Commerce and Institutions: Trade, Scope, and the Design of Regional Economic Organizations

Transcription

1 Commerce and Institutions: Trade, Scope, and the Design of Regional Economic Organizations Yoram Z. Haftel Department of Political Science University of Illinois at Chicago Paper prepared for the Workshop on the Politics of Trade Agreements: Theory, Measurement, and Empirical Applications. April 30-May 1, 2010, Princeton University. 1

2 Abstract: The design of current regional economic organizations (REOs) is remarkably diverse. Some REOs address numerous economic issues, while others have only limited mandates. Some REOs have an independent bureaucracy and a legalized dispute settlement mechanism (DSM), while others do not. What determines this institutional variation? Neofunctionalists maintain that commercial ties and the proposed level of economic cooperation are important drivers of regional institutionalization, institutional independence and legalization. A number of recent studies question this logic and argue that it is naïve. Empirical evidence on the links between commerce, economic scope, and regional institutions is scant, however. Using an original data set that contains detailed information on the activities and institutional structure of twenty-five REOs over two decades, this paper presents one of the first systematic analyses of these relationships. The empirical analysis indicates that the neofunctionalist wisdom is right after all. Controlling for several alternative explanations, I find that higher levels of regional trade are associated with greater REO institutionalization, scope, and more independent bureaucracies. In addition, greater economic scope is associated with more independent bureaucracies and more legalized DSMs. 2

3 The ever-growing number and the increasing political and economic significance of regional economic organizations (REOs) is one of the hallmarks of the current global economy. A glance over the landscape of these institutions suggests that their structure and functions are very diverse. Some organizations, such as the EU and ASEAN, address a variety of functional issues, while others, such as SAARC and ECO have only a limited scope. 1 Some REOs, such as ANCOM and WAEMU have an independent bureaucracy, but others, such as IOC and NAFTA, have weak or no corporate bureaucracy. Some REOs, such as the EU and COMESA, have a highly legalized dispute settlement mechanism (DSM), while others, such as SACU and LAIA, do not. What determines this institutional variation? The level of economic integration is perhaps the most cited explanation for the existence and institutionalization of these organizations. Most notably, neofunctionalists, who observed the process of European integration, maintained that increasing crossborder economic interactions result in growing societal demands for international regulation of such exchange (Caporaso 1998; Haas 1966). In particular, they argue that when the potential for economic gains from regional integration is significant market players will have an incentive to lobby for regional institutional arrangements that render the realization of these gains possible. The demand for regional rules, regulations, and policies by market players is a critical driving force of integration (Mattli 1999, 42). Consequently, governments in interdependent regions will recognize the necessity to manage their interactions through international institutions and will form regional organizations to fill this function. Logically, one expects that higher levels of interdependence should result in greater institutionalization of these regional arrangements. Several studies pay greater attention to specific design features and argue that high levels of interdependence lead to more independent organizations. As economic exchange increases and becomes more complex, societal actors demand such international governance in order to reduce transaction costs and lower risks of opportunism, uncertainty, and non-compliance (Abbott and Snidal 2000; Stone Sweet and Sandholtz 1998; Yarbrough and Yarbrough 1992, 86-88). Such independence is often embodied in more powerful regional secretariats or commissions. As these organs become more autonomous and have more resources, they are better equipped to monitor the behavior of member-states and provide neutral information on the degree of compliance with existing agreements (Abbott and Snidal 1998, 20; Haftel and Thompson 2006). In addition, more extensive economic interaction is likely to intensify disagreements over the distribution of costs and benefits of this exchange, and in turn strengthen the need for a disinterested body to address distributional concerns. Thus, greater economic interdependence is expected to result in a more independent corporate bureaucracy. Along similar lines, extant research suggests that higher levels of economic integration lead to the creation and legalization of regional DSMs. These adjudicating bodies are especially instrumental when members engage in heavy cross-border activity on a variety of economic issues. Moreover, greater economic scope is likely to increase 1 See the Appendix for the full name of these organizations. 3

4 the number of rules and their complexity. Under such circumstances, potential conflicts or disagreements on how to interpret the rules are more likely to erupt (Johns and Rosendorff 2010; Smith 2000; Stinnett 2007; Stone Sweet and Sandholtz 1998; Yarbrough and Yarbrough 1992; 1997). Governments, firms, and other social groups that contribute to and are affected by regional interdependence will therefore demand a neutral arbitration body to reduce the risk of doing business and to sort out unanticipated challenges that may arise from the integration process. Hence, greater economic interdependence is expected to result in a more legalized DSM. Despite the intuitive appeal of the purported positive effect of economic interdependence on regional institutions, this logic was challenged from a number of perspectives. In the 1970s, the neofunctionalist research program was heavily criticized, even by its own proponents, due to its inability to explain the institutional retreat of its most celebrated case, namely the European Community (Haas 1975). More recent studies argue that this functional approach fails to account for institutional differences across existing REOs. Miles Kahler, who labeled this approach naïve institutionalism contends that regional institution building demonstrates a cyclical pattern that bellies any simple explanation based on the demands of economic integration (1995, 80). Joseph Grieco finds that contrary to functionalist expectations, in some areas of the world there have been increases in intraregional trade without a corresponding increase in institutionalization Moreover, in some cases we can observe growth of institutionalization, in the absence of increased trade encapsulation (1997, 172). In the same vein, Acharya and Johnston (2007, 257) find little evidence of the salience of functional considerations [in the design of regional institutions]. These critical evaluations suffer from the same limitation they themselves underscore with respect to neofunctionalism. Just as the latter relied on the European experience to make sweeping generalizations, so did the former reach its conclusions based on a small number of case studies and without clear metrics of institutional design. This paper sheds new light on this debate. Using an original data set that contains detailed information on the activities and institutional structure of numerous REOs in the 1980s and 1990s, it presents one of the first systematic analyses of these relationships. The empirical results largely vindicate that the neofunctionalist logic. Accounting for several alternative explanations, I find that higher levels of economic integration are associated with greater regional institutionalization and scope, more independent bureaucracies, and more legalized DSMs. The next section elaborates on the nature of REOs, presents the measures used to evaluate institutional differences across these organizations, and describes this variation. The third section discusses matters of research design, the fourth section reports the results of the statistical analysis, and the final section concludes. Conceptualization and Measurement of REOs A glance at the landscape of regional economic organizations reveals a great deal of variation in the objectives and structures of these institutions. This section takes a closer look at the functions performed by and the institutional features contained in a large number of regional economic organizations. It, first, describes the type of institutions 4

5 considered in this study. It then discusses the design features purported to be affected by economic interdependence. It provides a definition for each organizational characteristic and justifies it, links this feature to specific quantifiable indicators, and elaborates on the manner by which their implementation is assessed. Next, it presents an original data set of twenty-five REOs coded according to these guidelines in the 1980s and the 1990s. This description underscores the significant institutional variation across the organizations included in the data set. The Appendix provides a list of these organizations and a table that summarizes the indicators included in the variables related to regional institutionalization. 2 What Are Regional Economic Organizations REOs are a class of international governmental organizations (IGOs). Unlike other international agreements, IGOs have a continuous institutional framework and some kind of formal structure. In addition, the membership of IGOs is restricted to nation-states and they include at least three such members (Pevehouse, Nordstrom, and Warnke 2004). Two aspects distinguish REOs from other IGOs. First, while they sometimes address non-economic issues, the primary objective of these organizations is the promotion of economic policy cooperation among their members (Mansfield and Milner 1999). Even though some REOs have security-related components, the Economic Community of Western African States (ECOWAS) and Mercosur for example, they are not alliances. Despite the involvement of many of these communities in social and political issues, their core mandate is economic. Second, membership in these organizations is restricted to geographically proximate states. While some REOs span sizable swaths of territory, like the Latin American Integration Association (LAIA), they are all regional rather than global in nature (Nye 1971). This conceptualization excludes several types of agreements that are related to economic regionalism but do not qualify as REOs. These exclusions are consistent with the conventional practice. 3 Many of the preferential trade agreements (PTAs) that were concluded in recent decades are bilateral and almost always lack a continuous institutional framework. These agreements, as well as similar agreements between two REOs or an REO and another country, do not meet the standard criteria of international organizations. Non-reciprocal agreements, such as the U.S. Caribbean Basin Initiative (CBI) and the Lomé Convention, are excluded on similar grounds. Finally, framework agreements, like the Asia Pacific Economic Cooperation (APEC), are also excluded. While these agreements may embrace the idea of regional cooperation, they lack concrete measures to achieve this goal. Twenty-five REOs that span most continents and include the majority of the states world-wide correspond to these criteria. To facilitate the systematic comparison of these organizations, I now turn to a detailed consideration of their institutional components. 2 For more detail on these organizations and indicators, see Haftel (2010), Chapter 3. 3 See Page 2000a; Pevehouse, Nordstrom, and Warnke 2004; Mansfield and Pevehouse 2000; Smith 2000; and UIA. 5

6 The Scope of Economic Activity Scope is widely recognized as an important dimension of institutional design (Koremenos, Lipson, and Snidal 2001) and regional integration (Acharya and Johnston 2007; Grieco 1997; Hicks and Kim 2009; Kahler 1995). It commonly refers to the range of issues incorporated in a given organization in order to meet its objectives. Scope therefore differs from and is subsumed by mandate, which refers to the broad purpose of the organization (Acharya and Johnston 2007, Boehmer, Gartzke, and Nordstrom 2004). Restricting the analysis to organizations with economic mandate, one can distinguish between those that address only a handful and those that touch on a variety of issues. A systematic evaluation of the scope of economic activity requires an identification and differentiation of the issues that REOs can potentially include. This task is complicated by the ambiguity of the boundaries between different subject-matters (Koremenos, Lipson, and Snidal 2001, 771). In spite of their broadly similar mandate, issue differentiation across REOs is still a significant obstacle. For example, are fiscal and monetary coordination two issues or one? Should one conceptualize trade liberalization according to the number of sectors covered in the agreement or the kind of liberalization (tariffs, quotas, non-tariff barriers)? A number of studies that conduct an in-depth qualitative assessment of the implications of economic scope for cooperation and integration in different regions dodge these issues (Acharya and Johnston 2007; Grieco 1997; Kahler 1995; Katzenstein 1997). They provide a number of insights into the significance of this institutional feature, but define this concept very broadly and fail to provide guidance on how to compare economic scope across time and space in a rigorous manner. As a result, evaluation of different REOs on this aspect appears impressionistic and lacks clear metrics. A more systematic method builds on the traditional distinction between different types of integration schemes. First developed by Bela Balassa (1961), this approach classifies REOs according to their commitment to economic integration. The four main categories, from the most limited to the broadest in scope, are free trade area, customs union, common market, and economic union. Consistent with the development of economic integration in Europe, on which this classification is founded, these categories are also conventionally used to describe the process of economic integration. In this view, REOs broaden and deepen their scope as they gradually advance from one scheme to the next. This classification is still widely used in economics (Frankel 1997; OECD 1993) and political science (Bearce and Omori 2004; Hicks and Kim 2009; Laursen 2003; Smith 2000). The clear and simple classification offered by this approach is very appealing. In addition, it captures some of the most important activities REOs usually cover, particularly those related to international trade. Nonetheless, this approach suffers from two important limitations. First, these different categories fail to capture the range of activities that REOs engage in, even in the supposedly exemplary case of European integration (Nye 1971, 28-30). Indeed, many REOs deal with a number of important issues such as foreign investment, development, and industrial cooperation, which do not fit in these traditional categories. Second, both theory and practice show that there are multiple paths to economic integration that diverge from this simple evolutionary model (Choi and Caporaso 2002, 483; Foroutan 1993; Nye 1971, 29; Page 2000a, 8). The CFA Franc 6

7 Zones in Sub-Saharan Africa, for example, have established a monetary union but not a customs union, and the GCC made considerable progress on free movement of labor and capital but not on a common external tariff. Third, this classification is formal and ignores the degree of implementation of these arrangements. The conceptualization of economic scope adopted in this study begins with the traditional approach, but adds a number of modifications to address the shortcomings outlined above. It identifies nine broad areas of economic cooperation and integration commonly tackled by REOs. The first six categories are largely compatible with the traditional categories. These are free movement of goods, customs union, free movement of services, free movement of capital and investment, free movement of labor, and monetary and fiscal cooperation. The remaining three areas include sectoral cooperation and harmonization, economic development, and efforts to enhance collective bargaining power. These are important issues that correspond to key goals of past and present REOs (Page 2000a). Each area consists of two to four specific indicators that capture the breadth and depth of cooperation on this issue. Taken together, these twenty-four indicators are general enough to travel across different regions and detailed enough not to miss essential economic issues that REOs typically address. Each and every indicator is measured along two dimensions. The first labeled DESIGNED SCOPE refers to approved agreements that specify the mandate of the organization. This is a necessary and important dimension of any institution (Acharya and Johnston 2007; Hicks and Kim 2009; Koremenos, Lipson, and Snidal 2001). Each indicator can obtain a value of 1, if present, and 0 if absent. Thus, the designed scope of economic activity can range from 0 to 24. This approach does not assume a gradual process of integration, but instead, allows any indicator to score a point independent of other indicators. In addition, consistent with conventional practice, all the indicators are weighted equally (Fortna 2004). Implementation The second dimension of economic scope is implementation, which refers to the actual steps that member-states take to realize the agreements they have reached. In most instances it involves the process by which governments transfer their international obligations to domestic laws, regulations, and policies (Victor, Raustiala, and Skolnikoff 1998, 4). 4 It captures the notion that international cooperation involves not only agreements and rules but also behavior according to them. As such, it is a key component of international institutions. As Lisa Martin (2000, 18) points out, if agreements are not implemented, and the necessary policies changed, no cooperation has taken place. So it is essential that we consider implementation of international agreements if we are to understand patterns of international cooperation. Notwithstanding the importance of this institutional aspect, it is widely overlooked by scholars of international politics, and measures of this concept are not 4 Implementation is related to but distinct from compliance. The former is usually a key step towards, but it is neither a necessary nor a sufficient condition for the latter (Raustiala and Slaughter 2002, 539). Implementation is also different from effectiveness, which refers to the relative success of an institution in obtaining its stated goals (Acharya and Johnston 2007, 24-25; Raustiala and Slaughter 2002; Victor, Raustiala, and Skolnikoff 1998). 7

8 readily available. The collection of comparable data on implementation is more difficult than on institutional design and is rarely done. As a first cut into this institutional dimension, I created an ordinal scale: zero, one-half, and one. If the REO did not implement the agreement regarding a specific indicator, or if implementation is low, the REO scores zero. If implementation is complete, or nearly complete, the REO scores one on the specific indicator. If implementation is partial, the REO scores half. For example, many REOs reduce barriers to trade and factors of movement in a gradual process. If they have completed only some of these steps, I code the degree of implementation as partial. The multiplication of the two dimensions produces a second variable, labeled IMPLEMENTED SCOPE. Member-states may implement all the planned activities, thus this variable ranges from 0 to 24. If the agreements are not fully implemented, the value on this variable will be lower than the one on institutional design alone. One might imagine, for example, an REO that has an agreement on twenty indicators, that is, DESIGNED SCOPE equals 20. Of these, ten indicators are fully implemented, four are partly implemented, and six remain only on paper. The value on IMPLEMENTED SCOPE is calculated as follows: (10*1) + (4*0.5) + (6*0) = 12. Institutional Independence Institutional independence includes three essential and tightly-linked dimensions: autonomy, neutrality, and delegation (Haftel and Thompson 2006). Autonomy refers to the organization s ability to function in a manner that is insulated from the interests of its members. Thus, greater autonomy reflects a more limited control of states over and a broader range of actions available to the organization (Abbott and Snidal 1998; Hawkins et al. 2006; Thompson 2006). 5 Neutrality reflects the notion that an independent actor is not prejudiced in favor of one party over another. Instead, a neutral actor is guided by professionalism and impartial judgment. As Keohane, Moravcsik and Slaughter (2000, 459) propose with respect to international courts and tribunals, independence measures the extent to which adjudication is rendered impartially with respect to concrete state interests. Even if an organization is autonomous and neutral, its influence depends on the responsibilities and power delegated to its institutions. As institutional properties, independence and delegation are tightly intertwined (Keohane, Moravcsik and Slaughter 2000). International institutions with a significant delegated authority have discretion with respect to agenda setting and dispute management (Abbott and Snidal 2000; Hawkins et al. 2006). The existence and discretion of regional bureaucracies and the existence and legalization of third-party dispute settlement mechanisms capture the three elements of REO independence. They are discussed in turn. Corporate Bureaucracy This administrative body, typically labeled secretariat or commission, is charged with everyday operations of the organization. The personnel is employed by the organization and thus intended to serve regional rather than national interests. Variation in the independence of regional bureaucracy is captured with two indicators that examine 5 Other observers conceptualized this dimension as institutional authority (Grieco 1997, ) and commitment institutions (Mattli 1999, 54). 8

9 whether an REO has a neutral bureaucracy and the extent to which this bureaucracy is delegated meaningful discretion. When states endow organizations with a secretariat or commission, they are delegating to a centralized authority and generating independence in the process. It is unsurprising, then, that governments sometimes resist the establishment of and that not all REOs have a secretariat. The Southern African Customs Union (SACU), for example, was managed through the Finance Department of South Africa until 2004 and the Bangkok Agreement is still coordinated by the Economic and Social Commission for Asia and the Pacific. Thus, one point is assigned to organizations that have a distinct permanent secretariat or its equivalent. When there is a permanent secretariat, the degree of responsibility that the member-states delegate to the organization s bureaucracy is an important sign of independence. While most secretariats perform informational and operational roles, some bureaucracies are provided with additional and much more active responsibilities. 6 In particular, some secretariats and commissions can initiate and recommend policies and thereby promote the goals of the organization, prerogatives that greatly enhance bureaucratic authority. The European Commission represents the ideal-type for secretariats with the power to initiate policies, and the bureaucracies of the Andean Community and WAEMU enjoy similar powers and prestige. Many other organizations, in contrast, have weak secretariats that play a rather passive role in the promotion of regional cooperation. REOs whose bureaucracy can initiate policies or make recommendations score another point. Thus, BUREAUCRACY is an ordinal variable that ranges from two for a highly independent bureaucracy to zero for no permanent regional secretariat. Dispute Settlement Mechanism The prevalence of dispute settlement mechanisms in regional and other organizations reflect the growing legalization of world politics. The imperfect design and the unforeseen consequences of international institutions require an impartial body to adjudicate when disagreements arise. Delegating this power to a centralized mechanism reduces uncertainty and fosters cooperation (Hawkins at al. 2006; Johns and Rosendorff 2010; Yarbrough and Yarbrough 1997, ). Not all DSMs are endowed with the same degree of autonomy and legal authority. Some REOs, like ASEAN and the GCC, adopted a non-binding third-party review process. Others, like COMESA and the CACM, established a standing tribunal that its rulings are binding. In addition, some organizations, such as ECCAS and AMU, have agreed to form a dispute settlement in principle, but have not followed up on their plans. James McCall Smith (2000) evaluates the legalization of different DSMs along five dimensions: the existence of a third-party review, the bindingness of ruling, the existence of a standing tribunal, private standing, and the effect of the ruling. Following Smith, these dimensions are collapsed into three ordered categories: none, low, and high levels of independence. A low level of independence refers to an instance in which thirdparty review is automatic and binding but judges are picked from an ad-hoc roster. This design prevents members from bypassing or ignoring the ruling of the DSM, but give 6 This categorization of functions builds on Jacobson (1984). 9

10 them power over the selection of the arbiter. A high level of independence entails a standing tribunal as well as automaticity and bindingness. With this design, states have little influence over the composition of the tribunal, ensuring its neutrality. I also consider whether the DSM actually exists. 7 Thus, to the extent that REOs have an operational DSM, DISPUTE SETTLEMENT scores one point for each level of legalization for a maximum of two points. Regional Institutionalization The combination of economic scope and high independence, discussed above, as well as regular meetings among high-level officials captures the aggregate level of regional institutionalization. 8 Two overarching variables sum up the degree of functional activity and political authority that states hand over to the organization. The first variable, labeled DESIGNED INSTITUTIONALIZATION, merges DESIGNED SCOPE with the remaining five indicators. It captures the structure and the intended functions of the REO. The second variable, labeled IMPLEMENTED INSTITUTIONALIZATION, takes into account the actual functioning of the organization. It substitutes DESIGNED SCOPE with IMPLEMENTED SCOPE. Both variables range from zero to twenty-nine. Coding and Data With definitions and measurements in hand, I code the twenty-five REOs listed in the Appendix on the institutional variables described above. Regional cooperation is a gradual process, and signing agreements, ratifying them, and implementing them, can be time consuming. Institutional change usually takes, at the very least, several years to accomplish. This reality has two implications for the design of the data set. First, since changes from one year to another can be incremental, I collected information every several years. More precisely, I coded the various REOs in a five-year interval: 1982, 1987, 1992, and 1997, which is conventional in the discipline. Second, because it is difficult to properly evaluate REOs that exist for a short period of time, only organizations that were formed before 1993 are examined. Eighteen out of the twentyfive (about 75%) REOs included in the sample were formed before 1982 and thus have four observations. Of the remaining organizations, four have three observations and three have two observations. When coding the degree of institutionalization of the REOs, I relied on various secondary sources. Several books that survey REOs were used to code the formal agreements. These include the IMF (1994), UNCTAD (1996), the African Development Bank (2000), and Page (2000a). In addition, I used the UIA s Yearbook of International Organizations (various years), the IMF Directory of Economic, Commodity, and Development Organizations website, and some of the REOs own websites. The coding of dispute settlement mechanisms follows Smith (2000). 9 Each coding is a result of crossexamination of these different sources. Similar secondary sources, as well as newspaper 7 This is different from utilization. DSMs may be operational but underutilized. The design of DSMs may affect their utilization and effectiveness (Haftel 2010; Kono 2007; Posner and Yoo 2005). 8 On the role of regular meeting among top-level officials, see Haftel (2007, 2010). 9 I coded cases that are excluded from Smith s study according to his own criteria. 10

11 articles and independent reports, were employed to determine the degree of implementation. In addition, in-depth case studies of specific REOs were surveyed. In order to increase the reliability of the coding and to reduce the risk of measurement errors, several sources were compared before the value of each indicator was determined. Haftel (2004) provides a list of these sources. The Landscape of Regional Economic Organizations This section builds on the data collection effort thus far described to assess the variation in the institutionalization and design of REOs included in the sample. It first examines general trends across regions and over time. It then considers specific design features and the relationships between them. The analysis highlights the substantial institutional variation across existing organizations and reinforces the call for a nuanced treatment of their structure and functions. It also indicates that the measures constructed for this study are compatible with extant research, providing them with face validity. The appendix presents descriptive statistics for all the variables. Regional Institutionalization To gauge the scale of regional institutionalization worldwide, each of the two overarching variables are divided into three groups along a continuum of institutionalization: low, medium, and high. Figure 3.1 reports the number of organizations that fall in each of these categories. 10 It substantiates the widespread intuition, shared by this study, that all REOs are not alike (Choi and Caporaso 2002; Mansfield and Milner 1999; Page 2000a). Many organizations are weakly institutionalized, but others exhibit intermediate or high levels of institutionalization. It also underscores the gap between institutional design and implementation. Considering only institutional design, it appears that a large number of REOs signed ambitious agreements with lofty goals. About seventy percent of all organizations fall in either the intermediate or high category. This trend is turned on its head when taking into account implementation, however. About seventy percent of the REOs fall in the low category and only one (the EU) falls in the high category. This observation underscores the reality that many organizations, especially those among developing countries, fall short of their stated objectives (de la Torre and Kelly 1992; Langhammer and Hiemenz 1990; OECD 1993, 65). 11 [Figure 1] An overview of the evolution of REOs throughout the 1980s and the 1990s indicates that economic regionalism is a dynamic phenomenon. This is apparent from the sheer number of organizations included in the sample in each time point, which increases from eighteen in 1982 to twenty-five in 1992, and the increasing level of institutionalization over this time-period. Figure 2 depicts the annual average of 10 For each REO, the value represents the average across the different time points. Even though the level of institutionalization changes over time, most REOs remain in the same category. 11 This gap reflects, in part, failed attempts by developing countries to import European agreements and institutions to their own regions. Langhammer and Hiemenz (1990, 2) label this as the fallacy of transposition. 11

12 institutionalization for the eighteen REOs that have all four observations, 12 and demonstrates that its average level has substantially increased over this fifteen-year interval. This is evident for both designed and implemented institutionalization, which exhibit a growth rate of twenty-five and thirty percent, respectively. REOs like ECOWAS, the Andean Community, and ASEAN are but a few examples of this development. 13 This description corresponds to other recent assessments of this trend. A number of observers point out that the wave of regionalism ensued in the early 1990s deals with more issues and strives for deeper cooperation, compared to previous efforts (Hettne 1999, 7-8; Kahler 1995, 81-82; Mansfield and Milner 1999). 14 [Figure 2] Turning to the institutionalization of specific REOs, Figure 3 presents the average designed and implemented levels for each organization. It further validates the observation that existing REOs exhibit considerable institutional variation. It also illustrates the uneven gap between design and implementation. For some REOs, such as ASEAN, SACU, and the EU, the former diverges from the latter only slightly. In other organizations, most notably AMU and ECCAS, implementation falls far short of institutional design. The variation reported in Figure 3 bestows the measure with a degree of face validity. Consistent with conventional wisdom, the EU emerges as the most institutionalized REO, far ahead all the other organizations (Acharya and Johnston 2007; Choi and Caporaso 2002, Grieco 1997; Katzenstein 1997). 15 Detailed comparisons that go beyond the EU offer additional indications that the measure described above is compatible with other assessments of cross-regional variation (Acharya and Johnston 2007, 245; Choi and Caporaso 2002; Grieco 1997; IMF 1994; Page 2000b). At the same time, it provides one of the first nuanced and systematic treatments of this variation and offers a clearer picture of institutional differences across regions and over time. [Figure 3] Design Features and Relationships among Them Moving beyond the overarching measures of regional institutionalization, it is worthwhile to consider the variation on specific design features. Patterns pertaining to the scope of economic activity largely resemble those of the broader institutionalization measures. This is not surprising as scope comprises about eighty-percent of regional institutionalization. Again, substantial variation exists on designed and implemented scope and the gap between them is often considerable. The average score on designed scope is more than twice the average score on implemented scope. 12 Newly formed REOs have low levels of regional institutionalization that may lead to an artificially low average. It is thus more informative to compare the same group of organizations across time. 13 Most, but not all, REOs have experienced increasing institutionalization. Some organizations, such as SACU and LAIA, showed little dynamism and others, such as MRU and CEPGL, have experienced declining levels of institutionalization. 14 This is sometimes dubbed as the new or open regionalism. 15 Nonetheless, the EU has much in common with other REOs and serves as a useful yardstick (Acharya and Johnston 2007; Kahler 1995). 12

13 The existence and power of corporate bureaucracy fluctuates across different organizations and over time. While almost all REOs have established a secretariat or a commission, only a handful of these bodies have the authority to make recommendations or initiate policies. Some, the ASEAN Secretariat and the WAEMU Commission for example, have obtained such power only in the 1990s. REOs institute a binding dispute settlement mechanism only infrequently. 16 About a third of the observations (thirty-two out of ninety) score at least one on this variable. Three-quarters of these mechanisms (twenty four out of thirty two) are standing tribunals, however. This feature, too, exhibits some temporal variation. REOs like EFTA and CACM have established their DSM only in recent decades. As Table A3 reports, the correlation between the different institutional features is always positive, suggesting that wider economic scope is frequently accompanied by greater institutional independence. At the same time, most bivariate correlations range from low to intermediate, underscoring the diverse nature of existing regional organizations. The correlation between economic scope and the two features pertaining to independence are quite modest, ranging from 0.3 to 0.5. Presenting the average score on implemented scope and institutional independence, Figure 4 illustrates this multiplicity of arrangements. Some organizations, such as WAEMU and IOC, score high and low on both aspects, respectively. Other REOs, in contrast, score high on one aspect but low on the other. ECOWAS is rather independent but has a limited economic scope. SACU and the GCC, on the other hand, have a wider scope but are less independent. Interestingly, the correlation between the two variables related to independence is only Thus, while some REOs have both an independent secretariat and a standing tribunal, others have one but not the other. LAIA, for example, includes an independent secretariat but not a binding DSM. In contrast, COMESA has a standing tribunal but a weak secretariat. Thus, institutional independence may develop in a number of different ways. Taken together, this descriptive analysis highlights the varied landscape of existing REOs with respect to their functions and structure, and indicates that each design feature captures a distinct aspect of regional institutionalization, which merits separate empirical analysis. Research Design In order to test the hypothesized effect of economic interdependence on regional institutions, I use the original data set described in the previous section. Because I am interested in the sources of the structure and functions of REOs, which are institutional traits of the organization and not of any particular dyad, the empirical analysis is conducted at the regional level of analysis, defined by organizational membership. The independent variables, described in more detail below, are also defined and measured at the regional level, since there are good reasons to believe that some types of interactions are not adequately captured by monadic or dyadic setup and are more amenable to regional ones. In particular, a number of recent studies point out that economic interdependence, security relations, and regime types tend to cluster geographically, suggesting that regional dynamics are at work (Buzan and Wæver 2003; Gleditsch 2002). 16 Many REOs have a non-binding DSM, see Smith (2000). 13

14 Such regional analysis serves as a useful complement (rather than a substitute) to the more conventional monadic and dyadic analyses. The dependent variables are the six variables related to regional institutionalization. These variables have different distributions and thereby handled with different statistical tools. The four variables related to the overarching level of regional institutionalization and economic scope DESIGNED INSTITUTIONALIZATION, IMPLEMENTED INSTITUTIONALIZATION, DESIGNED SCOPE, and IMPLEMENTED SCOPE are continuous. They are estimated with an ordinary least squares (OLS) regression technique. The estimations of these variables include a random-effects specification, which accounts for unobservable cross-regional variation. The values of the two remaining variables BUREAUCRACY and DISPUTE SETTLEMENT are ordered on a 0-2 scale. They are estimated with an ordered probit model. The random effects specification is not available with this technique. Instead, the standard errors of the estimates are robust and clustered by REO. Independent Variables The main independent variable, economic interdependence, is measured with TRADE SHARE, which is intra-regional trade as a percentage of the total regional trade (Grieco 1997; Page 2000a). A greater proportion of intra-regional trade indicates that REO members trade more among themselves relative to their trade with the rest of the world, which in turn suggests greater regional interdependence. UNCTAD provides information on trade share for most REOs, based on their exports. 17 Turning to some specific design features, extant research suggests that it is not necessarily the intensity of economic relationships, but rather the proposed level of institutionalized cooperation that determines the authority and centralization of regional institutions. Specifically, it is argued that greater economic scope calls for greater legalization and institutional strength (Acharya and Johnston 2007, 268; Johns and Rosendorff 2010; Kahler 2000; Smith 2000; Stinnett 2007). To account for this possibility IMPLEMENTED SCOPE is included in the models that examine the sources of top-level meetings and institutional independence. Because economic interdependence and economic scope are highly correlated, they are included in separate models. Two competing explanations emphasize the distribution of power among the group members and offer opposing views. First, according to assumptions of both realism and bargaining theory, powerful states should prefer weak international institutions designed to reflect the distribution of power among their members. Because they have viable unilateral and bilateral options, when it comes to creating or working through REOs, powerful states face greater opportunity costs than other states. Weaker states, on the other hand, might be expected to favor powerful organizations in order to constrain their more dominant counterparts. Insofar as the preferences of the powerful members prevail, greater power asymmetry between the members of an REO should be associated with lower levels of institutionalization. This argument is most applicable to the design features related to institutional independence because powerful states are especially 17 Trade data for SACU is not available, and most likely does not exist (Page 2000a, 117). 14

15 worried about excessive institutional autonomy and legalization (Haftel and Thompson 2006; Smith 2000). An alternative argument focuses on the importance of hegemonic power as a supplier of regional institutions. This view can be traced back to Karl Deutsch, who argued that successful political integration tends to develop around one core area that pushes it forward (Deutsch et al. 1957, 38). More recently, several scholars grounded this argument in theories of collective action and hegemonic leadership, arguing that hegemons bear the costs associated with higher levels of integration (Mattli 1999; Yarbrough and Yarbrough 1992, 61-66). This argument indicates that greater power asymmetry within a region is likely to be associated with greater institutionalization. The effect of power asymmetry on regional institutions is measured with the so-called concentration ratio, which takes into account both the relative economic size of all members and the number of members in the organization (Mansfield and Pevehouse 2000; Smith 2000). The value of this variable, labeled CONCENTRATION, increases as asymmetry grows and is bounded between 0 and 1. GDP data from the Penn World Tables is used to calculate this variable (Heston, Summers, and Aten 2002). A number of recent studies highlight the link between democracies, international organizations, and regional institutions. Mansfield, Milner, and Rosendorff (2002) argue that democratic governments are more likely to tie their hands to REOs, and find that democratic dyads are more likely to sign such agreement. Accordingly, one might expect REOs among democratic states to be more institutionalized than those among nondemocracies (Mansfield, Milner, and Pevehouse 2008). Regime type is claimed to have particular implications for institutional independence. Democracies, it is argued, are more likely to delegate authority to international organizations as well as to create and use international adjudicating bodies (Kahler 2000; Pevehouse and Russett 2006). To measure regional democraticness, I employ the widely used Polity IV definitions and data (Jaggers and Gurr 1995; Marshall and Jaggers 2002). 18 The composite regime type measure is the difference between the level of democracy and the level of autocracy, which ranges from -10 to 10 for strong autocracy and strong democracy, respectively. After calculating the five-year regional average of this composite variable, I distinguish between regions with an average Polity score of six or higher, indicating that the region is democratic, and those that fall short of this threshold, suggesting that the region is not inhabited by mature democracies. Thus, DEMOCRACY DUMMY is a dichotomous variable that scores one if the average level of regional democracy is greater than six, and zero otherwise. 19 The potential effect of domestic conflict on international cooperation remains largely unexplored. Nevertheless, a number of arguments suggest that civil wars are likely to inhibit regional institutionalization. First, states that experience domestic instability may have to devote their material and political resources to address these problems rather than to empower regional organizations (Nye 1971, 82). Second, insofar 18 This data set does not provide information on micro-states (population less than one million). Such states are not included in the calculation of regional measures. All members of the OECS are micro-states, thus values for this organization are missing. 19 Alternative specifications, such as the actual Polity average or the score of the least democratic member, do not change the results in a meaningful manner. 15

16 as REOs require the delegation of national autonomy, governments whose authority is challenged at home may be more jealous of their sovereignty, and thus less willing to entrust international institutions with political independence. Third, neighboring states may be reluctant to institutionalize their economic relationships with countries that experience domestic strife, fearing that the conflict will spill over national borders and destabilize the broader region. As one observer of regional integration in Africa notes, instability and insecurity have emerged as the greatest impediments to regional integration in Central Africa...The fear of a few states contagious warmongering hindered the desire to construct a common political and economic space (Ropivia 1999, 126). Domestic violence is measured with the count of all the incidents of domestic armed conflicts occurred in any of the organization s members, reported in the Uppsala Conflict Data Program s (UCDP) Armed Conflict Dataset (Gleditsch et al. 2002). This data set distinguishes among four types of wars: interstate armed conflict, extra-state armed conflicts, internationalized internal armed conflicts, and internal armed conflicts. It also divides armed conflicts into three levels of intensity: minor armed conflict, intermediate armed conflict, and war. 20 I count all the incidents among members of an REO that are defined as internal armed conflict. This variable, labeled CIVIL WAR MEMBERS, aggregates the number of conflicts over a five-year period, which is then divided by the number of member-states. Turning to international conflict, realists expect enmity and war to restrain cooperation (Grieco 1988), while institutionalists believe that governments are likely to invest in international institutions where conflict is more likely (Fortna 2004; Wallander and Keohane 1999, 30-32). I account for this factor with the amount of armed conflict within a given region. This is captured by the number of annual militarized interstate disputes (MIDs) between members of the organization over a five-year period (Ghosn, Palmer, and Bremer 2004; Jones, Bremer, and Singer 1996). The number of disputes is likely to increase as the number of REO members grows. To account for this possibility, the number of incidents is normalized by the number of members. Thus MIDS MEMBERS is the total number of armed disputes divided by the number of members. 21 The conventional wisdom holds that there is a tradeoff between the expansion and deepening of REOs. More inclusive organizations may face greater difficulties in institutionalizing their cooperation (Acharya and Johnston 2007; Langhammer and Hiemenz 1990, 69; Mansfield and Milner 1999, ). In addition, as the number of members increases, so is the possibility of greater diversity within the group (Haggard 1997, 24; Kahler 1995, 126; Koremonos, Lipson, and Snidal 2001). Thus, larger organizations may result in a more limited scope of economic activity. On the other hand, organizations that include numerous members can benefit from greater centralization and independence, which reduce the transaction costs associated with separate bilateral 20 A minor armed conflict involves at least 25 battle-related deaths per year and fewer than 1,000 battledeaths during the course of the conflict. An intermediate conflict involves at least 25 but less than 1,000 battle-related deaths per year and an accumulated total of at least 1,000 battle-deaths during the course of the conflict. Like in the COW data set, a war involves at least 1,000 battle-deaths per year. 21 Taking into account only violent disputes does not change the results. For an in-depth discussion of the link between regional conflict and regional institutionalization, see Haftel (2007, 2010). 16