European Foreign Policy as a Contested Institution

The differences between European integration in the economic domain on the one hand and integration in security affairs on the other are often emphasized by policymakers and scholars alike. Market-building policies and some market-correcting policies are governed according to what has become known as the “Community method,” that is, a complex set of institutional rules and practices that ensure a prominent role for supranational agencies and a high level of legalization. In foreign and security policy, on the contrary, supranational institutions have little or no power, and the obligations laid upon governments are vague or frequently ignored.¹

This seems to confirm the skepticism about functional spillovers that Stanley Hoffmann voiced almost forty years ago: “When the functions are concerned with the ineffable and intangible issues of Grosspolitik, when grandeur and prestige, rank and security, domination and dependence are at stake, we are fully within the realm of traditional interstate politics.”²

To many international relations scholars, the relatively feeble state of foreign policy integration is far from surprising. Foreign and security policies are pivotal to state sovereignty, and for these authors, sovereignty—understood as the survival of the state as a distinct political entity in world affairs—is a “good” that states value over all others. According to John J. Mearsheimer, “the most basic motive driving states is survival. States want to maintain their sovereignty.”³

If states are assumed to be “unitary actors who, at a minimum, seek their own preservation and, at a maximum, drive for universal domination,”⁴ then the absence of consensual political and military integration in Western Europe is easily understandable. Even if one allows political integration in exceptional circumstances, such as massive external threats or unique opportunities for territorial expansion,⁵ it can be argued that in Western Europe the benefits of an integrated foreign and military policy never outweighed the costs in terms of sovereignty.

While this perspective has some empirical merit, it neglects a crucial aspect of the Western European experience: there was, and still is, considerable variation in the preferences of governments concerning foreign policy integration. The optimal level of political unification in the EU is a matter of intense controversy not only among its citizens, but also among its member governments. In the light of these controversies, any interpretation of foreign and security policy integration that assumes a uniform concern for sovereignty and ignores differences between government preferences is bound to remain inadequate.

The institutional trajectory of European integration is decided mainly during Intergovernmental Conferences (IGCs), which since the 1980s have taken place approximately every five years. These conferences produce “grand bargains,” whose terms are generally written into the basic treaties of the EU. European governments have expressed divergent positions on the form and depth of foreign and security policy integration, and as a consequence long and difficult negotiations have ensued. For instance, during the IGC of 1996–97, which led to the Amsterdam Treaty, EU foreign ministers or their personal representatives met on more than twenty different occasions to discuss possible revisions of the Maastricht Treaty provisions pertaining to the Common Foreign and Security Policy (CFSP) and the defense policy of the Union. In this time, dozens of position papers, nonpapers, and draft treaty articles on the CFSP were presented and discussed by ministers and officials.⁶

Moreover, the Amsterdam Treaty negotiations were relatively simple compared to those that generated the foreign and security provisions of the Maastricht Treaty in the 1990–91 IGC, which occurred in the uncertain international context that followed the end of the Cold War.⁷ Those fierce negotiations on European treaty reform reflect the existence of significant disagreements among governments about the depth of integration in foreign and security policy, and make clear that the CFSP is a “contested institution.”⁸

Table 1 offers a synoptic presentation of these disagreements, with reference to the main treaty revisions on CFSP that were considered in the IGC of 1996–97. For each revision, the table indicates the position expressed by each member-state during the prenegotiation phase. If implemented, these treaty revisions would have meant a higher level of sovereignty pooling or sovereignty delegation in the institutional structure of the EU.⁹

Sovereignty pooling occurs when states agree to take collectively binding decisions without the possibility of national veto. In the context of the 1996–97 IGC, to pool sovereignty meant to allow qualified majority voting (QMV) in the Council of Ministers for all decisions pertaining to CFSP, or at least for all decisions of implementation. Sovereignty delegation occurs when states transfer some decisional power to supranational bodies. Specifically, in the mid-1990s to delegate sovereignty meant to increase the role and powers of the European Commission and the European Parliament in CFSP, to mitigate the intergovernmental nature of CFSP by unifying the “three-pillared” institutional structure of the EU, and to finance CFSP operations from the Community budget instead of ad hoc contributions by the member-states.

The diversity of government positions challenges the view that the preservation of sovereignty is a basic goal shared by all states. It also calls for an explanation: why are some European governments more willing to pool and delegate sovereignty than others? The next section reviews various possible answers to this puzzle.

Possible Explanatory Factors: Interests, Capabilities, Identities, and Constitutional Cultures

There are several research programs in international relations theory, and each of them stresses one particular set of factors as especially useful for explaining foreign and security policies in general and supranational integration in particular. While it is often assumed that foreign policy decisions require multicausal explanations, there is little clarity about when the predictions of different theoretical approaches should be seen as complementary and when they are competing. In this article I consider a number of explanatory factors that have been highlighted by different theoretical traditions, without trying to determine ex ante how their effects might be combined in each particular case.

Broadly speaking, international institution-building can be interpreted from two perspectives. The first assumes that governments have a purely instrumental attitude toward international institutions. According to this instrumentalist approach,¹⁰

institutions in general and specific aspects of institutional design are not valued per se, but only as means to attain exogenously defined goals. Institutions not only set constraints to strategic action, but they are themselves the object and outcome of strategic action. From the second perspective, the preferences of governments for certain institutions are not based primarily on their anticipated capacity to produce well-defined outcomes, but on their coherence with entrenched beliefs and normative commitments. In a sense, each of the two approaches stresses a distinct logic of action: a logic of expected consequences or a logic of appropriateness.¹¹ In this section I derive from these perspectives several more specific hypotheses concerning the CFSP.

In the instrumentalist perspective, the fact that governments prefer different institutional arrangements can be explained in reference to their varying interests or their varying resources. This consideration forms the basis for selecting the first two causal factors that will be considered in the examination of institution-building in CFSP.

The first causal factor to be considered is the congruence between the anticipated outcomes of a supranational foreign policy on the one hand, and each government's interests in world affairs on the other. The importance of this factor is stressed by rational institutionalism,¹²

which distinguishes preferences regarding substantive policy issues (or policy preferences) from preferences regarding institutional forms (or constitutional preferences) and explains the latter in terms of the former. Governments support or oppose the introduction of certain institutional and procedural rules depending on whether they believe that these rules will produce outcomes corresponding to their exogenously determined interests or not.¹³

In the context of CFSP, EU governments have preferences regarding a number of issues of world politics and can be expected to support those institutional forms that make their preferred outcome more likely to happen. According to rational institutionalism, the main determinant of a government's opposition to supranational institutions in CFSP should be its concern that, once they are introduced, the EU would make decisions that frequently run against the preferences of that government on particular international issues. Specifically, with regard to sovereignty pooling, the concern is that the government might often be outvoted in the Council of Ministers. With regard to sovereignty delegation, a government might be concerned that on specific policy questions, the supranational agencies would be more responsive to the preferences of a majority of member-states while the government itself is in a minority position.

Supranational integration implies the abolition of the national veto, and this might be especially difficult to accept for governments whose preferences can be expected to often be at odds with the view of the majority—that is, for preference outliers. Rational institutionalism expects that (1) the more a government fears to be outvoted, the less likely it is to accept supranational integration, and (2) the fear to be outvoted depends mainly on the extent to which its policy preferences conform to, or depart from, the expected policy preferences of a majority of governments. In this analysis, this aspect is called “policy conformity.”

When choosing institutional rules, of course, governments cannot know for certain what their policy preferences and those of the other governments will be once the rules are operative. However, it would be implausible to think that governments choose rules behind a “veil of ignorance”: because issues in world politics display a certain degree of continuity, the governments' knowledge about past and current preferences is sufficient to form reasonable expectations about future policy preferences. Hence the first hypothesis to be tested is the following:

H1: Governments that expect to be outliers with respect to policy preferences will be less supportive of supranational integration in the foreign and security field than governments with a high level of expected policy conformity.

The second causal factor to be considered in this article is also connected to a consequentialist approach to international politics, but this approach stresses differences in resources rather than differences in interests.¹⁴

Realist scholars in international relations are less interested in the specific content and variation of policy preferences than in the means states have to obtain what they want. While realism comes in many variants, all of them hold that the distribution of power capabilities among states is a crucial determinant of their behavior.¹⁵ A corollary is that differences in the behavior of states can be explained mainly by looking at their relative position in the state system.

Some realist authors, as I have pointed out in the previous section, make assumptions about state goals that are incompatible with the demonstrated willingness of various European governments to deepen their foreign and security policy integration. Other variants of realism, however, do not rule out the possibility of voluntary integration, but seek to explain it by referring to the distribution of power among states. In particular, Joseph Grieco has suggested a “neorealist” interpretation of the EU's Economic and Monetary Union, according to which the institutional arrangement agreed upon at Maastricht reflects the interest of France and Italy in gaining “voice opportunities” vis-à-vis Germany for determining monetary policy in Europe. Grieco's thesis is that, “especially for weaker but still salient states, institutionalization might constitute an effective second-best solution to the problem of working with, but not being dominated by, a stronger partner in the context of mutually beneficial joint action (the first-best solution would be to become more powerful and thus be on more equal terms with the stronger partner, or perhaps even not to require cooperation).”¹⁶

Realist theory leads to the expectation that the international distribution of power affects each government's interest in a common foreign and security policy in a different way. More specifically, governments whose power resources allow them to conduct an independent and effective foreign policy should see no need to relinquish their autonomy and have their hands tied by supranational institutions. Weaker countries, on the contrary, should be interested in an integrated foreign and security policy for at least two reasons. First, these countries can expect their influence on world affairs to increase when the EU acts as a unit. The autonomy they would lose would be offset by the collective power of the supranational polity of which they are a part. Second, a tight institutional structure would be a way to constrain the stronger member-states, whose independent foreign policy might become a threat to the interests of the smaller countries in the future. By this logic, supranational integration enhances the security of smaller states by augmenting their external influence and by constraining potential sources of tension.

The hypothesis deriving from realism's emphasis on relative power resources is thus the following:

H2: Governments with higher power capabilities will be less supportive of supranational integration in foreign and security matters than governments with lower capabilities.

Both explanatory factors considered so far—policy preferences and relative capabilities—suppose an instrumental and calculating attitude toward European institution-building in foreign and security matters. The other two factors considered in this article reflect a logic of appropriateness rather than of consequences. In international relations theory, this logic is generally emphasized by constructivism, which regards identities, culture, and norms as independent determinants of the behavior of actors in the international system. Some constructivists emphasize the processes of identity formation and transformation that derive from interactions among states,¹⁷

while others point to how the identities, values, and cultural attitudes of domestic social groups affect the behavior of their governments.¹⁸ The constructivist hypotheses considered in the rest of this article belong to the latter version of constructivism, which examines the domestic determinants of international behavior and thus takes a “bottom-up” perspective.

Whether constructivists stress the systemic or the domestic sources of state identity formation, they all hold that “variation in state identity, or changes in state identity, affect the national security interests or policies of states.”¹⁹

Furthermore, “changes in the collective identity of societal actors transform the interests of relevant collective actors that constitute the [international] system.”²⁰ Constructivists have described the changes in collective identities and values that occurred in Europe since the Middle Ages, and have argued that these developments substantially altered the way governments related to their external environment.²¹ For instance, Rodney Bruce Hall has argued that the nineteenth-century transition from a legitimating principle based on territorial sovereignty to the principles of national sovereignty and self-determination transformed the interests and practices of states.²²

Many think that since World War II, a further shift in collective identities has occurred in Europe, as nationalism has declined significantly among West European publics.²³

According to some authors, this decline is part of a general trend toward the endorsement of postmaterialist values.²⁴ Many observers of European affairs hold that European integration “has been accompanied by a weakening of exclusive nationalism and by what might be described as multiple identity, that is, the coexistence of identities to local, regional, and supranational territorial communities, alongside an identity with the nation.”²⁵

Asserting that a common European identity is progressively developing in Western Europe is not incompatible with the recognition of substantial differences among countries in this dimension. European identity formation can be promoted or hindered by a number of factors, both at the individual and collective level. The latter might include a diversity of historical experiences, especially those related to wars, and geographical features of countries, such as peripheral location or insularity. These differences can affect the degree to which collective identities are Europeanized.

Constructivists argue that the attitude of national government vis-à-vis European integration is shaped by the way their populations or their elites relate themselves to Europe. A widespread perception of belonging to a European entity, in addition to—or even as a substitute for—national belonging, is considered a major determinant of government policies on supranational integration.²⁶

A strong degree of European identity promotes the perception that authority is “shared” with other Europeans rather than “lost” to foreigners, and thus facilitates the transfer of competencies to the supranational level.

Generally speaking, collective identities might affect government policies toward European treaty reform through two types of causal mechanism. In the first, members of the political elite make choices on European political integration on the basis of their identities. In the second, members of the general public form preferences on European political integration on the basis of their identities, and political elites engaged in political competition adjust their stance toward the EU to what they perceive are the preferences of their potential voters. In the first case, identities influence policy directly; in the second case, their influence on policy is mediated by instrumental interests. In reality, it is plausible to assume that the two causal mechanisms tend to operate simultaneously, generating various patterns of interaction. As Thomas Risse notes, “[p]olitical elites (including party elites) try to promote ideas (including identity constructions) with an eye on gaining power or remaining in government.”²⁷

Political elites are constrained by public opinion, but the latter is malleable to the discourses propagated by the former, as research on public and elite support for European integration has shown.²⁸

The hypothesis deriving from the constructivist emphasis on collective identities is the following:

H3: Governments of countries with strong [mass/elite] identification with “Europe” will be more supportive of supranational integration in foreign and security policy than governments of countries with less European identification.

The text in brackets indicates that two versions of the hypothesis should be tested: one concerning the general public and the other concerning elite identities.

The balance between national and European identity may not be the only component of political culture that affects the willingness to create supranational political structures. Ideas about how the exercise of public authority is best organized might also exert an influence. Some scholars believe that “where states have an opportunity, they will seek to create international rules and institutions that are consistent with domestic principles of political order.”²⁹

For instance, Daniel Deudney has hypothesized that, when polities with multiple centers of authority—or “Philadelphian systems”—cannot avoid interacting with other polities, they find a policy of “binding”—that is, a reciprocal limitation of autonomy by means of institutional links—more “congenial” than balance-of-power practices.³⁰ Similarly, Anne-Marie Burley has argued that post–World War II multilateral institutions are projections of the U.S. New Deal regulatory state.³¹ Some proponents of the “democratic peace” thesis argue that democratic states tend to “export” the principle of peaceful conflict resolution when dealing with each other.³²

In the context of the EU, it has been noted that adaptation to supranational integration is easier for some countries than for others. In particular, the institutional structures of the EU strongly resemble those of the Federal Republic of Germany, ensuring a high degree of “congruence” between the two constitutional orders.³³

It can plausibly be argued that this “strikingly good fit”³⁴ makes European political integration more welcome to the German political and administrative elites than to other Europeans, such as the British, whose domestic state structures are considerably different from the EU. Wolfgang Wagner has shown that, during the negotiations on the Maastricht Treaty, the governments of countries with entrenched parliamentary representation at the regional level wanted a stronger European Parliament, whereas the countries with no regional parliaments were opposed.³⁵

Based on these considerations, the fourth causal factor examined in this article is the constitutional culture that prevails in a country, and specifically the legitimacy and practice of multilayered governance in the domestic context. Two typical constitutional cultures can be envisaged from this perspective. One conceives sovereignty as a unitary and indivisible attribute of a polity. With reference to France, Stanley Hoffmann has pointed to “the tendency to look at sovereignty not as a bundle of discrete powers and separable state functions, but rather as a talisman, indivisible and inalienable yet eminently losable.”³⁶

Similarly, J. C. D. Clark notes that “the United Kingdom's dynastic unification, crucially combined with the union of Church and State at the Reformation, endowed her with a unitary, absolutist doctrine of sovereignty, … within which the concept of ‘sharing sovereignty’ became a contradiction in terms.”³⁷ This idea of a “supreme, irresistible, absolute, uncontrolled authority, in which … the rights of sovereignty, reside” (as Sir William Blackstone's Commentaries formulated it) is still at the basis of the British legal system.³⁸ This conception is generally hostile to the vertical division of powers, which is seen as a threat to the integrity of public authority. According to a different type of constitutional culture, public power can be, and should be, distributed among multiple territorial levels, in accordance with the principle of subsidiarity or comparative efficiency. In the first type, the exercise of sovereign power is concentrated and “monocentric;” in the second type it is diffuse and “pluricentric.”

The fourth hypothesis considered in this article is thus:

H4: Governments of countries where a monocentric conception of political authority prevails will find it more difficult to accept a supranational transfer of powers in the CFSP than governments of countries where the pluricentric constitutional culture is stronger.

In other words, a constitutional culture that emphasizes multilayered governance within the boundaries of the state predisposes its leaders to accept and promote the creation of an additional layer of governance, if this seems to increase policy effectiveness. This predisposition facilitates institutional isomorphism. The idea of indivisible sovereignty, on the contrary, creates a conceptual obstacle to supranational political integration.

It is important to note that the analyses below are not based on measures of “culture,” but on an index measuring domestic institutional structures. In a later section, I will argue that this index is suitable for assessing the hypothesis formulated above. The validity of my conclusions about the impact of constitutional culture will depend crucially on the plausibility of that argument.

In this section I have identified four factors that might plausibly explain the variance in government support for a supranational foreign and security policy. Each of these factors is stressed by a different “school” in international relations theory, but this does not exclude the idea that they can complement each other in explaining governmental preferences concerning political integration.

Most importantly, each causal factor is formulated in such a general way that, taken together, they encompass a large number of more specific explanations advanced in the literature. For instance, national characteristics such as neutral status and a “special relationship” with the United States can be subsumed under preferences over policies; and factors such as the consequences of a peripheral or insular location, of historical experiences and memories, and socialization effects deriving from EU membership can be subsumed under the question of European identity. An explanatory framework that focuses on the four variables discussed above can hope to grasp many of the relevant differences between the member- states. Even so, the expectation is that the variables considered can explain a substantial part of the variance—not the total variance.

Explanatory Strategies

The previous section has identified a number of possible influences on the willingness of European governments to establish a supranational foreign and security policy. The hypotheses, however, have been formulated in such a general way as to leave unspecified the character of the causal links. Before the hypotheses can be tested empirically, this character must be made explicit. This article considers two possible interpretations of the causal link concept. According to the first interpretation, the causal factor increases the probability of the outcome (net of the effect of other relevant causes). According to the second interpretation, the cause (by itself or in combination with other causes) represents a necessary and/or sufficient condition for the outcome to occur. The first understanding underpins most quantitative research. The second understanding underpins, sometimes implicitly, many qualitative comparative, “small-N” investigations.³⁹

In this article I posit that both approaches are legitimate for the study of the problem at hand.⁴⁰

Therefore, in the following I will ask two related questions with regard to the four explanatory variables considered above: (1) Do these variables affect the probability that a government is supportive of a supranational reform of the CFSP, ceteris paribus? and (2) Are any of these causal conditions necessary and/or sufficient for supranationalism, by themselves or in combination with other causal conditions?

This article aims to answer the first question using multivariate logistic regression. Correlational methods, however, might not be equally suited for answering questions of the second kind, which are expressed in logical terms. Sufficiency and necessity can be conceptualized as set-theoretic relationships, and for this reason I will examine the second research question using a novel method that has been developed by Charles Ragin and is based on fuzzy-set logic.⁴¹

Both explanatory strategies involve a comparison between the member-states of the EU, although the logic of comparison is different.⁴² Moreover, both types of cross-national comparative investigation are intended as a complement to, not as a substitute for, detailed case studies at the country level.⁴³

Logistic regression is used frequently in the analysis of social and political data and does not require an introduction. Ragin's fuzzy-set method is less well known, and thus a brief description is appropriate. The method is a formal tool for qualitative comparisons that builds on a previous technique, Qualitative Comparative Analysis (QCA),⁴⁴

which is based on Boolean algebra and has been applied to various topics in political research, including international alliances, coercive diplomacy, revolutionary movements, the breakdown of democratic regimes, and the development of welfare states.⁴⁵ Similar to its predecessor, fuzzy-set analysis is specifically designed to deal with situations in which the number of cases is moderate (between about 5 and 50) and causality is multiple (different causes can generate the same outcome) and conjunctural (the impact of a factor on the outcome depends on its interaction with other factors—that is, what matters are combinations of causes).

Compared to QCA, the main strength of fuzzy-set analysis is that it is not restricted to situations in which the outcome and the causes are either present or absent; it can also be applied to observations displaying more or less of a certain property. This is because fuzzy sets embody both qualitative states (full membership and full nonmembership in a set) and variation by level (degrees of membership between 0 and 1). In addition, fuzzy-set analysis can incorporate a probabilistic interpretation of causality and take into account statistical significance.⁴⁶

Compared to quantitative methods, fuzzy-set analysis is designed for the discovery of patterns of causal necessity and sufficiency in intermediate-sized N's. It is very sensitive to causal complexity, that is, multiple and conjunctural causation. The use of interaction terms is subject to stringent limitations in most quantitative methods, whereas fuzzy-set analysis starts by examining all possible combinations of causal conditions (saturated interaction) and then simplifies the model in a top-down manner.⁴⁷

In a nutshell, researchers using fuzzy-set QCA do the following. First, they use substantive and theoretical knowledge to assign fuzzy-set membership scores to cases in both the outcome and the causal conditions in which they are interested. Second, they look for causal conditions with membership scores that are consistently greater than outcome membership scores. If there is a causal condition in which this happens in all cases, then this condition passes the test of necessity. Third, to examine sufficiency they compare the membership scores of the outcome with the score of all logically possible combinations of the causal conditions, and they look for combinations with membership scores that are consistently lower than outcome membership scores. If there is a combination for which this happens in all cases, then this condition passes the test of sufficiency. Researchers eliminate those expressions that pass the test of sufficiency but are logically redundant and obtain a logically parsimonious statement of causal sufficiency. This statement is then evaluated in terms of any simplifying assumption that it incorporates (simplifying assumptions are statements about the hypothetical outcome of combinations of causal conditions that do not occur in the population studied).⁴⁸

Operationalization and Data

In this article, logistic regression and the fuzzy-set method are used to analyze evidence collected from all the states that were EU members in 1996. This requires an operationalization of the causal factors that is sufficiently precise to allow comparisons between fifteen countries and that takes into account the difficulty of collecting in-depth homogeneous data for such a range of countries.

The first task is to operationalize the outcome variable, that is, the preferences of governments concerning the degree of supranationalism in their foreign and security policy cooperation. In most areas of political research, to identify the preferences of actors is more difficult than to ascertain their behavior, which is generally visible. To determine the preference ordering of collective actors such as national governments is even more difficult, given their composite nature and the possibility of internal disagreements and even conflicts. In this article, the problem of determining preferences is solved by relying on the public statements that the governments themselves issued to illustrate the position they intended to promote at the Intergovernmental Conferences for the institutional reform of the EU. The audiences for these statements were their parliaments and publics, as well as the governments of other member-states. The prenegotiation phase of the IGC held in 1996–97 seems a useful context to elicit these positions, as all fifteen governments prepared and publicized papers outlining their stance on a broad range of issues expected to be on the negotiating table.⁴⁹

As noted above, the position of each member-state on the most important treaty revisions concerning CFSP is summarized in Table 1. The explanation of how the information contained in that table provides the input (outcome variable) for the analyses is given in the next section, as the two methods employed differ in this respect.

The assessment of the first explanatory variable, “policy conformity,” requires a way to identify governments that tend to be preference outliers with respect to various international problems. Here I use voting behavior in the UN General Assembly as a proxy, because UN delegations vote on a number of issues of world politics that might be the topic of CFSP decisions. In fact, about 95 percent of roll-call votes in the period 1991–96 fell into four categories: disarmament, colonialism, the Middle East, and human and political rights (including the conflict in former Yugoslavia).⁵⁰

These issue areas either already belong to the CFSP's remit,⁵¹ or governments can reasonably expect them to be added to the EU's competencies in the future.⁵² I considered the five sessions held between the end of the Cold War and the start of the IGC of 1996–97 (Sessions 45th to 50th). During this period, the General Assembly adopted 434 resolutions by recorded or roll-call vote, and in 176 cases the vote of EU member countries⁵³ was not unanimous. For each of these 176 “contentious” resolutions, I determined how each EU member- state voted (three possibilities: “yes,” “no,” or abstention) and how many other member-states voted in the same way (“partners” henceforth).⁵⁴ Since in 1996 there were fifteen EU members, this number can range from a minimum of zero (the state is isolated) to a maximum of thirteen (the state is part of a majority of fourteen against one dissenter).⁵⁵ Each country is given a conformity score, which equals the average number of partners it had in the 176 resolutions. A low average is an indication that the state is a preference outlier, while a high average indicates that its policy preferences generally coincide with those of a majority of EU member-states.

The average number of partners, or “likeminded” states, determines the likelihood that a government would have its preferences overrun in hypothetical votes in the EU. This proxy certainly has several limitations. For instance, it can capture the risk, but not the cost, of being outvoted, because the latter depends on the relative importance of the issues for the governments. However, short of an indepth analysis of the foreign policy preferences of each EU government, this approach would seem the best way to systematically compare the disadvantage of supranationalism that each state has to bear because of disagreements about the best way to respond to international problems. This proxy, for instance, would indicate that a country such as Italy (average number of partners: 9.9) can be almost certain to be part of either a majority or a blocking minority, while the United Kingdom (5.3) would be much more uncertain about its ability to thwart undesired decisions in case of a vote.

To measure the second causal factor—power capabilities—I rely on the Composite Index of Material Capabilities (CIMC) developed by the Correlates of War (COW) Project, which is probably the most commonly used power index in the international relations literature. The CIMC results from two demographic indicators (total population and urban population), two industrial indicators (energy consumption and steel production), and two military indicators (military expenditures and armed forces size). For each of these indicators, the COW research team calculates the total score (in people, tons, dollars, etc.) for the international system in a given year, ascertains the percentage share held by each state, and calculates the average of these percentage shares.⁵⁶

In this article, however, I consider each state's share of the EU total, because I am interested in the differences in power among member-states only. The CIMC used here is for 1992, the last year for which data on EU countries are currently available in the National Material Capabilities Data set.⁵⁷

In comparison to other possible indicators of national power capabilities, the CIMC has the advantage of expressing economic power as well as military power, and potential military power (expressed by the economic indicators) as well as actual power. The inclusion of the economic dimension is useful because economic coercion and incentives are important tools of foreign and security policy even when the possibility of using military force remains essential. Considering potential military power is important to do justice to the neorealist position according to which structural factors ultimately govern international politics: in this logic, potential capabilities are crucial because states will sooner or later transform them into actual military capabilities as a result of the competitive nature of the international environment (a state that fails to do so is a clear anomaly from the point of view of neorealism).⁵⁸

However, the results reported in the next section are robust alternative measurements of national capabilities.⁵⁹

Concerning my third causal factor, the study of national and supranational identity is occasionally conducted by means of discourse analysis.⁶⁰

However, the state of the art in EU studies does not yet allow the use of this sophisticated tool to compare all EU member-states. Another common way to operationalize European identity is to rely on the results of opinion polls. The standard source on public attitudes toward European integration is the Eurobarometer survey, which is conducted periodically in all member-states on behalf of the European Commission. The survey question that will be used in this article is the following: “In the near future do you see yourself as … ?” where five answers are possible: (1) Austrian/Belgian/Danish/etc. only, (2) Austrian/Belgian/Danish/etc. and European, (3) European and Austrian/Belgian/Danish/etc., (4) European only, and (5) don't know. For each country, I construct a “European identity score,” which equals the sum of the percentages of all respondents indicating “European” as (part of) their identity minus the percentage of respondents indicating “[nationality] only.” I pooled the data from four different Eurobarometer surveys, which were conducted between March 1992 and December 1995.⁶¹

Two different European identity scores were computed: one for all respondents and one for the respondents that scored highest on the “opinion leadership index” developed by the Eurobarometer investigators.⁶²

Opinion leaders are a broader category than political elites, but their responses can be considered fairly good approximations of elite positions, as they are “the political stratum closest to the political elites.”⁶³

A comparison of the scores of the general public and the scores of the opinion leaders reveals three things: (1) In every country, the European identity score of opinion leaders is positive (that is, the number of respondents feeling European is larger than the number of “nationals only”), while in five out of fifteen countries the score of the general population is negative (that is, the respondents not feeling European outnumber those who do). (2) In every country, the opinion leaders score higher than the general population (the average difference is 23 points, with a minimum of 11.1 and a maximum of 33.8). This pattern clearly matches the results of a number of studies on public support for European integration.⁶⁴

(3) Cross-nationally, the opinion leaders' scores and the mass scores are almost perfectly correlated (Spearman's rho = .97). In other words, high levels of opinion leaders' identification with Europe correspond to high levels of mass identification, and vice versa. Because of this high correlation, two different models have been estimated in the next section: one with the scores of opinion leaders and the other with the scores of the general public.

In order to assess the fourth causal factor, it is necessary to estimate whether the constitutional culture of a country is closer to the monocentric or to the pluricentric ideal type. The actual institutional structure of the polity can be considered a proxy of that culture, as it is plausible to assume a broad correspondence between the predominant views on the legitimate distribution of public authority and the rules concerning regional governance in each country. Liesbet Hooghe and Gary Marks have developed a twelve-point index of regional governance in EU countries, based on the degree of constitutional federalism, the presence of special territorial autonomy, the role of regions in central government, and the existence of regional elections.⁶⁵

In the following analysis, I use the scores they assign to each EU member-state for 2000, with one modification: one of their dimensions, special territorial autonomy, is left out because it is often seen as an exceptional solution to a specific political problem rather than the institutional consequence of a distinctive way to conceive political authority. Later in this article, I elaborate on why constitutional culture is the most plausible causal mechanism linking subnational regional governance and supranational orientation in the case of CFSP reform.

This completes the description of the data that will be used in measuring the five variables involved in the following analyses. For the logistic regression analysis, no further manipulation of the data is necessary: they will be used as input as they appear in the original data sets.⁶⁶

The fuzzy-set analysis, however, requires the transformation of the raw data into fuzzy-set membership scores.

In the following fuzzy-set analysis, the relevant population consists of thirteen member-states of the EU that participated in the IGC of 1996–97. Ireland and Sweden have not been included because of too little information about their positions. On the basis of the data presented above, I have assigned to each EU member-state a fuzzy-set score in the outcome and in the four causal conditions. These scores are listed in Table 2. With regard to the outcome, the governments that supported all treaty revisions listed in Table 1 are considered fully in the “set of supranationalist governments” and consequently are assigned a fuzzy-set score of 1. The governments that rejected all treaty changes are considered to be fully out of the set of supranationalist governments and receive a fuzzy-set score of 0. The other governments are assigned intermediate scores exactly in proportion to the number of treaty changes they supported or opposed.

With regard to the causal conditions, I have assigned fuzzy-set scores to countries according to a standardized criterion: the country with the lowest value on each variable (as recorded in the data sources presented earlier in this section) is assigned a fuzzy-set membership score of 0, the country with the highest value is assigned a fuzzy-set score of 1, and all other countries receive intermediate scores.⁶⁷

Two comments on this procedure are necessary.

The first concerns the decision to retain all the variation to be found in the raw data. According to Ragin, one of the advantages of fuzzy sets over conventional ratio or interval scale measures is the possibility of eliminating variation in the data that is not meaningful from the point of view of the research question. For instance, if a researcher is interested in how democratic states conduct wars and has identified a threshold beyond which states can be considered fully democratic, then it can be advisable to ignore any further variation in levels of democracy beyond that threshold.

In the case of the four causal conditions examined here, however, no variation can be declared clearly irrelevant on the basis of prior information at the case level or theoretical considerations. In such situations, any truncation of the original data runs the risk of being arbitrary and controversial. The safest option is to retain all naturally occurring variation and invite readers to replicate the analysis with different fuzzy-set scores if they believe that more appropriate breakpoints can be identified.

The second comment concerns the countries that are classified as fully in or fully out of the set of countries with a relevant characteristic. I maintain that the countries that are assigned a score of 1 or 0 because they display the maximum or minimum value on the four variables correspond to the countries that can be considered fully in or fully out of each set on the basis of other substantive and theoretical knowledge. Because space constraints prevent a detailed discussion of those countries, the following offers a quick overview. Germany has been described by structural realists as the potentially dominant power in Europe⁶⁸

and can be considered to be fully in the “set of European countries with high material capabilities.” Conversely, Luxemburg can be seen as fully out of that set. On the other hand, Luxemburg and Italy can be seen as fully in (or virtually fully in) the “set of countries with Europeanized identities,”⁶⁹ while Britain can be considered fully out because “the prevailing English identity still perceives Europe as the (friendly) ‘other.’”⁷⁰ Germany is fully in the “set of countries with strong regional governance,” while a number of member-states do not display any federal features.⁷¹ Finally, Luxemburg can be seen as fully in the “set of policy-conformist countries,” because it lacks virtually all characteristics that could set it apart from most of its European partners: it has no “special relationship” with a non-EU power, no links to former colonies, no special geopolitical interests in other regions, and no military personnel permanently stationed abroad. The United Kingdom has opposite characteristics (notably a strong tendency to support U.S. policies) and, consequently, many opportunities for dissenting from its EU partners: for this reason it can be considered fully out of the set of conformist countries. In sum, I contend that the membership scores obtained through the standardized procedure described above, including those of full membership and full nonmembership, are corroborated by additional substantive and theoretical knowledge. Overall, the following analysis reflects the best estimate of membership scores by the present author.

Logistic Regression Analysis

The units of observation of the regression analysis are the governments' decisions to support or oppose specific treaty changes that were on the agenda of the 1996–97 IGC. As indicated above, six possible changes were particularly relevant for the creation of a supranational CFSP. The outcome variable is dichotomous (support of or opposition to the treaty change) and the number of observations is seventy.⁷²

The four hypotheses articulated earlier in this article can be reformulated with greater precision: the probability that a specific treaty change is supported by a government increases when (1) the policy conformity of the government is higher, (2) its level of material capabilities is lower, (3) European identification is stronger, and (4) domestic regional governance is stronger. These hypotheses are tested by estimating logistic models with robust standard errors and clustering by country.⁷³

Given the non-independence of same-country observations, the analysis is best seen as exploratory rather than definitive. Table 3 presents the results of the models.

Overall, the models perform well, predicting about 93 percent of all cases correctly, which is an 80 percent reduction in error from the null model that predicts about 64 percent correctly. The explanatory variables account for about 68 percent of the variation in outcome. Considering the individual variables, all have a statistically significant impact on the likelihood that a government will support a supranationalist treaty change, with the exception of policy conformity. That impact is in the expected direction.

To facilitate the interpretation of logit coefficients, I use statistical simulation to convert them into probabilities and confidence intervals, which provide a more intuitive illustration of the relationship between the variables.⁷⁴

Table 4 reports these probabilities and confidence intervals for three explanatory variables (policy conformity is not included as it did not reach statistical significance in the models). To save space, only the results for model 1 are reported, as the results of model 2 are similar. The values in the second column of Table 4 express the average probability that a particular treaty change is supported when one explanatory variable is set at its maximum or minimum in the sample, while all other explanatory variables are fixed at their mean. The values in the third column indicate the range in which 95 percent of the predicted values generated by the simulation fell.⁷⁵ The quantities in Table 4 represent counterfactual statements, indicating which probability of support could be expected in hypothetical cases that display certain values of the explanatory variables (that is, one variable at the maximum or minimum value and all others at their means).

Table 4 shows that, all else being equal, as material capabilities decrease from the maximum value found in the data (that of Germany) to the minimum value (that of Luxemburg), the probability of support rises from 0 to almost 1. In other words, the absence of support is certain when capabilities are at their highest, and the presence of support is virtually certain when they are at their lowest. Extremely narrow confidence intervals confirm that the margin of error is very low. Similarly, shifting the level of regional governance from its minimum (that of Ireland and other countries) to its maximum (Germany) increases the probability of support from almost 0 (certain absence) to 1 (certain presence), ceteris paribus. Again, extremely narrow confidence intervals indicate that error is very unlikely. Finally, all else being equal, as European identity increases from its minimum value (corresponding to Sweden) to its maximum value (Luxemburg), the probability of support rises from 0.59 to 0.99. This effect is smaller than the impact of capabilities and regional governance, but still substantial. The confidence intervals of the minimum are large: this means that, while one can be confident about the existence and direction of the relationship, one is less certain about the precise magnitude of the effect of European identity.

Fuzzy-Set Analysis

This section reports the results of the application of the fuzzy-set method to the scores in Table 2.⁷⁶

As in the previous section, two separate analyses are conducted, the first considering the European identity of the general public and the second considering the European identity of opinion leaders only (the other possible causal conditions remain the same).

The first result is that for none of the four causal conditions, fuzzy membership scores in the outcome were in all cases less than or equal to fuzzy membership scores in the causal condition. This suggests that none of the causal factors examined is necessary for supranationalism. This applies regardless of whether opinion leaders' identity or mass identity is considered.

The second result is that, after algebraic simplification, the analysis of sufficiency yields the following solution: there is one combination of causes that is sufficient for supranationalism—the combination of regional governance and policy conformity. In fuzzy-set notation, this solution can be expressed as follows:

where the symbol [bull ] indicates the logical “and,” and → means “is sufficient for.” Again, this result applies regardless of whether leaders' opinion or mass opinion is considered.

It should be noted that the sufficiency solution just reported can be described as conservative, because it does not use any “simplifying assumptions.” In QCA, simplifying assumptions are statements about the hypothetical outcome of combinations of causal conditions that do not occur in the population studied. They are a reflection of the limited diversity of naturally occurring social phenomena. On the basis of substantive knowledge about the object of study, the researcher may assume that a given combination of causal conditions, if it had occurred, would have been sufficient for the outcome. If such an assumption is included in the analysis, the result may be different than it would have been otherwise.

Various simplifying assumptions have been considered in the present analysis, but only one of them seems plausible enough to justify its inclusion in the final solution: the assumption that the combination of regional governance, low capabilities, strong European identity of the general public, and low policy conformity is sufficient for supranationalism.⁷⁷

This assumption seems plausible because the presence of the first three factors named, which are considered favorable to supranationalism on the basis of both theory and the preceding regression analysis, should be able to offset the single opposing factor (low policy conformity). If one incorporates this assumption into the analysis, two combinations of causes pass the test of sufficiency:

where the symbol + indicates the logical “or” and ∼ indicates the negation of a causal condition. In plain English, this means that either the combination of regional governance and policy conformity, or the combination of regional governance, European identity of the general public, and low power capabilities is sufficient for supranationalism. This indicates that there are two paths to supranationalism, both involving regional governance. Regional governance appears to be sufficient for supranationalism if it is combined either with policy conformity or with two other facilitating conditions: strong European identity of the general public and low material capabilities. If one looks at the European identity of opinion leaders rather than that of the general public, only the first causal combination appears sufficient.

I have shown that the main result of the fuzzy-set analysis is that the combination of regional governance and policy conformity is sufficient for supranationalism. Figure 1 presents this result in graphical form: for each one of the thirteen governments examined, it is the case that its membership in the fuzzy set of supranationalist governments (y-axis) is at least as high as its membership in the intersection of regional governance and policy conformity (x-axis).