Conceptualising regions: some methodological considerations

While it is beyond the scope of this section (or article) to survey all the ways to conceptualise an international region, we would like to briefly discuss major approaches, and in particular, those that have generated an actual division of the world into regions.

An important criterion for evaluating various concepts of international region is the extent to which they provide procedures for actually constructing regions. For example, if we want a systematic theory of regional security, for example, we need to be able to define our universe of cases, that is, regions.

One critical methodological point is whether we should allow countries to be members of more than one region. For example, David A. LakeFootnote ¹⁷ and Buzan and WæverFootnote ¹⁸ disagree on this point. Buzan and Wæver construct mutually exclusive regions while Lake allows a country to be a member of multiple regions. While this might not seem like such an important point (for example, Lake treats it in a footnote) this decision has important ramifications. Buzan and Wæver follow typical advice that regions, like typologies, should be mutually exclusive and exhaustive.Footnote ¹⁹ There are important reasons why this is often not good advice.Footnote ²⁰ In the case of regions one is forced to make difficult decisions about countries that lie on the boundary. For example, should Turkey be part of Europe or the Middle East? Should the USA be a part of the Middle East regional security complex or not? Our solution is to say that a country can be a member of multiple regions. In contrast, Buzan and Wæver are led by the mutually exclusive criterion to various special categories or concepts. They have ‘insulator’ states, which lie on the boundaries between RSCs. They have the concepts of ‘overlay’ and ‘penetration’ which allow major and superpowers to be involved in a RSC, but not be members. It also leads them to emphasise a certain level of analysis, typically higher, which allows boundaries to be more clearly drawn. One needs the ‘Middle East’ RSC because lower level RSCs within this geographical area would involve overlapping membership.

By far the most popular way to define a region is via some large, geographical and continent-sized areas, such as Africa, Europe, Latin America, etc. In large-N quantitative studies where region is a pretty common independent variable this is almost always the concept of region used. There is some variation because some would have a region ‘Asia’ while others would have ‘Southeast Asia’, ‘East Asia’, and ‘South Asia’. Similar problems arise with the Americas; for example, does the Caribbean count as part of Latin America?

This way of doing international regions is quite popular in books and anthologies on regional security. Often the first chapters deal with general theoretical issues, which are followed by chapters devoted to regions as big, geographical areas. Prominent anthologies such as Fawn, Patrick M. Morgan and Lake, or Amitav Acharya and Alastair Iain Johnston use this structure.Footnote ²¹ This not a problem per se and it is not necessary to formally define regions.

From a conceptual point of view one can ask what are the basic organising principles for this view of an international region. First, it is based purely on geographical contiguity. It does not allow, for example, superpowers to be members of distant regions. Second, it looks for ‘natural’ boundaries to draw lines. These are often quite large continents. Third, implicitly one uses the mutually exclusive and exhaustive guidelines to construct the regions. This means that every country is a member of some region and it cannot be a member of two regions.

The Correlates of War (COW) produced perhaps one of the first official lists of ‘states’. These were given numbers roughly corresponding to regions, for example, numbers 2–99 are the Americas. While we are not aware of any explicit discussion on the matter, the EUGENE softwareFootnote ²² does have an option for giving the user a region variable. It is important that this is just one variable; this means that a country cannot be a member of two regions. COW divides the world into five regions: (1) Europe, (2) Americas, (3) Middle East, (4) Africa, (5) Asia. Having fewer regions means there are fewer hard choices to make about countries on the border. Here the tricky ones are the boundaries between Africa and the Middle East (for example, Algeria is put in the Middle East) and between the Middle East and Asia (for example, Pakistan is in Asia).

To take another popular dataset, the Polity data on regime typeFootnote ²³ has a region variable as well. For this project the world is divided into Western Europe, Eastern Europe, Europe-settled, South America, Central America/Caribbean, South and Southeast Asia, Middle East/North Africa, and Sub-Saharan Africa. Again, Polity follows the rule of exhaustiveness and exclusivity.

Typically the region variable is not a controversial one, and usually not a critical one in the large-N theoretical or empirical analysis. As a result variation in these categorisations is not important. The rule of exhaustiveness and exclusivity is automatically followed because alternatives are not on the methodological map. We argue that the region variable is theoretically important and should be regarded as more than just a control variable.

In conceptualising region one needs to consider what is the opposite of a region. In most of international relations literature, the opposite of regional is global. We contrast three quite different levels: (1) region; (2) macro-region; and (3) global. At all levels we see institution construction and competition between institutions. While many define region in large, often continent-wide, geographical terms, we define regions for our purposes at the level below that, in large part, because this is the level of the most intense institutional activity. For example, Africa is a macro-region for the purposes of this article, and it has the African Union as its macro-regional institution. However, Africa is divided into about 10 regional economic institutions, such as ECOWAS, SADC, COMESA, etc.Footnote ²⁴

Regions in the literature of the 1960s and 1970s

The analysis of concepts typically involves comparing the new proposed concept against what Robert Adcock and David Collier call the ‘background concept’, which they define as ‘the broad constellation of meanings and understandings associated with a given concept’.Footnote ²⁵ Our case is a bit unusual since we have two separate conceptual universes to consider, the literature on ‘regions’ and the literature on regional institutions and organisations. In this section we consider the background concept of region, below we consider our alternative vis-à-vis the literature on IGOs.

There was significant work in the 1960s and 1970s on international regions, both in terms of security and economic integration. This was clearly motivated by the formation across the world of REIs. Led as always by Europe, regional integration institutions were formed in many areas, which naturally produced work trying to explain and theorise them. When regional integration schemes failed or stalled in the 1970s scholars lost interest in them until things got going again in the 1980s. For example, Joseph S. Nye, a major contributor to the regional integration and regional security literature, defined a region as ‘a limited number of states linked together by a geographical relationship and a by a degree of mutual interdependence’.Footnote ²⁶ He argued that regional economic institutions affect conflict involving member states differently than regional security institutions. We argue that multifunctional regional economic institutions serve both economic and security functions in order to more effectively manage conflict.

Bill Thompson conducted an extensive survey of the concept of region. He surveyed 21 authors and found 21 different attributes used to define region. By far the most popular of the twenty-one were: (1) ‘proximity or primary stress on geographic region’ and (2) ‘actors’ pattern of relations or interactions exhibit a particular degree of regularity or intensity’. Criterion (1) is clearly geography, while criterion (2) is some form of interdependence or interaction. Thompson summarises and synthesises the literature on regional subsystems with the following concept:

As we shall see this has not changed much in the intervening decades and the regionalism literature of the post-Cold War can be seen as a variation on these themes.

The one approach that Thompson does not really cover uses the homogeneity of countries as the central means for defining a region. Russett's work provides a good example of this: ‘How do we define a region? One possibility is simply to identify an area divided from another by geographic barriers […] But virtually all social scientists, including geographers, would reject this definition. A region, they might say, must be composed of units with common characteristics. Regions should be areas of relative homogeneity.’Footnote ²⁸

Homogeneity becomes in practice the use of factor analysis to look for countries correlated on some number of variables. For Russett this means dozens of characteristics of countries, such as language, religion, economic characteristics, etc. The factor analysis generates latent variables, which can be used to assign countries to regions. The key thing with the homogeneity approach is that it does not usually rely (at least directly) on interactions or dependence.Footnote ²⁹

When the regionalism literature gets going again in the 1990s it has a clear intellectual debt to those who worked in the 1950s through 1970s. Katzenstein has been one of the most prominent contributors to this resurgence and he makes his debt to Deutsch quite clear: ‘Together with Karl Deutsch this article defines a region as a set of countries markedly interdependent over a wide range of different dimensions. This is often, but not always, indicated by a flow of socio-economic transactions and communications and high political salience that differentiates a group of countries from others (Deutsch 1981, p. 54).’Footnote ³⁰ Here we see that interdependence remains core to conceptualising regions, and it is not surprising that the literature on regional security adopts some version of interdependence to define regions.Footnote ³¹

REIs and the construction of regions

Our concept of international region relies on states themselves organising territory and space. We believe that LakeFootnote ³² is on the right track with his emphasis on externalities. One classic response to negative externalities is to gain efficiency through internalising some aspects of production. Similarly, states can try to deal with various problems via regional institutions. They may try to increase trade, provide common services, deal with standardisation, or address environmental problems via regional institutions. These regional institutions have a geographic basis in part because the problems themselves have a geographic dimension, but also because states see other regions dealing with problems via regional international institutions.

This article claims that international regions are being constructed via institutions mostly devoted to economic development and integration, but which then form the platform for institutional expansion into a wide variety of areas such as human rights, democratisation, environment, and eventually security. In an interesting contrast with the Weberian state, which was created by war to wage war, economic issues drive the core of the institutional organisation of international space.

We focus on REIs as the institutional basis of the region. Economic integration comes in the form of REIs that can include free trade agreements, customs unions and economic unions.Footnote ³³ Free trade agreements specify that member states pledge not to place tariffs on each others? goods but are free to do so at any level with non-member states. Customs unions are free trade agreements that include a commitment to place the same tariff on the goods of non-member states. Finally, economic unions are rare but include a commitment to completely integrate economically. The EU comes the closest to fulfilling the requirement for this category.

These multifunctional regional economic institutions facilitate economic integration across a range of issue areas. According to economists, REIs facilitate trade cooperation as well as other forms of economic integration among a limited number of states.Footnote ³⁴ We employ, however, a much broader definition than that of economists. REIs are economic institutions that specify rules for trade liberalisation in the form of reduced tariff barriers, tariff policy harmonisation (for example, uniform custom policy), economic development (sustainable development policy, loan funds), the economic implications of shared natural resources (commercial fishing, hydroelectricity) and integration in infrastructure (maritime transport, immigration).

States join REIs for a plethora of reasons, such as improved access to foreign markets, removal or reduction of tariff barriers among member states, restrict members ability to raise trade barriers, insure against future disruptions in trade, insulate from the effects of hegemonic decline and global recession as well as to promote economic growth and development.Footnote ³⁵

During the 1990s, economists debated the nature of regions as well. They agreed that the world was transforming into a global economy based on trade flows generated by trading blocs. The debate centered on whether these blocs were competitive thus dividing the world into massive regional trading blocs that stifled globalisation or complementary thus being stepping-stones to multilateral trade liberalisation.Footnote ³⁶ We argue that this debate overlooks the institutional basis of such trade as well as the multifunctional nature of REIs. Our argument focuses on how the world is moving toward regions defined largely by regional economic institutions that serve a range of economic purposes.

Concept and operationalisation of a regional economic institution (REI)

We show below that most of the world's territory has been regionalised via institution creation in the form of REIs. REIs deal with a wide range of problems and they are often the go-to place when new problems arise. For example, when Zimbabwean elections in 2008 proved controversial, there were wide-spread calls for SADC to get involved. REIs are typically the most important regional organisation (though perhaps not strong in an absolute sense). They are ‘regional’ because they group together proximate and usually contiguous states. They are ‘institutions’ because they have formal legal status; they are constituted by binding treaties, and often have large numbers of treaties, protocols and agreements that constitute the institution itself.

We focus on economic institutions because regional integration has typically started from dealing with economic problems and issues. While security issues are not absent, and in fact, many REIs do eventually become involved in security, economic integration is the core. REI member states view economics and security issues as inherently linked. It seems like the normal evolution has been to focus on economic integration first even if this economic integration has some security motivations, as illustrated by the history of the EU. It is certainly conceivable that security organisations and alliances could move into economics, but that seems quite rare.Footnote ³⁷ Thus in an interesting contrast with the Weberian state which was formed for and by war, the political organisation of regional space is mostly informed by economic issues of various sorts.

To be included as an REI there must be a clear document (or usually set of documents) that provides the binding legal foundation for the institution. The requirement of international legal status, that is, treaties signed by member nations, distinguishes ‘institution’ from informal forms of interaction, which we call ‘discussion forums’. As with all concepts there is a gray zone.Footnote ³⁸ This is not to say that these forums are not important, but that they constitute a kind of proto-institution, they lie in the institutional grey zone. APEC, for example, has regular meetings of its member states, but it has no legal status; there are no legally binding agreements signed by its members. ASEAN+3 (South Korea, China, and Japan) is also a discussion forum. APEC could become a REI; ASEAN could expand to include South Korea, China, and Japan. Discussion forums are often formed quite explicitly because states are not willing to formalise the institution.Footnote ³⁹

We are interested in the multifunctional regional institutions. As such we exclude from our list all IGOs that are single-issue in nature. If one takes, for example, the Correlates of War database of IGOsFootnote ⁴⁰ the vast majority are excluded on this basis. We exclude purely security organisations as well, such as NATO or the CSCE on this basis. We are interested in the creation of institutional governance in regional space. We thus limit ourselves to those IGOs and REIs that might assume governance over a variety of economic activities.

States create and design international organisations to perform particular tasks. Kenneth Abbott and Duncan SnidalFootnote ⁴¹ argue that formal IOs perform administrative, information-gathering and dispute settlement functions among other tasks for states. For instance, they can help small and developing states pool together scarce resources to attain mutual ends. We argue that important REIs are more than just treaties. They possess an organisational structure that includes at least an executive body composed of heads of state that meets regularly, a secretariat and permanent headquarters. Such structure is necessary to implement, monitor, and eventually enforce treaty rules that govern REIs.

The basic criteria for an REI to be included in our analysis are the following: (1) multi-functional, economic integration; (2) treaty-based; (3) basic IGO structure (for example, executive body that meets regularly, secretariat, headquarters); and (4) active between 1980 and 2005, though we only include those active in 2005 in this article.

REIs: criteria for importance

We have proposed using regional institutions to identify how states have constituted regions. So another ‘background concept’Footnote ⁴² of relevance is IGO, or more specifically the subset of important regional IGOs. Are we just picking out some important IGOs and then calling them regions? If our concept has important substantive importance then it should differ in some important ways from just equating regions with IGOs?

We are interested in the institutional construction of regions, which means we are interested in the ‘important’ IGOs. At an intuitive level these are not hard to identify. For example, in an anthology on regional economics or security these are the institutions that actually get discussed in the case study chapters. These are the ones that have well-known acronyms, such as NAFTA, ASEAN, MERCOSUR. This is important because often the IGO variables are independent variables that are hypothesised to influence state behaviour. The logic is that only important international institutions with some resources will have influence, and it seems reasonable to exclude small, functional IGOs from consideration.

One fairly popular approach to defining, implicitly, an important IGO is to use some yearbook or data source likes Arthur S. BanksFootnote ⁴³ as the selection device. For example, Jon C. PevehouseFootnote ⁴⁴ uses Banks and Beckfield'sFootnote ⁴⁵ the Stateman's yearbook and the Europa yearbook. It is not very clear what the reference books' criteria are but they are certainly related to perceptions of importance of the IGO. In Table 1 we have a columns for Pevehouse and Beckfield to illustrate the extent to which our REIs are included in their lists compiled in this manner.

Frequently the criteria for inclusion of important IGOs in a study are not clear.Footnote ⁴⁶ For many scholars it is easier to include all IGOs (typically as defined by COW) than to come up with explicit criteria for the inclusion of some and the exclusion of others. In this article, we construct criteria for important IGOs and apply it to one class of institutions, REIs, in order to isolate the ones that matter most in different regions. This initial step allows us to better understand the institutional basis of regions. Since we make an argument that regions are created via economic institutions, identifying and evaluating the important IGOs that constitute a region is important.

We propose various criteria that allow us to identify the most important regional economic institutions. If most of these institutions are absent then an IGO or PTA would not be considered important in a general sense (though it might be important in a given issue area). However, not all important REIs will have all these characteristics, but certainly will have two thirds or three quarters of them. We think that these include most of the implicit criteria used by reference books that discuss individual IGOs, which form the basis of the choices made by scholars like Pevehouse and Beckfield. These characteristics are: (1) treaty-nesting and linkage; (2) international legal personality; and (3) primary or important IGO used in region as identified by scholars or international bodies (4) emanations; and (5) annual or more frequent meetings of Presidents, Prime Ministers and/or cabinet level officials.

Treaty nesting and linkage is a characteristic of important REIs. International treaties are often linked in different ways. Linked and nested treaties constitute literally what we, and others, mean by an international institution. REIs, as institutions, can essentially be described as interconnected sets of treaties that bridge numerous concerns. One treaty may be nested or connected to other treaties. For instance, Treaty B may implement the treaty obligations in Treaty A while Treaty D can change the scope of treaty obligations in Treaty C.Footnote ⁴⁷ Most regional economic institutions are composed of dozens of treaties (for example, CIS, EU, ECOWAS, CARICOM). These nested and linked treaties are closely related to the multifunctional character of important regional institutions, since an individual treaty might deal with only one issue area.

International legal personality is a key characteristic of an important REI. It is the legal right and capacity to possess rights and duties as a legal person under international law. International legal personality affords IOs the right to make and challenge international law through signing treaties and bringing claims to international courts.Footnote ⁴⁸ States automatically possess such status because they have sovereignty. It is this right to legal personality, inherent in sovereignty, that has allowed states to be the sole creators of international treaties and therefore international law. Now, international organisations either acquire such a status on their own or states grant it to them via delegation.Footnote ⁴⁹ Examples of REIs that have signed trade agreements with states and other IGOs are the EU, MERCOSUR and ASEAN.

Various actors by their actions and policies identify the important regional organisations. For example, the UN Charter talks about regional peacekeeping. So if the UN or other macro-regional organisations decide to engage a regional institution that implies that it is an important one. The African Union Charter gives special attention and importance to some REIs. Scholars as well by their studies and analyses informally vote on which are the most important regional institutions. Buzan and WæverFootnote ⁵⁰ when discussing regional security bring some regional organisations into the analyses which they think are relevant. In anthologies on regional economics or security, these are the organisations that are discussed. These are the organisations often included in reference books, such as Banks.Footnote ⁵¹

Because important REIs are engaged a wide variety of substantive areas, they create specialised organisations, bureaucracies, etc. Just as the UN has specialised agencies so do important REIs. In IGO literature these are often called ‘emanations’ and many datasets, for example, COW, exclude them because they are not independent entities. However, for us the existence of numerous emanations is a clear signal of importance.

Another way that states vote and give importance is via the attention devoted to the IGO by key government leaders, either presidents, prime ministers or cabinet level officials. If the presidents of the member countries meet every year, make decisions, etc., that signals to domestic and international audiences that an institution is important. If regular meetings just involve lower-level bureaucrats then that signals the organisation is less important.

One could certainly come up with additional criteria of importance. However, we think that these six are more than enough.Footnote ⁵² Typically, unimportant, single-issue IGOs have none of these characteristics. At the same time, important REIs do not need to have all characteristics. For example, NAFTA is clearly an important REI, but it has few treaties (only three).

The economic-institutional regions of the world

One of our main goals is to investigate the extent to which there has been an economic-institutional organisation of international space. That space has already been organised via the creation of states. So we take that existing organisation as a starting point: to divide the world's states into regions first requires a list of states of the world. We take the popular COW list of nations.Footnote ⁵³ This excludes very small microstates but includes all states of any size, for example, most of the small Pacific island states are on the COW list. We also exclude a few small European territories, Monaco, Liechtenstein, Andorra, and San Marino (which are not considered independent members of the EU). We have thus an international system with 188 members. These states can belong to no REI, one REI, or multiple ones, that is, from our perspective they can belong to 0, 1, or multiple regions.

Table 1 then provides our list of economic-institutional regions as constructed by the states of the world for 2005, in other words Table 1 is our list of REIs. Hence, we find the world divided into 34 regions.

The data in Table 1 show that regions as constituted by REIs are not just a relabelling of existing concepts such as PTA, RTA, RIA, FIGO, etc. Typically only about half of our REIs are to be found in these other datasets. The case of the most overlap is Volgy et al.s FIGOs. This is not surprising given that they have by far the most organisational set of requirements. Of course they have many FIGOs, which are not REIs – only about 15 per cent of their FIGOs are REIs. Conceptually they stress the organisational and administrative dimensions while we stress legally-binding treaties, protocols, etc.

Table 1 also includes a comparison of REIs with other studies that only include ‘important’ IGOs or more generally all preferential trade agreements (PTAs). The zeros in the table indicate that the REI is not included in the study. Given that our criteria for inclusion seem quite strong compared to others it is interesting to note the relative lack of overlap. Notably, most of the central Asian REIs are not included in these various datasets. Some of the less important African ones are also commonly excluded. However, the most rigorously defined dataset, Volgy et al., includes almost all of REIs with only a couple of exceptions. Of course, it is beyond the scope of this article to investigate these differences further; the key point is that our conception of region based on REIs is not just a relabelling of related concepts such as IGO, PTA, etc.

Liesbet Hooghe and Gary Marks have a project very similar to ours. They distinguish between general purpose and specialised IGOs and focus on what they call ‘authoritative’ IGOs: ‘Some 50 of these [COW IGOs] can be described as authoritative, having a formal constitution, a supreme legislative body, a standing executive, a permanent professional administration, and some formal mechanisms for enforcing decisions and settling disputes. Of these, 13 are responsible for a range of policies and might be described as general-purpose.’Footnote ⁵⁴ Clearly this sets the bar significantly higher than we do. But what is relevant is that their criteria are very similar to ours.

Table 1. Political-economic construction of regions: REIs as regions

Note: M-Mansfield (1994), MS-McCall Smith (1995), H-Haftel (1992), P-Pevehouse (1992), B-Beckfield (2000), V-Volgy (2004), HM-Hooghe and Marks (2009).The year is the most recent year included in the dataset.1= included dataset, 0=not included in dataset.

Weak and small REIs might be quite vulnerable to more dynamic REIs. At one point in time the EU and the European Free Trade Area had roughly the same number of members. Today the EFTA is on life-support. Venezuela has recently applied to join MERCOSUR, and there may be a natural attraction of states toward stronger REIs. Part of the dynamics of regions is how weak REIs may get weaker as states jump to stronger REIs. One might predict that states in weaker REIs will eventually join MERCOSUR.

We have already discussed the issue of whether a country can be a member of more than one region. Since the states themselves select into REIs we let them decide. In some cases, particularly big countries, it is quite likely that they belong to multiple regions. For example, Russia is a member of the CIS, the Eurasian Economic Community, and the Shanghai Cooperation Organisation (SCO). There are a number of hypothesis one might make about countries and REIs where there is significant overlap. Sometimes countries form a new REI because they want further integration, this was the case for Russia and the Eurasian Economic Community. A good number of CIS member states are not happy with the CIS and Russian dominance, they are looking to join other REIs (for example, Moldova and Ukraine looking westward) or even can create competitive REIs, such as the creation of GUAM by states unhappy with the CIS.

One interesting aspect of the REI membership data as of 2005 is that the world has already gone quite far in constituting regions on an institutional basis. Out of the 188 states, only 15 are not members of an REI. We think it is quite amazing how far the institutional construction of regions has gone in just a few decades. This suggests a number of hypotheses about why states have been active in region creation over the last 20 years. We engage in some speculations on this point in the Conclusion.

We call those states that are members of no REI, such as Israel or North Korea ‘regionless’ states. It is interesting to note that these are usually regionally concentrated and not randomly scattered around the globe. There are some geographic areas that have no REI, such as East Asia, for example, Japan, South Korea, Taiwan and the Middle East. While it might seem obvious that serious conflict has prevented the formation of regional institutions in the Middle East, if one looks comparatively it is worth noting that other regions such as Africa with serious conflicts have experienced significant institutional creation.

Many analyses of regional security explicitly have an underlying dimension of anarchy to institutionalised, or even to security community.Footnote ⁵⁵ The degree of institutionalisation of a region has wide-ranging implications for almost all policy areas. Our economic-institutional map of the world has some clear frontier regions, which are unstable and which are often the sites of serious conflict. For example, the EU has been successful in internalising many of the security externalities produced by Balkan countries as they have become members of the EU. The EU has in fact occupied countries like Serbia and Kosovo. ASEAN+3 has the potential to become a REI for Japan, China and South Korea.

In contrast with the staticness of traditional geographic definitions of regions, a social constructivist view of regions is quite dynamic. We see a range of regions from those that are weakly organised to those who have security communities. In other words, the organisation of space can vary from strong to weak. Similarly, we see stronger institutions growing and taking over more space particularly in the frontiers.

Security-based approaches to conceptualising regions

The standard methodology of concept analysis typically involves two stages. The first stage explores the theoretical issues involved in choosing the specific factors used to define the concept. In the context of regions, one can look at Thompson'sFootnote ⁵⁶ list of 21 different factors scholars have used to define region. It could be Russett'sFootnote ⁵⁷ analysis that looks at various criteria of homogeneity to define regions such as culture, language, trade, etc. The second stage, if possible, compares empirically the results of different conceptualisations in practice. For example, in the literature on democracy one correlates the various democracy scores to ascertain how different the concepts are in practice.

Unlike most analyses, which involve closely related notions of the same concept, for example, democracy, we compare our institutional view of regions with one, which on the face of it is quite different. With very few exceptions, scholars of regional economic integration or security have used some kind of interaction, dependency, or externality approach. The only major exception to this is work like Russett's that uses factor analysis on attributes of states. Hence it is of interest to compare how a social constructivist view of regions which is based on the states' own behaviour in institution creation with regions that are constructed by the analyst based on some substantive criteria.

Adler and Patricia Greve stress the importance of ‘overlap’ in the analysis of regional mechanisms of security governance. Instead of seeing regional governance as an either-or, they argue that we see different mechanisms working within the same region. Our comparison of regions based on institutions versus regions based on security dependencies works in an analogous way. When they argue that ‘“regions” boundaries are determined by the practices that constitute regions’Footnote ⁵⁸ we can see this in terms of institution construction as well as security practices. Our contrast of institutions versus RSC thus finds a close parallel in their contrast between security communities and balance of power systems.

As we suggested in the introduction, perhaps the strongest contrast is between an economic-institutional view of regions and a conflict-security notion. Given that conflict and trade are both strongly geographically-based it is not implausible that there is significant overlap.Footnote ⁵⁹ This overlap might have implications for economic as well as conflict relationships. It might mean that REIs might take on conflict management roles.Footnote ⁶⁰

In this section we focus on the regional security literature in general, and in particular, on those few works that have moved to an operationalisation of region based on security criteria. For example, while Lake's work is central to a theoretical analysis of regional security, he does not produce a list of regions. Wæver and BuzanFootnote ⁶¹ have been perhaps the most prominent researchers in regional security, and have divided the world into regional security complexes. In addition, we discuss more briefly Douglas Lemke'sFootnote ⁶² analysis of local hierarchies (which excludes the developed areas).

The literature on RSCs takes the core notions of interdependence and geography to work out a notion of region that is based on security. Buzan and Wæver have worked out the regional security complex idea in a series of important books.Footnote ⁶³ The formal definition has shifted a bit over time, and they have described RSCs in somewhat different terms in different places, but the core has remained pretty constant. We take their 2003 book as the current statement of their concept:

To summarise, we see four attributes of the Buzan and Wæver conceptualisation of regions: (1) interdependence; (2) security-based; (3) geographic proximity; and (4) separation from other regions and the global system.

Lake has engaged in a dialogue with Buzan and Wæver regarding regional security and has proposed his own definition of a RSC:

Geography is clearly present with the requirement of a ‘local’ security externality in a ‘particular geographic area’. The concept of ‘security externality’ replaces the popular idea of interdependence. By definition if there is a security externality, countries are interdependent.

Externalities can be generated across big distances. For example, the election of Barack Obama is a security externality for the countries in the Middle East. Conversely, the success of the Taliban in Pakistan is an externality for the USA. It is not unreasonable to include all states active in a region as members of the region. If these geographically distant states are excluded from the definition of the region then they will have to be brought into the framework via some other means. This is the theoretical role of penetration and overlap for Buzan and Wæver. One way or another the framework has to be able to allow for the participation of major and superpowers in regions outside their geographic home.

LemkeFootnote ⁶⁹ is another effort to conceptualise regions on security grounds and provide operational criteria. He is interested in developing a regional version of the power transition theory,Footnote ⁷⁰ which was originally only a system-level theory. His ‘local hierarchies’ are close to RSCs. He defines region via the idea of the possibility of military interaction between states: ‘Specifically, rather than proximity and observed interaction being central, the essential characteristic of interstate relations assigning states to the same local hierarchy is the ability to interact militarily.’Footnote ⁷¹ He first constructs dyads that are able to interact militarily. Then he constructs regions based on all those countries that can interact militarily with the local, dominant powers.Footnote ⁷²

Buzan and WæverFootnote ⁷³ provide an extensive analysis of the RSCs of the world. They have a rough scale of RSC size, from ‘mini’ complexes to ‘super’ ones. In general, they tend to privilege what we call the macro region. They do not provide an explicit list of RSCs and their member states. However, they do construct a series of maps that we have used to code states into RSCs, (available from the authors on request). Most of the time this is not too difficult, but there are occasions when the boundary lines in their maps go through the middle of countries, for example, in Africa. For purposes of illustration we have reproduced one of their maps as Figure 1.

Figure 1.

In terms of scale, a RSC generally corresponds to our macro-region. Our REI regions fit most often with what they call a subcomplex. The left column of Table 2 gives the Buzan and Wæver RSC and the right column gives the REIs which correspond to that RSC. If either column is empty that means there is no corresponding RSC or REI. Of course the matches here are not necessarily perfect, there may be individual countries that are in an RSC but not an REI, or vice versa. For example, Chile is a member of the Southern Cone subcomplex, but not a member of MERCOSUR.

What are the criteria of comparison between regions defined by REIs and those defined as RSC? The comparison is tricky between the various kinds of RSC. High overlap is when the RSC subcomplex matches the REI. In a significant minority of cases we find this, such as in Central America or the southern cone (MERCOSUR). A less good match is when the RSC has boundaries that match quite well with two or more REIs, often this occurs in areas where there is complex but no subcomplexes. The North American Complex includes both NAFTA and the Caribbean REIs. This suggests that one might want to break the complex into two subcomplexes.

More common, and probably more interesting, is when the complex includes several REIs in a position of competition, such as the Western European complex or the Central Asian subcomplex. Here is where the methodological point of allowing states to be members of multiple regions has real substantive impact. REIs with a high degree of overlap with RSCs are probably unstable in various ways, we might expect to see significant movement in terms of institutional strength and membership. It is the rule allowing multiple memberships that helps us see that some macroregions might be quite different dynamics than others.

In many cases the match between REIs and RSCs is very good, notably in the Americas. In some cases it is quite bad, for example, the Balkans subcomplex (Albania, Macedonia, Croatia, Bosnia-Herzegovina, Slovenia, Greece, Romania) is composed of half EU countries and half CEFTA countries. At the same time the expressed purpose of the CEFTA is to prepare countries for EU membership. Sometimes the overlap is quite good, but a few notable countries are not members of the REI. For example, the Gulf subcomplex overlaps to a large degree with the Gulf Cooperation Council (GCC), however Iran and Iraq are members of the subcomplex, but not members of the GCC.

Table 2. Comparison of REI and RSC approaches to regions

Note: see Table 1 for acronyms. We do not include RSC supercomplexes in this table.

Another form of mismatch is when the same REI appears in multiple RSCs. Central Asia and the post-Soviet space, along with Central Africa, provide a good example of this. Our institutional analysis of regions suggests then that the RSC boundaries are unstable since one has institutions covering more than one RSC.

Finally, we have 15 states, which are regionless. It is perhaps not surprising also that our regionless states of the Middle East, that is, Israel, Lebanon, Syria and Iraq form a subcomplex as do the regionless states of North Asia, that is, Koreas, and Japan. It would be a useful study to see when conflict prevents the creation of REIs versus when conflict is an incentive for institutional creation and development. A second group of regionless states are those that have nearby regional institutions, which they will probably join. This includes states such as Chile or East Timor (likely to join ASEAN).Footnote ⁷⁴

In summary, there is important overlap between REIs and RSC and some important differences. While it is beyond the scope of this article, these both suggest some important hypotheses both about the dynamics of REIs as well as RSCs. In general, one might propose that when there is significant overlap between REIs and different RSCs that the boundaries of both would be likely to change. When the match is quite good between REIs and RSCs one might expect the regional system to be more stable. In some RSCs with multiple REIs there is clearly a dominant one, which one might expect to grow at the expense of the weaker ones, such as in the history of Western Europe.

Concepts of region and causal relationships

As we briefly discussed in the Introduction, new concepts prove their value, in part at least, by producing new hypotheses, theories, and ideas about causal relationships. Regions as purely geographic entities rarely incited new theoretical or empirical hypotheses. In contrast, regions as defined in terms of interactions or dependencies have proved quite influential. For example, the functionalist literature on the EU along with the work of Deutsch, Nye, Russett and others has had an important impact the analysis of international regions.

The previous section explored two quite different ways to think about regions. Buzan and Wæver have used security dependence to construct RSCs. We propose REIs as an alternative. While the previous section looked at the two in terms of the empirical comparison of two different concepts, in this section we very briefly explore one possible causal connection between the two.

Is there a causal relationship between REIs and RSCs? To put it more generally is there some sort of causal relationship between security dependence and the development of REIs? One way to think about this is to ask to what extent security issues were central to the founding of the REI, or were central to major changes in the institution. For example, while the initial European common market was economic in content, security issues were clearing a motivating factor. European countries moved to economic integration after the failure of the European defense community proposal in the early 1950s. As we saw in the previous section, there is significant overlap between RSCs and REIs; this suggests that security might be casually related to the development of institutionalised regions.

A classic realpolitik strategy involves using military alliances to manage security affairs. Alliances as traditionally defined include defense pacts, ententes, non-aggression pacts and neutrality agreements. Classic realist scholarship has not linked regional institutions with alliances in a general way, beyond what is seen as the special case of NATO. However, NATO suggests that alliances might be incorporated into a regional institution. REIs are multifunctional institutions so it is quite possible that they move into the security sphere, just as states have moved over time into new functional areas, such as the environment. To make the hypothesis more concrete we ask about the extent to which multilateral alliances signed since 1989 fall within REI regions. Table 3 lists all multilateral alliances signed since 1989. These have been coded by one of the standard databases of military alliancesFootnote ⁷⁵ which use conventional concepts and definitions of alliances.

Table 3. Military alliances in REIs, 1989–2002

It is quite stunning to see that virtually all of the multilateral alliances since 1989 have been formed within REI regions. REIs have gotten involved in other kinds of security behaviour as well, such as peacekeeping, drug-related security, small arms control, etc. While many have focused on the USA's global campaign against armed Islamic groups, there has been much security cooperation on other issues as well as within regional institutions. It seems quite possible that these REI-based alliances are motivated by different concerns than power- or threat-balancing alliances. Our analysis implies that alliance, and potentially other security, policies have become quite regionalised and institutionalised.Footnote ⁷⁶

If one looks at the many security problems of Africa, the connection between REIs and conflict management is quite clear. African REIs have extensive treaties dealing with conflict management, peacekeeping, and the like. There is a clear sense in which the African Union is supposed to coordinate with regional bodies in peacekeeping activities. Africa is not alone, the CIS has been involved in the conflicts in Caucasus, and ASEAN has played a role in dealing with various conflicts in Southeast Asia.Footnote ⁷⁷

This very cursory analysis suggests that alliance behaviour has changed in important ways since 1989, and has taken a strong regional, institutional focus. The point of this little analysis is to suggest that an institutional conceptualisation of regions can lead to new hypotheses about alliance behaviour.