The concept of output legitimacy

Legitimacy is, as Bernard Crick had it, ‘the master question of political science’ (Crick Reference Crick1959, 150). At the same time, legitimacy remains an ‘essentially contested concept’, whose very meaning and usefulness are debated vigorously (Hurrelmann et al. Reference Hurrelmann, Schneider and Steffek2007). Legitimacy, as is well-known, can be conceptualized in two very different ways. In a normative perspective, legitimacy is the scholar’s yardstick for the evaluation of existing or imagined political regimes. This yardstick may be derived from moral philosophy, democratic theory, or legal theory. In the sociological perspective, which is usually attributed to Max Weber, legitimacy is an empirical fact that scholars can investigate and report. It resides in the beliefs of citizens about the rightfulness of political authority.

Few conceptual contributions have left more of an imprint on the academic debate about legitimacy than the distinction between ‘input’ and ‘output’ legitimacy that was coined by Fritz Scharpf. It has become almost canonical, in the sense that the meaning and usefulness of this distinction are rarely contested (but see Gaus Reference Gaus2010). Input and output legitimacy figure in academic discourses of both the normative and the empirical-analytical kind, but the empirical-analytical perspective seems to prevail today. The use of the term ‘input legitimacy’ is widely consistent in the IR and European Studies literature. Input legitimacy refers to institutional arrangements that allow citizens to communicate their interests to political decision-makers; or, as in the case of direct democracy, to take decisions themselves. Authors may quibble over the functionality, democratic quality, and empirical feasibility of this or that particular arrangement of input, but by and large there does not seem to be serious disagreement over what input legitimacy means, within the state and beyond. If democracy is to be governance ‘by the people’, mechanisms enabling citizen input are imperative and must form part of the institutional set-up of the polity.

In the context of international and European governance, output legitimacy is normally related to ‘effective’ or ‘efficient’ problem-solving (see, among others, Höreth Reference Höreth1999, 251; Bäckstrand Reference Bäckstrand2006, 292; Curtin and Meijer Reference Curtin and Meijer2006, 112; Risse Reference Risse2006, 180; Lindgren and Persson Reference Lindgren and Persson2010, 451). Authors adopting such a perspective stress the benefits that citizens reap from the functioning of political institutions. It is also often suggested that deficits in institutional performance will decrease output legitimacy: ‘A political order that does not perform well will ultimately be considered illegitimate no matter how democratic the policymaking process’ (Risse and Kleine Reference Risse and Kleine2007, 74). The conjecture here is that output and input legitimacy are generated through different institutional mechanisms, and that democracy is unrelated to institutional performance. In the now almost canonical view, output-legitimacy is ‘derived from the capacity of a government or institution to solve collective problems and to meet the expectations of the governed citizens’ (Mayntz Reference Mayntz2010, 10). With a focus on organizational performance and efficiency often comes the idea that output legitimacy is, conceptually speaking, the antonym of democratic legitimacy. As Gaus (Reference Gaus2010) remarks, it has become normal in the disciplines of IR and European Studies to contrast output legitimacy tout court with democratic legitimacy. The idea is that democracy and institutional performance are unrelated and that performance may be able to compensate for a lack of democratic credentials. Some contributors have pushed this line of reasoning even further, saying that ‘[o]utput legitimacy implies that a political system and specific policies are legitimated by their success’ (Van Kersbergen and Van Waarden Reference Van Kersbergen and Van Waarden2004, 158). And in the most extreme version of such an approach, the term output legitimacy simply denotes empirical acceptance of governance, unrelated to any explicit standard of evaluation (Take Reference Take2012, 6).

Most of the studies cited here refer to output legitimacy in an empirical-analytical and not in a normative sense. The normative perspective on output legitimacy seems to have widely disappeared from the picture, some notable exceptions notwithstanding (see Kriesi Reference Kriesi2013). The dominance of the empirical-analytical view on output legitimacy has led to a certain asymmetry in concept formation. In the current IR and European studies literature, democratic procedures of citizen ‘input’ figure both as normative criterion of assessment and as an (assumed) empirical motivation for citizens’ rule-following. Normative treatments of output legitimacy, on the other hand, are rare in the contemporary IR, European Studies and international theory literature. It is my purpose here to investigate what, in a distinctly normative perspective, the essence of output legitimacy is.

To answer this question I go back to the original input/output distinction in the work of Fritz Scharpf, as laid out in his inaugural lecture at the University of Konstanz. The purpose of that lecture was to take stock of contemporary developments in democratic theory (Scharpf Reference Scharpf1970, 8). Drawing on David Easton’s analysis of political systems (Easton Reference Easton1965), Scharpf distinguished two perspectives on the political process that democratic theory can adopt. One perspective focuses on the mechanisms of citizen input into the system, understood in the sense of Easton’s ‘demands’ as ‘articulated interests’ (Scharpf Reference Scharpf1970, 21). The second perspective is concerned with the quality of the output that the system delivers. According to Scharpf’s conceptualization, the key problem here is how the design of the democratic polity can ensure that its output serves ‘the public interest’Footnote ¹ of the citizenry and not the particular interests of some well-organized or vociferous groups. Scharpf discusses a wide range of specific institutional mechanisms under the output heading. As output-oriented he classifies democratic theories focusing on the separation of powers, on parliamentary and public deliberation, and on the welfare state (Scharpf Reference Scharpf1970, 21–24).

The oddity in concept-formation here is that in Easton’s systems theory, the term ‘output legitimacy’ cannot be found and in his theoretical universe would not make much sense either. Outputs, for Easton (who uses this word mostly in the plural) are tangible benefits for discernible individuals, which result from political decisions made in the system (Easton Reference Easton1975, 438). Outputs ‘offer the members [of the polity, J.S.] some benefit in return for which they can be expected to offer support; or the outputs may impose upon the members some identifiable disadvantage, such as an onerous tax or restriction on function. In this case they might be expected to become antagonistic toward the political objects and to extend negative support’ (Easton Reference Easton1965, 382).

Outputs as benefits can hence create what Easton calls ‘specific support’ of the system and failure to deliver such benefits may erode it. But specific support, for Easton, is not sufficient to secure political stability. It needs to be complemented with ‘diffuse support’, which is created by beliefs in the ‘legitimacy’ of the system. Easton describes several types and sources of legitimacy at length, but none of them is related to the material outputs of the system (Easton Reference Easton1965, 286–310). Accordingly, outputs for Easton can foster empirical support for a political regime, in fact any kind of political regime, but they cannot be the source of its legitimacy (see also Bolleyer and Reh Reference Bolleyer and Reh2012). Easton’s association of legitimacy with diffuse support and of outputs with specific support is perfectly in line with Max Weber’s classic conceptualization of legitimacy. Weber also excluded that ‘personal advantage’ could be a basis of legitimacy (Weber Reference Weber1978, 213).

In his adaptation of Easton’s political systems perspective, Fritz Scharpf thus named a dimension of democratic legitimacy by a term that in Easton’s conceptual universe denotes an alternative to democratic legitimacy. What is more, Scharpf himself has been edging away from his normative conception of democratic output and towards a more empirical-analytical notion of output legitimacy as effective/efficient problem-solving (he uses both adjectives in this context). His more recent writings somehow straddle the line between the normative and the empirical perspective on legitimacy. In his work on the EU, for example, Scharpf still associates preventing abuses of power and securing the epistemic quality of decisions with output legitimacy, thus keeping the normative connotations of the concept alive (Scharpf Reference Scharpf1999, 13). On the other hand, Scharpf also argues that output-legitimacy of a system is derived ‘from its capacity to solve problems requiring collective solutions’ (Scharpf Reference Scharpf1999, 11; see also Scharpf Reference Scharpf2000, 104). The capacity of a political system to solve problems is different, however, from the capacity of a political system to solve them in a manner respectful of democratic standards. Efficacy and efficiency of governance may be well defended on pragmatic grounds (for IOs, see Gutner and Thompson Reference Gutner and Thompson2010). However, not every efficient or efficacious political measure is democratically acceptable.

What I adopt from Scharpf’s early writings for the purposes of this article is the insistence that ‘output legitimacy’ is related to the democratic quality of government. It is situated in the democratic dimension of government ‘for the people’ and closely related to the notion of public interest. Scharpf’s own conception of democracy is based on ‘the normative premise that legitimate government must serve the “common good” of the respective constituency, and that this function must be protected against both the self-interest of governors and the rent-seeking strategies of special interests’ (Scharpf Reference Scharpf2003, 3). As an approach to democracy, we would probably relate this to the family of ‘republican’ theories. In the next two sections I will, first, systematically develop this conception of democratic output legitimacy further and then, in a second step, apply it to the context of IOs. While it is easy to see what democratic output legitimacy of government means within the state, the particular features of the international level of governance require a more sustained defense of my idea. I should also like to stress that my insistence on democratic output legitimacy is not intended to marginalize pragmatic concerns with efficiency and efficacy of governance. The aim is to highlight that also with regard to governance output, there is a democratic dimension that needs to complement efficiency concerns.

One last conceptual clarification is in order. How does a conception of democratic output legitimacy relate to the idea of ‘throughput legitimacy’ that has gained clout in the literature over the last years? ‘Throughput legitimacy’ is intended to complement the input and output dimensions with a focus on process and procedure (Haus et al. Reference Haus, Heinelt and Murray2005; Wolf Reference Wolf2006, 214). The notion of throughput refers once again to Easton’s political system imagery, and in particular to how inputs into the system are processed on their way to the output side. As Schmidt argues with regard to European politics, ‘[t]hroughput is process-oriented, and based on the interactions […] of all actors engaged in EU governance’ (Schmidt Reference Schmidt2013, 5). Crucial normative aspects of throughput are transparency, deliberative quality of the policy process, etc. In this way, the procedural dimension of policy-making is separated both from the input and the output side.

I fully agree that procedure is important but have two objections to introducing a third dimension of legitimacy to the typology: first, if we use Scharpf’s original method and allocate traditions of normative democratic theory to types of political legitimacy, it appears that there is nothing like a genuine throughput tradition in the history of normative democratic theory. Rather, such procedural aspects are emphasized heavily in the fields of management and public administration, which have turned to problems of democratic quality of governance over the last decades (Pierre Reference Pierre2000; Vigoda Reference Vigoda2002; Box Reference Box2006). My second point is more important for concept formation than this genealogical note. To contain procedure in a separate dimension of (throughput) legitimacy suggests that input and output do not have a procedural dimension. As we will see below, especially the epistemic side of output generation hinges upon the quality of deliberative procedures. And the same seems to be true for input. There is no democratic input legitimacy when citizens’ interests are not treated fairly in the subsequent political process. I therefore think that it would not be an advantage to introduce ‘throughput’ as a separate type of legitimacy. Neither input nor output legitimacy are intelligible without reference to procedures. Why procedures are crucially important for democratic output legitimacy will become clear in the next section.

Democratic output legitimacy and the public interest

The purpose of this section is to establish and defend a conception of democratic output legitimacy. This requires, in a first step, to clarify the notion of democracy used here and its relation to the quality of political decisions. Democracy is both a philosophical ideal and a social practice. According to Abraham Lincoln’s almost proverbial definition, democracy is government of the people, by the people, and for the people. Government by the people implies that citizens should be able to select political office holders and to hold them to account. All persons in power should have a mandate that can be traced back to an electoral entitlement, at least through chains of delegation. Moreover, in a democratic polity citizens should be able to participate in the political process and make their voices heard in public. This is the input side of democracy.

Governance for the people implies that decisions taken in a democratic political system should benefit all citizens, and safeguard their human and civil rights. This is the domain of output legitimacy. Input mechanisms, such as elections, referenda, or majority voting, are no guarantee that this will be the case. Typical dangers of representative democracy are the nepotism of rulers and the pathologies of party politics, when, for instance, power games marginalize substantive considerations. Direct democracy based on referenda is particularly vulnerable to populism. Some recent referenda outcomes, such as the Swiss ban on minarets and the assault on gay rights in Croatia and Slovakia, can illustrate this point.Footnote ²

Moreover, complex problems of governance may overburden voters and elected personnel when they require levels of scientific or technical expertise that laypeople and members of parliament are unlikely to have. A normative conception of democratic output legitimacy therefore must focus on these two problems of democratic government: keeping majorities and powerful factions in check to (inter alia) protect human rights; and safeguarding the epistemic quality of decisions under conditions of complexity. Democratic output legitimacy hence is, quite like input legitimacy, a necessary component of democratic legitimacy (which is the wider, two-dimensional concept). A political system needs both democratic input and democratic output legitimacy when it claims to represent governance by and for the people.

Most institutions that secure democratic output legitimacy are non-majoritarian in character (Majone Reference Majone1998). These include courts, especially constitutional courts, to hedge the power of elected majorities, as well as constitutional guarantees of citizen rights that cannot be abolished with simple majorities, or not changed at all. The delegation of tasks to independent agencies is justified by the quality of decisions and the need to limit the power of elected representatives and parochial interests (Mashaw Reference Mashaw1997). A textbook example to illustrate this democratic rationale for delegation is the independent central bank. The problem that independent central banks are designed to resolve in a democratic polity is that elected politicians have incentives to manipulate interest rates for their own short-term goals, especially when re-election is imminent. To counter this threat, the task of determining interest rates is transferred to non-elected specialists who are guided by technical considerations. These expert economists need to be shielded from the pushing and shoving of everyday politics to guarantee independence of their judgment. Hence central banking is de-politicized for democratic (in the sense of being for the people) and not just managerial reasons. The democratic justification for the exemption of the central bank from parliamentary control and direct political input is that this will be in the long-term interests of citizens. Note that these long-term interests are assumed, not in any way empirically ascertained. I will come back to this crucial point below.

That democratic systems of government need to have non-majoritarian safeguards in place to protect the human and civil rights of citizens is almost uncontroversial. Laws that violate the rights of ethnic or religious minorities, of women, homosexuals or homeless people are democratically illegitimate in the output dimension, even if they are passed in conformity with input-democratic procedure. More contested is the idea that democracy requires independent, non-majoritarian institutions to safeguard the public interest (see also Moravcsik and Sangiovanni Reference Moravcsik and Sangiovanni2003, 127). What exactly is the public interest, and why do we need depoliticized institutions to take care of it?

The public interest is an elusive idea, but at the same time it is ubiquitous in debates about government, law, and economics. In academia, especially in political science, it is often used to construct a ‘normative theory about the desirability of reflecting the preferences of a general polity over special interests’ (Levine and Forrence Reference Levine and Forrence1990, 172). Being in the public interest hence denotes a particular quality of the output of the political system: the warranted presumption that political decisions benefit every citizen and not just a specific group of them. Implicit in this construction is the assumption that the abstract collective of citizens can have a discernable common interest, and that it is possible to know what that interest is. Especially the latter point sparks controversy.

Kenneth Arrow quite effectively debunked the idea that one could deduce the substance of the public interest from the policy preferences of voters (Arrow Reference Arrow1963). Other economists pointed out that all public policy choices, no matter how even-handed they may seem, are destined to privilege some private interests over others (Mahoney et al. Reference Mahoney, McGahan and Pitelis2009, 1041). Political scientist highlighted that the public interest is constantly re-negotiated in society. Its content is fluid, and there is no archimedic point from which the public interest could be defined unambiguously. In such a view, ‘it is only as political datum that the public interest has a definable relevance to the study of politics and public policy’ (Sorauf Reference Sorauf1962, 190). Skeptics hence concluded that the ‘public interest’ should be discarded as a theoretical concept. Critical arguments concerning the constructed and partial nature of the public interest are made also in the realm of international studies. IOs may pursue the political agenda of the hegemonic United States, or the West in general (Drezner Reference Drezner2007); or the material interests of a global business elite (Soederberg Reference Soederberg2003, 16). What comes in the guise of a global public interest are then, in reality, partial and self-serving decisions. Rhetorical appeal to the public interest in such a view would be just a figure of speech that makes unpleasant policy choices more palatable to dissenters (Eriksen and Sending Reference Eriksen and Sending2013, 231; see also Downs Reference Downs1962, 4).

Against the cynic’s assertion that the public interest is just what the cunning and powerful say it is, I contend that the public interest is a counterfactual ideal with practical implications. Walter Lippmann famously argued that ‘the public interest may be presumed to be what men would choose if they saw clearly, thought rationally, acted disinterestedly and benevolently’ (Lippmann Reference Lippmann1955, 40). Thus defined, the public interest is a counterfactual contrast foil. As is the case with similar counterfactual constructions (such as Habermas’s ideal speech situation), it is easier to spot instances where this ideal is not realized than stating with confidence that it is. Accordingly, the public interest is often defined ‘ex negativo’, that is, by denouncing decisions that are clearly not in the interest of every member of a society. Mattli and Woods, for instance, follow such a strategy when contrasting a vaguely defined ‘common interest regulation’ with very concrete forms of regulatory ‘capture’ (Mattli and Woods Reference Mattli and Woods2009, Ch. 1).

If the public interest is a counterfactual ideal it may be approached by way of thought experiment, as Lippmann suggested. Some even jump to the conclusion that with the help of such experiments it is possible to know ex ante what the public interest in a given situation would be, and which policy would cater to it. Ho, for example, argues with reference to Rawls’s ‘veil of ignorance’ that the public interest can be determined by ‘taking a thought experiment and momentarily forgetting one’s own identity and interests’, and that this would give us a clear indication of what policies to choose (Ho Reference Ho2012, 10). If this was so easy, we could safely rely on competent and impartial experts to make political choices on our behalf. Technocrats could, in the end, replace politicians. Most of us, however, would be less optimistic than Ho that the substance of the public interest can be found foro interno by way of thought experiment. ‘Proceduralists’, as they are often called, therefore suggest that the public interest is approached best through processes of intersubjective communication. John Dewey and, in more recent years, theorists of deliberative democracy, stand for such an approach (Bozeman Reference Bozeman2007, 110–12). For proceduralists, the public interest is best pursued through a process of reason-giving, in which all arguments pertaining to the issue at stake are weighted and participants’ will to impartiality can be assumed.

We can hence imagine the public interest of a polity as an ideal that can be approximated (even if never reached with certainty) through an inclusive political conversation. The normative point then would be to ingeniously devise institutions that can approximate this ideal situation to the greatest extent possible. Creating deliberative settings in the real world is not an easy task. There is a slippery slope towards technocratic arrangements, which have the advantage of providing shelter from the pressure of factions, thus ‘rationalizing’ the conversation. In contemporary political theory, Philip Pettit makes a particularly strong case for de-politicizing institutions to enhance the deliberative quality of their proceedings, arguing that ‘if deliberation is really supposed to rule in public life, then there is no option but to depoliticize public decisions in various ways’ (Pettit Reference Pettit2004, 64).

However, partisans of de-politicized governance are often charged with elitism and paternalism. To quote Kant, under paternal rule ‘(…) the subjects, as immature children who cannot distinguish what is truly useful or harmful to themselves, would be obliged to behave purely passively, and to rely upon the judgement of the head of state as to how they ought to be happy (…)’ (Kant Reference Kant1991, 74, emphasis in the original). In contemporary circumstances, it is probably not the head of a state to rule paternally but rather the state’s executive apparatus, with non-majoritarian institutions figuring prominently. These institutions typically function on the basis of assumed citizen interests. They are geared towards approximating a counterfactual public interest by way of thought experiment and with a strong will to impartiality. Yet these long-term interests are not necessarily articulated by citizens themselves and they are not established by collecting the actual views of citizens on the issue. Rather, they are arrived at by inferring what citizens should rationally prefer from some theoretical model.

Let us come back to the example of the independent central bank to illustrate this point. The assumption here is that all citizens should have a rational self-interest in the welfare gains that independent central banking will produce. Individuals are imagined as sharing a way of reasoning according to which more economic welfare is, ceteris paribus, better than less. Hence, having an independent central bank can be defended as being in the best interest of all citizens without permanently consulting those very citizens, or their elected representatives, over the decisions concerning interest rates. Only the assumption that such a public interest exists, that it is largely uncontroversial and stable over time, can justify the de-politicization of decision-making by way of delegation. In a very similar vein, assumed citizen interests can justify the delegation of tasks to IOs.

At any level of government, institutions geared towards securing output legitimacy thus need to confront the problem of a potential gap between what citizens should want, according to some model, and what they actually say they want. A first theoretical problem here are competing values and value rankings. Technocratic, non-majoritarian bodies not only work on the basis of assumed citizen interests but also on the assumption that it is possible to distinguish superior from inferior policies, requiring some uncontroversial scale of assessment. The standards of assessment may be contested in practice as much as the ranking of competing values. If these standards are simply imposed, without any possibility for the subjects of governance to challenge them, this constitutes a problem of technocratic paternalism. Technocratic bodies can assess, for instance, if genetically modified organisms may be patented under an existing legal regime regulating intellectual property rights but they cannot determine whether intellectual property rights should have precedence over ethical concerns. This is not a question of techné.

Second, there is the problem of shifts in the interests of the constituency. ‘Since the public mood is in a continual state of flux, it is axiomatic that the public interest must be viewed as changing also, rather than static or constant’ (Harmon Reference Harmon1969, 485). Even if it is possible for technocratic organizations to ascertain the interests of the constituency at one point in time it is unclear how subsequent transformations could be taken into account (Føllesdal/Hix Reference Føllesdal and Hix2006, 554). Locking in certain aims or policies of institutions forever therefore becomes a threat to democracy, rather than an asset. This has repercussions on the conception of democratic output legitimacy and the non-majoritarian institutions in charge of securing it. If political goals and respective institutional missions consensual once are unlikely to remain consensual forever, institutions with an output orientation need mechanisms of regular goal adjustment. They need to remain responsive to societal change and to challenges formulated by citizens. There must be some interface, in other words, between the input-oriented and the output-oriented side of the political system.

On the other hand, with regard to contestation and shifts in interests we should be careful not to overstate the case. Not all citizen interests shift rapidly and not all values are constantly contested. It seems perfectly legitimate to have technocratic agencies supervise the airspace; determine and enforce standards of hygiene in restaurants or hospitals; or devise standards of construction that protect buildings from earthquakes or fire hazards. Such missions seem unproblematic because it can be plausibly assumed that citizens have a consistent and enduring aversion to plane crashes, food poisoning and buildings on fire. The public interest does not seem to be terribly opaque here and having independent agencies operating at arm’s length from the political day-to-day business seems legitimate.

One may of course object that even apparently uncontroversial tasks may, for one reason or another, move to the center of contestation, and it may become necessary to ascertain and renegotiate, rather than simply assume, the public interest. A distinction between a routine and a crisis mode of governance may be helpful to theorize this (Peters Reference Peters1993, 348–52). A crisis occurs in our context when technocratic agencies that rely on output legitimacy are publicly challenged because their conception of the public interest is outdated, or because their claim to public interest orientation itself has become doubtful (for instance through charges of corruption). The important thing then is that their routine mode of governance can be shifted to the crisis mode. Hence, there need to be ways to open up cloistered bodies towards input from the outside so that their tasks, policies, and mission can be re-negotiated. I engage with the question of how this might be achieved in practice in the fourth section below. The theoretical point may be formulated as follows: the input-oriented and the output-oriented mechanisms of governance need not and should not be permanently intertwined. And at the conceptual level they can be kept neatly distinct. What we need in practice, however, are avenues of communication that can be activated, reliably and with ease, when the conception of the public interest on which the output side works, is called into question.

To summarize, in this section I have established and defended a notion of democratic output legitimacy of governance. Democratic output legitimacy refers to the quality of governance, and in particular its ability to cater to the public interest. A conception of output legitimacy with democratic credentials hence is intimately tied to the notion of a public interest. What that public interest is can in some instances be legitimately assumed. However, democratic output legitimacy needs to be coupled with input-oriented mechanisms whenever it becomes uncertain and contested what the public interest is. The relevant theoretical literature mostly refers to the context of the democratic nation-state, because such have been the terms of debate in political theory and public administration over the last century. In the next section I will apply the conception of democratic output legitimacy to the realm of IR, and more precisely, to public IOs.

The democratic output legitimacy of IOs

In this section I will spell out the potential of IOs with regard to democratic output legitimacy. I will argue that, under ideal circumstances, IOs can and should provide two assets in terms of public-interest orientation: another layer of checks and balances to control state power, and enhanced deliberative quality of governance with a transnational reach. Before doing so, however, I need to explore the relations between IOs and democracy more generally. That IOs can and should be democratic is far from uncontroversial. In fact, a good number of scholars perceive IOs as a formidable threat to democratic governance, rather than an asset. In this section, I prepare the ground for applying the notion of democratic output legitimacy to public IOs by addressing the relationship between democracy and IOs more generally. I will first discuss two key questions. Does democracy, transnationally conceived, require the existence of IOs? And can IOs themselves be democratic at all?

Let us start thinking about the relation between IOs and democracy by imagining our present world without them, a counterfactual status quo ante. It seems that such a world would be democratically deficient even if all states within it were perfectly democratic. National democracy unfolds within contingent boundaries that are the result of incidental historical developments (Tilly Reference Tilly1985), and that cannot be determined by democratic procedures (Näsström Reference Näsström2003, Reference Näsström2007). The political division of the world into separate territories spawns a problem that one may call ‘democratic externalities’ (Morgan Reference Morgan2003, 176). They arise when decisions that are democratically taken by a territorially defined demos have non-trivial consequences for the inhabitants of other territories (Held Reference Held1991, 142). The decision of a nationally bounded demos to pursue a high-carbon lifestyle, for instance, has external effects on others who are denied the possibility to impact the decision.

The traditional doctrine of state sovereignty suggested that every state, and implicitly also every national democratic community, is justified in not taking effects of its decisions on non-members into account. Such notions of sovereignty as non-accountability to the outside are losing ground. In the field of international law, Eyal Benvenisti recently proposed a radical reframing of sovereignty. He re-interprets ‘sovereignty and the “inherent” rights of peoples to self-determination as requiring states to assume certain underlying obligations toward strangers situated beyond national boundaries’ (Benvenisti Reference Benvenisti2013, 297). Cosmopolitan political theorists also defend the view that there is no normatively sound justification for disregarding the externalities that (democratic) national policies produce (Archibugi Reference Archibugi2004, 444), and conclude that transnational political institutions are needed to deal with them (Zürn Reference Zürn2000, 189). By implication, ‘democracy itself may no longer be possible except as a project of state interdependence and global cooperation’ (Benhabib Reference Benhabib2012, 10).

If externalities of domestic governance give rise to a ‘democratic deficit’ in IR, this deficit exists before any cooperation among states is institutionalized. It is exacerbated by traditional forms of international politics, in which powerful states not just passively constrain the options of others by way of unintended externality, but actively manipulate their choices by way of threat. In fact, domination exists in the international system before the advent of any supranational structure of authority. At the transnational level, there is no institution to prevent economically and militarily powerful actors from bullying and coercing other states, and the citizens living within them. The potential for abuses of power is particularly high in a social situation where there is little effective control over powerful actors and even less potential for sanctioning them. Quite obviously, the ‘anarchy’ (Dickinson Reference Dickinson1916) of the international system is such a setting. The manifest absence of central authority invites abuses of power (Grant and Keohane Reference Grant and Keohane2005, 30).

In much of traditional IR theory it is not even assumed that states take into account a transnational public interest (however defined) but that they maximize their national self-interest. Of course, self-interested behavior of political actors is not a unique feature of the international level of politics. Elected representatives promote the interests of their local constituencies in national parliaments as well. However, national parliamentarians are embedded into social structures that constrain the pursuit of parochial interests significantly. Their parties force politicians to take the interests of other constituencies into account because they are normally operating nation-wide and thus already need to balance parochial concerns internally. Coalition governments or multi-chamber legislatures may impose similar restraints on rent-seeking. Not least, the national media, whose reporting elected politicians cannot ignore, are there to expose ruthless pursuit of parochial interests. All these constraints are not present, or starkly underdeveloped, at the international level. I therefore argue that democratic externalities and the unchecked domination of some states over others and their citizens constitute an ‘original democratic deficit’ in international affairs, covered and legitimated by notions of sovereignty and self-determination. Public IOs with a multilateral design can be justified as a way of addressing, and mitigating, the original democratic deficit in world politics. There is a democratic, and not just a pragmatic, case to be made for IOs.

Several groups of scholars are unlikely to subscribe to this view. Some IR theorists may question if existing IOs are capable of acting as guardians of a transnational public interest. Call this the realist objection. ‘For realists, institutions reflect state calculations of self-interest based primarily on concerns about relative power; as a result, institutional outcomes invariably reflect the balance of power’ (Mearsheimer Reference Mearsheimer1995, 82). Accordingly, realists conceptualize IOs as handmaidens of hegemonic states, created to serve their parochial political purposes. In this view, IOs are biased by nature and do not have any independent political agency. They cannot act as neutral institutions oriented toward a transnational public interest even if they try.

Other schools of international theory disagree with the realist picture of IOs. Rational institutionalists approach IOs from a functional perspective as enabling states to jointly achieve certain ends, mainly by reducing transaction costs (Abbott and Snidal Reference Abbott and Snidal1998). One important function of IOs in this respect is to act as a neutral arbiter and provider of information. This already presupposes agency and some degree of autonomy for IOs and is more amenable to the idea that IOs can act as guardians of a transnational public interest. Constructivist IR scholars highlight the foundational norms and principles guiding the operation of IOs. Multilateralism is a particular organizational form that stresses state equality and the universal validity of international norms, thus hedging state power (Ruggie Reference Ruggie1992). Moreover, constructivists have shown how IO Secretariats and expert staff are able to act quite independently from the wishes of the states that created the organization (Barnett and Finnemore Reference Barnett and Finnemore2004).

To the extent that these accounts are complementary rather than mutually exclusive, the three theoretical perspectives can be brought together in a more nuanced analysis of the present international order and its institutions. John Ikenberry suggests that although the current international order was established and shaped by the hegemonic United States, it is not an empire in the traditional sense. It rather is a specific type of liberal hegemony, ‘built around political bargains, diffuse reciprocity, provision of public goods, and mutually agreeable institutions and working relationships. (…) The liberal hegemonic state dominates the order by establishing its rules and institutions – but in doing so it operates to a greater or lesser extent within those rules and institutions’ (Ikenberry Reference Ikenberry2011, 26). The norms and principles of the multilateral order constrain not only weaker states but also the hegemon itself. Thus, even if the United States remains a hegemon in the multilateral system it is not excluded that today’s multilateral IOs can act as guardians of a transnational public interest and constrain the actions of even the most powerful state significantly.

Let us now turn to those critics who argue from a democratic theory perspective that IOs are powerful actors in their own right but undermine democracy, rather than furthering it. Call this the democratic deficit objection. On closer inspection it consists of two different arguments; (i) that IOs undermine domestic democracy; and (ii) that they are structurally unable to democratize their own proceedings. As for the first argument, Karl Kaiser remarked already in the early 1970s that the internationalization of political decision-making had tilted the national balance of power towards the executive and thus compromised the primacy of parliament at the national level (Kaiser Reference Kaiser1971, 715). Parliaments had little influence on international negotiations in IOs and no control over what these organizations did. Governments were able to bypass parliamentary scrutiny and opposition by shifting certain policy decisions to the international level. IOs hence emerged as a threat to national democracy. Robert Dahl subscribed to this line of argument but added that IOs themselves can never be democratic, diagnosing an unavoidable trade-off between the scale of government and the possibility of citizen participation, concluding that ‘we should openly recognize that international decision-making will not be democratic’ (Dahl Reference Dahl1999, 23).

I consider the two varieties of the democratic deficit objection in turn. In response to the first type of criticism, Keohane et al. argued that multilateral IOs may pose a threat to parliamentarism but nevertheless can be an asset to national democracy. They can help in off-setting factions, protecting minority rights, and enhancing the quality of democratic deliberation. Keohane et al. claim that ‘properly authorized multilateral institutions, such as other commonplace constitutional institutions, may be justified in imposing checks, constraints, and corrections on majorities that are not well-informed, rights-regarding, or fairly represented’ (Keohane et al. Reference Keohane, Macedo and Moravcsik2009, 15). IOs hence are a remedy to the deficits of national democracy. That Keohane et al. come to more optimistic conclusions than Dahl regarding the effects of IOs on the quality of national democracy is at least in part a consequence of their definition of what democracy actually is. Whereas Dahl puts his emphasis squarely on the input side, Robert Keohane and other neoliberal institutionalists adopt a notion of ‘constitutional democracy’, emphasizing non-majoritarian and non-electoral institutions: the rule of law, the provision of public goods and the protection of minority interests. In my preferred terminology, Keohane et al. stress the output side of the democratic system over the input side and also suggest that there is a trade-off between the input and the output dimension (Keohane et al. Reference Keohane, Macedo and Moravcsik2009, 2).

Keohane et al. defend inter-state multilateralism on democratic grounds but for them the proper place for democracy is within the confines of the nation-state, not beyond them. The democratic benefits of IOs, in terms of output, manifest themselves nationally. Among political cosmopolitans who support the creation of global democratic institutions, Mathias Koenig-Archibugi has taken most care to reject the impossibility arguments made by Dahl and other skeptics (Koenig-Archibugi Reference Koenig-Archibugi2011). Mustering empirical evidence on domestic democratization he quite effectively refutes all claims that modest size of the territory, cultural homogeneity, or certain levels of socio-economic equality are necessary conditions for having democratic institutions. The example of India, the largest democratic state in terms of population size, is his case in point – a huge country that is multiethnic, multilingual and religiously diverse, still poor by all human development indicators and rife with socio-economic inequality. If India managed to become a quite stable democracy against all odds how can we be so sure that global democracy is unviable?

Koenig-Archibugi’s notion of democracy is complex but the emphasis clearly is on mechanisms of the input type that allow for political representation of and accountability to the citizens of the world. In fact, the cosmopolitan literature so far has discussed mainly input aspects of global democracy. Input mechanisms can take a variety of forms: global parliament (Falk and Strauss Reference Falk and Strauss2001); global stakeholder democracy (Macdonald Reference Macdonald2008); polycentric global deliberation (Dryzek Reference Dryzek2006); but this article is not the place to discuss the varieties of cosmopolitan political theory in any detail. What imports here is that cosmopolitans, with their emphasis on input and ‘government by the people’ so far have rarely engaged with democratic mechanisms of the output type.

Let us now turn to democratic output legitimacy and IOs. I begin with human rights that all states and, by way of extrapolation IOs, are to protect (Rawls Reference Rawls1999, 37). In that respect, it has been argued that international agreements monitored by IOs can help protect human and civil rights against assaults by state governments and other social groups (Hafner-Burton and Tsutsui Reference Hafner-Burton and Tsutsui2005; Keohane et al. Reference Keohane, Macedo and Moravcsik2009, 7–8). IOs also promote state compliance with international treaties in the field of human rights (Raustiala and Slaughter Reference Raustiala and Slaughter2002). Recent scholarship, however, has pointed out that IOs themselves under certain circumstances may take actions or decisions that put human rights at risk (Wouters et al. Reference Wouters2010). Rather than assuming that the effect of IOs on human rights is always beneficial, we therefore need to formulate a normative requirement that IOs should safeguard human rights in their actions if they are to avail of democratic output legitimacy. Buchanan and Keohane called this the ‘minimal moral acceptability’ of global governance arrangements (Buchanan and Keohane Reference Zürn2006, 419). This requirement is in no way IO-specific but applies to all political institutions from the local to the global level.

The second aspect of output legitimacy is the ability of an institution to keep powerful factions and private interests in check. To clarify this important point let us confront a world in which IOs are present with an imagined world from which IOs are absent and international politics still follows the mode of inter-state bargaining – issue-specific and mostly bilateral. In such a constellation, transnational problems will be resolved and conflicts settled by quid pro quo deals. These deals will reflect the threatening potential and the ability to make promises of the states involved, and they will have externalities for non-parties. In a ‘Westphalian’ state system without IOs, there is no actor that could plausibly claim to take the high view, in particular when political problems affect a great number of countries. The particular promise of IOs is to be able to do just that. ‘States establish IOs to act as a representative or embodiment of a community of states. This was a central aspiration in the postwar organizational boom and remains an important, if only partially fulfilled, aspect of IO operations today’ (Abbott and Snidal Reference Abbott and Snidal1998, 24). IOs offer a place where a global public interest can be discussed and where public reason can be at work transnationally (Mitzen Reference Mitzen2013, 53/54).

Due to their inclusiveness, multilateral IOs are the place where the global public interest can be debated and views from very different territorial perspectives can be put forward. I have argued in the section on transnational democracy above that preventing abuses of power becomes particularly salient in the context of international politics. In the international system, states are ‘private’ actors, accountable chiefly to their domestic constituency (Eriksen and Sending Reference Eriksen and Sending2013, 227). If the external unaccountability of the state in general, and the excessive power of some states in particular, is the problem, then IOs can be a solution insofar as they limit the ability of nation-states to impose their preferred policy options on others: ‘The creators and defenders of these organizations in the 20th century were not unaware of power politics. On the contrary, they conceived of these organizations as ways to reduce the impact of unequal military and economic resources on policy’ (Keohane Reference Keohane2006, 5–6). Advocates of IO in the singular, and of IOs in the plural, always highlighted the need to hedge and limit the power of military capacities and economic wealth.

Historically, this function has been allotted to global forum organizations such as the League of Nations and the United Nations. They were to ‘take public policy away from the few overstrained centers of excessive power, and to base it boldly and broadly on the general wishes and will of the peoples of the world’ (Salter Reference Salter1921, 255). Even if that might have been too high an aspiration, the League and other IOs have certainly begun to hold states publicly to account for their international and, increasingly so, also for their domestic behavior. We could hence say that IOs have the genuinely democratic potential to create an additional, vertical layer of checks and balances on state power.

IOs of the forum type can expose state conduct to the scrutiny of diplomatic peers and a ‘global public’, searching for collective approval of their policies (Claude Reference Claude1966). Such attempts may fail, as in the case when US Secretary of State, Colin Powell, in February 2003 produced bogus ‘evidence’ of Iraqi weapons of mass destruction in the United Nations Security Council. While the Security Council was unable to ultimately stop the US operation in Iraq it reinforced the legal principles of the Charter and raised the political costs of the hegemon’s unilateralism (Hurd Reference Hurd2007, 191–3). The Security Council is not shielded against state power in the sense of being financially or organizationally autonomous. As a global forum it facilitates checking states by forcing them to explain and justify their conduct before a global audience. Forum organizations can thus tackle the democratic externalities that territorial fragmentation brings about and keep powerful states in check.

The specific promise of functional IOs, on the other hand, is in enhancing the epistemic quality of governance. Collecting, verifying and comparing data is one of the chief accountability techniques of IOs vis-á-vis their member states (Ward Reference Ward2004). However, the decisive criterion, as explained above, is to what extent the decisions produced in IOs are in the global public interest. This aspect of output legitimacy is more relevant for rule-making organizations of the functional agency type. These IOs develop norms and regulations for a rather narrow policy fields, and as specialized bodies usually work at arms-length from elected politicians and governmental representatives. Here civil servants and expert consultants have a more prominent role in developing the agenda, preparing norms or recommendations, and monitoring their implementation. This lends a technocratic character to these organizations, and renders the ‘domestic analogy’ with independent agencies within the state particularly plausible.

Functional IOs with a predominantly technocratic and rule-making character exist since the 19th century (Reinalda Reference Reinalda2009). The purpose of these ‘public unions’, as they were usually called in the early days, was to manage the mounting interdependence that resulted from economic globalization and massive advances in transportation and communication technologies. Early advocates suggested IOs to tackle existing interdependencies among states and by doing so create new ones (Reinsch Reference Reinsch1911; Woolf Reference Woolf1916; Mitrany Reference Mitrany1933). Unlike ‘forum organizations’, functional organizations often take over tasks from the state and impose collective decisions in their particular functional area upon it. They are intrusive by design, which renders the question of their legitimacy particularly pressing. Functional IOs make rules, and in order for these to be ‘rules for the world’ certain conditions must apply. Historically, part of the project of IO was to shift the negotiation mode from diplomatic horse-trading to ‘arguing’, a communicative practice in which participants try to persuade each other only by the force of the better argument (Elster Reference Elster1986; Risse Reference Risse2000). To quote James Arthur Salter once again, this implies ‘(…) that even a particular negotiation should not be of the nature of a bargain; that there is for most questions somewhere a just solution independent of the relative strength of the contending parties, and that the question should be settled on these its intrinsic merits’ (Salter Reference Salter1921, 257).

Legal and scientific discourses are types of communication in which one would generally presume an orientation of speakers towards the better argument. Therefore, IOs of the functional kind are suggested as organizational environments facilitating such a shift of discourse, due to their de-politicized nature in which much of the agenda-setting is handed over to civil servants and independent experts. Hence, IOs of the functional kind (ideally) provide an environment in which the epistemic quality of decisions can be enhanced by virtue of the expertise of the specialized staff employed by the organizations and/or by the scientific expertise they solicit (Haas Reference Haas1978). Largely technocratic IOs, such as for instance the OECD, collect and aggregate data, and allow for a comparison of political experience from different parts of the world (Martens and Jakobi Reference Martens and Jakobi2010). Ideally, this will allow for a collaborative search for best practices, with a wider horizon of experiences at hand than national policy-makers would normally have. An official of the European Commission may not know more about a sector in a member state than a civil servant from that state, but will probably know more about that sector across 28 member states than any national official (Kassim and Menon Reference Kassim and Menon2003, 128). The public interest orientation of IOs may hence be defined with Andrew Baker as ‘the extent to which IO bureaucracies deliberately catalyze and facilitate participatory and consultative mechanisms for the formulation and implementation of regulatory principles, in ways which restrain powerful, vested sectional interests, while attaining a balance between a range of perspectives’ (Baker Reference Baker2012, 394).

To summarize, under ideal circumstances IOs can and should provide two assets in terms of public-interest orientation: another layer of checks and balances to control state power, and enhanced deliberative quality. Forum organizations like the UN Security Council, if they function properly, can serve especially the first goal. Functional standard-setting organizations serve the second. The latter type of organization thrives on a specific combination of organizational autonomy and expertise, which suggests that the character of a functional IO that can claim high democratic output-legitimacy would inevitably be technocratic. It seems important, therefore, to relate the critical discussion of technocratic paternalism from the second section to the debate on IOs. This is the task of the next section.

Input, output and global democracy

In the section on output legitimacy and the public interest above I argued that thought experiments are potentially insufficient to approach the public interest of a constituency. Approximating the public interest in cases of crisis requires inclusive procedures of communication, a give and take of reasons. This is a blind spot of the technocratic IO model. While reasonably accountable to states, IOs still are much less responsive to the transnational citizenry for the political programs they pursue and the consequences that their decisions have (Papadopoulos Reference Papadopoulos2010). They lack the electoral mechanism that in a national democratic polity triggers sensitivity for public opinion and citizen preferences within the political elite (Manin et al. Reference Manin, Przeworski and Stokes1999, 45; Ashworth Reference Ashworth2012). Even if international technocrats are benevolent, the epistemic problem with the construction of the public interest by experts remains.

Interestingly, some early advocates of IOs were aware of such hazards and critically discussed them. Already in 1930, liberal internationalist Alfred Zimmern argued that ‘the most striking achievements in international policy in the last decade have not been due to the decisions of statesmen but to the “recommendations” of experts’, and he found the trade-off between expertise in government and democratic participation particularly pronounced at the international level (Zimmern Reference Zimmern1930, 11; see also Mitrany Reference Mitrany1955). Translated into the terminology of input and output legitimacy adopted here, the question is the following: how much input, and what kind of, do IOs need that are specialized in creating democratic output? To what extent can this be reconciled with the requirement that they be shielded against political pressures?

First of all, there may be ‘pathologies’ of IOs that prevent them from realizing their democratic potential as guardians of the transnational public interest (Barnett and Finnemore Reference Barnett and Finnemore2004, 34–41). Officials may pursue their own ideas or self-interest and evade external accountability; dynamics of group think may prompt entire IOs to dig themselves in ideologically, becoming impermeable to outside criticism and advice. These (potential) pathologies of IOs refer us back to the trade-off between organizational autonomy and openness to criticism. As argued in the last section, IOs need a certain degree of autonomy if they are to maximize their output-democratic functions. Only relatively independent IOs can hold states to account and decide issues ‘on their merits’. I have argued that autonomy and de-politicization are, from the democratic theory point of view, a potential advantage in the ‘output’-dimension of a multi-level system of governance. On the other hand, the links of technocratic bodies to their constituency must not be severed completely, so as to allow for contestation and conversation regarding the global public interest.

In the institutional framework of the democratic nation-state, technocratic bodies are overseen by elected politicians. That mechanism can confront technocracies with the oscillating wishes of the citizenry. However, at the international level an executive with a direct electoral mandate does not exist. The intergovernmental legislative (conference or council of state parties) is a collective bargaining apparatus in which many, if not all, actors enjoy veto power and that finds it hard to act swiftly and decisively. Vis-á-vis IO-bureaucracies states face typical coordination problems of ‘collective principals’ (Nielson and Tierney Reference Nielson and Tierney2003, 242). Politicians therefore pressure IOs and other transnational technocratic bodies individually. This leads to a characteristic re-configuration of the relationship between technocracy and democracy that Martin Shapiro describes as follows: ‘In a transnational setting, […] attempts at political intervention in ‘technical’ regulatory decisions largely will be attempts by politicians representing particular nation-states. They will be seen not as democratic interventions against technocracy but as national interventions intended to gain national advantage at the expense of other members of the transnational regime. Therefore, in a transnational regulatory regime, politics and politicians tend to be identified with bad national self-interest, and international technicians with the common good’ (Shapiro Reference Shapiro2005, 349).

Therefore, a ‘statist’ strategy to make IOs more responsive to individual member states would not really resolve the conundrum. Rather, input-oriented measures to counter the threat of technocratic paternalism need to be cosmopolitan in nature. Transnational parliamentary assemblies would be one way to organize accountability of technocracies in analogy to the democratic nation-state. Such assemblies would, it seems, need to feature genuinely transnational forms of political organization, such as transnational political parties, in order to structure citizen input along the lines of transnational political cleavages, rather than national ones (on this, see Bartolini Reference Bartolini2005; Noël and Thérien Reference Noël and Thérien2008; skeptical Thomassen and Schmitt Reference Thomassen and Schmitt2004).

Yet parliamentarization at the global level is a utopia for the time being. It is, however, not the only potential remedy to the democratic deficit in the input dimension. More modest proposals abound that seek to create additional avenues of participation through transnational civil society access (overview in Steffek et al. Reference Steffek, Claudia and Nanz2008; Omelicheva Reference Omelicheva2009; Bexell et al. Reference Bexell, Tallberg and Uhlin2010). Civil society organizations are contributing to a conversation about global governance by monitoring public governance; by translating highly technical discourse into a language that citizens can understand; and by re-framing issues and promoting alternatives. There is a broad variety of actors, including social movements, religious congregations, special interest groups, and public interest NGOs involved in identifying and flagging problematic topics and decisions. These non-state actors feed new pieces of information, ‘local knowledge’ and critical commentary to IOs. They can bring new arguments to bear, yet without the threatening potential and veto power of influential states. In my view they are in a better position to challenge and debate the notion of a public good on which an IO operates.

There are some hazards associated with civil society participation in that it may aggravate, rather than reduce, current imbalances in access to global policy-making. Empirical studies have shown that the question of who actually represents ‘global civil society’ and which groups enjoy better input opportunities is a non-trivial one at the European and the global level (Kohler-Koch Reference Kohler-Koch2010). Business interests tend to be better represented in global governance, and the North-Atlantic area better than the global South. But detailed analyses of civil society participation in global governance reveal more complex patterns. For instance, advocacy NGOs from Africa and Asia are better represented at the World Trade Organization than business groups from the same regions (Piewitt Reference Piewitt2010, 480). Inequalities in global civil society are a cause for concern and they will never be eliminated completely. On the other hand, civil society organizations still have a non-negligible ability to voice political concerns and to act as watchdogs vis-á-vis states and IOs (Steffek Reference Steffek2010, 63/64).

For reasons of space, I cannot discuss such avenues of citizen input in more detail here. The point of this section was to show how the gap between IOs and citizens could be bridged in principle to allow for contestation of the public interest definitions that IOs use. As electoral democracy beyond the state is not in sight, the challenge is to ingenuously strengthen established instruments of participation to make IOs responsive to criticisms arising from the transnational public, and to thus shift them to the crisis mode of operation.