Accountability and Multi-level Governance: More Accountability, Less Democracy?

Abstract

This paper seeks to explain why the trend towards more cooperative forms of policy-making, though in all likelihood necessary for policy efficiency and even at first glance promising with respect to inclusiveness and pluralism, can have negative consequences for democratic accountability. The paper first explores the properties of multi-level governance that lead to a deficit in democratic accountability (lack of visibility, uncoupling from representative institutions, composition of networks, and ‘multi-levelness’ itself) before coming to more general conclusions on the characteristics and limits of accountability mechanisms in multi-level governance and on their consequences for democracy.

Though the amplitude of the trend varies, it is hardly possible to dispute today that policy styles in different policy areas, and in very diverse national and local settings, converge towards cooperation among governmental levels as well as between public and non-public actors. In the European Union, the ‘Open Method of Coordination’ is today emblematic of such a cooperative ethos, and its advent is related to the shortcomings and failures of the less flexible Community Method. Relying, for instance, on a cross-national comparison of the use of more or less ‘dirigist’ policy instruments in the environmental sector, Jordan et al. (2005) find that traditional regulatory instruments coexist or compete today to a varying degree with newer cooperative or voluntary instruments. Apart from benchmarking, all other ‘new’ policy instruments listed by the authors – such as co-regulation, negotiated agreements, or voluntary codes of conduct – imply a more or less pronounced role for civil society actors (Jordan et al. 2005: 481). Further, a comparative study of three policy sectors in seven European democracies shows that, even in the United Kingdom, which ‘constitutes Lijphart's paradigmatic case for a majoritarian democracy, its policy networks have a fragmented character which actually resembles the characteristics of Swiss networks, the paradigmatic case of a consensus democracy’ (Kriesi et al. 2006: 350).

This change in policy style is described as a shift from ‘government’ to ‘governance’. The latter is characterised by cooperative relations in policy-making between public and non-public actors, often embodied in ‘public–private partnerships’.1 This means that policies are formulated or implemented by networks involving public actors (politicians and administrators) together with non-public actors of different nature (firms, interest representatives and stakeholders, and experts). Deliberation, bargaining, and compromise-seeking is the main modus operandi of network forms of governance. With ‘downwards’ devolution and decentralisation, and ‘upwards’ Europeanisation, governance is today frequently coupled to ‘multi-levelness.’ Though ‘multi-level governance’ may take different forms (Hooghe and Marks 2003), multi-levelness means that policy-making requires the cooperation of distinct governmental levels (local, subnational/regional, national, European, transnational), in what one might be tempted to call rather multi-level government. EU structural and regional policies exemplify multi-level governance, as they are based on both the cooperation of public actors across levels and on the cooperation with non-public actors in partnership forms.

An indication of the correlation between multi-levelness, which blurs the centre–periphery divide, and network governance, which blurs the state–society divide, is provided by the fact that subnational governments have an interest in showing that they stand close to civil society actors in order to prove the authenticity of their representational claims (Piattoni 2009: 174). Multi-level governance manifests itself, for instance, in regional Monitoring Committees, which are deliberative bodies supervising the administration of EU Structural Funds (Kamlage 2008). More technocratic versions of multi-level governance include ‘Comitology’, or the ‘Open Method of Coordination’, where national experts work in cooperation with the EU administration, as well as policy networks that involve the Commission services together with national agencies (Martens 2008), such as the European Competition Network. Though such networks may formally have no binding decision-making power, information is exchanged therein so that they acquire a coordinating function and are expected to favour cross-national convergence on policy practice. Recent empirical research shows that networks deliver a variety of outputs: among others, decisions, standards, or merely knowledge (Marcussen and Olsen 2007: 286). Clearly, the lack of formal authorisation is more problematic if authority to issue binding decisions is delegated to governance networks than if networks offer only consultative advice (‘epistemic communities’).

Based on the existing literature, this paper seeks to explain why the trend towards more cooperative forms of policy-making, though in all likelihood necessary for policy efficiency and even at first glance promising with respect to inclusiveness and pluralism, can have negative consequences for democratic accountability.2 This conceptual discussion has so far been neglected, to a large extent, because managerial concerns about governance performance tend to prevail in that research field:3 based on a survey of about 1,600 projects included in a ‘Connex’ database on EU governance (GOVDATA), for example, Kohler-Koch (2006: 5) concludes that not more than 17 per cent of them address questions of democracy or legitimacy. Accountability is viewed here as a social relationship between an actor A and a forum B. A is accountable to B if A is obligated to inform B about A's decisions and actions, to justify them, and to face different consequences depending on whether or not B is satisfied with A's performance or A's justification.4 As suggested by Barnett and Finnemore (2006: 171), ‘accountability matters because of the presumption that its absence means that those in power have the capacity to act without regard for those who authorize their actions and for those whose lives are affected by those actions’ (Barnett and Finnemore 2006: 171). Democratic accountability is the accountability of decision-makers to the electoral forum: if voters are satisfied with governmental performance, they will renew their mandate to the incumbents (positive consequence); if not, they will ‘throw the rascals out’ (negative consequence).

This paper first explores the properties of multi-level governance that lead to a deficit in democratic accountability before coming to more general conclusions on the characteristics of accountability mechanisms in multi-level governance and on their consequences for democracy. It concludes that multi-level governance is, in a sense, characterised by more accountability, but also by less democracy.

A Preliminary Note of Caution

Accountability is usually considered as a virtue (Bovens 2010). Should we, however, always seek to maximise it? This is not self-evident, because accountability can be a double-edged sword for a number of reasons (March and Olsen 1995). Increased accountability can produce unintended negative consequences with respect to the expected goals, summarised by Hood (2007: 202–3) under Albert Hirschman's concepts of futility, jeopardy, and perversity. Bovens (2005) emphasises the risks of ‘excess accountability’, and similarly Heald (2006: 60) signals ‘the danger of over-exposure’. A feeling of suffocation from the necessity to account may lead to various ‘subterfuges’ and forms of blame-avoidance behaviour; examples of a defensive attitude animated by risk-avoidance and by the desire to shield oneself from the threat of accountability are attitudes of excessive proceduralism, which inhibit creative thinking, or sheer ‘appearance of conformity’ with the expectations of the forum (Philp 2009: 43). In a context of an increasingly ‘audience’ democracy (Manin 1997), politicians are, for instance, likely to care more – with the help of marketing consultants (‘spin doctors’) – about their self-presentation than about genuine accountability.

As regards democratic accountability, it may be inimical to compromise-seeking, which is necessary in differentiated and fragmented societies, and which may require that negotiations take place behind closed doors. As accountability to voters may inhibit solutions that cannot be ‘sold’ with populist justifications, ‘informalisation’ strategies may be preferred by policy-makers to avoid public and media scrutiny. A different issue is whether such strategies are preferable; which goal should prevail between policy efficiency and democratic accountability when they collide is a matter of normative controversy. In his justification of delegation of regulatory power to independent agencies, Majone (2005), for instance, privileges the necessity of credible commitments and expertise of policy-makers, because democratic accountability may lead to suboptimal policy outcomes, mainly due to the short-term horizon of democratically accountable politicians. Yet there is no consensus on such a preference.5

One may then wish to optimise accountability rather than simply maximising it; yet, balancing the advantages and drawbacks of accountability remains a controversial matter: there is no agreement on the amplitude of negative effects of ‘excess’ accountability on policy performance, nor is there an agreement on how much the effectiveness of mechanisms of democratic accountability contributes to ‘diffuse support’ for political systems. Notwithstanding that difficulty, the reader will note that this paper views democratic accountability as, indeed, a virtue for political systems. This derives not only from the author's normative preferences, but also (and more importantly) from an argument on governability: it is assumed that political systems suffering from an atrophy of mechanisms of democratic accountability are more subject to attacks by anti-establishment political entrepreneurs.6

The Weak ‘Democratic Anchorage’ of Network Forms of Governance

According to Sørensen (2005),7 the democratic anchorage of network forms of governance should be ensured if they display the following attributes: control by elected politicians, representativeness of participating collective actors, possibility of contestation for stakeholders thanks to dissemination of information to them, and to the availability of channels for ‘voice’, respect of the rule of law, and procedural fairness.8 Such a democratic anchorage – with democratic accountability as a core feature – is nevertheless questionable, for a number of reasons that are presented now.

The Lack of Visibility of Governance Networks

A number of factors make what happens in policy networks hardly visible to outsiders. How many among the lay citizens are aware, for instance, that policy-making takes that shape and involves several categories of actors who are not the ‘usual suspects’ in terms of forging political decisions? Moreover, as mentioned above, network activities are often deliberately informal and opaque, in order to facilitate the achievement of compromise. Besides, since networks involve several actors, it is hard to identify those who are responsible for decisions. This is the ‘paradox of shared responsibility’ (Bovens 1998: 45–52) which can be aggravated the more networks are pluralistic; interestingly, network pluralism favours representation but may inhibit accountability. Responsibility is then diluted, as is also the case in public–private partnerships where, in order to escape criticism for poor performance, public and private actors are likely to engage in ‘blame-shift games’ that are prejudicial to accountability (Hood 2007: 200; see also Flinders 2005: 231). It should be noted that, even if the problem of weak visibility is attenuated through provisions for transparency and access to information, the latter are no substitute for genuine accountability mechanisms. With transparency (and with public debate) there is no guarantee for sanctions: accounts may be given, discussion may follow, and then nothing happens.9 Hence, even though transparency and publicity are often cited as a remedy for accountability problems, although necessary, they are not sufficient.

The lack of visibility of networks is a symptom of the broader divorce (Leca 1996) between the sphere of ‘politique des problèmes’ (dominated by problem-solving governance arrangements) and the sphere of ‘politique d'opinion’ (the arena of party competition). Even if ‘audience’ democracy means that, today, politicians' conduct has become the object of increased (and feared) media scrutiny, this does not apply to the day-to day policy practice of multi-level governance – for which the media usually do not display any interest – and neither do journalists possess enough expertise to delve into it. For instance, two case studies on the EU Open Method of Coordination showed that the media did not accomplish their control function (de la Porte and Nanz 2004: 277). More generally, ‘politique d'opinion’ is virtually absent in the EU system; and, unless politicisation of the public opinion occurs through infrequent referendums, there is no real ‘communicative discourse’ of political elites on European matters (Schmidt 2006). Such a divorce is also perceptible in the related uncoupling of networks from democratic institutions.

The Uncoupling of Governance Networks from the Democratic Circuit

Governance networks are often, to a large extent, uncoupled from the official representative bodies, and reference has even been made to the advent of a ‘post-parliamentary’ or ‘post-liberal’ governance. If decisions are prepared by policy networks the legislative function of parliaments is affected; if they are implemented by them, it is their control function that is weakened. As a result, one can speak of a ‘loosening grip of representative democracy on acts of governing’ (Bekkers et al. 2007: 308). Besides, in spite of being decisive for policy outputs, ‘meta-governance’, i.e. network management (Klijn 2008) or the governance of networks themselves (Torfing 2007: 13), is largely delegated to the administration, which usually makes the decisions regarding their design, their participants, their attributions, the framing of issues on their agenda, and their management.10 True, members of the bureaucracy are accountable to their political superiors who are subject to electoral sanctions; however, the length of the chain of delegation combined with the magnitude of administrative discretion makes their democratic accountability fictitious.

Another problem is that, even though parliaments have the formal right to overrule decisions formulated in networks or to supervise how decisions are implemented by them (in that sense, the concept of ‘post-parliamentary’ governance is not entirely adequate), it is questionable if this ‘shadow of hierarchy’ represents a credible menace. One may have doubts on the capacity of parliaments to monitor and exert effective oversight over the action of these decisional circuits. Lack of time and expertise is a real problem here: ‘agents’ can benefit from informational asymmetries at the prejudice of democratic ‘principals’, who may be professional politicians and at the same time dilettantes as regards quite a number of policy issues. Auel (2007), for instance, found considerable cross-national variation on the ability of individual parliaments to control the executive when it is involved in European policy-making. Interestingly, she concluded that parliaments are better informed if they succeed in becoming involved in informal negotiations with the government; in order for parliaments to become more influential, they must then adopt opaque practices. This implies a trade-off: governments are more accountable to parliaments on EU matters if the latter operate outside public scrutiny, and this causes prejudice regarding the accountability of parliamentarians on the part of the citizenry. Though Auel's work is on classic intergovernmental decision-making in the EU, there is no reason why its conclusions would not apply equally (if not more) to the day-to-day level of multi-level governance. Raunio (2007: 169) refers, for instance, to – admittedly so far scarce – evidence on the Open Method of Coordination, according to which ‘national parliaments have not scrutinized OMC documents in the same way as they process EU laws’.11 It probably matters, too, whether decisions are formulated by networks or if they are only implemented by them. In the first case, it is more likely that elected politicians will remain influential, or even be included in networks, because they are the target group to be convinced. In the latter case, by contrast, it is the administration (many times at the local or regional level) that cooperates with civil society organisations, both being remote from the world of elected politicians that formally made the decisions.12

If multi-level governance is uncoupled and remote from representative government, there are risks of attribution errors in responsibility:13 decisions are made in reality by actors other than those (the most visible) regarded as authorised decision-makers by the people or the affected communities. The effectiveness of the democratic feedback loop is thus undermined: the retrospective evaluation of office-holders on the grounds of their policy achievements, and the prospective evaluation of candidates on the grounds of their pledges, become – to a large extent – fictitious. The incumbent parties are held responsible for political decisions whose formulation or implementation escapes their control, at least partly, and candidates standing for election make promises that they structurally will not be in a position to fulfil.

The Composition of Policy Networks

Actors composing policy networks become part of them for instrumental reasons: public authorities think that governance through networks can contribute to governability by adding expertise or providing legitimacy through ‘stakeholder’ participation to the policy process. They are also afraid that the lack of inclusion of ‘stakeholders’ or economic interests in the decision-making process would undermine governability, because ‘stakeholders’ may have a blackmailing potential through the threat of ‘voice’, and firms may have a similar potential through their ‘exit’ option. If networks are pluralistic, this is then more a side effect of the search for political support than the result of normative concerns. In fact, policy networks are largely composed of bureaucrats and other policy experts, on the one hand, and of interest representatives, NGOs, and purely private actors on the other. Elected politicians are not frequently at the core of networks (Klijn 2008: 27); Skelcher (2007: 34), for instance, found that they remain very much absent from public–private partnerships in the UK.

As noted above, members of the bureaucracy are only indirectly accountable to the citizenry due to the lengthy chain of delegation. This is even more the case in the administrative structure of the European Commission, or in the case of the blossoming regulatory agencies whose members participate in governance networks. As for experts, interestingly, in a sense, they must be unaccountable to constituencies: in order to be credible as such, they have to convince others of their independence. They have no ‘masters’ or ‘principals’ and are accountable only to their professional community, which is a form of ‘peer’ accountability (see below). Forums of professional control usually lack broad publicity, and they are mainly of ‘soft’ nature, which means that – using the distinction by March and Olsen (1989) – experts' conduct will be more driven by a ‘logic of appropriateness’ (i.e. they internalise professional norms as to appropriate conduct) than by a ‘logic of consequentiality’ (conduct driven by the anticipation of consequences such as sanctions). In that sense, this ‘social accountability regime’ is very different from the traditional accountability regime in public governance, based on the legally codified threat of electoral sanctions for politicians, and the equally codified threat of administrative sanctions for members of public bureaucracies (Mashaw 2006).

Representatives of interest groups and NGOs are only accountable to limited constituencies: primarily to donors, and also to their members, but this is true only if organisations have formal membership structures.14 This is partial accountability; they are not accountable either to the general public or to the communities affected by their actions. Though NGOs act as ‘surrogates’ (Rubenstein 2007) for the populations whose well-being is of concern to them, and as instances of surveillance in the name of these populations, the latter do not delegate this task to them; nor are they able to sanction them if they are not satisfied with the way they perform it.15 This is typically a problem of deficiency because of ‘self-authorised’ representation, which Urbinati and Warren (2008: 403) define in the following terms: a multitude of self-proclaimed representatives ‘claim to represent a wide variety of goods: human rights and security, health, education, animals, rainforests, community, spirituality, safety, peace, economic development, and so on’. These people may be sincere in their beliefs and claims about representation, and Trenz (2009: 12–13) maintains that representation by civil society organisations goes ‘beyond the traditional principal–agent model of a linear aggregation of individual preferences from the constituents to the representatives’. It is grounded on generalised ‘trust in expertise, reflexive capacities, moral integrity or simply advocacy and advertisement by mostly self-appointed civil society representatives’ (Trenz 2009: 10), and on the ‘general resonance’ created ‘within a wider audience’ by the claims made by these representatives. However, one wonders how generalised trust on organisations or broad resonance of their claims can be attested in an undisputed way. If representation is not based on any form of formal authorisation ex ante, it requires at least some form of consent ex post on behalf of the ‘represented’16 (similar to charismatic authority, which is not an essential property of its holder but a relational property, and must be acknowledged by those subject to it). In reality, no procedures are foreseen for the purpose of subjecting self-proclaimed representatives to a test of ‘recognition’ of their claims. The more an organisation claims to represent diffuse interests – and thus legitimises itself by asserting that it is immune to capture by particular interests – the more its accountability to the population of reference becomes fictitious.

Deficiencies in representation are also visible in the fact that not all social interests are able to organise themselves sufficiently in powerful civil society organisations to be listened to by decision-makers.17 Further, NGOs do not escape the problems of elitism that reduce internal accountability of their leaderships even to the rank-and-file who authorise them to act in the name of broad populations. For instance, empirical work (Saurugger 2008; Warleigh 2006) shows that NGOs involved in EU policy-making lack adequate internal democratic structures and that their supporters do not manifest a will to monitor their action. These problems become more acute with the professionalisation of organisations that is often required by their incorporation into policy-making circuits (as suggested already quite a long time ago by the literature of neo-corporatist inspiration on the attribution of public status to interest groups). Ultimately, although NGOs can pressure power-holders to take the requirement for accountable behaviour seriously, again such pressure can only be ‘soft’ (Kohler-Koch 2008: 5). In order not to remain ‘toothless’ (Schedler 1999: 16–17), soft pressure necessitates strong norms (Löfgren and Agger 2007: 43). Norms of appropriate conduct must be unchallenged by a community displaying strong cohesion – or, at least, persuasively defended by their advocates – so that power-holders would have no other serious option than adhering to them. This is not always the case.

Private corporations, for their part, are primarily accountable to their shareholders: this poses again the problem of partiality and lack of external accountability on the basis of affectedness, as these firms are not accountable to those who are subject to their externalities (workers, residents in neighbouring areas, etc.). If firms are then accountable on the basis of the ‘principle of ownership’, they are not accountable on the basis of the ‘principle of affected rights and interests’ (Mulgan 2003: 139). Firms are also accountable through the market, and NGOs, for instance, threaten with boycotts those among them who are reluctant to comply with social or environmental standards (such as in the footwear or clothing sector).Yet the success of such boycotts depends on the behaviour of actors with purchasing power, and it is doubtful that the majority of the latter will be motivated by social or environmental concerns (Hale 2008: 80). It seems in fact that neither internal (through shareholder action) nor external (through public opinion alerts) pressures have a significant positive effect on corporate responsibility with respect to social or environmental concerns (Vogel 2005), and even if such effects are produced they very much depend on the degree of market competition, the type of issue or company involved (highly visible brands are more vulnerable to pressure), or the existence of strong social expectations on firms to behave in a normatively acceptable way (Graz and Nölke 2008: 4–5). Ultimately, the market accountability regime is very different from that of public governance (Mashaw 2006).

‘Multi-levelness’ as an Aggravating Factor

In contemporary governance, cooperation across governmental levels is necessary because, in spite of the formal division of competencies between decisional levels in federalist systems (or quasi-federalist like the EU), in reality several competencies overlap across them and because the resources for effective policy-making must be pooled from different levels as well. However, accountability is further inhibited by the multi-level aspect of governance: there is a duplication of accountability problems due to the coexistence of a cooperative logic between vertical levels with the ‘horizontal’ logic of public–private cooperation.

In order to avoid policy blockade due to situations of possible mutual veto between actors involved in federalist or quasi-federalist systems (the so-called ‘joint-decision trap’), informal cooperation between them is required, which happens at the prejudice of transparency (‘Politikverflechtung’). Even in the absence of non-public actors, cooperation across levels operates along an intergovernmental logic: several executives are involved in decision-making. This generates problems of delegation – intergovernmental negotiations are made, or at least prepared, by administrators who can enjoy considerable discretion – and of dilution of responsibility (propitious to blame-shifting) due to negotiation and compromise between several actors. Further, even in principle democratically accountable actors such as national or regional governments are only theoretically accountable for their participation in intergovernmental policy coordination because of lack of information to outsiders on their positions and decisions. Only those who are close to negotiators are aware of the meanders of intergovernmental negotiations.

Finally, participants in these negotiations are themselves caught in an accountability dilemma: they must satisfy multiple ‘forums’ with different preferences. Even actors who are directly subject to the control of their electorates are, in reality, subject to ‘two-level’ accountability: they must account for their actions not only to their constituencies, but also to their negotiation partners. This leads to a trade-off, and it is probable that the latter accountability constraint will prevail, precisely because democratic ‘principals’ (citizens and even members of parliaments) may lack information that would allow them to detect blame-avoidance strategies on behalf of ‘agents’ who are called to account. Even if account-holders would be able to open up the ‘black box’ of networks (which requires a significant amount of information), their representatives therein would be in a position to justify shifts in their preferences through the claims and menaces made by their partners. Remoteness from ‘principals’, combined with proximity to peers in multi-level governance, is likely to lead then to the resolution of the accountability dilemma at the cost of accountability ‘at home’ (very similar to the prevalence of ‘logic of influence’ in neo-corporatist arrangements to the prejudice of ‘logic of members’: see Schmitter and Streeck 1999).18

The Prevalence of ‘Peer’ Accountability over Accountability ‘at Home’

A feeling of ‘common fate’ may develop within networks. Majone (1997: 262), for example, writes about their impact on regulatory agencies:

An agency that sees itself as part of a transnational network is more motivated to defend its independence and professional integrity: unprofessional or politically motivated behaviour would compromise its international reputation and make co-operation more difficult in the future.

The concept of ‘peer’ (or ‘interdependence’: Scott 2000: 50–52) accountability is here appropriate, ‘based on mutual monitoring of one another's performance within a network of groups, public and private, sharing common concerns’ (Goodin 2003: 378). In peer accountability, network participants become primarily accountable to their network partners in soft and horizontal accountability relations. The fear of ‘naming and shaming’ yields disciplining effects on them because ‘free riders’ or unreliable actors risk loss of reputation in the network, and their partners will not continue to trust them in the future, or might even ostracise them (Scott 2006: 180).19

For peer accountability to function effectively between network partners committed to common goals, these goals must enjoy broad social support. Only if policy networks are sufficiently representative and pluralist can this mutual and soft form of accountability operate to the profit of the common good. In order to consider the preferences of all interested parties, networks should not exclude either weaker interests or actors whose preferences do not coincide with the network's ‘mainstream’ orientation.20 Yet this is challenged by two bodies of literature, each focusing on distinct limits to pluralism. First, the literature on collective action suggests that lack of pluralism may result from strategic behaviour by ‘insiders’ whose interest lies in using benefits from network participation as exclusive goods, and in externalising the costs generated by their choices (‘rent-seeking’ behaviour). If participation in networks entails advantages taking the form of ‘club goods’, there are incentives for network members to behave as partners in minimum winning coalitions who have no interest in extending the group of beneficiaries to which they belong.21 Second, the literature on deliberative politics reveals that the lack of cognitive variety in policy networks –too strong a sense of community, too intense ties between members, too much self-referentiality (all very helpful for mutual trust-building) – can lead to closure and the formation of ‘group-think’ or to ‘enclave deliberation’ (Sunstein 2001). Deliberation between people with initially similar opinions tends to reinforce their commonality through a positive feedback loop mechanism. This impedes learning based on mutual critical scrutiny, which is necessary for effective peer accountability, and which requires in a sense mutual trust not to be blind. Ultimately, horizontal peer accountability is no functional equivalent of vertical democratic accountability: mutual learning by policy-makers is quite another objective than popular control by ‘policy-takers’, and control in the network does not substitute control of the network.

Accountability in Multi-level Governance: Effective and Democratic?

In the light of the previous developments, it appears that in network and multi-level forms of governance the democratic accountability of policy-makers is weakened. However, this is not unaccountable governance. Quite to the contrary, it is characterised by a ‘multiplication of control mechanisms’ (Costa et al. 2003: 670), composite as well as diffuse, leading ‘to a more diversified and pluralistic set of accountability relationships’ (Bovens 2007b: 110). It suffices to think about the role as accountability forums of user councils, administrative courts, ombudsmen institutions, or performance reviews by external experts (Michels and Meijer 2008: 169). These mechanisms establish various forms of accountability: political, but also legal, administrative, or financial. They may be viewed as an adaptation of accountability to the complexities of network governance: multi-level accountability networks are, for instance, established between courts or between ombudsmen, thus involving the ECJ or the European ombudsman together with their national counterparts (Harlow and Rawlings 2007).

Regarding the efficiency of accountability mechanisms, the effects of their multiplication are uncertain:

•	One may argue that redundancy improves control (Scott 2000), that being watched by multiple controllers has a disciplining effect, and that the pluralism of critical perspectives brought about by a diversity of accountability forums is welcome. Mulgan (2003: 220), for instance, writes: ‘Providing a framework for different accountability agencies to examine each other critically helps to create a virtuous circle of compounded accountability in which each agency is subject to scrutiny from at least one other agency.’
•	In addition, redundancy provides multiple venues to account-holders and increases their blackmailing potential by making the environment of decision-makers less predictable.
•	On the other hand, being placed under the scrutiny of ‘too many eyes’ may induce risk-averse behaviour and blame-avoidance strategies on behalf of the controlled. We may add that surveillance by too many eyes may lead in the end to fatalism or indifference, as it increases the randomness of control (Hood 1998). In short, how actors will behave in a context of indeterminacy may well be indeterminate, too.
•	Additionally, the action of accountability mechanisms may lose efficiency for lack of coordination. In multi-level governance, mechanisms of generalised and ‘comprehensive’ accountability are to a large extent replaced by disaggregated and ‘compartmentalised’ modes of issue accountability (Tsakatika 2007). Forums of accountability – network peers, stakeholder panels, courts, ombudsmen, the media – are dispersed and do not form a coherent and comprehensive accountability system. The latter would require that forums regularly communicate with each other to coordinate their action and divide tasks between them. The picture of a forum ‘patchwork’ might be an adequate description here. As Scott (2000: 57) writes: ‘These mechanisms are in tension with one another, in the sense of having different concerns, power, procedures, and culture, which generate competing agendas and capacities.’
•	Further, part of the accountability mechanisms at work are of the ‘light’ or ‘soft’ type: indirect, not institutionalised or weakly codified, operating through moral commitments and social pressure at the individual level, through exposure to the public sphere, etc.22 Hence there is a risk of them remaining toothless: although the efficiency of ‘hard’ sanctions is disputed in the literature on cooperative governance, the efficiency of ‘soft’ sanctions is not established either.

For democracy, the implications are:

•	The accountability mechanisms in multi-level governance perform a different control function than mechanisms of democratic accountability, which allow citizens to be confident that their preferences (input) will be mirrored in decision-making (output). They are, therefore, no adequate substitute that would ensure responsiveness. As a matter of fact, accountability forums such as review panels, courts, ombudsmen, or the media may well not include any ‘principals’ who have delegated their authority to ‘agents’ who would then act as their representatives. Hence, actors in these forums may have a distinct agenda from democratic principals with whom policy-makers are in a relationship of delegation and representation: the relationship between accountability and representation is loosened. In other words, ‘third parties’ who control ex post are not necessarily the same as those who formulate mandates ex ante, and they may themselves be weakly accountable. As the task of holding the rulers accountable is one of their reasons for being, there is no certainty at all that they will do much to empower those in whose name they act. Organised civil society actors acting as ‘surrogate’ accountability-holders, such as NGOs, are not spared such a risk.
•	It is also worth mentioning that the highly praised ‘peer’ accountability mechanisms within networks lack transparency themselves.
•	Ultimately, in network and multi-level forms of governance there may not only be a trade-off between democratic accountability and policy efficiency (the well-known ‘input–output’ dilemma regarding the search for legitimacy), but also a trade-off between democratic accountability and other forms of accountability. Even if accountability mechanisms are characterised by openness or are participatory (which should not be taken for granted),23 even if there is a plurality of them, they are no substitute for the weakening of accountability through the electoral circuit of representative democracy. The next section elaborates on this crucial argument.

Conclusion: Accountable Governance is Not Equal to Democratic Government

In her book on Democracy in Europe, Vivien Schmidt (2006: 28–9) writes that ‘governance with some of the people’ cannot make up for ‘the lack of government by and of the people’ (emphasis original). This means that, even if processes of multi-level governance are participatory and inclusive, broad organised pluralism cannot be a corrective to the uncoupling of governance networks from the democratic (representative) circuit (Tsakatika 2007). Accountability to the citizenry at large is weakened in these forms of governance. What is emphasised, at best, is accountability to particular stakeholder groups which claim to represent segmented publics subject to spatially concentrated negative externalities of policy-making: ‘Holders in a functional perspective replace citizens in a normative perspective’ (Greven 2007: 241). This is part of a more general trend towards an ‘advocacy-democracy’ (Cain et al. 2003), wherein cause groups play an increasingly important role, while the role of the actors and institutions of the democratic/representative circuit – parties and parliaments – declines.24

It may of course be argued that ‘the value to democracy of electoral and non-electoral representation is positive sum’ (Saward 2009: 21), yet the fact remains that the disciplining effect of the formal right to sanction through the vote is reduced. Moreover, being a ‘stakeholder’ is – again – not an essential property, but the outcome of a process of symbolic construction. ‘Stakeholders’ are authorised to act as account-holders if they manage to be considered as bearers of strong, and thus legitimate, preferences on policy issues: the formal egalitarian dimension of the ‘one man, one vote’ principle disappears. Worse, it is frequently those who have to provide accounts who also define whose preferences are strong enough, so that they are authorised in a top-down manner to act as account-holders: le fait du Prince. In addition, stakeholders must possess resources – organisational capacity, expertise, or blackmailing potential – to constrain policy-makers to act under the shadow of accountability to them. The impact of resources in the political life of our democracies is not new, and formal equality seldom coincided with political equality:25 a long time ago political scientist Stein Rokkan (1966: 105) argued that ‘votes count, but resources decide’. The literature on ‘iron triangles’ and the like has shown that access to networks where bureaucrats and organised interests are core actors has been for decades a decisive resource for policy influence. However, with the proliferation of network forms of governance at different stages of the policy process (including in their multi-level variant), votes come now to count less, whereas resources decide more.

What is more, other trends in governance such as judicialisation, delegation to independent agencies, or administrative reform inspired by ‘New Public Management’ principles further weaken the effect of vote by shifting power to unrepresentative and non-democratically accountable institutions. Sometimes deliberately, sometimes not, democratic accountability is undermined today. Can fragmented and largely horizontal accountability mechanisms, often based on a checks-and-balances logic, and sometimes operating informally, then be viewed as an acceptable option given the limits of democratic accountability in complex systems of multi-level governance? Rubenstein (2007: 631) correctly holds that ‘standard’ (democratic) accountability is superior to its different surrogates, which should be viewed as not more than second-best alternatives. In spite of the current proliferation of accountability mechanisms, and of the normative discourse on accountability as a virtue, ‘simply’ accountable governance is no synonym for democratic government, similar to (more or less limited) pluralist decision-making that can supplement democratic government but not be a substitute for it (whereas it is largely conceived as such beyond the nation state with ‘civil society’ participation and the like).

We may well be in the presence of a paradox: network and multi-level governance is characterised by more accountability, but less democracy. And perhaps another manifestation of its uncoupling from the sphere of ‘politique d'opinion’ lies in the fact that the latter seems to evolve (but further empirical research is required to substantiate that hypothesis) towards more democracy, but less accountability. We observe a growing use of referendum devices, especially on European integration, and national elections themselves become more competitive. Citizens more often make use of their capacity to change the team that will govern them and above all its leader, whose role becomes crucial in a context of ‘presidentialisation’ of politics (Poguntke and Webb 2005). However, determining who will be in office does not necessarily mean that one also has a real possibility to select from among a very diverse supply of policy programmes, especially as denationalisation and counter-majoritarian institutions limit policy discretion at national levels. Under highly mediatised and ‘audience’ democracies, ‘image and style increasingly [push] policies and substance aside’ (Farrell and Webb 2002: 122); or, as formulated by Mair (2008: 227):

To be sure, there is a choice between the competing teams of leaders and, given the growing evidence of bipolarity, that particular choice is becoming more sharply defined. But there is less and less choice in policy terms, suggesting that political competition is drifting towards an opposition of form rather than of content.

Not only ‘multi-level governance’ in the domain of policy-making, but also such phenomena in the field of traditional party politics, are not good news as regards the durability of a genuine and substantial democratic accountability.

Acknowledgements

Research for this paper benefited from generous funding by the Swiss federal government (Secrétariat d'Etat à la recherche). Previous versions were presented at the ‘Connex’ workshop on Accountability, European University Institute, Florence, 21 April 2008, and at the IPSA World Congress, Santiago Chile, 12–16 July 2010. For helpful comments I am indebted to workshop and Congress discussants and participants (particularly Christopher Hood, Peter Mair, and Eva Soerensen), to Alan Scott, and to two anonymous reviewers.

Notes

1. See, for example, Ansell and Gash (2008), who review a large number of cases of collaborative governance.

2. Starting from a ‘post-liberal’ point of view, Esmark (2007) comes to more positive conclusions on the accountability of network governance.

3. However, problems of democratic accountability in multi-level governance are not too dissimilar from problems that have been addressed in the literature in the past, such as those related to functional representation. Unlike a significant part of current research on governance, past research on neo-corporatism showed familiarity with problems of lack of coupling between the circuit of interest and the circuit of partisan (democratic) representation, or with problems of internal democracy and accountability of collective actors participating in corporatist arrangements.

4. This definition is adapted from Bovens (2007a).

5. For an argument on the negative consequences of informalisation for democratic accountability and legitimacy, see Greven (2005).

6. See, for instance, Papadopoulos and Magnette (2010) on the link between the ‘democratic deficit’ of the EU and ‘Euroscepticism’.

7. For a similar list of criteria see also Löfgren and Agger (2007).

8. As observed in empirical work by Tyler (2006) and Grimes (2006), perceptions of procedural fairness seem indeed to be a condition for (so-called ‘throughput’) political legitimacy.

9. Philp (2009) does not include the availability of sanctions as a constitutive element of accountability. It appears that this is not purely a conceptual matter, but derives from the author's worldview that emphasises the perverse consequences of accountability. Philp favours the design of ‘integrity-based systems’ resting on a high level of generalised trust, because he considers that, in the end, ‘compliance-based systems’– i.e. including a set of positive or negative incentives expected to induce virtuous behaviour –‘kill’ accountability (or at least reduce it to a ritual). But such a worldview is neither more nor less plausible than the view of those who precisely do not trust ‘integrity-based systems’ per se, unless they operate under the shadow of formal sanctions.

10. Klijn (2008: 26) holds that ‘the political authority as network manager is more of an actor among the actors, rather than the big leader’. This, however, should not lead us to underestimate the importance of the network management function: between a big leader and a simple primus inter pares there is ample gradation of leadership forms. For example, Skelcher et al. (2005) found in their empirical research in the UK that public administrators played an important role in network design.

11. This may have to do with the fact that formal decision-making on the OMC remains at national level (see, however, also de la Porta and Nanz 2004: 278; Tsakatika 2007: 552–3).

12. I owe that distinction to Beate Kohler-Koch.

13. A survey of service provision by non-profit organisations in the US by Van Slyke and Roch (2004: 203) showed that the service users who are more likely not to be able to correctly identify the service provider are those who are dissatisfied with the services provided. This has important implications for attribution of responsibility: those who are most likely to express their discontent are also likely to select the wrong target.

14. For an analysis of different forms of NGO accountability depending on NGO type and NGO interactions with their environment, see Ebrahim (2007).

15. For example, global environmental concerns of NGOs from the North are viewed with suspicion by actors from the South who feel that such concerns are an obstacle to the growth of developing countries. Rubenstein (2007: 625) distinguishes ‘surrogate accountability’ from ‘mediated standard accountability’. In the latter the ultimate accountability-holder delegates the tasks of surveillance or sanction to an agent (to the judicial system for instance), while in surrogate accountability there is no such delegation.

16. ‘Those who recognize someone as representative are the ultimate source of its legitimacy’, write Pollak et al. (2009: 23), and their silence should not be interpreted as tacit acquiescence, because the ‘represented’ may be unaware of the existence of ‘representatives’ of their claims.

17. One might apply here the famous critique of pluralism by Schattschneider (1960: 34–5): ‘the flaw in the pluralist heaven is that the heavenly choir sings with a strong upper-class accent’. For collective action, wealth as such matters probably less than the organisational capacity; and, for inclusion in governance networks, ‘like-mindedness’ with their designers may help.

18. The narrowing of the preference gap between collective actors in ‘summit’ negotiations and deliberations is the result of the intense relations between them which, in turn, lead to a loosening of their ties with the rank-and-file.

19. Note, however, that probably the most powerful network members will not be particularly sensitive to that kind of pressure, if they believe that things cannot be done without their acquiescence. According to Blatter (2007: 278), ‘naming and shaming’ impacts on the acquisition and deprivation of ‘symbolic capital’; but if actors possess other forms of capital they may not bother much about that.

20. To give an example of limited pluralism, German NGOs involved in European governance through consultation procedures are in favour of more ‘positive’ integration, and do not mirror the preferences of ‘Eurosceptic’ segments of the electorate (Trenz et al. 2009: 20–21).

21. I owe the comparison to Christopher Lord.

22. Public stigmatisation of inappropriate conduct is, for instance, a core strategy of NGOs functioning as account-holders in transnational governance.

23. Bekkers et al. (2007: 311) find that alternative models of democracy (such as deliberative or consumer) are useful for the democratic anchorage of network governance. However, on the basis of a number of case studies conducted at local, European, and transnational level, they had to conclude that the empirical manifestations of these models proved to be insufficient as a compensation for the ‘original’ democratic deficit that is a consequence of the models' uncoupling from representative democracy.

24. For this argument applied to the Open Method of Coordination in the EU, see Büchs (2008).

25. In that sense, and for several other reasons, representative democracy should not be idealised, and this also applies to its contribution to the accountability of rulers (Papadopoulos 2003: 486–92).