What does the AAP entail?

In a landmark analysis of the AAP, Goodin shows that some common interpretations of the principle are incoherent. Understanding the principle as requiring the inclusion only of those interests that are actually affected by the actual decision is incoherent, Goodin shows, because what the decision actually turns out to be depends upon who actually makes the decision. Seen in that way, the principle would be unable to identify who should participate in making a decision until after that very decision has been made, creating a paradox. Therefore, what matters for inclusion is who would be affected by any possible outcome of a decision process. Moreover, we need to ask not only who is affected by any possible outcome of any options that are on the agenda, but also by outcomes of options that might be on the agenda. In other words, participation in agenda-setting itself has to be governed by the AAP. Goodin concludes that the most coherent and cogent interpretation of the AAP is that a say should be given to anyone who might possibly be affected by any possible decision arising out of any possible agenda. He notes that this ‘possibilist’ interpretation of the AAP provides ‘good grounds for thinking that (at least in principle) we should give virtually everyone a vote on virtually everything virtually everywhere in the world’ (Goodin Reference Goodin2007, 64).

Goodin expects many readers to recoil from such a slippery slope, and discusses ways in which the expansionary implications of the possibilistic AAP could be reined in. The only qualification that he regards as cogent is to amend the principle as follows: a say should be given to anyone who might probably – instead of possibly – be affected by any possible decision arising out of any possible agenda. Requiring the impact on one’s interests to be probable rather than merely possible reduces the expansionary thrust of the AAP to some extent, but Goodin notes that it is still very substantial.

Since many consider it wildly impractical to give virtually everyone a vote on virtually everything virtually everywhere in the world, Goodin mentions two arrangements that ‘might best approximate that ideal in some practice that is feasible’ (Goodin Reference Goodin2007, 64). The second-best solution is to overlay existing territorially defined states with a limited ‘world government’ specialized in hearing appeals from citizens affected by the actions of states in which they have no vote. The third-best solution is to create a system of ‘international law’ that obliges states to pay compensation to people whose interests have been affected by decisions but had no opportunity to participate in making them.

Responses to the expansive implications of the AAP

The apparent metamorphosis of the AAP into an all-included principle raises an important question: can the AAP provide the theoretical foundation for an empirical assessment of democratic deficits in global politics? The goal of using the principle to gauge democratic deficits in global governance would be undermined if the yardstick became impossibly demanding. Since neither Goodin’s first-best, nor his second-best, nor his third-best arrangements are realized in the real world, we are left wondering whether all existing modes of political governance fare equally bad with respect to the demands of the principle.

Not surprisingly, the expansive implications of the AAP lead some political theorists to reject it in favor of more restrictive criteria. For instance, Fraser endorses an alternative principle of inclusion partly because ‘the all-affected principle falls prey to the reductio ab absurdum of the butterfly effect, which holds that everyone is affected by everything’ (Reference Fraser2008, 64).Footnote ⁴ According to the alternative favored by Fraser, which she calls the ‘all-subjected principle’, what confers moral standing to individuals is not ‘the sheer fact of causal interdependence, but rather their joint subjection to a structure of governance that sets the ground rules that govern their interaction’ (Reference Fraser2008, 65). But this move away from the AAP raises two questions. First, is it philosophically persuasive? Second, is it practically necessary?

In relation to the first question, the all-subjected principle seems to come with problems of its own. If subjection is defined as being under a legal obligation to comply with rules, then perversely the principle would fail to enfranchise those who are subject to coercion without being expected to abide by laws (Abizadeh Reference Abizadeh2012, 12). If, by contrast, subjection means being subject to coercion, the restriction seems normatively problematic. If we are interested in the ability of people to lead their lives in pursuit of their conceptions of the good, that is their autonomy, then we should take into account that this ability is affected not only coercion but also non-coercive interventions; hence, they should be entitled to have a say on the latter (Valentini Reference Valentini2014, 792–93). Moreover, requiring the inclusion only of those people who are actually coerced is incoherent for the reason Goodin has identified in relation to the AAP: whether coercion is used or not depends upon who makes the decision, and whether coercion is on the agenda or not depends on who participates in determining the agenda. What matters is therefore if the decision-maker has the capacity to coerce (Koenig-Archibugi Reference Koenig-Archibugi2012). But it can be very difficult to distinguish between the capacity to coerce and the capacity to affect interests more generally.Footnote ⁵ Determining what the all-subjected principle requires in practice may not be any easier than applying the AAP.

The debate is on-going, but these points suggest that replacing the AAP with the all-subjected principle is not likely to provide a straightforward solution to the expansiveness problem. Fortunately, however, abandoning the AAP is not necessary to solve that expansiveness problem, as the remainder of this article will show. The version of the AAP formulated by Goodin can be retained as the normative lodestar in its original form, but its implications deserve closer consideration. There are two reasons for retaining it as a useful yardstick for assessing the democratic legitimacy of global policy-making.

The first reason is that, like other general normative principles, the AAP can serve as the basis for both absolute and comparative judgments of legitimacy. In other words, it can be used to identify the contours of social arrangements that would conform to the principle, and it can also be used to compare different (actual and hypothetical) arrangements in terms of their relative distance from the ideal. While Sen (Reference Sen2009) and Valentini (Reference Valentini2011) disagree on the usefulness of absolute (‘transcendental’) judgments of social arrangements, they agree that criteria for comparative judgments are useful even if none of the social arrangements being compared perfectly meets standards of justice. The comparative use involves doing with the AAP what Pogge urges with reference to human rights: ‘a comparative judgment is required about how well the existing institutional design does in terms of realizing human rights relative to its best feasible alternative’ (Reference Pogge2010, 31). Readers may regard this point as relatively straightforward, so I will move on to discuss the next point more extensively.

The second reason why the AAP can function as a guide for empirical assessments is that even the ‘possibilistic’ interpretation of the AAP yields implications that are less expansive than those suggested by Goodin himself. This is because, in making decisions, actors often shape and, crucially, restrict the agenda of other decision-makers. What can be called ‘upstream’ decision-makers can restrict the agenda of ‘downstream’ decision-makers by imposing a variety of constraints. Among the forms that such constraints can take are limits on the transfer of material resources necessary for action, the imposition of legal responsibilities or limits to delegated authority, the shaping of categories of thought or beliefs about what is desirable and feasible, or others.Footnote ⁶ In turn, restricting someone else’s agenda typically reduces the size of the group that is affected by any decision taken given that agenda. Since such restrictions can involve multiple links in a chain, the group of people that the AAP requires to be involved in a downstream decision can become relatively small and, crucially for our purposes, identifiable through empirical research.

Dimensions of decision-making

Thus, the key argument of this article is that ‘upstream’ decision-makers often shape and, crucially, restrict the agenda of ‘downstream’ decision-makers, and that restricting someone else’s agenda typically reduces the size of the group that can be affected by decisions taken given that agenda. Consequently, we need to take into account such restrictions when we ask who could possibly be affected by a decision and thus should have a voice. Which decisions are constraining (upstream) for which decisions is ultimately an empirical question. However, as Grant and Keohane note, ‘having the right to participate in politics as an affected party is ambiguous at the global level’ (Reference Grant and Keohane2005, 33). Thus, empirical research cannot proceed entirely inductively and needs a heuristic framework that somehow tells researchers where to start and where to focus their attention.

There are various ways in which the relationship, and specifically hierarchical relationships, between decisions can be conceptualized. For instance, the focus could be on varying spatial scales or levels, so the analysis focus on decision-making sites within what is usefully described as multilevel governance (Papadopoulos Reference Papadopoulos2010; Zürn Reference Zürn2010). But a particularly promising way of conceptualizing relationships between decisions draws on a way of seeing the policy process that is widespread in the public policy literature. One example is the work of Abbott and Snidal (Reference Abbott and Snidal2009), who identify five components of the regulatory process: agenda-setting, negotiation of standards, implementation, monitoring, and enforcement.

Inspired by this literature, I maintain that global policy-making entails at least four main dimensions that generate different types of decision: (a) decisions on whether and how to frame an issue as a policy problem that requires urgent action, (b) decisions on which governance arrangement, either existing or to be established, should be used to address that policy problem, (c) decisions on which specific policies should be adopted in the context of that governance arrangement, and (d) decisions concerning the implementation of those policies. I understand this as a minimal list of decision types: each item is both essential and irreducible to other items. To count as comprehensive, an assessment of global policy-making cannot avoid examining them explicitly. But some researchers may well find it useful to consider additional dimensions beyond the four covered here.

While this list is inspired by the literature on ‘policy stages’, it is important to bear in mind two things. First, those dimensions do not operate in neat sequences and that feedback loops are possible. Second, the policy-stages literature usually assumes the prior existence of an organization (often a government) where agenda-setting, decision-making, and implementation occur, whereas it is more appropriate to assume that the creation and selection of organizational settings is often the result of prior decisions to frame policy problems in certain ways – in other words, the organizational setting of policy-making is endogenous to the policy process (Jupille, Mattli, and Snidal Reference Jupille, Mattli and Snidal2013).

What is important for our purposes is less the exact delimitation and number of dimensions in the policy process than the insight that, in different ways, decisions made in relation of one dimension of policy-making constrain the agenda in relation to other dimensions, and that this relationship provides analysts with some guidelines on applying the affectedness criterion to the assessment of the democratic credentials of actual instances of global policy-making. Three sets of relationships are especially relevant. (1) The agenda of decision-makers charged with implementation is restricted by prior choices of policy instruments. For instance, if policy-makers decide to achieve their objectives by legally prohibiting certain behaviors rather than materially rewarding other behaviors, those in charge of implementation are likely to have the legal authority to impose sanctions and perhaps some resources to detect violations, but not the ability to distribute material resources widely across the target population. (2) The agenda of those choosing among policy instruments is constrained by prior decisions on the features of the governance arrangements within which those choices are taken. For instance, if an organization has not explicitly been endowed with the authority to adopt binding international law, normally this option will not be on the agenda of policy-makers working within that organization. (3) Decisions on the existence, nature and severity of the policy problem place significant constraints on all other dimensions of policy, and specifically on the creation or selection of the governance institutions meant to address it. For instance, no international organization is likely to be created to address issues that are predominantly framed as non-problems, or as non-transnational problems, or as transnational problems impervious to the kind of interventions typically associated with international organizations (Dimitrov et al. Reference Dimitrov, Sprinz, DiGiusto and Kelle2007).

Implications for the empirical study of affectedness and democratic control

Recall Goodin’s variant of the AAP: a say should be given to anyone who might probably be affected by any possible decision arising out of any possible agenda. The ‘any possible agenda’ clause in this formulation should thus be interpreted as ‘any agenda that is possible after the constraints set by other decisions’. Elsewhere (Koenig-Archibugi Reference Koenig-Archibugi2012), I accept Goodin’s point that the content of decisions cannot be used to differentiate between persons or groups entitled to participation and those who are not, but I argue that the power resources controlled by decision-maker do matter for that purpose. If a decision-maker’s agenda is heavily constrained, then whatever decision she takes, it will probably (see Goodin’s modified formula stated above) affect only a limited circle of persons. Such a focus on power and constraints allows us to develop an approach to specifying participatory entitlements that preserves the intuitively appealing aspects of the AAP while avoiding the slippery slope for which virtually everyone in the world should be equally able to control any person involved in global governance for any aspect of their work. The approach is based on the following propositions:

• ∙ The process of framing policy problems determines the definition of global priorities and thus the allocation of global efforts and resources among a wide variety of possible uses. Global priority setting presumably affects virtually everyone in the world; hence the AAP requires participation to be as inclusive as possible. This obligation concerns less the ‘invention’ of policy frames than the selection of priorities for global action. However, problem framing and prioritization tends to be diffuse processes, and this characteristic makes it exceedingly difficult to identify who is powerful in any particular circumstances. It is challenging to imagine what democratic institutions focused on this dimension may look like.Footnote ⁷

• ∙ The process of creating and selecting governance institutions is influenced by how policy problems have been framed, but power-wielders can still choose among a wide range of institutional designs. The creation and selection of governance arrangements affects very large sets of persons in a diffuse way, hence the AAP would require participation to be very inclusive. Compared to problem framing and agenda-setting, however, the activity of selecting governance institutions is less diffuse and the identification of power-wielders is easier. Compared to the democratic control of framing, institutionalized mechanisms of democratic control are easier to conceive and, perhaps, create.

• ∙ The design of governance institutions places procedural and substantive constraints on the range of policies that can be chosen by policy-makers. These constraints typically restrict the set of people who will probably be affected by the policies decided within the governance arrangement. One reason for this is that the agenda of an organization is often restricted to a limited number of policy issues, and thus any decision that might be taken is likely to have a significant impact only on certain communities but not on others. Another reason is that the organization may be under tight constraints with regard to its ability to extract resources from members and other actors, which in turn limits its ability to affect the interests of many people. The narrower the range of policies that decision-makers within a governance arrangement can effectively choose, the smaller the circle of those who have a valid claim to participation.

• ∙ Which policies are chosen limits the range of options for those responsible for policy implementation. For instance, the latter may only control a limited amount of material resources to be spent to the benefit of a predefined category of beneficiaries according to specific guidelines. In such cases, the set of people who are likely to be directly affected by their decisions may well be small.

The argument so far distinguishes between individuals who are affected by the choice among the options on the agenda and those who are not, but of course often people are affected by choices to different degrees. In other words, individuals can have different ‘stakes’ in the choice, where a stake can be defined as the ‘the pay-off difference between the better option from the individual’s perspective and the worse one’ (Conradt and List Reference Conradt and List2009, 730). Some authors argue persuasively that differences in stakes should be taken into account when allocating participatory entitlements (Macdonald Reference Macdonald2008; Brighouse and Fleurbaey Reference Brighouse and Fleurbaey2010; Valentini Reference Valentini2014). The framework presented here is compatible with this position. A synthesis of the two perspectives would entail that the assessment of democratic deficits should take into account simultaneously differences in power among decision-makers and differences in stakes among stakeholders (Koenig-Archibugi Reference Koenig-Archibugi2012). Consider the example given above: those responsible for policy implementation of a development project control a limited amount of material resources to be spent to the benefit of a predefined category of beneficiaries according to specific guidelines. A vast number of tax-payers across many countries may be affected by bad implementation choices that result in their contributions being wasted. But the intended beneficiaries of the development project are likely to be affected to a much larger degree by its success or failure, and therefore ensuring that they have a say in the relevant implementation decision-making processes should arguably have priority over including the more diffuse interests of tax-payers.Footnote ⁸

What do the abstract points made so far mean for the empirical study of the democratic credentials of global governance? They yield some practical guidelines: (1) researchers should examine who is (most) affected by the selection and design of governance institutions given the policy problem as it has been defined. Then they should assess to what extent (and how) those stakeholders can exercise control over power-wielders in that dimension. (2) Researchers should examine who is (most) affected by policy choice given the governance institution as it has been created or selected. Then they should assess to what extent (and how) those stakeholders can exercise democratic control over power-wielders in that dimension. (3) Researchers should examine who is (most) affected by policy implementation given the policies as they have been chosen. Then they should assess to what extent (and how) those stakeholders can exercise democratic control over power-wielders involved in policy implementation.

Establishing guidelines for the empirical assessment of the democratic quality of problem framing and prioritization is more difficult compared to the other dimensions, because of the highly diffuse and fluid nature of the process. Researchers from a variety of traditions agree that influence on policy framing is highly unequal and concentrated in ‘western’ elites,Footnote ⁹ but in practice it may be impossible to identify ex ante actors with a disproportionate capacity to influence global discourses and norms before they are instantiated in specific governance arrangements and policies. Even the ex post determination of the locus of actual authority may be problematic. We can state a further guideline: (4) researchers should identify systematic biases in the formation of a global discourse, which may have allowed it to become dominant without being thoroughly challenged by alternative discourses in an efficient ‘marketplace of ideas’ or an ‘ideal speech situation’. But we should bear in mind that the empirical challenges for this task are substantial.

All four guidelines entail that researchers should aim to identify groups of stakeholders by tracking the power that could be exercised over them, rather than assuming that stakeholder groups are identical to predefined categories such as the population of a country, but also defined by gender, ethnicity, or class (Macdonald Reference Macdonald2003).

An important point needs to be clarified in relation to the normative assumptions underlying these guidelines. To give participation opportunities to people affected by the options on a given agenda cannot redress a prior injustice stemming from excluding others from shaping that agenda.Footnote ¹⁰ The approach developed here is not ‘compensatory’ in the sense that an entire policy process (from upstream to downstream) could become democratically legitimate by virtue of some parts of it satisfying the AAP. The approach has rather a diagnostic function: to help pinpoint where deficits occur, and possibly provide insights into which institutional mechanisms might reduce them. This diagnostic function is of course compatible with finding that all dimensions of policy-making suffer from significant deficits (as in the case study below). However, it is also compatible with finding that, given similar undemocratic constraints placed on the set of decisions that, say, two international organizations (IOs) could take, one IO involves a much wider range of affected interests than the other when it comes to picking policies from that set. Such a finding could prompt institutional reformers to focus their attention on what happens upstream from the relatively inclusive IO rather than what happens within it.

In sum, this section has developed two key arguments: (1) we can accept the principle that a say should be given to anyone who might probably be affected by any possible decision arising out of any possible agenda without having to conclude that virtually everyone should have a vote on virtually everything in the world. This is because the ‘any possible agenda’ clause in this formulation should be interpreted as ‘any agenda that is possible after the constraints set by prior decisions’. (2) Agendas are likely to be progressively more constrained as the policy process moves from problem framing to institutional selection and design, to the selection of policy instruments, to the implementation of those policy instruments. Correspondingly, whichever decision is taken in relation to each dimension, it is likely to affect a progressively narrower circle of persons, which facilitates the identification of those who must have a say in order for the AAP to be satisfied.

Democratic control over problem definition and agenda-setting

Most child labor policy initiatives are driven by what can be called the ‘abolitionist’ policy paradigm. The roots of the abolitionist paradigm are in the early 19th century, when the work of children started to be perceived as a policy problem and as an international problem. Sensitivity toward the negative effect of children’s work on their development and health, and a range of other policy goals – for example, the supply of healthy recruits for national armies – increased with the European industrial revolution. Several countries adopted legislation aimed at reducing child work either directly, in the form of minimum age for employment laws, and/or indirectly, in the form of compulsory education legislation. At the same time, social reformers and policy-makers perceived child work as a problem that required international cooperation. This occurred for two main reasons. First, there was a widespread belief that a unilateral ban on children’s work would have put domestic industries at a disadvantage vis-à-vis foreign competitors. Governments would be reluctant to eliminate child labor if other governments were not taking similar steps. The second rationale for international action on children’s work in particular is that at stake was a fundamental human interests and promoting children’s welfare should be a global and not purely national responsibility. This rationale gained strength in the 20th century as part of a general movement toward a discourse of human rights and the weakening of exclusive state sovereignty in these matters. Children’s work became framed as a violation of universally valid human rights (Baccini and Koenig-Archibugi Reference Baccini and Koenig-Archibugi2014). Importantly, the abolitionist paradigm prescribes not only policies aimed at reducing the supply of child labor, such as poverty reduction strategies, but also policies aimed at discouraging the demand of child labor by imposing sanctions on potential employers and monitoring workplaces.

Various elements of the abolitionist paradigm are contested and resisted, but no single coherent alternative paradigm has emerged. Some alternative perspectives depart from the fundamental values underpinning abolitionism. Some maintain that the idea of a work-free childhood is a narrowly western idea that does not resonate in many non-western contexts. For instance, Nordtveit reports that in Namibia, Swaziland, and Benin ‘[s]ome community members totally reject the western discourse of education and protection, and state that child labor (albeit not sex work) is a normal part of the local culture and thus a natural part of the socialization process’ (Reference Nordtveit2010, 707) (see also Bass Reference Bass2004; Abebe and Bessell Reference Abebe and Bessell2011). In some cases, beliefs in the legitimacy of child work are linked to wider beliefs about social stratification, and specifically ideas that work by children of specific classes, castes, or groups (e.g. girls) is part of the ‘natural order of things’ (Weiner Reference Weiner1991). A different normative perspective maintains that children are capable of self-determination and should decide whether work is in their interest or not, and that working in good conditions can contribute to their development into free and responsible human beings (Liebel Reference Liebel2004).

Opposition to the abolitionist paradigm stems not only from alternative values, but also from considerations of practical feasibility. Interviews conducted by the author of this article with about 60 parents in villages receiving assistance from ILO child labor projects in India show that their understanding of children’s work departs in significant ways from abolitionist beliefs. Their priority was the availability of schools with no teacher absenteeism and no discrimination on grounds of poverty and caste, as well as the provision of food, clothes, shoes, and books to children attending schools, without the need to pay fees. They also stressed the role of better employment opportunities for adults and financial help for the purchase of farming animals. By contrast, they did not favor attempt to reduce the demand for child labor through sanctions and inspections. A research team from the Institute for Human Development (IHD), which collected information on a sample of 5545 individuals working in the same area, similarly found that villagers often expressed skepticism toward, or outright opposition to, key components of the abolitionist policy paradigm, and in particular the legal prohibition of child labor and related enforcement measures. In the focus groups conducted by the IHD team, parents ‘complained of restrictions on preventing children from working. According to them, interventions without taking care of their woes and means of livelihood did not offer a sustainable solution, and children would continue to work either openly or covertly’ (Sharma, Sharma, and Raj Reference Sharma, Sharma and Raj2000, 67).Footnote ¹¹

A multitude of power-wielders develop, promote, and adopt the abolitionist policy paradigm and its alternatives. As noted above, the discursive dimension of global governance is particularly fluid. First, virtually everyone in the world is affected in some way or another by decisions on which issues should be considered problems that require (urgent) policy action across borders. This is partly because political attention and resources are scarce and easily shifted by perceptions of relative importance. Even once child labor is defined as an issue that deserves priority attention and resources compared to other issues, a wide set of people is affected by the dominance of a particular policy paradigm rather than a different one, and the extent to which a globally dominant paradigm is resisted and challenged at various governance levels. Second, the relative influence of various power-wielders on global discourses is quite difficult to estimate and especially very difficult to predict in advance. It is therefore difficult to determine which sites of decision-making should be open to the participation of whom.

Despite these difficulties, it is reasonable to conclude that the way child labor was framed as an international policy problem does not satisfy the AAP. The capacity to shape the relevant discourses was unevenly distributed across social groups. This imbalance can be seen between the governments that supported the abolitionist paradigm (mostly in the rich West) and those that were more skeptical (mostly in low-income countries), but also between civil society groups that promoted competing framings. Two developments since the 1980s were particularly important. First, activists from developing countries were successful in mobilizing support in western countries against what they considered inadequate levels of protection of children on the part of their own governments. This resulted in North–South coalitions involving both NGOs such as the National Consumers League and Anti-Slavery International and government representatives, notably in the ministries of labor in the United States and Germany. Second, there was a resurgence of concern in developed countries about purported unfair competitive advantages gained by developing countries thanks to the exploitation of their workers. Child labor became linked to broader anxieties about increasing economic globalization and competition. Two components of the abolitionist paradigm that were first combined in the 19th century – child labor as a moral evil and as a source of unfair trade advantages – became closely intertwined again at the end of the 20th century.

The resulting pressure placed governments of developing countries, many of which were unwilling to adopt the abolitionist paradigm in their own policy-making, on the defensive. National government representatives occasionally voiced opposition to abolitionist policy initiatives, usually by referring to economic constraints in the poorest countries.Footnote ¹² Opposition to the abolitionist paradigm was voiced also by some civil society groups, notably organizations of working children, sometimes known as NATs (Niños, Niñas y Adolescentes Trabajadores), and their adult supporters (Liebel Reference Liebel2004). However, these working children’s organizations – let alone unorganized working children – never had the same opportunities to influence public debate as the transnational abolitionist coalition.

There thus is evidence of a gap between the globally predominant policy paradigm and the perceptions of the groups that are most intensely affected by the resulting policies. To be sure, children everywhere face substantial obstacles in making their voices heard beyond their immediate social circle, and the usually disadvantaged and marginalized status of parents of working children hinders the public reception of their views as well. But activists and scholars have criticized the ILO and abolitionist NGOs cooperating with it, notably the Global March coalition, for failing to provide opportunities for genuine participation by children expressing a variety of views. Invernizzi and Milne (Reference Invernizzi and Milne2002) and Hahn and Holzscheiter (Reference Hahn and Holzscheiter2013) note that the main international forums for the development of child labor policies, notably the conferences organized by the ILO, offer opportunities for participation to abolitionist NGOs such as the Global March, but not to organizations of working children. According to Invernizzi and Milne, the marginalization of working children in the public debate results from an ‘abhorrence of child labour that is largely constructed by a highly efficient media and public relations process’ (Reference Invernizzi and Milne2002, 405). This asymmetry points at a significant democratic deficit in the discursive construction of children’s work as a policy problem and of viable solutions.

Democratic control over forum selection

The decision to adopt and promote the abolitionist paradigm ensured that powerful actors consider the work of children as (a) a policy problem that (b) has to be addressed internationally through (c) a governance arrangement that can nudge states to make a binding commitment (d) to outlaw child labor in the domestic economy. This dominant framing constrained the institutional design options of policy-makers. For instance, policy-makers never seriously considered the option of building a global welfare state guaranteeing food, health, and other basic needs to all human beings by levying supranational taxes on the use of flight paths, sea lanes, ocean fishing areas, or transnational economic transactions, as advocated by some scholars (e.g. Parekh Reference Parekh2003). But decision-makers retained the ability to choose from a range of designs.

The choice was made to preserve the ILO’s central role in the organizational field of child labor policy, which it occupied since its creation in 1919. The ILO provides the institutional context for the drafting and adoption of international labor conventions, which are legally binding on states that choose to ratify them and thus accept to be subject to the ILO’s monitoring and supervision system. Between 1919 and 1973 the ILO adopted several minimum age conventions, to which the Convention on the Worst Forms of Child Labor was added in 1999. While the ILO’s supervisory system is relatively thorough and demanding, the organization itself cannot impose sanctions for non-compliance, apart from threatening expulsion to persistent violators of ratified conventions. The ILO bureaucracy is mainly a promotional authority that attempts to ‘socialize’ states into accepting a certain set of labor standards as normatively appropriate and instrumentally beneficial. Promotional activities are linked to the provision of ‘technical assistance’, and in the field of child labor the ILO’s International Program on the Elimination of Child Labor occupies a prominent role.

The centrality of the ILO in international policy-making on child labor should not be seen as inevitable, but as the result of decisions taken by political actors. In the past 30 years, numerous actors have proposed different, if not necessarily conflicting, institutional arrangements for addressing the problem of child labor at the transnational level. These include: (1) unilateral governmental trade measures, such as US Generalized System of Preferences program which includes a ‘labor rights clause’ from 1984; (2) the inclusion of child labor concerns in multilateral trade agreements, and specifically the World Trade Organization (WTO); and (3) private-sector initiatives that certify to consumers that certain goods have been produced without child labor and/or by companies that help adult workers to raise family incomes. In this sub-section, I will focus on the choice between the ILO and the WTO as main institutional options compatible with the abolitionist paradigm.

Especially from the mid-1990s to the mid-2000s, the possibility of using the regulative and enforcement mechanisms of the WTO to promote the elimination of child labor as part of a broader ‘social clause’ was frequently demanded and debated in national and international contexts. The US government made some attempts to include the possibility of trade sanctions for labor rights violations in the rules of the WTO at various ministerial meetings, notably the Singapore meeting in 1996 and the Seattle meeting in 1999. The European Union was divided on the issue, with the government of France leading a coalition in favor of creating a linkage between trade and labor standards and the government of United Kingdom leading an anti-linkage coalition (Burgoon Reference Burgoon2004). The attempts to include labor standards in the WTO were thwarted by the virtually unanimous opposition of governments of developing countries, with the Indian government being a particularly vociferous critic of the social clause proposal (Kolben Reference Kolben2006).

From the point of view of developing countries, the ILO was more inclusive than the US domestic trade sanctioning mechanisms, but also less unequal in terms of bargaining power than the WTO. The ILO lacks the enforcement powers of the WTO dispute resolution mechanisms and was thus less effective as a potential tool for protectionism. The ILO was acceptable to governments of developed countries as it could be, and has been, used to legitimize decisions on granting and withdrawing special trade privileges to specific countries (Orbie and Tortell Reference Orbie and Tortell2009).

At first sight, the process may seem an example of Southern governments that, especially in the case of democratic India, effectively represented the most directly affected populations and succeeded in thwarting the designs of Northern governments representing less intensely affected groups. However, the degree to which the governments of developing countries are accountable to, and/or represents the interests of, the sections of the populations who are most affected by policies targeting child labor is not obvious. While India’s political system provides more channels for democratic accountability than most, scholars as well as activists argue that it is structurally biased against defending the interests of poor children (Weiner Reference Weiner1991).

It is at least conceivable that civil society groups may articulate the preferences and interests of children and their families more accurately than governments. One of the main supporters of a social clause was the International Confederation of Free Trade Unions (ICFTU) (Anner Reference Anner2001; Roozendaal Reference Roozendaal2002; Burgoon Reference Burgoon2004). ‘The ICFTU sees itself as a democratic, mass-based organization representing 156 million members who indirectly pay membership dues to the organization. That is, for international labor unions there is a clear sense of to whom they are accountable, which is not always the case with many NGOs’ (Anner Reference Anner2001, 7). While some critics allege that the ICFTU acted as the mouthpiece of Northern trade unions when demanding a social clause, surveys suggest that the proposal was supported by large majorities of southern delegates to international trade union conferences (Griffin, Nyland, and O’Rourke Reference Griffin, Nyland and O’Rourke2003). However, the support given to the social clause proposal by southern trade unions may be the result of complex relationships and dependencies between them and northern unions, within the ICFTU and elsewhere, rather than a genuine belief that the social clause proposal was or would be supported by their own constituencies (Anner Reference Anner2001). As Burgoon noted, ‘The truth is that some developing-country labor groups ambivalently support linkage and are disfranchised in government policy-making, but face other “poor peoples” groups opposed to linkage. And the politically relevant fact is that social support for linkage is not strong enough to dent the anti-linkage façade of developing-country governments’ (Reference Burgoon2004, 41). Moreover, the ability of trade union federations to represent workers in the informal sector is being increasingly questioned. Cooney points out that ‘It is difficult to argue now that delegates drawn from trade union federations can give an effective institutional voice to the majority of the workforce in most countries: Those workers in the informal sector are unorganized and/or are unemployed’ (Reference Cooney1999, 371). This assertion is perhaps true especially in relation to working children and their families.

In sum: even taking into account the restriction of the agenda caused by problem framing, decisions on which kind of institution should make child labor policies still affects most people in the world, because – depending on the institutional design – these policies can have knock-on effects on trade flows, employment, incomes, and other outcomes of interest to a wide constituency. The AAP thus requires a highly inclusive decision-making process. The fact that the governments of developing countries were able to prevent the WTO from playing the main role in this policy domain indicates that intergovernmental power relations were not highly imbalanced. However, this is not sufficient to conclude that the AAP was satisfied. Children who (are likely to) work and their families are disproportionally affected by institutional design decisions, but they have relatively little influence over the outcomes. Even in electoral democracies, their own governments are often unresponsive to their preferences. Moreover, civil society actors involved in decision-making are often unsuited to close the democratic control gap: as we saw, (a) trade union federations are, however, tenuously, accountable to real constituencies, but these do not generally include working children and their families; and (b) numerous groups purport to speak on behalf of working children and their communities, but they are generally not accountable to them. The result is that, also on this dimension, levels of democratic control of the relevant power-wielders fall significantly short of the requirements of the AAP.

Democratic control over policy design

The decision to tackle child labor through the ILO has important implications for the kind of policies that can be adopted and, consequently, for where the boundaries of the most affected groups lie. While the WTO option would have linked child labor decisions to trade policy decisions and hence affected people in most economic sectors, the choice for the ILO means that policies have to be relatively focused on matters that are more immediately relevant to working children and their families, narrowing the circle of people who are most affected. Moreover, the rules and structures of the ILO place significant substantive and procedural constraints on the decisions about policies that can be adopted in the context of that organization. Specifically, the agenda of decision-makers does not include options that would substantially compromise the sovereignty of member states. Governments can refuse to ratify conventions and to participate in ILO child labor programs (although they cannot completely escape scrutiny, because of the global reports following the 1998 Declaration of Fundamental Principles and Rights at Work).

To illustrate what was possible in practice, and who participated in the decision, I will refer to one specific set of ILO projects, the Integrated Area-Specific Projects (IASP) against Hazardous and Exploitative Forms of Child Labour that were implemented in six districts in four Indian states during the period 1999–2003. The main goal of the IASP was to withdraw 10,000 children aged 7–12 from hazardous employment, provide them with non-formal education for up to 18 months and then mainstream them into regular schools. The project also aimed at ensuring the school enrollment 20,000 children aged 5–7 who would have otherwise not been enrolled, partly through the creation of community-level support structures. The project design combined regulative, allocative, and ideational components. The regulative dimensions consisted in the establishment of child labor law enforcement task forces at the district level, which reflected the ILO’s traditional emphasis on legal strategies and enforcement machineries. The allocative dimensions consisted in the creation and administration of transitional educational centers for former child laborers and the provision of matching grants for self-help groups of mothers. Children attending IASP Transitory Education Centers (TECs) were expected to do so for between 6 and 18 months, and then join the regular state school system. The ideational dimension of the project was important as well. It aimed at changing perceptions of families, employers, local administrators, and other target groups about the costs and benefits of various options regarding education and child work. Ultimately, the project aimed at experimenting with integrated interventions and persuading decision-makers in the Indian government to consider the lessons learned from IASP and adopt them in their own policy interventions.

It is relatively easy to identify who was responsible for taking the key decisions on the content of these interventions. The design of the key features of the IASP resulted from trilateral negotiations between the ILO, the Government of India and the donors, notably the Swedish and British governments. It is also relatively easy to identify those who were affected most directly and intensely by them: the children who were targeted by the program, their families, and immediate communities. I will refer to them as ‘beneficiaries’, without implying that they necessarily benefitted from it. In order to assess whether the AAP was satisfied, we need to ask (a) to what extent the decision-makers who designed and approved the policies of the IASP were authorized by, or accountable or accessible to, the putative beneficiaries, and (b) whether the latter were involved in designing the basic features of IASP. The answer is negative for both questions. First, India is an electoral democracy, but this does not guarantee that the most affected will have influence on specific decisions, especially when they are very poor (Kosack Reference Kosack2014). Second, the IASP lacked any significant mechanism for beneficiaries to provide inputs into the design of the project. The final evaluation report for the project implemented in the Mirzapur district notes that the IASP project ‘has been more or less a “packaged”, “top-down” programme as it was not able to fully assess and mobilize the primary stakeholders through comprehensive “Needs Assessment” exercises in order to ensure active participation of all its stakeholders’ (Kumar Reference Kumar2003, 15). The report concludes that ‘The community or the parents never felt “participation” in IASP programmes and its decisions’ (Kumar Reference Kumar2003, 88).

Did the decisions of the IASP power-wielders reflect the implicit preferences of the putative beneficiaries, if not their expressed will? To answer this question, we need to ask whether participation by the intended beneficiaries would have produced a substantially different policy mix. There are reasons for believing so. An example is the question of the transitional character of the intervention. A key feature of the project was the emphasis on TECs as bridging schools aimed at transferring children into mainstream government schools as soon as possible, rather than creating a parallel education system. A least in the Mirzapur district, where I interviewed numerous beneficiaries of the IASP, the parents would have preferred a parallel system, given the difficulties of securing access to the state school system and the limited benefits expected from enrolling their children in them. Similarly, the Mirzapur monitoring and evaluation report stated that ‘the parents and the community do not accept this role of TEC as bridge schools as most of the parents wish them to continue for longer and further longer period. There is dearth of good quality schools in all the villages of the district’ (Kumar Reference Kumar2003, 53). In contrast to government schools, there was little or no teacher absenteeism in IASP TECs and children received regular midday meals, educational and vocational material, and uniforms and sandals. Parents in the villages that had IASP TECs told me that the closure of the schools had been sudden and regrettable. They reported to have reacted with anger but also a sense of powerlessness. Several interviewees stated that their children were attending regular government schools, but at the time of my visit they still hoped that the TEC would reopen, as it had been of better quality.

The relationship between education in TECs and in mainstream schools is a complex subject, on which parents are likely to have different views both within and across villages and districts, not least because of local knowledge of the availability and quality of government schools in the vicinity of the village. If this component of the IASP policy mix had been opened up for debate in the affected communities, with the results taken into account in the policy-making process, it is probable that the content of policies would have been different, and left more flexibility in relation to local conditions. More generally, the donors, the ILO, and the Government of India agreed on a limited time frame for the project, but the beneficiaries were in little doubt that it was too short to address their needs in a satisfactory way.

We can conclude that not only the persons most directly affected by the content of the IASP program did not have sufficient voice in determining it, but also that this exclusion had significant consequences on which policies were due to be implemented. The way decisions were taken at this stage does not satisfy the AAP.

Democratic control over policy implementation

In this section, I continue to examine the IASP project and focus on its implementation in one of the six target districts, Mirzapur, which I visited. We have seen that the agreement establishing the IASP placed significant constraints on the agenda of the decision-makers in charge of project implementation. However, a range of important issues remained for them to decide. Who were those decision-makers? The main responsibility for implementation lay with the director of the IASP program based in the ILO offices in New Delhi and his staff and, above all, the district-level Child Labour Project Societies, which for all practical purposes meant the District Collectors as their chairperson. In India, the District Collector is the most powerful government official in the district under his/her responsibility and is appointed by the central government to administer law and order and perform a wide range of judicial and executive tasks. The District Collector of each district appointed a project director to manage the operations of the IASP, which in three IASP districts was the district assistant labor commissioner. The Collector and the project director selected other project personnel, normally through local NGOs. From the point of view of the beneficiaries, the most important roles were the teachers of the TECs and the community mobilization teams. The community mobilization teams had the task of persuading parents that the long-term returns from education outweigh the short-term loss of income from child labor, to help mothers to set up and use the credit scheme for self-help groups (SHGs) and to work with teachers in state schools to facilitate the admission and retention of children who attended TECs. The district project directors, the field officers, the community teams, and the implementing NGOs were all vertically accountable to the District Collector, who had the power to dismiss personnel and terminate contracts. The IASP Director for the ILO mainly exercised financial accountability.

To what extent were these decision-makers subject to control mechanisms exercised by the people who were most directly and intensely affected by how the IASP was implemented? These intensely affected people were the children whose transition from work to schooling the project was meant to facilitate, their families, and other members of their villages. The IASP was virtually devoid of channels through which they could participate in decision-making. In certain parts of India, such as Karnataka, institutional reforms had made teachers accountable to the local village community (panchayat), with significant improvements in teaching standards (Drèze and Sen Reference Drèze and Sen2002, 174–75).Footnote ¹³ Such mechanisms were not incorporated in the IASP. Community staff and TEC teachers conducted parents meetings, but their aim was to stimulate the interest of parents for the continued education of the children, rather than providing feedback on the project or provide an accountability forum.

My interviews with parents indicate a favorable assessment of the IASP TEC in their villages. Teachers attended regularly and teaching quality was considered better than in mainstream government schools. This positive assessment also emerges from a systematic survey conducted by the monitoring and evaluation institution (Kumar Reference Kumar2003). But when I asked the parents what they could have done if (hypothetically) the teacher had underperformed, the answer was invariably ‘nothing’. The inclusion of ‘user’ accountability mechanisms in the design of the IASP would have provided some safeguards against the risk of non- or under-performing teachers, but they were not created.

Ultimately, the main officials responsible for implementation were not accountable to the intended beneficiaries of the IASP. The overall IASP Director was appointed by the ILO and insulated from the project beneficiaries. That an element of such user participation might have made a difference in practice is suggested by the misgivings that some SHG members expressed to me about the ‘rule’ that members withdrawing money from the common SHG bank accounts needed to pay 2% monthly back into the account as interest. The IASP project director acknowledged that he had ‘encouraged’ such a rule to protect the interests of those who did not take out money. Some interviewees opposed the rule, but felt too uneducated to ask their local community worker to change it. It appears that the community workers did not let the SHGs decide whether to adopt the rule or not, and no participatory mechanism was put in place to correct for that. Nor was the District Collector accountable to the beneficiaries of the IASP activities. Collectors are accountable to state politicians, who exercise control over them mainly through the tool of frequent transfers (Banik Reference Banik2001). ‘As a consequence of its colonial heritage, as well as the hierarchical social system, administrative accountability in India was always internal and upwards, and the civil service’s accountability to the public had been very limited’ (Saxena Reference Saxena2010, 453). This lack of accountability to beneficiaries facilitated the alleged misuse of IASP funds by Mirzapur’s district administration, which triggered a freeze of IASP funds by the IASP Director. The affair delayed the start of activities in the Mirzapur district, and because of the fixed termination date for the project, only two-thirds of the total budget was spent and the rest was returned to the donor. The villagers lost a significant amount of schooling resources because of these problems.

In contravention of the AAP, therefore, the people most affected by implementation decisions faced significant barriers both in participating in taking those decisions and in holding decision-makers accountable.