Empirical and normative logic of two-level games

In his 1988 article, Putnam addressed the question of when and how domestic politics and international negotiations affect one another. His key claim for there being a two-level logic to international negotiation is the need simultaneously to account for the interaction of domestic and international considerations. To capture this interaction, Putnam introduced the metaphor of a single game that is played at the same time on two game boards (Putnam, Reference Putnam1988: 434). At the Level I (international) board, the players are national negotiators and the aim of the game is to achieve agreement, while maximizing domestic political gains and minimizing costs; at the Level II (domestic) board, these same negotiators, in their role as national politicians, play the domestic political game of aggregating interests and constructing coalitions. Moves that are rational on one board may incur unacceptable losses on the other, often in situations of substantial complexity. However, players may spot a move on one board that will trigger realignments on the other, enabling them to achieve otherwise unachievable goals.

A crucial element of the model is the necessity for domestic ratification of the agreement reached internationally (Putnam, Reference Putnam1988: 435). Analytically, of course, negotiation and ratification are distinct. However, in practice, negotiation and ratification are linked in the strategies of players. Given the possibility of non-ratification, negotiators need to convince their negotiating partners that they can credibly deliver on their international commitments. Such potentially ratifiable outcomes are in the win-set of the status quo, as outcomes that are preferable to negotiators over non-agreement. What Putnam calls the ‘contours’ of the negotiators’ respective win-sets is the critical explanatory variable for the model (Putnam, Reference Putnam1988: 437).

Of central importance in these contours is the size of the win-set. Ceteris paribus, a larger win-set makes Level I agreement easier to achieve. Yet a smaller ratifiable win-set can give a bargaining advantage at the international level, by enabling players to maximize their Level I gains by pointing to their domestically constrained freedom of action as a means of inducing further concessions. On the other hand, and offsetting this effect, smaller win-sets increase the risk of involuntary defection. Win-set size is determined by a number of factors (Putnam, Reference Putnam1988: 437–445), including the distribution of power, the preferences and the possible coalitions among Level II constituents. Win-set size also depends on the Level II political institutions involved in ratification, so that the greater the autonomy of central decision-makers from constituents, the larger their win-set. Within the framework of a two-level analysis, this carries the otherwise paradoxical implication that greater domestic freedom can weaken international bargaining power.

Side-payments can be made to attract marginal supporters and in two-level games these can be sourced internationally. Thus, what counts at Level II is not total national costs and benefits but their ‘incidence, relative to existing coalitions’ (Putnam, Reference Putnam1988: 450). The implication of this is that international negotiators may gain some freedom from their domestic constituents by virtue of the international coalitions they form. These international coalitions may facilitate change both economically and in terms of public opinion. Issues under discussion at the international level may mobilize or change public opinion at home and this ‘reverberation’ may then affect the international outcome in turn. ‘Suasive reverberation’ can change domestic minds but reverberation could also trigger a domestic backlash. The phenomena of reverberation and synergistic issue-linkage preclude modelling the domestic game separately and using the outputs as inputs to the international game (Putnam, Reference Putnam1988: 456).

Finally, in empirical terms, the role of the chief negotiator (Putnam, Reference Putnam1988: 456–457), who formally links the two levels, needs consideration. Construing his role as agent of his constituents unrealistically simplifies the analysis: items on his agenda may include how his personal Level II standing or domestic political coalition is improved or damaged by the negotiated outcome; achievement of personal conceptions of national interest or using the necessities of international agreements to push through domestic policies that would otherwise be unachievable. The chief negotiator is thus a veto-player in the negotiation process, and his identity may widen or narrow the win-set.

To develop a normative logic of such games, we begin by noting that normatively governable relationships exist between actors who stand in structured patterns of relations with one another (MacCormick, Reference MacCormick1999: 3–6). In a two-level game, a central way of understanding these normative dimensions of roles and relations is in terms of the notion of representation, since chiefs of government and other diplomats are operating at Level I as representatives of their constituents whose ratification is necessary at Level II. Negotiations at Level I are thus to be understood as taking place among representatives of states. Yet the relationship of representation also has implications for how relationships at Level II are construed. It is precisely this dual feature of the role of representative that maps onto international negotiation understood as a two-level game. If international negotiators acquire their powers at Level I by being the authorized representatives of political associations, they also incur the obligations of representatives in relationship to their constituents at Level II.

Representation is not a univocal concept, as Pitkin (Reference Pitkin1967) stressed. Pitkin explicitly identified five different senses in which we can meaningfully talk about representation. There is authorization, when an agent is empowered to make commitments on behalf of a principal. There is accountability, when elected officials explain their tenure of office. Representation may take a descriptive form, as it will do if a sample is statistically representative of a population. Representation may be symbolic, as in the case of a flag that represents a country. And representation may be substantive, as when an agent acts in the interests of a principal. Pitkin also mentions a principle of responsiveness, which she associates with accountability, but which is clearly logically distinct, since a willingness to give an account for one’s actions need only imply a readiness to accept censure and possible dismissal should the justification not be found acceptable, whereas responsiveness implies a willingness to alter a proposed course of action if it is unfavourably received.

We can thus identify at least six types of representation. Pitkin showed that these different senses defined different normatively governed relationships among agents. Thus, she pointed out that one influential strand of thinking, inaugurated by Hobbes (Reference Hobbes1651), sees representatives in terms of authorization. On this understanding, a representative is someone authorized to act for others. Yet this sense of representation carries no implications about the extent to which the representative is accountable and indeed, on Hobbes’s own account, authorized sovereigns are not accountable to the members of the community that they govern. By a parallel argument, we cannot assimilate a substantive notion of representation, in which someone acts in the interests of another, either to authorization or to accountability, since one may act for someone’s interests without being either authorized by or accountable to that person. To say that views of representation are distinct is not to say that they cannot be found together empirically in particular circumstances, as Pitkin (Reference Pitkin1967: 225–226) herself pointed out. Heads of states or elected representatives may be treated as authorized agents for certain purposes, yet in terms of democratic norms they are only representative when they account for their actions. Some heads of state, such as the French president, have a role both in terms of authorization as when they sign international treaties and symbolically as standing for the nation on ceremonial occasions. And there is an obvious sense in which political representatives in a democracy are supposed to act in the substantive interests of their constituents. Different conceptions of representation can therefore be defined by the way in which they combine the distinct elements of representation that Pitkin distinguished.

Pitkin’s (Reference Pitkin1967: Ch. 10) own conception defined democratic representation as a mixture of authorization, accountability, and substantive representation. However, her general logical analysis can be distinguished from this particular account, as has been pointed out (Lloyd Thomas, Reference Lloyd Thomas1969; Rehfeld, Reference Rehfeld2005: 180–192) and that analysis establishes that no one sense of representation can stand duty for the others. Rather we should understand particular practices of representation as combining these conceptual elements in different ways.

How then are we to conceptualize the role of representative in international relations? Clearly, the notion of authorization has to be central. If this were not so, politicians not democratically elected or mandated would be unable to act as representatives in treaty negotiations. Yet, as Rehfeld (Reference Rehfeld2006) has stressed, undemocratic representation in international relations is common and rests upon there being some rule of recognition in terms of which the representatives can speak for the countries whom they represent. The core logic of authorization in international relations is not democratic because of requirements and expectations generated at the level of the international system. Negotiators must be able to enter into collectively binding agreements with one another if there are to be meaningful international agreements. There would be no point to an international agreement if the parties to that agreement did not have the authority to commit the state they were representing to a particular course of action. In the absence of any state’s ability to make authorized commitment, no other potential party to the agreement would have reason for thinking that the agreement was credible and therefore its effects would be nugatory. In short, the logic of assurance requires that states as representatives be authorized to make agreements.

However, to say that our notion of representation in international relations need not presuppose domestic democratic arrangements is not to deny that there is a distinctively democratic conception of international representation. For such a conception, we suggest that the crucial notion, particularly if we are thinking about the two-level logic of ratification, is representation as accountability. We pick accountability in preference to either representation as responsiveness or substantive representation for similar reasons in each case. In situations where there are conflicts of view or interests, no government can be responsive to all its domestic constituents, and by the same token there will be domestic disagreements as to what the substance is of the national interest. The circumstances of politics (Weale, Reference Weale2007: 12–18) preclude there being any easy resolution of these disagreements. Another way of putting this point is to say that representatives define their own audiences since they stand in singular–multiple relationships to their constituents (Saward, Reference Saward2006: 307–308).

From the point of view of normative evaluation therefore, we are led back to the notion of representation as emerging from the political processes that define a country’s position in international negotiations and that hold to account those who are charged with pursuing the national interest in the international field. Representatives have to construct interests, but their construction is always open to contestation within democratic practices and institutions. In terms of democratic theory this involves representation as accountability. This does not mean that we abandon the idea that in international affairs the task of governments is to advance the interests of their populations. It is rather to say that the only way of ascertaining how those interests are advanced is through political processes in which there is some accountability.

Accountability is conventionally conceived as having two components, a deliberative element and a control element. Bovens’s (Reference Bovens2007a: 450) formulation captures this well: ‘accountability is a relationship between an actor and a forum, in which the actor has an obligation to explain and justify his or her conduct, the forum can pose questions and pass judgment, and the actor may face consequences’ (compare Rehfeld, Reference Rehfeld2005: 189; Bovens, Reference Bovens2007b: 107; Castiglione, Reference Castiglione2007). In the deliberative aspect those who are accountable have to explain and justify their behaviour to others (compare Gutmann and Thompson, Reference Gutmann and Thompson2004: 3). The forms of justification range from government statements to parliamentary debates to public discussion in the media and electoral campaigning. By whatever means, accountability is a form of public reasoning, the presupposition of which is that when accountability is discharged there is an intelligible connection between the action for which an account is being rendered and the putative reasons that are offered to explain and justify that course of action.

In democratic systems this deliberative aspect of accountability is conjoined with mechanisms of political control, whether in parliament or through electoral competition. Note that the connection between any particular form of control and deliberative accountability is a contingent one in the sense that it is an empirical matter whether the control in question is an effective incentive to provide reasoned explanations of policy choices and to sanction office holders if the explanation is judged inadequate. To take an obvious example, electoral competition will not be an effective control if negative campaigning accompanied only by sound bites predominates. In two-level games, the ratification of international agreements, either by legislatures or by popular referendums, is the usual institutional mechanism of democratic control. Again, much empirical work is needed to examine how far and under what circumstances these mechanisms are effective.

Accountability and reasonable acceptability

If accountability involves explanation and justification, then we can conceptualize the notion in terms of public reasoning. Public reasoning takes place when citizens and their representatives seek to justify to one another their preferences for alternative forms of public policy. In the particular case in which governments are seeking to justify the ratification of an international agreement, this means in normative terms that governments have to offer a justification to citizens involving considerations in the form of public reasons that are acceptable to citizens and their domestic representatives.

Rawls (Reference Rawls1996: 48–54) distinguishes the reasonable from the rational by virtue of the fact that a reasonable person is willing to propose and abide by reasonable terms of co-operation. The reasons that governments offer as justifying the ratification of an international agreement require such an agreement to be to the general advantage where the burdens and benefits of that general advantage are fairly shared. To illustrate what this might mean in practice, consider international agreements in matters of trade policy or the environment. Such agreements typically aim at the general advantage of those nations making the agreement, for example greater prosperity or a cleaner environment. However, from the point of view of the reasonable, it is not sufficient to show that there is a general advantage to be gained from such an agreement; governments also have to able to say to their citizens that the domestic burdens and benefits arising from this general advantage will be fairly shared. Normatively speaking, democratic accountability as reasonable acceptability is imposed on the ratification of international agreements.

Reasonable justification is clearly distinct from rational choice in many instances of potential international agreement. Eichengreen and Uzan (Reference Eichengreen and Uzan1993) for example show how the US, the UK, and France could not reach agreement in 1933 over international trade and financial arrangements because the Deladier government in France, with a precarious parliamentary majority, was vulnerable to the lobbying of French farmers who would have lost from the reduction of tariffs that the US and the UK favoured. On the assumption that the agreement would have been generally favourable, the failure to achieve the joint gains was blocked by the inability to find a way of securing side-payments to those who would lose from the agreement. From the point of view of the French farmers, the blocking of the agreement may have been perfectly rational; the question of whether it was reasonable has to be determined by reference to the norms of fairness that can be applied to such situations.

International agreements on trade and the environment normally only meet the test of being a potential Pareto improvement, that is to say an improvement that passes the Kaldor–Hicks test of social welfare (see Hicks, Reference Hicks1939, Reference Hicks1941; Kaldor, Reference Kaldor1939). According to the Kaldor–Hicks test, a measure improves social welfare not because everyone is better off but because it creates sufficient improvement such that the gainers could compensate the losers. In conventional welfare economics, the Kaldor–Hicks test says that the change in policy should be made, even if no compensation is actually paid. Typical circumstances in which this sort of situation arises are those in which producer groups are capturing rent from trade restrictions, whilst consumers are paying unnecessarily high prices for goods. In these circumstances, an international agreement to lower tariff barriers will worsen the situation of those earning rent, but will improve the lot of consumers in general. If the gains are large enough, then the winners ought to be able to compensate the losers, although in practice such compensation need not be paid. (Kaldor’s own original example was the 1844 repeal of the Corn Laws in the UK in which domestic grain suppliers had their produce opened up to international competition.)

A norm of fairness will qualify the welfare economic application of the Kaldor–Hicks test. Within a broad Rawlsian framework, the difference principle is an obvious candidate for such a norm. According to the difference principle, if there is an aggregate gain, the burdens and benefits of that gain should be shared in such a way that the worst-off group in society would be as well off as it can be. Brian Barry (Reference Barry1989: Ch. vi) has constructed an argument for the difference principle in terms of reasonable agreement – a reconstruction that would seem especially relevant to situations of political accountability. Barry considers what the representatives of social groups could reasonably say to one another when contemplating how to balance economic prosperity in general with fair distributive shares for all. In a two-group society, representatives of the worst-off group would have reason not to agree to changes beyond a situation in which their prospects are as high as they can be. In a multi-group society, Barry suggests (Reference Barry1989: 232) that the worst-off still have an argument for the difference principle, provided that everyone accepts the principle that no inequalities should arise from ‘morally arbitrary’ advantages. Situations exist in which, in order to make the worst-off group as well off as possible, those in the middle of the income distribution may have to accept less. Yet, the worst-off can argue in favour of the difference principle that these advantages should be forgone because they would represent inequalities that were arbitrary from the moral point of view.

Although usually applied to comparisons of alternative social states, the difference principle can be thought of as a constraint on acceptable changes involved in moving from one state to another, of the sort arising from international negotiations. Understood in this way, such a norm would typically entail a substantial set of compensatory payments from gainers to losers. But is it reasonable to apply the difference principle to such situations? What grounds are there for thinking that the principle might provide a justifiable norm of fairness that would provide a constraint in applying the Kaldor–Hicks compensation test?

For us, a decisive argument against the difference principle in general, which also applies to the case of international agreements, is that it could lead to a situation in which very substantial gains to middle-income groups were sacrificed in order to make very small compensatory payments to the worst-off group. Of course, opinions will vary as to whether it is right or wrong to make such sacrifices, but it is difficult to see in a situation of political accountability that there would not be a reasonable case to be advanced to the effect that large gains should not be sacrificed in this way, particularly if the worst-off group were above some (perhaps high) minimum level of social welfare. The inequality could be characterized as arbitrary from a moral point of view, but equally the distinction between a morally arbitrary inequality and a piece of good fortune is not hard and fast.

A different case arises if the worst-off group, and poor groups more generally, are liable to suffer a setback to their interests as a result of the international agreement being ratified. Their representatives can reasonably argue that the compensation should be paid, otherwise they would be carrying an unfair share of the burden of the change. Even if they could not employ the full force of the difference principle, representatives of losers in this situation could argue that those in their position should not receive a setback to their interests or economic expectations. Moreover, those doing well from the international agreement cannot reasonably expect that the poor would be willing to have their situation made even worse as a result of the agreement.

Conversely, those losing from an international agreement might suffer disadvantage, but where pre-existing advantages were unjustified in the first place. An example of such a group could be producers who benefited from protectionist trade restrictions or received a government subsidy to support exports. It would be possible for political representatives from other groups to say that compensation would be unreasonable because those harmed by the international negotiation were losing privileges that they should never have had and that they should not be compensated for earlier ‘unjust enrichment’. Moreover, since changes in law and policy are a predictable part of business life, no producer can reasonably expect to carry on with business as usual in any case.

Note too that these lines of argument would also block a ‘proportionate gain’ principle, according to which compensation should be paid in such a way that all groups in society should gain an equal amount from the general advantage, relative to the base-line before the international agreement arose. Such a principle might be thought to be implied by a ‘reasonable expectations’ test. Sidgwick (Reference Sidgwick1901: 271) famously pointed out that one maxim of justice might seem to be that laws should avoid running counter to ‘natural and normal expectations’, since those who are disappointed in their expectations by a change in the law complain of injustice. If this principle applies, people could complain about being treated unreasonably if their expectations of livelihood were adversely affected by a change in policy, including policy brought about by international negotiation. Moreover, one meaning of ‘reasonable’ is that of demanding only modest compromises of people, and large drops in income arising from trade negotiations would run counter to reasonableness in that sense. However, and contrary to this line of argument, there is no general reason on distributive grounds for saying that compensation should always be paid. Expectations need to be not just ‘natural and normal’ but also reasonable, in the sense that when asked to justify those expectations the justification is framed in terms that others can accept. Were this not the case, licence would be given to those who enjoy historic but unjustifiable privileges to go on enjoying them (Weale, Reference Weale1998: 23–24).

So there are potentially two principles of fairness, both weaker than the difference principle, that have a plausible prima facie claim to be normative constraints on how the burdens and benefits of a general advantage might be distributed. One is that reasonable acceptability means that already poor groups should not be further disadvantaged as the result of an international agreement, so that the gainers should compensate the losers. The second is that where groups were in a privileged situation, enjoying ‘unjust enrichment’, compensation should not be paid. Could these groups overlap, however, and if so, what should be done in that case?

There can clearly be situations in which these two groups will overlap. Poor farmers may be the beneficiaries of regimes of agricultural protection and find that free trade in agriculture will adversely affect their livelihoods. Similarly, members of poor fishing communities can find their livelihoods damaged by internationally agreed conservation measures that are to the general advantage. In this situation, we suggest that the point of reasonable acceptability would be one that protected the poor from further deprivation and that demands for compensation would be reasonable.

Representation and international partners

So far our normative model of two-level games is one in which representatives of states negotiate with one another under constraints of accountability to their domestic constituents. However, international negotiators also stand in normatively governed relationships with respect to one another. To understand the two-level features of these relationships, we need to build upon those accounts in the Grotian tradition that see international relations in terms of regimes understood as ‘social institutions consisting of agreed upon principles, norms, rules, procedures, and other programs that govern the interactions of actors in specific issue areas’ (Levy, Young, and Zürn, Reference Levy, Young and Zürn1995: 274, cited also in Young and Levy, Reference Young and Levy1999: 1). The relevant norm of fairness in such an approach is that of justice as reciprocity, in the sense that states restrain their aggression towards one another and are willing to enter into international agreements when it is to their mutual advantage.

Why take a Grotian approach? The Grotian norms can be thought of as the minimum set of norms that could govern relations among agents. As Sidgwick (Reference Sidgwick1891) argued many years ago, the Grotian norms correspond at the international level to a regime of an individualistic minimum at the domestic level, that is to say a regime in which the only duties that agents owe to one another are non-aggression and non-interference, an idea that has recently been noted by Krasner (Reference Krasner1999: 14). If we are looking for the minimum duties that political representatives owe to one another, then we need to begin with the Grotian norms. Under Grotian norms, all states are equal, are secured their territorial integrity, and are governed by the norm of pacta sunt servanda.

Even within this attenuated conception of fairness typical of the international order, there is a need for regimes, understood as ‘identifiable practices consisting of recognized roles linked by clusters of roles or conventions governing relations among the occupants of those roles’ (Young, Reference Young1989: 5), particularly when states move from mutual non-aggression to common collective action. The reason for this is that reciprocal relations have to be maintained over time in a world in which it is impossible for everyone to act simultaneously. However, in any serious international agreement, ratification and implementation will take place at different speeds for different actors, and this will give rise to the monitoring and reporting requirements that are familiar features of international regimes.

When regimes become highly institutionalized, the simple notion of international fairness as reciprocity can become extended in various ways. For example, some actors may be willing to invest more in the co-operation than other actors because the public good that is provided by the co-operative action is worth more to them than it is to those other actors. In this case we have the principle of asymmetric reciprocity, in which one actor gives more than other actors in order to achieve the public good. This is the phenomenon that Olson (Reference Olson1965: 29) calls the ‘exploitation of the great by the small’ and can be illustrated in the case of the US contribution to NATO. Similarly, in dense and extended regimes, some actors may be willing to modify the initial conditions within which agreements are made. An example of this would be agreements on technology transfer under the climate change regime. The structural funds in the EU can also be thought of in this context, providing some (modest) redistribution of resources from the better off to the worse off. However, the core notion of fairness is that of reciprocity to mutual advantage within the constraints of a mutual respect for state integrity.

So far this is merely to say that international regimes rest upon norms of fairness at the international level. However, the point of the two-level game analysis is that Level I and Level II interact. This is the significance of saying that negotiators are playing simultaneously at different tables. Thus, concessions made at Level I carry implications for the understanding of representation at Level II, just as conditions at Level II – for example institutional requirements of ratification – influence the agreements that can be made at Level I. Within a normative interpretation, we need to consider what the counterparts are in terms of the principles and rules that should apply.

Consider first the influence on Level II norms deriving from obligations that arise at Level I. At Level II, as we have seen, the principal relationship is cast in terms of accountability. That is to say, those representing a country at the international level have to be able to justify to their domestic constituents the agreement that is being proposed, even when that agreement involves a rebalancing of domestic interests. However, if the logic of justification is genuinely two-level, then representatives ought properly to refer in their justifications to the obligations that states owe to one another by virtue of their joint participation in a regime or set of regimes. Thus, if an obligation arises from previous international agreements, representatives may properly say to their domestic constituents that these obligations ought to weigh in the decision, even though weighing them may prove detrimental to the national interest narrowly conceived. In a related paper, we have suggested that Tony Blair’s defence of increased UK contributions to the EU budget as the price for enlargement fell into this category (Savage and Weale, Reference Savage and Weale2006).

If the logic of two-level games is genuinely operative in this way, however, the interaction of international and domestic bargains must work in both directions, so that Level II considerations should be admissible at Level I. That is to say, it is not simply that international agreements carry implications for the way in which chiefs of governments should meet their obligations of accountability to their domestic constituents, but also that these obligations of accountability should be anticipated by international partners in negotiation. In the empirical theory of two-level games, this shows up in the role that ratification plays in international negotiations. The argument here is that international negotiators will be constrained in what they can push for by the knowledge that those whose agreement they are seeking will in turn have to seek agreement, formal or political, from their domestic constituents, and by the same token, those from whom concessions are being asked will be able to claim domestic constraints as a reason for resisting concessions.

The normative equivalent of this pattern is one in which international negotiators will have to ask not only about the fairness or reasonable acceptability of the agreement they are proposing, but also whether this fairness can be rendered accountable to the domestic constituents of those international representatives whose agreement they are seeking. Thus, the two-level feature of the normative structure can be brought out in the following double-barrelled conditions:

1. 1. International negotiators ought to think not only of their own legitimate interests but also whether they are making reasonably acceptable demands on partners accountable to their domestic constituents.

2. 2. Domestic constituents will have to think not only about the reasonableness of the burdens that they are expected to bear but also about the burdens that the protection of their interests imposes on others, given the feasible set of international options.

The general principle that guides this normative account is that all agents simultaneously have to ask whether the expectations that they have under any agreement are reasonable given the expectations of others under that agreement, just as in the empirical analysis we ask whether actors have a rational expectation of achieving win-sets by inducing reconfigurations of domestic coalitions in the light of possible international coalitions. The reasonable and the rational therefore mirror one another in the logical structure of norms and expectations, although they are distinct in their practical implications.

To show that this specification is not merely formal but finds expression in the world of international relations, consider international negotiations over air pollution in Europe since the 1970s. The first international agreement for the prevention of air pollution was the Convention on Long-Range Transboundary Air Pollution (LRTAP) agreed in November 1979. This regime arose from the Scandinavian concern over air pollution over sulphur dioxide emissions and other gases leading to the acidification of Scandinavian lakes. Those who have studied the evolution and development of this regime agree both that the rational bargaining position of the affected countries was weak, since they were downwind of polluting sources with no retaliatory capacity, and that the regime has led to changes in behaviour in polluting countries. Thus, Munton et al. (Reference Munton, Soroos, Nikita and Levy1999: 215) write that the influence of LRTAP ‘can be traced to its ability to induce states to treat its norms and outputs as authoritative, beyond simply calculation of interest’ and Underdal (Reference Underdal2000: 377) writes that ‘we see evidence of research findings influencing beliefs and positions of governments as well as the general public’ in most of the relevant countries.

Central to the success of these argumentative effects were a number of features. Firstly, there was the technical acceptability of a mathematical and statistical model of European air pollution developed by IIASA (see Weale, Reference Weale1992: 198). Though not perfect, for example in the degree of its resolution, the model underwent extensive peer review in order to establish its technical credibility (Horndijk, Reference Horndijk1991). Secondly, those responsible for the regime were able to innovate intellectually, most notably in the concept of the ‘critical load’, which defined the point at which pollution started to damage eco-systems, and which became an acceptable concept to those negotiating over abatement. Thirdly, as Oran Young (Reference Young1999: 266) has pointed out, LRTAP gave credibility to a broader set of principles, most especially the principle of precaution, which was advanced not only to serve the narrow interests of the affected countries (legitimate though they might be) but also as general principles that ought to be acceptable to governments and democratic publics.

This example thus reveals one aspect of the normative pattern of two-level games to which we have sought to draw attention. In developing both a robust analytical and evidential framework and advancing principles like that of precaution that have validity to broader publics, the Scandinavian countries were meeting our first condition of the normative logic of two-level games, namely that international negotiators were making reasonably acceptable demands on partners who were in turn accountable to domestic constituents. They did not seek to make requests that were unreasonable in the sense that they were able to provide evidence, arguments, and principles that could be used by representatives negotiating on behalf of countries required to undertake costly compliance in explaining and justifying the burdens incurred to their own domestic constituents. They were conducting their international negotiation under the shadow of domestic accountability.

Implications for democratic theory

So far we have argued that the requirements on representatives of accountability and reasonableness provide the basis for a normative logic of two-level games. If representatives have to be able to make credible commitments in situations in which they are accountable to their domestic constituents, and in which they recognize that their negotiating partners are also accountable to their domestic constituents, then the range of considerations that they can advance are limited to those that are reasonably acceptable. The interlocking norms of fairness and accountability therefore provide restrictions on what arguments can be legitimately advanced in two-level games. Here we comment on two implications of our approach for the democratic theory more generally.

There has been a long-standing discussion in democratic theory between trustee and delegate theories of representation. Delegate theories insist that representatives should not only be institutionally under the control of their constituents but also think of themselves as the transmission mechanism for their constituents’ wishes. On trustee theories, representatives always need some freedom from their constituents if they are properly to represent them. Clearly the empirical logic of two-level games works against the delegate account of representation, since the logic of such games bestows more freedom on representatives than the delegate account would allow. How does the matter stand in relation to the normative logic?

In situations in which coalitions of domestic actors can rationally anticipate the successful exercise of their veto power, there is latent in the distinction of the reasonable and the rational an alternative conception of political accountability in which Putnam’s freedom of chiefs of governments in two-level games plays a part. Domestic interest groups are typically well attuned to international negotiations that have implications for their welfare and prosperity. Unless the international negotiations produce a result that so rearranges domestic coalitions that ratification is secured, the losers can use a public process of ratification to block agreements that are to the general advantage – even when they would only be carrying a fair share of the burdens of securing that general advantage. If we say that the normative logic is one of domestic accountability, we would seem to be providing the conditions under which interests could exploit their veto power without justification.

Accountability, however, need not mean day-to-day accountability, in which the hands of international negotiators are tied. In Putnam’s account, a significant consequence of two-level games is that chiefs of government acquire more power relative to their domestic constituents than they would have were it just a one-level decision. This outcome arises because they can use the international level to gain some discursive leverage at the domestic level. If accountability were purely a matter of controlling representatives on a day-to-day basis, tying their hands in international negotiations, the freedom of representatives would always mark a loss on the democratic metric. No doubt the freedom of negotiators does often mark such a loss on the democratic metric. However, there are circumstances in which executive freedom can be construed as a democratic gain, for it enables time and opportunity for governments to explain and justify the generally beneficial character of their policies even when politically significant interests are adversely affected. In other words, a trustee conception of political accountability may be one of the conditions preventing the rational pursuit of self-interest by organized groups who wish to supplant the reasonable by the rational.

We have discussed accountability in relation to the obligations of states and their domestic constituents. This understanding of accountability at Level I confines itself to responsibilities of negotiators to negotiating peers. Might there be responsibilities also to those peers’ constituents or to emerging ‘global publics’? Grant and Keohane (Reference Grant and Keohane2005) have explored the problems of and necessity for mechanisms of accountability over global governance. They stress the absence of a global public in a structured political relationship to international institutions as key to the problem of accountability globally (Grant and Keohane, Reference Grant and Keohane2005: 34) and note that non-democratic as well as democratic accountability mechanisms need to be considered. They identify emerging standards of legitimacy in world politics such as human rights norms and standards derived from international law such as human rights instruments and WTO rulings (Grant and Keohane, Reference Grant and Keohane2005: 34–35) and suggest that ‘peer accountability’ and ‘reputational accountability’ are the main mechanisms by which states can be held to account (39). While our argument does not speak directly to this issue, it may be that recognition of the normative logic of a norm of reasonable acceptability could become a factor in peer and reputational accountability.

One further answer to this question is that there are particular reasons why democracies should be open to principles articulated at the international level. Democracies can be characterized as open societies, an idea that goes back to Popper. The concept of an open society is normally associated with an account of the institutions and practices that are characteristic of democracy, for example the free association of political movements or the absence of restrictions upon communication among citizens. However, the notion of democracy as an open society can also be understood in terms of the character of its mode of deliberation in decision-making. This involves the thought that in a democracy decision-makers are open to arguments, whatever their source (Popper, Reference Popper1945: 225). Popper was originally thinking of the requirements of political equality within domestic political debate and the normal pattern is indeed that decision-makers in a democracy deliberate on arguments and principles that are advanced in domestic forums of decision-making, whatever these may be. However, relevant arguments may be articulated not only domestically but also internationally. If only domestically generated arguments count, then this would be incompatible with the notion that arguments should be weighed wherever they come from, and so would be incompatible with the idea of democracy as an open society.

To see the force of this argument, consider a situation suggested by Pogge (Reference Pogge2003). Suppose we were to say that an important cause of the poverty of developing countries was due to the shaping of international trade rules by wealthy countries. Then we would, in effect, be saying that when rich countries make the rules, they are not making them in a form in which they are anticipating the requirements of reasonable acceptability they are imposing on their negotiating partners with respect to their domestic populations. To be open to arguments in the way required by the principles of democracy is to say that international obligations should be shaped by a sense of fairness generated through the normative logic of two-level games. Of course, in any particular case, that logic may or may not be effective. But it will always be relevant if the logic of two-level games is taken seriously.