International Organizations

Transcription

1 6/8/2017 Categorization in International Organizations Draft paper for ISA Hong Kong Doron Ella INTERNATIONAL RELATIONS DEPARTMENT, HEBREW UNIVERSITY OF JERUSALEM

2 Introduction In December of 2016 the European Union and the United States refused to grant China a market economy status within the framework of the World Trade Organization (WTO). China s categorization as a non-market economy within the WTO allows other members to apply antidumping measures against its steel exports, with the goal of making it reconsider its export policies, as was agreed to during the Uruguay Round of negotiations (GATT, 1994). In the case of the United Nations Framework Convention on Climate Change (UNFCCC), Turkey had been initially categorized both as an Annex I and as an Annex II Party, meaning that if it decided to ratify the Kyoto Protocol, then it had to legally abide by a set of stricter rules than countries that were categorized as non-annex Parties. Turkey refused to ratify the Kyoto protocol unless it would be removed from its Annex II category-group, and in 2002 it was removed, after members of the organization reached a compromise which left Turkey only as an Annex I Party with favorable conditions (UNFCCC, 2001). These two examples illustrate how international organizations (IOs) employ different categorization mechanisms, and how these mechanisms subsequently affect member-states institutional behavior and policy-making. Nevertheless, many IOs do not employ categorization mechanisms, and those that do, usually vary in the type and utilization of these mechanisms. This variation could be explained by IOs different agendas towards the desired level of cooperation, and their desire to achieve balance between different members in controlling various institutional bodies. This paper investigates why some IOs classify member-states in different categories, but others do not. It also examines differences in the type of categorization and the specific cooperation problems these mechanisms are intended to solve. I assume that states aim to establish IOs which will able to attract the most relevant international actors. Consequently, the aspiration for inclusiveness exacerbates certain cooperation problems due to the heterogeneity of member-states. Therefore, as I argue in further detail below, categorization mechanisms are employed by actors as a rational solution for cooperation problems that hinder the successful operation of IOs that include heterogeneous states, while also deterring possible defections of participating members. In this regard, I present two main arguments. First, categorization is intended to solve cooperation problems in IOs that deal with issues characterized by high uncertainty about the state of the world. And second, categorization is intended to solve cooperation problems in IOs that deal with issues requiring deep cooperation, and thus are more institutionalized. These cooperation problems 1

3 intensify in IOs that are more inclusive in nature and include members with heterogeneous capabilities and preferences about the ideal treaty level. While categorization can also be treated as an independent variable (for example, see: Ella 2017), this paper focuses on the factors leading to the choice of categorization as a rational solution for cooperation problems, thus, as a dependent variable. As such, it could be considered valuable to learn under what circumstances actors decide to incorporate categorization mechanisms within newly established IOs. Furthermore, why states decide to incorporate certain types of categorization mechanisms and not others. These questions are important since categorization is a common institutional feature within various contemporary IOs, and it constitutes a rather distinctive aspect of their membership rules. Lately, states and their delegates have started focusing on categorization as a part of negotiations in various IOs (for example, see the debates at the conferences of the climate regime: IISD 2014, 35 36). However, there has not been a serious attempt at mapping and explaining the empirical and theoretical meaning of institutional categorization. Classic institutional literature had focused on how IOs promote inter-state cooperation by disseminating information and prolonging the shadow of the future (for example, see: Krasner 1982; Keohane 1984). Rational design, as the most relevant body of literature which should address the issues of categorization, has expanded the classic institutionalists arguments and focused on examining how certain institutional design aspects affect member-states cooperation patterns, compliance and overall institutional efficacy (for a comprehensive account of institutional design theory, see: Koremenos et al. 2001; also see: Koremenos 2005; Johnston and Acharya 2006; Fearon 1998; Rosendorff and Milner 2001; Mitchell and Keilbach 2001; Thompson 2010). However, rational design literature has not given any scholarly attention to the design feature of categorization, or to the factors that might determine this mechanism s incorporation within IOs. Still, some studies focus on these factors in isolation. For instance, Martin (1992; 1994) discusses the effects of members heterogeneity in IOs, and Koremenos (2005) focuses on the interaction between international uncertainty and flexibility, while Downs et al. (1996; 1998) investigate how the depth of agreements affect members compliance within IOs. Yet, these factors and others have not been examined in interaction with the design aspect of categorization, and 2

4 therefore considering their interactive effect could prove as a meaningful contribution to current IR literature. In order to empirically test these hypotheses, I have created a new data-set on categorization within 162 IOs dealing with various issues, ranging from trade liberalization, to telecommunications, climate change and non-proliferation. This data-set encompasses information on IOs categorization mechanisms and their types, alongside other variables which will be discussed further-on. This article starts by defining categorization and placing it within the rational design theoretical context. The theoretical section introduces the aspect of members heterogeneity - how it affects matters of institutional efficiency, heightens cooperation problems which are caused by uncertainty and degrees of institutionalization (as a manifestation of the depth of cooperation), and how the incorporation of categorization mechanisms can ameliorate these problems. Next, the categorization data-set will be introduced, alongside with the proposed model, and the main variables are operationalized. Finally, the results will be discussed in length and several directions for further research are suggested. Institutional Categorization: A Theoretical Framework Institutional categorization is a process by which states are classified into different groups, according to criteria relevant to the IO s issue-area, or in some cases, according to the establishing states political agenda. For example, the WTO promotes differentiated treatment for developing countries, as some of its members acknowledged that developing countries are at a very different stages of economic, financial and technological developments and therefore have entirely different capacities as compared to developed countries in taking on multilateral commitments and obligations (WTO 2001a, 9; 2001b, 1-2). Currently, the WTO grants developing countries with longer time periods for implementing agreements and commitments, alongside technical assistance (WTO 2001b, 2). The process of categorization usually takes place when members join the organization; it is formally documented and constitutes a part of the IO's basic structure (see, for example: UNFCCC 1992). The outcome of the categorization process is the creation of several formal groups of states that are usually expected to adhere to certain, differentiated, rules of behavior that correspond with their institutional category-group. Hence, whether to employ a categorization mechanism is decided upon when states establish a new IO, and then this 3

5 mechanism is utilized to classify newly joined members and regulate their ongoing cooperation within the IO s framework. Categorization can be considered as a flexibility provision of IOs membership rules, as it partly constitutes the terms on which new members are accepted (and into which category), and how these can be adjusted in the future as conditions change (on flexibility, see: Koremenos et al. 2001; Koremenos 2005; Thompson 2010). As such, categorization is an instance of adaptive flexibility, similar to escape clauses, because it allows members to respond to new circumstances and adjust their position within the institutional framework, corresponding to new international and domestic realities, while preserving existing institutional arrangements. This can be done either by rearranging members' positions within existing categories, or by changing the nature of current categories so that they can better correspond to new realities. Categorization mechanisms can isolate a specific problem - such as inconsistencies between category-members provisions and responsibilities and their capabilities and/or preferences - and insulate the broader organization from its impact, thus preventing potential defections (see: Koremenos et al. 2001, 773). Categories and their types usually vary from one IO to another, can center upon different subjects, and assign different institutional rights and obligations to their members. As a part of the IO s membership rules, the employment of categorization is intended to determine how certain states should conduct themselves within its framework and regulate various aspects of their participation according to their category-placement. I divide categorization mechanisms into two types, based on the mechanism s criteria for the classification of members which are: 1. various aspects of members' material capabilities, and/or 2. Permanent circumstances, such as members' geographic location in certain regions, or other rigid criteria, such as being classified as founding or non-founding members. I label the former as hierarchic or vertical, and the latter as nominal or horizontal categorization. IOs with hierarchic category-systems have their members classified according to various formal material indexes. These might include GDP, GNP per capita, credit ratings, military capabilities - such as air, maritime and nuclear capabilities, certain technological capacities (for example, capacities to tackle climate change effects), or other characteristics such as economysystem types (such as the differentiation between market and non-market economies). Hierarchical categorization mechanisms take into consideration variation in members capabilities, and how 4

6 these might affect their ability and willingness to comply. Additionally, as a platform that distinguishes members according to their material capabilities, hierarchical categorization is employed with the thought in mind that without it certain members may be rendered incapable of fully complying with the IO s rules and treaties, and consequently would have to defect. Therefore, hierarchical mechanisms apply differentiated rights and obligations to members according to their institutional classification. Moreover, as also pointed above, hierarchical categorization mechanisms often encompass flexibility provisions (as opposed to nominal categorization mechanisms, which are permanent in nature). The flexibility of IOs categorization mechanisms is demonstrated through modification, but mainly through process of category-migration. IOs with hierarchic categorization mechanisms may allow members-states to migrate between category-groups. The establishing states take into consideration that international and domestic realities, such as members capabilities and preferences, are prone to change over time, and therefore they design flexible categorization mechanisms which allow movement between categories to accommodate future developments (for an account of the migration process within the UNFCCC for instance, see: Depledge 2009). Migration between categories could also be regarded as a graduation process. The use of this term is primarily applied in Multilateral Development Banks (MDBs) which provide loans for developing countries. Countries graduate when they reach a certain development level and no longer need financial aid. For example, since its inception in 1960, 41 member-states have graduated from the World Bank s International Development Association (IDA) and were no longer categorized as borrowing countries. Nevertheless, since then, 11 countries, including many African countries such as Cameron, Zimbabwe and the Republic of Congo, have encountered reoccurring economic difficulties and became reverse graduates, returning to their previous status as borrowers. Some countries, China being the most notable example, have succeeded in graduating from being borrowing countries and attain the prestigious status of a contributor country (IDA 2016). IOs with nominal/horizontal category-systems, however, have their members classified into groups according to rigid criteria, for the most part by their regional and geographic location, usually by using members' formal classification under UN Regional Groups (see: DGACM 2014), or simply by declaring states as regional or non-regional members. Although at first glance this type of categorization seems trivial, it actually enables members to put candidates for elections, 5

7 corresponding with their regional quota, to various governing bodies within the organization, such as the Board of Directors, and thus has the potential to make the process and its outcome quite political. Indeed, nominal categorization can be perceived as a device which allows institutional representation, which supports collective action by like-minded groups of decision-makers (also see: Kahler 1992, 705). In the Asian Infrastructure Investments Bank (AIIB), for example, the Board of Directors is composed of 12 members, of whom 9 are elected by the governors representing regional (Asian) members, and only 3 by governors representing non-regional members (AIIB 2015, 14). This division guarantees that regional members would have more power over decision making within the organization. Nominal categorization, in some cases, could determine whether certain members would be eligible for assistance. Within the Inter-American Development Bank (IDB) only regional (American) members can be considered as borrowing members, and non-borrowing members benefit in that only suppliers from member-states can provide goods and services for IDB-financed projects and the Bank can only employ citizens from those countries (IDB 2016). While most of the global IOs categorize their members according to regional groups defined by the UN, other IOs, dealing with issues that relate to specific areas which are not necessarily described as regions, such as Antarctica or the Mediterranean Sea, might categorize their members accordingly. For instance, the General Fisheries Commission for the Mediterranean (GFCM) defines its members as coastal states whose vessels engage in fishing in its area of application, and Japan is classified in a group of its own, due to the fact that it is not an actual Mediterranean coastal state (GFCM, 2017). In sum, when the establishing states decide to incorporate a categorization mechanism into IOs, they have to decide what type of mechanism should it be (hierarchic, nominal, or both). This means that they would have to decide on what basis (and material index, if applicable) the organization will categorize its member-states, how many categories will it include, and whether these categories will assign differentiated rights, provisions and obligations to their members. The establishing states also need to decide whether, and under what conditions, to allow members to move between categories, or whether to allow amendments and modifications to be made to these categories under certain circumstances, thus including flexibility provisions within these mechanisms. 6

8 Inclusiveness, Heterogeneity and Institutional Efficiency Because states use IOs to further their own goals (Koremenos et al. 2001, 762), they aim to establish organizations which will be efficient, sustainable, and politically viable. Consequently, for IOs to achieve their goals within the scope of their specific issue-areas, they should be able to attract the most relevant international actors, or if they require large-scale cooperation, they should be able to attract as many actors as possible, while maintaining a multilateral framework which is both inclusive and relatively efficient. For example, when designing environmental IOs, it is imperative to include the U.S. and China, as they are the world s most polluting states, and leaving them out would render these type of organizations ineffective. Therefore, states choose to incorporate categorization mechanisms within IOs with the general goal of encouraging relevant states to join, and later on prevent the possible defection of participating members - either by not complying with the IO s rules and treaties or by abandoning it altogether. In other words, categorization is intended to make initial agreements easier to reach and maintain over time. This is done by solving cooperation problems that are exacerbated by the heterogeneity of members. Thompson and Verdier (2014, 15-16) assume that most regimes are organized by a small group of founders who try to enlist the cooperation of a larger group of regular members. The founders take the lead on account of their larger resources, greater interest in the regime, or principled concern over the issue. Arguments regarding the effects of members heterogeneity on cooperation are based on this assumption. Indeed, cooperation is harder to achieve in large numbers. Having many actors can increase the conflicts of interests among them, uncertainty about others preferences, and opportunities for undetected free riding (Martin 1992, 773). Large numbers are associated with members heterogeneity. States vary greatly in their capabilities and preferences, and such variation introduces heterogeneity into problems of international cooperation (Martin 1994, ). Indeed, heterogeneity inhibits efforts to cooperate unless unusual conditions, such as the existence of a privileged group, are met within the scope of individual issues, heterogeneity can create conflicts of interest that reduce available gains from cooperation (Ibid). This is mainly due to asymmetries in states capabilities to successfully comply with multilateral agreements and also due to asymmetries in preference intensities, which are built into most models of international bargaining (Ibid, 483). Preference heterogeneity can also be understood as a distributional conflict, since it means that different members prefer that 7

9 different outcomes prevail (Martin 2006, 142). Therefore, inclusive organizations suffer from cooperation problems that are related to the heterogeneity of members capabilities and preferences, and these in turn may affect their efficiency. As noted, aspiring for inclusiveness might depress the overall efficiency of IOs. Different states, with distinctive capabilities and preferences, might not be able, or willing to fully comply with the organization s rules and treaties, if these are adjusted to correspond mainly with the capabilities and preferences of other, more capable members. Mirroring that, more capable members may not be willing to invest more in order to compensate for the inability or unwillingness of laggards to fully comply, or they will design delegated monitoring and/or dispute settlement mechanisms (see: Koremenos 2016, 6). Indeed, compliance problems are caused by the capacity limitations of states and uncontrollable social and economic changes (Downs, et al. 1996, 392). Some have argued that this problem could be dealt with by devoting resources to surveillance and sanctioning of free riders, or by limiting the number of actors for specific issues, thus avoiding cooperation problems of numerous participants (Martin 1992, 773). This is directly related to the breadth-depth trade-off discussion. It is commonly argued that multilateral agreements that are more inclusive will necessarily be shallower in their level of cooperation (Gilligan 2004, 459), and when adding less capable states to an organization, it necessities reducing the level of cooperation of other members (Ibid). In some cases, this is similar to what Thompson and Verdier (2014, 15) call the member surplus, where incentives to comply are calculated to elicit the participation of the state that is burdened with the highest cost of compliance, thereby creating rents for the other members. Therefore, by excluding states that are not committed to cooperation, the level of cooperation by current members can be deeper (Gilligan 2004, 460). However, it is argued that the breadth-versus-depth dilemma disappears if IOs suspend the requirement of uniformity of obligations among their members (Thompson and Verdier 2014, 16). Thus, a suggested solution would be the customization of members obligations within a multilateral framework, in order to create a hybrid which allows countries to choose their level of cooperation, resulting in a framework which is both broad and deep (Ibid, 15-16). Hierarchical categorization is a rational solution for the member surplus problem, as its outcome is exactly a customized agreement made for the needs of different groups of members, by defining differentiated compliance spectrums. This mechanism lowers the costs of continuous 8

10 cooperation by members, as they are allowed to comply according to their current capabilities and preferences, while preserving the material benefits the organization has to offer for its participants. Categorization is meant to decompose conflictual issues that stem from members heterogeneity, albeit at the cost of somewhat lowering the overall efficiency of IOs. For example, the principle of common but differentiated responsibilities, which has been implemented with the goal of differentiating between developing and developed countries within the UNFCCC, was meant to decompose the conflictual issue of historical environmental responsibilities and current burden sharing (UNFCCC 1992). Nevertheless, categorization functions as a mechanism that allows a rather efficient equilibrium between inclusiveness and depth, because incorporating such mechanisms is less wasteful in transaction costs, due to the fact that only a few different agreements are negotiated for each category and not for each and every member. Therefore, incorporating categorization mechanisms allows for the creation of an inclusiveness-depth equilibrium, while lowering the incentives for members to defect. Nominal categorization, however, is also related to heterogeneity. Inclusive and global IOs might have to categorize their members in regional groupings so that each group would be granted with accommodating representatives within the IO s Board of Governors/Directors. This type of nominal categorization allows for a broad representation of all members, without taking into account their material capabilities or preferences. In sum, states are heterogeneous in terms of their capabilities and national interests, and these vary across organizations that deal with diverse issues and encompass different cooperation problems. Therefore, states might decline to sign and/or ratify multilateral agreements which they will not be able, or want to fully implement, unless they are willing to incur the consequences of their non-compliance in order to maintain their membership (except for cases where punishment for non-compliance is exclusion). This is while other states might have to contribute more in order to compensate for the non-compliance of laggards. Indeed, the more diverse the preferences of states, the less likely they are to agree on a common policy. And, as the distribution of power diverge, it is less likely that states will agree to comply by the same institutional rules (see: Martin 2006, 23). Categorization is intended to attract the most relevant international actors by ameliorating cooperation problems that are amplified by members heterogeneity. Categorization is also intended to regulate members cooperation, by allowing them to comply in accordance with 9

11 their capabilities and preferences, making IOs more sustainable and politically viable, while maintaining an efficient and flexible institutional environment. Categorization as a Rational Solution for Cooperation Problems As a debatable working assumption, I rely on Koremenos (2016, 2), who argues that states tailor international law to their cooperation problems and its design is consistent with the goal of effectiveness in the face of harsh political realities. Successful compliance can only be achieved if the design of an agreement corresponds to its underlying cooperation problems and its members characteristics. And, without suitable design provisions, states will often not even ratify the agreement, regardless of how heavily involved they were in the discourse leading to it (Ibid, 13). Indeed, when establishing new IOs, it is imperative to ameliorate cooperation problems that might impede effectiveness. Mitchell and Keilbach (2001, 891) argue that in order to solve cooperation problems within multilateral organizations, actors produce various institutional mechanisms that aim to deter defection without deterring participation. Categorization is one of those mechanisms. Although it might generate inequalities between members compliance spectrums, categorization is a rational solution for cooperation problems that are apparent in IOs that deal with issues that incorporate high levels of uncertainty and/or require deep cooperation. Uncertainty in IOs with Heterogeneous Membership H1: IOs that deal with issues which encompass high uncertainty, and have heterogeneous members, are more likely to incorporate hierarchical categorization mechanisms, compared to IOs with low uncertainty and/or a rather homogenous membership. Table 1: Interaction between uncertainty and members heterogeneity Uncertainty Low High Heterogeneity Low No categorization No categorization High Nominal / Hierarchical categorization no categorization 10

12 Because categorization is treated as a flexibility provision, it is logical to follow Koremenos et al. (2001) conjecture F1 of flexibility increases with uncertainty. The type of uncertainty addressed here is uncertainty about the state of the world, which refers to uncertainty regarding the consequences of cooperation. This kind of uncertainty obscures the true and current distribution of benefits, which can only be known in the future (Koremenos 2016, 39). Furthermore, participating in IOs with high uncertainty about the (future) state of the world may result in unanticipated costs or advance consequences, and states usually know that a given agreement yields substantial benefits in the current state of the world but could impose substantial losses in other possible future states of the world (Ibid, 46-47). IOs that deal with issues which are characterized by high uncertainty raise the costs of participation by heterogeneous members, thus generating cooperation problems that could be dealt with by incorporating categorization mechanisms. In certain issues it is difficult to determine member-states' capabilities and willingness to successfully implement the organization's decisions, and to determine the collective outcome, i.e. gains and losses from multilateral cooperation. Therefore, uncertainty costs multiply contracting costs for new members (Hawkins and Jacoby 2006, 203). As a result, not all states would agree to the same degree of treaty implementation and intervention by the IO. Indeed, uncertainty about whether cooperation will be beneficial in all possible future conditions, might make states forego current cooperation and the long-term gains it could bring (Koremenos 2016, 4). For example, Rosendorff and Milner (2001, 844) argue that it is widely believed that trade is an area where governments face domestic uncertainty that has significant costs: price or supply shocks, technological changes, and/or foreign government policy changes. Other issues, such as climate change, are considered as new and their global effects and ways to tackle them are largely contested subjects. The scientific complexity of issues such as climate change could hamper cooperation by the pervasive uncertainty behind their causes and effects, and the distributive implications of choosing different approaches to addressing them (Thompson and Verdier 2014, 22; see also: Depledge 2009). Under conditions of high uncertainty, participating states are more likely to embrace a behavior which Mitchell (2001, 231) describes as contingent compliance. States will comply once they are assured that enough additional 11

13 members will comply and that doing so themselves will not put them at a significant material disadvantage, thus promoting the creation of a shallower organization and impeding its efficiency. However, states can design IOs to cope with cooperation problems caused by uncertainty. Kucik and Reinhardt (2008, 481) argue that when states enter into a multilateral agreement under conditions of uncertainty, they should aspire to create an agreement with formal flexibility provisions which should enjoy greater, more sustained levels of overall compliance. Best (2012, ) argues that IOs can deal with uncertain policy environment by drafting more ambiguous policies, which leave room for interpretation. This is while Koremenos (2005, 549) suggests that IOs incorporate specific flexibility provisions that serve as an "international insurance scheme", which is a limited duration agreement that can be renegotiated. A similar solution, suggested by Thompson (2010, 272), is that states, uncertain over their future economic, technological and domestic political circumstances, accommodate this uncertainty with institutional designs that allow the parties to adjust the agreement as prevailing circumstances change, or that allow individual members to temporarily adjust their behavior as their own circumstances change. I argue that actors incorporate categorization mechanisms in IOs, in part so that they can deal with these uncertainties in advance, and lower the costs of changing or modifying the organization's agreements in ways that could impair the final desired outcome. By incorporating flexible hierarchical categorization mechanisms, each member can, at first, cooperate within its designated category, and later on, as international and domestic realties change, move to other categories, which could better accommodate its capabilities and preferences. This is possible as categorization mechanisms keep the distribution of costs relatively constant for each category-group. Applying hierarchic categorization mechanisms, in this sense, function as a collective safeguard for a group of participating members, similar to escape clauses, with the goal of ameliorating cooperation problems that are generated due to uncertainties about the state of the world. Safeguards are provisions of an international agreement that allow countries to suspend the concessions they previously negotiated without violating or abrogating the terms of the agreement, albeit at varying costs imposed on countries that decide to use them (Rosendorff and Milner 2001, ). Categorization mechanisms and escape clauses are similar in several ways. Both categorization and escape clauses can be considered as rational responses of states to their institutional environment; both function as flexibility provisions; both are argued to ameliorate 12

14 uncertainty concerns; and both should make initial agreements easier to reach. Nevertheless, categorization mechanisms are different from escape clauses because they are not applied as adhoc devices for escaping obligations. Categorization, as opposed to escape clauses, is intended to function as a permanent provision for members that are currently unable or unwilling to abide by the same rules and obligations as other members. While escape clauses may be used by members as they see fit, categorization attaches different obligations to members, according to their category-groups. And while in many cases members can migrate between categories, they can t do it often and on a unilateral basis. Escape clauses and categorization mechanisms, however, are not mutually exclusive, and IOs can incorporate both. Depth of Cooperation (Institutionalization) H2: IOs with heterogeneous membership which require deep cooperation (and thus are more institutionalized) are more likely to incorporate categorization mechanisms, as opposed to IOs with rather homogenous membership and/or IOs that require shallow cooperation. Table 2: Interactions between heterogeneity, depth of cooperation and categorization Depth of Cooperation Low High Heterogeneity Low No categorization No categorization High Nominal / Hierarchical categorization no categorization Complying with the rules of certain IOs may require significant domestic (re)adjustments and broad domestic political backing by member-states. Such adjustments are more prevalent in IOs which are characterized by collaboration games situations in which equilibrium outcomes are collectively suboptimal. Thus, these are more likely to generate collaboration problems - they contain strong incentives to defect from established cooperative patterns of behavior, since defection might result in immediate payoffs, and therefore require delegated mechanisms that can focus on maintenance of agreements (Martin 1992, ). Resolving the dilemma of 13

15 collaboration problems, as Martin (Ibid, 769) argues, is a matter of mutual policy adjustment, since all players must agree to move away from the suboptimal equilibrium, thus rejecting their dominant strategy and agreeing to deepen the signed agreements. This is while in other cases, for example, IOs that concern telecommunications and aviation, usually only require some degree of coordination and standardization by members (see, for example: Krasner 1991). These generate coordination problems that do not require strong delegated maintenance mechanisms, since no state would gain by deviating from the established outcome, once an equilibrium has been established, and each need to devote little attention to the prevention of cheating (Martin 1992, ). Hence, I argue that IOs with heterogeneous membership that deal with issues which are characterized by collaboration problems, and thus require deep cooperation in order to achieve their goal, are more likely to incorporate categorization mechanisms. These mechanisms allow for states with different capabilities and preferences to initially join the organization and be classified accordingly. Following that, categorization mechanisms assist in maintaining cooperation within the institutional framework by lowering the incentives of less capable or willing states to defect. Under a categorized institutional environment, the continuation of cooperative behavior would become more worthwhile, since payoffs from defection could not match the benefits that are given to members of suitable categories. The element of heterogeneity is significant due to the fact that depth is only relevant to members with high compliance costs. Therefore, depth calls for exclusion and narrowness (Thompson and Verdier 2014, 18). However, many issues that are dealt with by IOs require both inclusiveness and certain degrees of depth. Consequently, the incorporation of categorization mechanisms within these IOs is a rational solution for cooperation problems that arise due to the requirement of high degrees of depth. Depth of cooperation varies within IOs. Downs et al. (1996, 383) define the depth of cooperation as the extent to which an agreement requires behind-the-border integration with regard to the issue at hand. In other words, an agreement's depth of cooperation is the extent to which it requires states to depart from what they would have done in its absence, and how it would actually alter the current status-quo. Some states, for example, might have greater difficulties adjusting their domestic market in accordance with the WTO's trade regulations, while other states might find it more difficult to tackle climate change effects. Some IOs deal with issues that require 14

16 maximum or broad cooperation by members in order to achieve a common goal, while other IOs deal with issues which may require that several actors invest more than others. Therefore, IOs that have diverse and heterogeneous membership might encounter cooperation problems due to the inability or unwillingness of some members to make the required adjustments, a situation which may lead to shallower cooperation or defections. Depth of cooperation corresponds with the organization s degree of institutionalization (see: Haftel 2006). Certain organizations might only possess few rules on procedures, such as voting, that are nonbinding or do not compel states to take action, and their ability to enforce agreements and implement decisions is minimal and optional for member-states. These organizations are considered as having shallow cooperation, since they do not require states to agree to binding agreements which would compel them to depart from current policies. This is while other organizations, which are characterized by having deep cooperation, possess clear mechanisms for coercing or influencing state behavior. In addition, highly institutionalized organizations have the means to observe and collect information on their member-states. These organizations typically possess many of the institutional characteristics regarding extensive codification of rules and procedures between and within the executive, legislative, and judicial structures (see: Boehmer et al. 2004, 37-38). Therefore, deeper cooperation requires a more institutionalized organization - creating cooperation problems in regards to achieving maximum compliance by all relevant actors. Many states would not agree to join and participate in such institutionalized and invasive IOs and lawfully abide by their agreements. Consequently, due to their binding and largely enforcing character, institutionalized organizations are more likely to apply categorization mechanisms in order to attract key states which would not have joined otherwise. The CIO Dataset and Research Design Analyzing the possible causes for the incorporation of categorization within IOs requires a systematic collection of information on the processes of IOs design. For this purpose, this paper relies on an original data-set, named CIO ( Categorization in International Organizations ), composing design information on 162 IOs, established between 1875 and First, the dependent variable - whether IOs incorporate a categorization mechanism, and its type (hierarchical, nominal, or both), was coded by exploring the IOs official websites, as well as 15

17 examining their Articles of Agreement and the Rules of Procedure for their Board of Governors / Directors, alongside documents related to certain amendments made to the agreements through the years. This is a binary variable, meaning that whenever an IO incorporated any type of a categorization mechanism it receives the value 1, and otherwise 0. Furthermore, the type of categorization was coded in an additional item and received 1 for nominal, 2 for hierarchical, and 3 for both. The type of categorization was determined by examining the actual consequences of the categorization process. This means that whenever an IO classified its members according to various material indexes, and members consequently were grouped in different categories that differentiated their rights and obligations towards the IO, or granted them with more or less control over decision-making processes and outcomes (for example, by granting veto power), then this mechanism was considered as having hierarchical consequences. In other cases, whenever the IO classified its members according to rigid criteria, and this categorization could only affect elections to the IO s various governing bodies, then it was considered as nominal. Among 162 IOs, 49 incorporated either nominal or hierarchical categorization mechanisms, while 113 did not. Within the categorized IOs, 23 had nominal categorization mechanisms, 19 had hierarchical, and 7 had both (see: tables 3 and 4). Table 3: Variation between categorized and non-categorized IOs Categorization Mechanisms within IOs IOs that employ categorization mechanisms, (49) IOs that do not employ categorization mechansism, (113) Table 4: Variation within categorized IOs 16

18 Type of Categorization Mechanisms 7, 14% Both Hierarchical and Nominal categorization 23, 47% Nominal Categorization 19, 39% Hierarchical Categorization Several examples illustrate how prominent IOs hierarchically categorize their members, apply differentiated rights and obligations, and include flexibility provisions within their categorization mechanisms. First, the WTO employs a hierarchic categorization system which is composed of two sets of parallel mechanisms: Initially, members who join the WTO declare for themselves whether they want to be classified as developed, developing or as least developed countries (LDCs). Developing countries and LDCs are granted with longer transition periods and technical assistance. Furthermore, the WTO allows its member-states to classify other members as market or non-market economies. Consequently, members that are classified under the latter will have difficulties in regulating the exports of goods under the organization's framework, due to its anti-dumping measures (see: WTO 2017a; 2017b). The WTO takes into consideration that the position of its members within these development indexes is ever-evolving and is prone to change over time, thus incorporating a flexibility provision within its categories which allows members to move between categories. A Second example is the UNFCCC. Up until the Paris Agreement in 2015, the UNFCCC employed a hierarchical categorization system to classify members into three main groups based on their economic development status and technological capabilities to tackle climate change, as Annex I, Annex II, and non-annex I Parties. Annex I Parties included the industrialized countries that were members of the OECD in 1992, including countries with economies in transition (the EIT Parties). Annex II Parties consisted of the OECD members of Annex I, excluding the EIT Parties. And, Non-Annex I Parties were mostly developing countries and LDCs (UNFCCC 2014). 17

19 In terms of rights and obligations, members of the Annex I category-group were expected, under the Kyoto Protocol, to reduce their overall emissions by an average of 5% below 1990 levels during the first commitment period ( ), and were obliged to publically communicate their reduction measures and results to the Convention on a regular, non-voluntary basis. Members of Annex II category-group, however, in addition to their emission reduction requirements, were required to provide financial resources to enable developing countries to undertake emission reduction activities under the Convention and to help them adapt to adverse effects of climate change by assisting in environmental mitigation and in putting adaptation measures into effect. Also, they had to take all practicable steps to promote the development and transfer of environmentally friendly technologies to EIT Parties and developing countries. This is while non- Annex category-members were only expected to actively participate in the Convention's Conferences of Parties (COPs) and cooperate with other members in implementing adaptation and mitigation measures (UNFCCC 2014). The UNFCCC includes flexibility provisions within its categorization mechanisms, as illustrated by the Turkish case. And indeed, under certain circumstances, members of the UNFCCC can submit a proposal to change their category. However, this decision is subjected to a formal review by a special committee, which decides whether this proposal should go for majority vote. The UNFCCC s flexibility provisions were also illustrated by the outcome of the Paris Agreement in 2015, as it was decided to abolish current Annex and non-annex categories and employ a mechanism which would differentiate members in a more dynamic manner (see: UNFCCC 2015; also see: Ella 2017). In the security area, the UN Security Council (UNSC) categorizes its members as permanent and non-permanent members. Permanent members of the UNSC consist of the victors World War II, and were also chosen according to their contemporaneous status as "great powers". However, this classification ignored material realities, as can be seen from France s classification as a great power although at that time it did not have the corresponding capabilities or future potential (see: Heimann 2015). In terms of institutional rights, permanent members of the UNSC are granted veto power, and thus are awarded with more control over decision making processes (United Nations 1945). Similarly, members which are categorized as nuclear-weapon states within the Non-Proliferation Treaty (NPT) receive veto rights when it comes to modifying the treaty, since only a majority that includes all nuclear-weapon states is required to approve amendments (IAEA 1970). 18

20 The AIIB is a very instructive example of an IO that incorporates nominal categorization mechanisms, which have hierarchical consequences. Firstly, the AIIB classifies its members according to their geographic location as Asian or non-asian countries, along with classifying them as founding or non-founding members - depending on whether these states joined the organization during a specific time-window. Although its categories are permanent in nature, their hierarchical consequences are manifested by the fact that members that are categorized as founding members are allocated with additional 600 votes (AIIB 2015, 2-3). Furthermore, 75% of the Bank s total shares are reserved for regional members and 25% for non-regional members (Ibid, 29-31), thus allowing regional members more control over the Bank s decision making and resource allocation. Indeed, certain IOs employ both nominal and hierarchical categorization mechanisms. The Asian Development Bank (ADB), for example, firstly classifies its members as Asian or non-asian states. This categorization, however, does not affect voting power directly but only affects the election of states' representatives for the organization's Board of Directors. Additionally, the ADB applies a hierarchical categorization mechanism, which classifies its members according to their economic development levels as developed or developing countries, while non-asian members are most, if not all, developed countries. The ADB classifies and clusters its developing countries members into 3 groups, and this classification is based on two indexes: 1. The World Bank's gross national income (GNI) per capita estimates, and 2. Creditworthiness, assessed by an internal committee of the organization. This categorization is done for the purpose of assessing the eligibility of certain members to get loans from the Bank, and later on, successfully go through the process of graduation (ADB 2013, 1-3). However, most regional and exclusive organizations choose not to include categorization mechanisms. The most notable example is the European Union (EU) and its related bodies. Due to its exclusivity which allows only European members to join and participate, the need for nominal categorization based on geographic location is unnecessary. Also, due to relative homogeneity of European states in terms of preferences and to a lesser extent, their capabilities homogeneity (and the consent of more capable members, such as Germany, to incur some of the costs of including less capable members within the Union), the incorporation of hierarchical categorization mechanisms seems redundant. 19

21 Ideal Point score In order to determine the heterogeneity of members preferences in every IO, I relied on Baily, Strezhnev and Voeten (2015), who calculated ideal-points for country-years, based on the UNGA voting records. This was calculated for each country individually (and not as a part of a dyad). Nevertheless, as Koremenos (2016, 115) points out, this measure is calculated the same for all issue-areas, although states might differ in their preferences in regards to different issues. In the CIO dataset, for each IO, the standard deviation of all member-states ideal-points was calculated in the year of the IO s establishment. Calculating ideal-points standard deviation illustrates how far members preferences are from each other, meaning that the higher the score for any specific IO, the more heterogeneous are its members. Third, to determine the heterogeneity of members capabilities in the IOs examined, I relied on the Correlates of War CINC index (Pevehouse et al. 2004; see also: Singer 1987). This index contains annual values for total population, urban population, iron and steel production, energy consumption, military personnel, and military expenditure of all state members (COW 2017). Similarly to preference heterogeneity, for each IO, the standard deviation of member-states CINC scores were calculated in the year of the IO s establishment - the higher the score, the more heterogeneous are the IO s members (see: table 3 and 4). Table 3: Preference Heterogeneity within Categorized and non-categorized IOs Preference Heterogeneity IOs 20