public policy and administration, and international relations). This is a significant increase, yet still less than half of the sample. It is interesting to note that within the broad area of political science, there is a clear prevalence of research in the field of public policy and administration. This might suggest that of the two possible directions identified by Tortola (2017) for MLG scholarship – i.e. developing a theory of public policy-making or a theory of state transformation – the former road has so far proven the most popular among scholars.

Figure 4. MLG static disciplinary breakdown.

That much of MLG research so far has been devoted to understanding the functioning of policy-making and implementation in multilevel systems seems confirmed by the other two major slices of the pie chart, namely regional and urban studies, and energy and environmental studies – two predominantly policy-oriented fields of study. The latter of the two is particularly notable as it represents a plurality of existing MLG scholarship. Multilevel governance seems to provide an important framework for analyzing and tackling questions in the growing and inherently multilevel and multi-actor energy and environmental fields. We will return to this topic later in the article, while analyzing the policy area breakdown of empirical MLG research.

As noted in the introduction, the expansion of MLG scholarship over time and across fields is often ascribed to, among other factors, the malleability of multilevel governance as a concept. In introducing a recent symposium on MLG, Jeffery and Peterson (2020b, 763) argue that MLG has become an expansive research programme. Taking Hooghe and Marks (2001) as a pivotal moment in the development of the theoretical framework of MLG, they state that the current state of MLG thinking ‘bears little resemblance to its original conceptualization’. While MLG has been praised as ‘a powerful theoretical account of the distribution of public authority across different territorial scales’ (Jeffery and Peterson 2020b, 762), its ‘impressive extension’ (Stephenson 2013, 821) may have been facilitated by a ‘definitional murkiness’ that has, moreover, increased over time (Tortola 2017, 236). To convey the complexity and ambiguity of the MLG idea, Piattoni (2010, 2) describes it as ‘an umbrella under which many disparate phenomena are subsumed, to the point that it may lose all denotative precision and become ‘over-stretched’’.¹² Being a versatile concept that lends itself to being defined, interpreted, and even stretched in different ways is therefore a double-edged sword: if, on the one hand, it drives the popularity of MLG, on the other it may afflict this research programme with conceptual vagueness, and hinder scholarly dialogue and cumulation – or at best limit it within separate islands of theorizing. This issue has accompanied MLG scholarship virtually from the start, and is reflected in the substantial amount of MLG research devoted to conceptual reflections, criticism, and/or refinement.

To capture this aspect of the MLG literature we have separated articles applying MLG to empirical cases (for descriptive, prescriptive, or explanatory purposes), from research that primarily reflects on and problematizes the concept of multilevel governance.¹³ A total of 110 articles can be classified as conceptual: almost one fifth of our sample. The yearly share of this type of research is shown in Figure 5. No clear overall trend emerges from this chart, except a seeming stabilization of the portion of conceptual scholarship at around 15–20% of the total in the last decade under examination. In general, the figure suggests that the conceptual and theoretical debate around MLG has been rather lively almost from the inception of this literature.¹⁴

Figure 5. Conceptual and empirical literature.

Interestingly, the disciplinary split within the conceptual MLG literature deviates from the overall figures shown above: articles in the field of EU studies and political science broadly defined represent, respectively, 17% (19 items) and 51% (56 items) of the conceptual sub-sample, whereas energy and environmental studies and regional and urban studies count for 13% (14 articles) and 10% (11 articles) respectively. This in turn indicates that the conceptual debate on MLG remains relatively confined within those traditional areas from which the research programme as a whole has been moving away. What this means vis-à-vis the conceptual shortcomings described above – above all whether the MLG research programme is converging towards a shared definitional core or, conversely, keeps circling around the same conceptual questions – can only be established through a closer look at the conceptual part of our sample. This will be the subject of the fifth section of the article. Before that, in the next section, we delve deeper into the empirical portion of our sample.

The empirical MLG literature: gauging normalization

As noted at the beginning of the article, the idea of multilevel governance originates in a very specific empirical context, namely the regional and cohesion policy of the European Union. In exploring the empirical portion of our sample, therefore, we take that context as an initial benchmark to see the extent to which MLG research has departed from it over the years. To capture this question, we refer to the idea of ‘normalization’. Within the context of EU studies, the term normalization has long been employed to indicate the EU’s increasing resemblance to a proper polity as a consequence of greater integration, and its resulting amenability to being studied and compared with other political systems (primarily federal ones) through the conventional categories and theories of comparative politics (Caporaso et al. 1997; Kreppel 2012; Tortola 2014). By analogy, we define the normalization of multilevel governance, and the MLG research programme, as the degree to which this concept has transcended the idiosyncratic institutional and policy context in which it was coined, and been applied to a broader range of political referents, hence joining, at least prima facie, the ‘normal’ analytical toolkit of political studies.

As a first step in our analysis of MLG normalization, we explore the degree to which the MLG scholarship has travelled outside the sui generis context of the European Union, building on Tortola (2017), who has identified this as a major source of ambiguity in MLG research. This exploration entails a twofold question: the first is whether MLG has been applied to political systems geographically outside (totally or partially) of the EU.¹⁵ The second, institutional in nature, is whether MLG has been applied to political systems other than the EU. The two questions overlap to a significant extent but not entirely. The MLG framework may, for instance, be used to examine multilevel policy-making in a federal country that happens to be a EU member state, without involving the Union as a relevant institutional level. On the other hand, the MLG lens may be applied to a non-EU political system which is nonetheless linked to the Union as regards the policy dynamics under examination (as it may happen, e.g. in studies of the EU neighbourhood policy). What matters for our purposes is that both cases constitute departures from the original empirical context of MLG.

Of the 480 empirical articles in our subsample, 210 (44%) explore multilevel governance relationships whose highest institutional level is not the EU, but either an international organization (122 articles), or a nation-state (88 articles). As for the geographic criterion, 194 articles apply MLG outside the borders of the EU, and of these, 126 include only non-EU political systems.¹⁶ Among the non-EU countries most frequently studied through the lens of MLG are federations such as Canada (37 articles), the United States (29) and Australia (13), but also unitary systems like China (14) and Norway (12). On the EU side, the most popular empirical cases are the United Kingdom (68 articles), Germany (61), Sweden (28), Italy (27), Spain (24), the Netherlands (21), and Belgium (19).¹⁷ Figure 6 shows the yearly split between EU-centred MLG studies, and articles that transcend the EU geographically and/or institutionally. As the figure shows, the latter have steadily gained importance within MLG scholarship, and became a majority over the last decade under examination.

Figure 6. MLG outside the EU.

As a second step we examine the extent to which MLG so far has travelled outside of its original policy field, that is regional policy. Looking at the distribution of our empirical subsample, shown in Figure 7, regional policy emerges as an important, yet not the largest group, containing less than one in five articles. Significantly larger is the subset of publications in the field of environmental policy – a finding consistent with the popularity of MLG in energy and environmental studies, discussed in Section 3. Institutional and administrative matters, and social and health policy come third and fourth, respectively. Taken together, these four policy areas represent three fourths of MLG research to date.

Figure 7. MLG empirical research by policy area.

Figure 8 shows how the policy split within MLG literature has evolved over time, by tracing the four policy areas just mentioned, plus a residual category containing the remaining areas. Here too, the first trend to be noted is a historical decline of regional policy as a subject of MLG-oriented investigation. A similar, yet less marked relative decline concerns the field of institutional and administrative matters. On the other hand, the relative weight of the remaining two areas analyzed has grown approximately from the mid-2000s. This is particularly the case for environmental policy, which has established itself rather stably as the dominant policy area within MLG scholarship in the last decade under examination.

Figure 8. MLG policy areas over time.

All considered, the above indicates that, while there has certainly been an ‘emancipation’ of MLG research from regional policy, the extent to which this is leading to empirical normalization (in the way we have defined the term here) is uncertain at best, for there are signals of a possible re-concentration of this literature within the study of environmental issues. This may in turn result from two, not necessarily exclusive factors. On the one hand, MLG may be providing a particularly helpful analytical lens, and probably also a prescriptive idea, with regard to the distinctive aspects – particularly the multilevelness and multi-actorness – of environmental problems and policy. A special affinity of this sort between the MLG concept and the field of environmental studies might be leading to re-establishment of a new sui generis connection replacing the old one with the EU’s regional policy. On the other hand, however, the increasing application of MLG to the study of environmental issues may just reflect the growing importance of environmental policy as a societal problem and a field of study in recent years (e.g. UNESCO 2021). Seen this way, what emerges from the data is not so much an idiosyncratic connection but more simply the statistical result of a broader increase in publications on environmental issues.

As a third and final step, we examine the methodological make-up of the empirical MLG literature, separating, for simplicity, single case from comparative studies. The idea behind this distinction is twofold. First, the frequency of comparative studies will give us an (albeit imperfect) indication of the degree to which the MLG is formulated as a general concept, able to travel across time and space, as opposed to being applied to single and potentially idiosyncratic politico-institutional settings. Second, because comparison is often used as a way to control for variables, the frequency of this method will also give us some initial information on the causal vs descriptive/prescriptive nature of MLG literature.

Approximately two thirds of our empirical sub-sample (315 articles) is composed of singe case studies. Interestingly, of the remaining 165 comparative studies, 154 compare territorial units, while only 11 compare policy areas – a finding perhaps consistent with the tendency for MLG research to be concentrated in very few policy areas, discussed above. Figure 9 shows the relative distribution of single case and comparative studies over time. To note here is the absence of any discernible trend in the MLG methodological split: apart from a few short-lived exceptions, single case studies have generally remained the main mode of analysis in MLG research. One possible interpretation of these results, besides the ones sketched above, is that the complexity required in a MLG analysis even of a single case – tracing institutions and policies across sequences and levels – might discourage cross-systems comparisons.¹⁸

Figure 9. Single case and comparative studies over time.

To summarize the above findings, and going back to our initial question, MLG research has certainly become increasingly detached from its original empirical context, namely EU regional policy. The extent to which this implies normalization, however, is not as clear. For one thing, the degree of empirical normalization found is not symmetrically distributed across the three aspects examined in this section – namely the geographic/institutional, policy, and methodological aspect – but is instead concentrated predominantly in the first, in the form of an increasing application of MLG outside of the EU. For another, as already discussed, there seem to be indications of a regrouping of MLG within the area of environmental studies. Whether this will lead MLG to a new sui generis equilibrium (from EU studies to environmental issues), remains to be seen.

Zooming into the conceptual approaches to MLG

Based on a reading of our 110 conceptual articles, this section explores the theoretical debate on MLG asking, broadly, to what extent the proliferation of contributions to MLG thinking across time and disciplines has reduced blind spots, yielded conceptual refinement or, conversely, added to theoretical confusion. For convenience, in what follows we refer only to a part of our conceptual sub-sample. A more comprehensive list of references is included as Appendix B to this article.

MLG across time: conceptual development and remaining blind spots

The first two conceptual articles in our sample were published in the mid 1990s, respectively by Liesbet Hooghe (1995) and Gary Marks (1996). The next conceptual articles did not come out until 2000. This early group of publications was mainly concerned with the drivers of the mobilization of subnational governmental actors in the EU, and their influence on the institutional balance of power in the Union. A highly cited piece by Charlie Jeffery (2000) reflects critically on MLG, expressing doubts about its generalizability beyond structural policy, and at the same time advancing hypotheses on factors likely to condition the differential impact of MLG on decision-making.¹⁹ Responding to Jeffery’s critique, Marks and Hooghe (2000) operate a junction between MLG and economic thinking on the optimal scale for the allocation of authority and the conduct of policy, based on works on fiscal federalism.²⁰ In this period, legal scholarship also started expressing interest in MLG, particularly as the concept marked ‘a move away from the claims of exclusive authority and political legitimacy on the parts of sovereign states towards a more horizontal structure of interlocking and interdependent systems of regulation’ (Harding 2000, 144).

The mid-2000s saw an acceleration in the number of conceptual articles on MLG, as well as its circulation outside EU scholarship – e.g. in the policy evaluation community (Stame 2004). About ten years after the initial formulation of MLG, Bache and Flinders (2004a) – who published in the same year an edited collection entitled Multi-Level Governance that included a chapter by Marks and Hooghe – used MLG to challenge the standard Westminster model of British politics. Aalberts (2004) critically discusses the concept through the lens of social constructivism in international politics. Bulkeley and Betsill (2005; Betsill and Bulkeley 2006) popularized MLG in two influential articles on the local governance of climate change. In 2008 Mahoney and Baumgartner (2008), two prominent North American policy analysts familiar with European politics, further expanded the MLG paradigm in a comparative study of the EU multilevel system and US federalism, focusing on opportunities for ‘venue shopping’ by interest groups in the two systems. Equally innovative in this article was the connection of MLG with the different variants of neo-institutionalist scholarship. In parallel, some scholars further developed the MLG concept in the analysis of the EU multilevel system: van Tatenhove, Mak and Liefferink (2006), for instance, tease out different informal MLG strategies,²¹ while Newig and Koontz (2014) make the connection with participatory implementation. Dickson (2014) connects MLG to the ‘paradiplomacy’ of regions, Benz, Corcaci, and Doser (2016) apply the concept to administrative coordination, and Guderjan and Miles (2016) shift the focus to the role of local governments in MLG. Generally speaking, this is a period in which the concept is applied to new polities, actors (the bureaucracy, interest groups, cities), and processes (informal, evaluation).

In the second half of the 2000s a more distinctly normative research agenda emerged, responding to the perceived neglect of issues of democratic legitimacy in the debate sketched above. Hunold (2005) highlights the risk that participation mechanisms in MLG inhibit contestation by co-opting civil society organizations, while Kronsell (2005) presents a feminist critique of integration theories that deplores the neglect of gender dynamics in MLG. Finally, Papadopoulos (2005) raises issues of social justice and democratic accountability, further developing the latter topic in a well-received piece published two years later with other articles on the topic in a journal special issue dedicated to ‘Accountability in EU Multilevel Governance’ (Papadopoulos 2007).²² Legitimacy issues related to the democratic deficit of the EU were further taken up by Hurrelmann and Debardeleben (2009) and Piattoni (2009), who anchored her reflections on MLG to the debate on ‘input’ legitimacy in the EU. Piattoni shortly thereafter published a comprehensive and very influential book on MLG, in which she stated that MLG ‘invites normative reflections on the conditions under which binding decisions gain widespread acceptance and bestow legitimacy on the institutions that produce them’ (Piattoni 2010, 1). However, issues of pragmatic legitimacy in the context of MLG – trust in rule-making authorities in the face of authority shifts and consent to policy outputs likely to generate winners and losers – are only marginally addressed in the literature, primarily through the angle of compliance of lower jurisdictional levels in processes of policy implementation. More broadly, we did not find any strong connection to the ‘post-functionalist’ view of socio-political conflict triggered by supranational integration that Hooghe and Marks (2009) developed more recently. Politics in the partisan and the social movements arenas and the role – if any – of the mass public remains a marginal topic in the conceptual literature on MLG.²³ Functional approaches to MLG are predominant, even though the identity logic – defined as ‘an expression of the desire for self-rule by a group that sees itself as a distinct community’ (Hooghe, Marks, and Schakel 2020, 196) – is seen as a major driving force of MLG too.

MLG in new territories: increased polysemy

Jeffery and Peterson (2020b) ask if multilevel governance gives analytical purchase beyond the European Union. The influence of findings based on research on local collaborations, national federalist systems, and international regimes became visible in the distinction between type I and type II of MLG (Hooghe and Marks 2003). Piattoni (2010) referred in many places to similarities between federalist thinking and conceptualizations of MLG, and the concept became increasingly popular among scholars of federalism (Tortola 2017).²⁴ At times, MLG is also used in scholarship on implementation and discussed in conjunction with other parent frameworks of governance ‘with adjectives’ – such as territorial, polycentric, participatory, adaptive, network or hybrid governance. The conceptual ‘translations’ are usually accurate, and misinterpretations are rare. This is unsurprising, as the ‘brokers’ are sometimes political scientists or specialists of public administration who publish outside their discipline. For instance, in a reflective piece published in an interdisciplinary regional studies journal, political scientists Alcantara, Broschek, and Nelles (2016, 34) acknowledge that ‘as scholars began to export the concept outside of the EU, significant conceptual confusion and muddiness ensued’, and proceed to an exercise in clarification and in ‘concept shrinking’. They present MLG and intergovernmental relations as distinct instances of multilevel politics – differentiated by actor constellations, scales, and decision-making processes²⁵ – that can be simultaneously present within the same polity, and even within the same policy sector. However, with some exceptions, most notably in legal scholarship,²⁶ we notice no significant advances in the paradigm when MLG is used outside of political science at large (including public policy and public administration). More precisely, one can distinguish between three categories of texts with an unequal contribution to theory development.

A first set of authors develop largely critical reflections on MLG. For instance, in a theoretically well-informed piece, Faludi (2012) highlights the ambiguities of the concept and critically takes stock of the penetration of MLG narratives in planning discourse. In a similar vein, Bulkeley and Betsill (2013, 144–145) criticize the ‘overtly ‘statist’ focus of many multilevel governance analyses’ and claim that research should ‘engage more critically with where the authority and capability for addressing climate change as an urban problem lie’.²⁷ A second group of texts demonstrates familiarity with the political science debates on MLG without, however, bringing new arguments. Büchs (2009), for example, reflects on the role of non-public actors in MLG, while Termeer, Dewulf, and van Lieshout (2010) compare MLG with ‘monocentric’ and ‘adaptive’ governance as conceptual lenses to apprehend ‘multiscale problems in today’s network society’. Finally, a number of studies, frequently on environmental matters, tend to use MLG in a way that is only loosely connected to debates in political scholarship – thereby tending to confirm the more general validity of the assessment by Rykkja, Neby, and Hope (2014) on the lack of vital insights from political science and public administration research into research on climate change policy. For instance, Lidskog and Elander (2010) examine problems of representation, participation and deliberation in the governance of climate change in an article that is well-informed by scholarship on democratic theory but does not really engage in a conversation with MLG scholarship on these problems. In another study, Mwangi and Wardell (2013, 81) reflect on how far the concept of MLG needs to be ‘stretched’ when applying it to coordination issues across scales in natural resource management, but surprisingly relate the emergence of MLG not only to ‘mainstream political science’ but also to ‘the critical influence of Foucault’s seminal concept of ‘governmentality’’. It even happened that references to MLG appeared in a theoretical text on environmental governance without any mention of Hooghe and Marks’s work (Farrell 2004).

To summarize, MLG as a concept has undeniably expanded not only beyond EU studies, but also beyond studies of politics in general. This can undoubtedly be considered as a sign of the concept’s broad appeal, as it inspired research and generated empirical expectations in new areas. When exported from its initial ‘home’, MLG is often blended with parent concepts and adjacent research programmes, and the concept tends to become increasingly polysemic, together with its stretching: MLG has become more encompassing and more broadly applicable, but also less specific. Besides, conceptual expansion has not led to major theoretical breakthroughs, although one should consider especially legal scholarship more favourably in that respect. MLG appears sometimes unquestioningly as a ready-to-use conceptual lens, and even – in more applied research – as the optimal design for policy, while critical debates on the concept remain the exception.

Enduring ambiguities on the conceptual front

In his conceptual analysis, Tortola (2017) suggests that MLG remains ambiguous not only about its applicability outside of the EU – an aspect that we have examined empirically in section four – but also regarding two other dimensions.

The first is the role of non-state actors: MLG is often associated with the ‘governance turn’ in policy studies emphasizing the collaboration of public authorities with non-public actors in the formulation and implementation of policy: ‘The multi-level concept then contained both vertical and horizontal dimensions. ‘Multi-level’ referred to the increased interdependence of governments operating at different territorial levels, while ‘governance’ signalled the growing interdependence between governments and non-governmental actors at various territorial levels’ (Bache and Flinders 2004b, 3).²⁸ It has nevertheless been argued that MLG research depicts above all linkages between tiers of government (Börzel 2020) whose spheres of competence are ‘ill-defined and shifting’ (Marks 1993, 407), while the non-state component of MLG is ‘relegated to a secondary role vis-à-vis the territorial one, if not ignored altogether’ (Tortola 2017, 237).²⁹

Second, and actually relatedly as we shall see, it remains unclear whether MLG focuses on the transformation of formal institutional structures (for example through partnerships across governmental levels) or more broadly on policy processes, including their informal component (see also Piattoni 2009). The latter aspect can also be inferred by the fact that MLG is ‘replete with network metaphors’ (Bache, Bartle, and Flinders 2016, 528): relatively fluid – but also elusive – networks are assigned a crucial role in negotiation processes that defy the formal hierarchies of power. Despite this, the connection with the literature on collaborative modes of governance – empirically mostly dealing with networks of policymaking at the regional or local level – remained loose and exceptional.³⁰

The public authority and institutionalist biases were already visible in Marks’s original definition of MLG: ‘I suggest that we are seeing the emergence of multilevel governance, a system of continuous negotiation among nested governments at several territorial tiers – supranational, national, regional, and local – as the result of a broad process of institutional creation and decisional reallocation that has pulled some previously centralized functions of the state up to the supranational level and some down to the local/regional level’ (Marks 1993, 392; original emphasis). This was reiterated in the preface of Hooghe’ and Mark’s monograph that was partly based on previous work: MLG ‘describes the dispersion of authoritative decision making across multiple territorial levels’ (Hooghe and Marks 2001, xi). On the other hand, Marks et al. (1996, 167) merged interest groups and subnational governments when talking about subnational actors, and formal and informal consultation of NGOs as well as public-private partnerships are recurrent features in the MLG literature on global policy-making.³¹ Let us also take the case of the European Union’s cohesion policy, which is the prototypical example of MLG: it is ‘based on ‘partnership’ among the European Commission, national governments, subnational governments, and private actors’ (Hooghe and Marks 2001, xi; emphasis added). This form of partnership, that includes ‘social actors, such as trade unions, firms, and local public interest groups’, is even seen as ‘the chief institutional innovation of cohesion policy’ (Hooghe and Marks 2001, 108). However, when empirically studying policymaking in cohesion policy, Hooghe and Marks only assess the influence of official public authorities – the European Commission, and national, regional as well as local governments (Hooghe and Marks 2001, 93–104). The distinction between types of MLG does not help for clarification either: Hooghe and Marks (2003) see partnerships and self-rule by diverse voluntary groups as more common in so-called MLG Type II (fluid, problem-driven and task-specific) governance structures. Nevertheless, for Hassel (2010), highly institutionalized private actors such as associations can be formally integrated in Type I regimes too, and the difference with Type II regimes is then just that organized interests tend to be less institutionalized and more ad hoc, such as fluid coalitions and movements that gain access through ‘voice’ strategies.

We examined more closely how the articles of the conceptual subsample address the following three dimensions: focus on EU governance system vs focus on other systems; concentration on public actors vs inclusion of non-public actors in the analysis; institutional-formal vs processual approach. Although Tortola simplifies these as dichotomous oppositions, we observed that the difference is frequently one of degree: for instance, some analyses privilege public actors but do not overlook the role of non-private ones, or mainly focus on formal processes without ignoring informal aspects altogether.³² Our examination confirms the enduring ambiguities across publications, while suggesting that Börzel’s criticism on a statist bias appears exaggerated after more than two decades of research on MLG: although a significant amount of work focuses indeed on public authorities, most conceptual pieces consider non-public actors as well. Our study also reveals that two patterns prevail: studies on public actors tend to adopt an institutional-formal approach, while studies including also other actors opt for a more processual approach, with this latter pattern dominating in studies of governance systems other than the EU. This suggests that the ambiguities detected by Tortola (2017) are, at least to some extent, correlated.

Conclusion: an unfinished intellectual journey

Our systematic literature review clearly shows that the idea of multilevel governance has travelled well outside its original territorial, disciplinary and policy confines: the EU as a polity and region of interest, European studies as an academic discipline, and regional and cohesion policy as an empirical anchorage. Since the 2010s, publications on the EU became the exception, and the MLG framework is now not only applied to many countries with different political systems, but also to global decision-making. This denotes how the concept has expanded from EU studies to political science scholarship more in general but, perhaps more importantly, one should note the diffusion of MLG to other disciplines such as urban, regional, or environmental studies. Furthermore, the steady growth of MLG scholarship over time confirms this to be a research agenda with ever increasing appeal that now goes much beyond its original ‘nest’ within EU regional policy to cover several new policy areas: most notably environmental, but also administrative and other matters. MLG is a carrier of inspiration for research in many disciplines and is capable of shedding light into very diverse processes in significantly different governance systems. These are major achievements, and it is difficult to think of other conceptual constructs generated within the framework of the European Union that have successfully departed from their original empirical context and acquired so much resonance and recognition. However, this exceptional expansion also entails some limits with regards to the ‘normalization’ of the concept, it did not clarify and perhaps even amplified some original ambiguities, and it has not prevented some blind spots from persisting.

With regards to empirical limits, although MLG research has greatly increased its geographical and institutional reach, a global north bias persists, which is however a more general problem of political science research. Μoreover, the prevalence of single case studies, together with the progressive re-concentration of MLG literature on environmental issues, might not be optimal for further ambitious theoretical developments. As we have also seen the concept is also not devoid of ambiguities, and its extension to a wide array of constitutional forms – among which federalist as well as centralized polities – did not make it analytically sharper, although it is surely in line with an understanding of these forms not as opposites but as part of a continuum (Hooghe, Marks, and Schakel 2020, 199; Toonen 2010). A closer look into the theoretical scholarship on MLG reveals an enduring ambivalence about what precisely constitutes MLG in terms of actors, institutions, and processes. It is mainly two perspectives that prevail: scholarship that concentrates on public actors tends to adopt an institutional-formal approach, while scholarship considering also other actors opts for a more processual approach, with this latter pattern dominating in studies of governance systems other than the EU. This is in line with the ambiguities detected by Tortola (2017), but it also suggests that they are, at least to some extent, correlated. Finally, with regards to blind spots, the predominant functionalist problem-solving angle on MLG has not left much space to fine-grain the assessment of the democratic properties of MLG and to scrutinize whether it is regarded – both normatively and empirically – as a legitimate mode of governance.³³ More generally, despite the ‘post-functionalist’ turn in European integration studies initiated by Hooghe and Marks, the ‘politics’ of MLG – such as policy feedback and the related cleavages between winners and losers – are not at the core of conceptual discussions. Furthermore, in areas outside the original home of MLG work, critical pieces and debates remain the exception.

Let us conclude then by suggesting a few avenues for future research. The first direction can be to extend and deepen the present exercise of a systematic literature review. A qualitative study of the large empirical sample of works on MLG is warranted to acquire a more comprehensive picture of the uses of the concept. Questions remain as to what extent the ambiguities identified time and again by scholars – which were confirmed by our scrutiny of the conceptual subsample – are also those that are visible in empirical works. Further, do we encounter in empirical MLG research the same blind spots as we have observed in the theoretical development of the concept? Kleider (2020), for instance, sees three research programmes emanating from the original conceptualization of MLG: the role of identity, of political contestation, and the impact of MLG on policy. We have seen that the first two of these topics – actually interrelated in part – hardly penetrate conceptual debates: are they maybe more present in empirical studies? As for the third of Kleider’s themes, after almost three decades of empirical works on MLG, it is worth systematically scrutinizing how the concept is used as an explanans. MLG has been criticized for not being conducive to proper causal statements. Although Marks et al. (1996, 167) cautioned that the concept did not have such ambitions, in terms of policy it is expected that MLG improves policy-makers’ problem-solving capacity, experimentation and learning through diffusion across jurisdictions, but MLG is also likely to cause coordination dilemmas, policy blockade, and to lead to implementation gaps (Benz 2021, chapter 2; Trein 2022, 65–66). Are there any regularities, prevailing patterns and trends in the causal processes observed, and what lessons can be drawn from them? We have also seen that the more MLG diffuses, the more its original meaning and the connections with the foundational pieces on the topic become hard to discern. It would be highly interesting to further test and refine such claims through an analysis of citation networks that can shed more light on the intellectual filiations and the elective affinities between scholars working with the concept of MLG.

The second direction derives from the fact that MLG may not be the only game in town. The thesis of dispersion of authority and power-sharing must be tested against rival narratives of power (re-)concentration, be it within domestic executives, or within supranational institutions and technocratic bodies. Most notably, a competing narrative emphasizes the ‘presidentialization’ of politics (Poguntke and Webb 2005), i.e. executive dominance and the concentration of power around the heads of government, a small and informal group of their favourite ministers, and expert staff of policy professionals, public servants as well as PR specialists (regardless of any formal authority shift beyond and below the national level). An equivalent narrative at EU level is the ‘intergovernmentalist’ paradigm, which regained currency in recent years with the advent of the Economic and Monetary Union (Fabbrini 2015) and with ‘emergency politics’ to cope with successive crises (Schmidt 2022). Further, the power of international bureaucracies newly received increased attention in the study of global policy-making (e.g. Knill and Steinebach 2022). The authors of this contribution plead for a conversation between scholarship on MLG and scholarship relying on rival paradigms.

Our systematic review also suggests a third direction: in view of the remaining ambiguities and blind spots of the concept there is still space for creativity, or at least refinement, to make MLG theoretically more fruitful and enhance its heuristic value. We believe it is perfectly legitimate to apply the concept to a variety of settings characterized by ‘multilevelness’ outside and beyond the EU, including centralized countries and transnational governance systems. As to the processes that the concept should capture, they should clearly not be limited to those that are formal since we know that informal cooperation and negotiations do matter for policy outcomes. We think however that it is confusing to subsume under the MLG umbrella so different things as intergovernmental relations and public-private partnerships. Therefore, depending on the actors involved, it is necessary to differentiate between multilevel government and multilevel governance. The wide currency of MLG proves that it is able to capture something that other concepts cannot: above all ‘antagonistic cooperation’ (Grodzins 1966) across jurisdictional levels. The latter might be becoming ever more relevant – for instance because of the increasing salience of environmental issues. Nevertheless, the expansion of MLG has not been followed by more intense cross-disciplinary debates. Therefore, we also plead for deepening the exchanges between political science work on MLG and research in other disciplines, which so far have remained rather marginal. Such a cross-disciplinary conversation is necessary to remedy knowledge fragmentation, to avoid as much as possible the pitfalls of conceptual stretching, and to develop MLG as a coherent research programme.

Acknowledgements

The authors would like to thank three anonymous reviewers and participants in the University of Groningen European Politics and Society research colloquium for helpful comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 See recently Trein, Thomann, and Maggetti (2019); Jeffery and Peterson (2020a); Benz, Broschek, and Lederer (2021a); Trein (2022); and systematic reviews of the related concepts of polycentric (Baldwin 2023) and network (van den Oord et al. 2013) governance.

2 For example, on the applicability of MLG: according to Piattoni (2010, 255) ‘possible instances of MLG include not just cohesion, environment, and higher education, but also, for example, agriculture, transportation, and tourism’. In a similar vein, Jeffery and Peterson (2020b) ask if multi-level governance ‘travels’ as an analytical framework beyond cohesion policy. Their tentative answer is that it does, albeit to a varying extent, depending on policy sectors. Stephenson (2013, 822–823) also lists many policy sectors to which MLG has been applied after its inception. As shown below, our rigorous method of literature selection allows us not only to describe MLG’s policy span more accurately, but also trace its evolution over time.

3 In a previously published work, Marks (1992, 192) prefigures the concept with mentions of a ‘multilevel perspective’ and of a ‘complex, multilayered, decisionmaking process stretching beneath the state as well as above’ (221).

4 Restricting our selection to articles means, among other things, that Marks’s 1993 piece – a book chapter – was excluded from our sample. This might be ironic, but it is not a substantial problem for our analysis, which is concerned primarily with the way MLG research has evolved after Marks’s introduction of the concept.

5 For instance, German-speaking authors use the MLG lens to study federalism and to make comparisons with the compound polity of the European Union; Hooghe and Marks (2001, xiii) write that they are motivated by similar research questions.

6 More precisely, we used the following WoS collections of peer-reviewed articles: Science Citation Index Expanded (SCI-EXPANDED); Social Sciences Citation Index (SSCI); Arts & Humanities Citation Index (A&HCI); Emerging Sources Citation Index (ESCI).

7 In both Scopus and WoS, searching for ‘multi-level governance’’ also captures articles in which the term is not hyphenated (‘multi level governance’). On the other hand, our search terms excluded cases in which MLG is referred to via different expressions: for instance, when the terms ‘multilevel’ (or ‘multi-level’) and ‘governance’ are not next to each other (as in the example of ‘multi-level intervention and governance’), or when multi-level governance is not explicitly mentioned but implied in the wording (as in the case of ‘multi-level approach’). These are, however, idiosyncratic and marginal cases, which do not affect the overall quality of our sample.

8 To compare this figure with some previous estimates, Enderlein, Wälti, and Zürn (2010) find 150 articles in 15 different academic journals from 2000 to 2009, while Tortola (2017) estimates 365 articles on MLG between 1995 and 2014.

9 In some cases, the term multi-level or multilevel governance appears in the running head of the article, thereby artificially inflating the number of occurrences. To control for that, we manually scanned our sample and eliminated articles that meet the threshold of 10 mentions only as a result of their running head.

10 Clearly, the flip side of this strategy is the risk of excluding some ‘false negatives’, i.e. articles that are substantively about MLG even though they mention the term fewer than 10 times. On balance, we deem this a price worth paying in order to have a homogeneous sample for our analysis. This is, however, yet another reason to regard our study as complementary to more qualitative reviews, as explained in the introduction.

11 This is a manifest variable – essentially the list of countries mentioned in the text under examination – for which intercoder reliability concerns are negligible. For the same reason, we decided not to run reliability tests on Countries, as shown in Table 1.

12 Stephenson (2013), for instance, sees five types of uses of MLG and ten focal points in research on MLG, while Tortola (2017) identifies three axes of ambiguity, generating eight definitional possibilities for MLG.

13 We interpret this ‘conceptual’ category extensively, including not just research focusing on the meaning and definition of MLG, but also articles reflecting on the normative implications of the concept, overviews of MLG scholarship, and a few introductions to special issues.

14 Of course, this sort of assessment is inevitably relative to the novelty of the concept under examination. In a similar analysis conducted on the concept of ‘coopetition’ in management literature Dorn, Schweiger, and Albers (2016) find that as much as 36% of the surveyed articles are conceptual in nature.

15 Because our sample does not include articles published after Brexit, for the purposes of our analysis the United Kingdom is counted as a member of the EU.

16 In 17 articles, we found no lists or other indications of the countries analyzed in the text.

17 Overall, this confirms MLG as a research agenda focused primarily on the global north, as noted by Benz, Broschek, and Lederer (2021b, 281–282) and by Giraudy and Niedzwiecki (2022). Similar geographic biases were found in a systematic review of studies dealing with the concept of policy entrepreneurship in the related field of policy studies (Frisch-Aviram, Beeri, and Cohen 2020), and in political science research in general (Wilson and Knutsen 2022). Note that, to some extent, the geographic coverage observed in our sample might also result from our exclusion of non-English language articles, as discussed above.

18 We should note, however, that this does not seem to be a peculiarity of the MLG literature: for instance, the systematic review of studies on policy entrepreneurship found that 87.9% of the articles relate to a single country case study (Frisch-Aviram, Beeri, and Cohen 2020).

19 In another often-cited article published soon thereafter, Jordan (2001, 201) also takes issue with the concept of MLG, arguing that it offers only a partial, and ‘not especially novel’, account of European integration, which became ‘popular because it captures the mood of the times’. Further, ‘MLG provides an appealing picture of what the EU looks like but is weak at explaining which levels are the most important and why, and what actually motivated the experiment in governance in the first place’ (Jordan 2001, 194, original emphasis).

20 Börzel and Hosli (2003) further elaborate on the contribution of federalism research for the understanding of the EU as a multilevel system.

21 A book was also published on the topic a few years before: Christiansen and Piattoni (2003).

22 See also the volume edited by Benz and Papadopoulos (2006).

23 By contrast, research on German federalism for example reveals the existence of frictions between the cooperative requirements of intergovernmental negotiations across levels and the competitive logic of the party system: see Lehmbruch (2000). On the de-politicizing effect of MLG (and the limits thereof) see Papadopoulos (2017).

24 Giraudy and Niedzwiecki (2022) suggest that MLG shares some commonalities as a research programme with the subnational research (SNR) programme, however there is not much communication between them. Consequently, each of them has blind spots that become apparent if one is familiar with the other research strand.

25 In a nutshell, MLG is described as closer to collaborative governance than to hierarchic government. See also Fossum (2017), who presents ‘poly-cephalous’ federalism as a better description of the EU than MLG.

26 Petersmann (2011) bridges political science and international law by arguing that the multilevel governance of global ‘interdependent public goods’ should be embedded in a legal framework of ‘multilevel (and later cosmopolitan) constitutionalism’.

27 See Eckersley (2017) for a similar criticism regarding local policy-making.

28 Similar definitions can be found for instance in Enderlein, Wälti, and Zürn 2010, 4; Kohler-Koch and Rittberger 2009, 8; Stephenson 2013, 828; Trein 2022, 64–65. This is also the dominant approach in German scholarship on MLG. See Benz (2010, 214): ‘In contrast to terms like federal or confederal, the notion of governance refers to the fact that more often than not private actors are involved in policy-making’; ‘The inclusion of private actors has been said to distinguish MLG from the traditional concept of ‘intergovernmental relations’ among executives, and should imply a broader perspective on actors and structures’ (Benz 2019, 392); see also Behnke, Broschek, and Sonnicksen (2019, 3–4): ‘Actors involved represent formal political authorities as well as often quasi-governmental or even non-governmental actors representing society-based interests’. Tatham (2022) has an even broader view of MLG that includes authority transfers to non-majoritarian institutions (central banks, independent agencies and courts), to firms through privatization, and to lay citizens (through participatory mechanisms).

29 Tatham (2022) estimates that only few scholars study the different dimensions of power dispersion at the same time, and Benz (2021, 15) finds the ‘governance’ aspect in MLG more difficult to define than the ‘multilevel’ aspect.

30 See among book-length contributions Klijn and Koppenjan (2016), Skelcher, Sullivan, and Jeffares (2013) and Sørensen (2012).

31 See for instance the contributions by Beisheim, Campe, and Schäferhoff (2010) and by Scholte (2010) in the Handbook on Multi-level Governance.

32 For example, Bache, Bartle, and Flinders (2016, 532; original emphasis) deplore the absence of a ‘critical reflection on the internal or external consistency’ of the binary divide between MLG Type I and Type II structures. They provide convincing empirical examples of hybrid types among the entities that they study.

33 For instance, the initial reflections on the normative implications of MLG have not triggered much empirical research on the topic, unlike the developments in the adjacent research programme on ‘network’ governance following initial work by Klijn and Skelcher (2007) among others.