Does the European Union
an n of 1?
Written especially for the ECSA Review Vol. X, No. 3 (Fall 1997), pp.1-5.
Four ECSA members tackle the debate over whether or not the European Union is a unique case, and whether it matters.
James A. Caporaso
BEFORE TURNING TO the central question of this forum, it may be useful to remind ourselves that the study of regional integration began on a distinctly comparative and historical note. The work of Haas and Deutsch, the main co-founders of regional integration studies in the United States, strikingly illustrate this point. In Political Community and the North Atlantic Area (1957), Karl Deutsch and his collaborators undertook a historical-comparative study of state and community formation, drawing on the experiences of multinational empires such as the Austro-Hungarian Empire, nation states such as the United Kingdom, Switzerland, Germany, and Italy, and international organizations such as the North Atlantic Treaty Organization. Similarly, Ernst Haas, who is remembered mostly for The Uniting of Europe (1958), also wrote extensively about the Nordic Council, the North Atlantic Treaty Organization, the Western European Union, the Council of Europe, and the European Free Trade Association. The title of his essay, "International Integration: The European and the Universal Process," highlights Haas' search for similarities and differences between integration in Europe and the rest of the world. Philippe Schmitter extended the evolving integration framework to Latin America. Finally, Joseph Nye, in Peace in Parts (1971), set out a conceptual framework for the study of comparative regional integration and then proceeded to apply the framework to Africa, the Middle East, Latin America and Europe (East and West). Early integration theorists, while certainly aware of important contextual differences among regional units, did not clamor for "sui generis theory."
Yet, as the European Community (EC) developed, thickened its institutional base, expanded the scope of its policy competences, and in general became more complex, it also came to be studied more narrowly, in isolation from other regional integration processes and some would argue in isolation from social theory itself. Which brings me to the questions posed for this ECSA Review forum: Does the EC represent an n of 1? Is it unique? If so, does it require a theory of its own?
What is the case for uniqueness? It is possible to argue that the processes of integration in Europe are specialized, and qualitatively different from processes elsewhere. The historical thrust of the EC is so novel that it truly represents a Hegelian moment, a novelty that, however prescient in terms of future developments, has no current analogies. If this is so, it is best to describe these developments as accurately as possible, and not to engage in the futile task of shoehorning them into some preconceived theoretical framework.
How can one respond to this position? The first point, drawing inspiration from language philosophy, argues that unique or general are not properties of phenomena in the empirical world, but properties of the language we use to describe them. Few things resist being described one way or the other, depending on the level of generalization used. To say that the EC is unique is simply a shorthand for saying that we have not yet developed the categories, abstract enough, to see the EC as an instance of a more general class of phenomena. Nevertheless, generalization by itself, without an improvement in explanatory capacity, would be a hollow victory. Thus the central question is, "What categories are abstract enough to generate comparable cases, and not so general that they prevent useful comparisons?"
Accepting the above, we can now ask what prevents us from developing general theories that incorporate the EC. I argue that three kinds of differences among regional organizations are regularly treated as obstacles to general theory, that several or all of these conditions are often confounded, and that this retards generalization more than is warranted. These three types of differences are descriptive differences among regions, parametric differences (i.e. in observed relationships), and theoretical differences. I will make some comments about each type of difference, argue that the first and second should not concern us at all, and that the third category is worth extended discussion.
Descriptive differences among regional organizations abound. The conditions surrounding integration in the EEC, the Council on Mutual Economic Assistance, the Organization of African Unity, the Arab League, the Central American Common Market, and the East African Common Market are manifestly different. There were and are important differences with regard to level of economic development, societal pluralism, autonomy of key interest groups, types of economies being integrated, and the role of economic and governmental elites. If differences such as these rule out generalization, then generalization clearly is not possible. Few scholars would argue that this is the case. Descriptive differences simply provide important variation to be explained. They are the raw material of explanatory theory.
Parametric differences are differences in functional relationships among variables across regions, differences not merely in facts, but in how those facts are organized into lawful relationships. Thus, in Western Europe the process of integration was led by key interest groups and statesmen, while in Latin America technocrats were more important. Or in Western Europe, the key integrating sectors were those with a high level of autonomy from government, while in certain less developed regions integrating sectors were under strong governmental control. These statements seem to strike at the core of generalization since they undermine the invariance of relationships among variables. However, parametric differences can also be treated as "data," i.e. observed variation that needs to be explained. The only difference is that there the variation lies in the strength, direction, and even quality of the functional relationship, rather than in each variable per se. By treating these differences as something to be explained, we supply a bridge between idiographic and nomothetic positions, a position espoused long ago by Adam Przeworski and Henry Teune (1970).
Now suppose that observed differences across systems cannot be accounted for by the stock of theories at our disposal. In other words, from the standpoint of our present knowledge, we simply cannot explain the variation across different regions. How do we respond? I admit the answer to this question is partly a matter of faith, or if consciously worked out, of deeply-rooted philosophical beliefs. If we believe that socio-political reality is contextual and that generalizations are bounded by historical-comparative contexts, we will be content to let these differences rest. If we believe that differences in context can themselves ultimately be explained, and that such differences reflect "merely" the state of our knowledge, we will continue to search for even more general theory. My position is the latter. One reason for staking my bets on the nomothetic view is that I don't think there is any way we can know what we are capable or incapable of knowing in advance (how can we know what we can't know?). To argue otherwise is to advance a very dubious epistemological claim. If this is true, it makes sense to keep up pressure to develop even more general explanations.
My main argument has been that treatment of the EC as a special case has been driven mostly by disciplinary pressures, the increasing academic division of labor, and the growing complexity of the EC itself, rather than by explicit philosophical argument. Indeed, our philosophical priors have not been put on the table. And it is philosophical commitments, rather than empirical or even theoretical beliefs, that are at issue. If the EC is sui generis, this argument should be advanced in explicit terms, so that the debate can be joined.
James A. Caporaso is professor of political science at the University
of Washington and past chair of ECSA.
THE EUROPEAN UNION is the most complex, densely institutionalized and authoritative supranational regime in the world. It is unique in many respects. But its uniqueness does not invalidate our efforts to understand it from a comparative perspective. As Max Weber argued in Wirtschaft und Gesellschaft (1925), every social phenomenon may be viewed from the standpoint of its unique character and as raw matter for generalization. Because the European Union (EU) is an unusual regime, comparison between it and other political arrangements can be particularly revealing. By comparing phenomena that share similarities along causally relevant dimensions, we open up the possibility of systematically controlling for contending causal influences.
The bulk of comparative work on the EU does not compare it with anything else, but conceptualizes a wide variety of lower level units for comparison--e.g., social movements, political parties, interest groups, policy sectors, policy networks, legislative decisions. This strategy multiplies units for comparison by slicing polities into subsystemic parts. At the Seattle ECSA meeting, almost all comparative papers pursued such a strategy. A far less common approach--which I will defend here--is to compare the EU at the macro level with other regimes or polities. Such comparison was a chief preoccupation of first and second generation EU scholars, including Karl Deutsch, Ernst Haas, and Philippe Schmitter, who explicitly thought of European integration as part of a larger universe of integrative processes in other parts of the world.
Does such comparison make sense today, when the EU has developed in ways that are different from any other regional regime or domestic polity? This question is prior to issues of the strengths and weaknesses of small-n comparison, for only after one has determined the comparability of social phenomena and, hence, criteria for selecting units for comparison, does one need to discuss methods of comparison.
If one wishes to compare the EU as a whole with something else one must use concepts that stretch across the relevant cases. But as Giovanni Sartori pointed out in his 1970 essay in American Political Science Review, there is a trade-off between conceptual generality and substantive content. In order to generalize about social phenomena, one needs to abstract, but in order to say something meaningful one's concepts should have transparent empirical referents. This tension lies at the core of the n = 1 issue, but it is not insoluble. There are several ways to conceptualize the EU meaningfully as part of a larger universe of cases.
First, and most obviously, the EU can be conceptualized as an international regime, or more specifically, as an example of a regional regime oriented to economic integration. As a regional regime, the EU may be compared to other free trade areas and customs unions, such as NAFTA, Mercosur, or ASEAN. There are currently around fifty such regimes around the world. A common misperception is that on a scale of supranationality, the EU is the only regime that scores positively. But NAFTA, Mercosur, and the WTO each exercise some supranational authority in relations with individual member states and can be placed along a continuum with the EU. Such comparison lends itself to investigation of the relationship between level/type of economic integration and level/type of political integration. To what extent (if at all) does economic integration give rise to pressure for political integration?
More recently, some writers have begun to conceptualize the EU as a polity, i.e., a regime responsible for authoritative decisions concerning the allocation of values in a society. This opens up a variety of comparative perspectives. As a multi-level polity, the EU may be compared to other polities in which authority is dispersed among constituent governments at two or more levels. A variety of federal and confederal polities--Switzerland, Germany, Canada and the USA chief among them--share this characteristic. The EU is more diverse than any of these polities, but once again, the issue for comparison is whether one may conceptualize the differences as variations along some meaningful underlying dimension.
As a process of fundamental institutional change, European integration may be compared to previous reallocations of authority, from the diffusion of authority in the break-up of the Carolingian Empire in the ninth century to state building. From this standpoint, one might attempt to explain two-sided shifts in authoritative competencies from central states to subnational and supranational governments in Western Europe over the past forty years by comparison both across space and across time.
Each of these perspectives poses multiple cases for systemic comparison. In this field, as in many others in the social sciences, even the best research design will not allow one to control for the range of plausible causal influences in explaining outcomes. In none of the examples above is it sensible to assume that the cases are independent. Although n does not equal one, the number of comparable cases is unlikely to be great enough to give one sufficient degrees of freedom to control for all, or even most, relevant variables. In short, macro comparison of the EU is imprisoned in a small-n world in which statistical controls are impossible or of limited value. But the explanatory leverage of qualitative comparison is potentially great, and here one may hook into the ongoing debate about case study methods among John Goldthorpe, Charles Ragin, Robert Keohane, Sidney Verba and others.
Comparison is entirely feasible even assuming that the EU is unique. What matters is second order similarity, that is, the existence of underlying dimensions on which one may place the EU alongside other cases. The key questions are then whether these dimensions are connotatively precise and descriptively meaningful and whether they provide a useful basis for generalizing. The n = 1 issue is really a red herring, for the goal of comparison is to find intelligible patterns of commonality beneath apparent diversity.
Gary Marks is professor of political science at the University
of North Carolina at Chapel Hill, and current chair of ECSA.
IF THE EUROPEAN Union (EU) is a unique, sui generis political institution, how can social scientists generalize about its dynamics? Which theories are appropriate? How can sufficient empirical variation for testing be generated?
The skeptic's response is to reject generalization outright. In their classic article, Wayne Sandholtz and John Zysman reject "tests of competing explanations" aimed at determining "which one is better" as engendering "a false sense of scientism." Sandholtz and others have since reiterated this stance even more forcefully, linking it to "neo-functionalist," "historical institutionalist" and "multi-level governance" interpretations of integration, which focus on unintended consequences and endogenous causes in complex systems that undermine our ability to generalize.
Such skepticism is respectable but in my view exaggerated. The "n = 1 problem" is not unique to the EU; it is a foundational characteristic of social science. Unlike collisions among elementary particles, complex social interactions are in fundamental ways unique. Yet useful theories and results in international and comparative politics exist because scholars have circumvented the n = 1 problem by employing "quasi-experimental" methods: Alternative hypotheses derived from general theories are subjected to potential empirical disconfirmation within a research design that controls for certain variables and permits others to vary. The number of cases can be increased by judicious comparison or by disaggregating cases to isolate comparable elements, either within EU politics or between EU cases and others-the latter a method of qualitative inference recently elaborated by Gary King, Robert Keohane and Sidney Verba. Skeptics offer no fundamental reason why such methods are inappropriate to the study of the EU.
Hence the n = 1 problem is not a counsel of despair, but a challenge to be more self-conscious and rigorous in the selection of proper theoretical and methodological tools. The theoretical challenge is to select theories appropriate to the scope and nature of the phenomenon being studied. The methodological challenge is to generate sufficient observations to confirm or disconfirm competing causal hypotheses about that phenomenon.
Theoretically, EU studies have made much progress in this direction. Perhaps the most promising trend over the past decade has been the acceptance of Ernst Haas's self-criticism, over a quarter century old, that a sui generis theory of "regional integration" can only be a "pre-theory." Such "grand theories" were discarded for various reasons, the most fundamental of which is their lack of microfoundations in general theories of societal interests, state power or institutional delegation. Donald Puchala's oft-cited elephantine metaphor reminds us that multiple and more fine-grained theories are required: A theory of regional integration makes no more sense than a theory of American or comparative politics. The result of ignoring this was deadly: as late as 1990, "regional integration" was with few exceptions a discipline closed unto itself, uninfluenced and unable to influence rich theoretical developments in international and comparative politics.
Today the study of the EU rests on firmer theoretical foundations. Concepts and theories drawn from general social science theory (e.g. "optimum currency area," "regulatory state," "preference intensity" and "conditional agenda-setting") are the norm. Central approaches to integration (e.g. "intergovernmentalism" and "historical institutionalism") are no longer sui generis, but designate sets of pan-disciplinary theoretical instruments designed to explain specific types of policy outputs within narrower, but more generalizable institutional and functional contexts. Increasingly-though this practice could be more widespread-theories employed in the EU literature rest on explicit assumptions that permit us both to specify precisely the empirical domain in which it does (and does not!) apply, as well as its proper relationship to existing theories. Mark Pollack's recent demonstration how supranational autonomy emerges within specific conditions of formal delegation defined by intergovernmental theory is a model.
Methodologically, however, EU studies has further to go to meet the n = 1 challenge. The primary weakness of EU studies today, I submit, is not a lack of theoretical innovation. To the contrary, numerous sophisticated theoretical conjectures exist to explain most important aspects of integration. The primary weakness of EU studies is instead its unwillingness to subject theories to potential disconfirmation through rigorous empirical testing. There are four fundamental reasons for this:
Few studies rest on explicit hypotheses with specified standards of disconfirmation. It is generally unclear what such how we would recognize disconfirming evidence; hence such evidence is almost never reported.
Few studies test alternative theories. With the exception of "straw men" positions that follow from no coherent theory-e.g. the expectation of "lowest common denominator" bargaining outcomes in which every government always achieves its ideal-most studies simply assemble evidence supporting a preferred explanation and ignore alternative explanations.
Few studies engage in comparative analysis. Even though some (though few) studies examine more than one case or disaggregate single cases, such efforts are almost never part of a systematic research design.
Few studies are based on "hard" primary sources, such as government documents, multiple oral histories and reliable reconstructions of confidential decision-making. In contrast to the rich European tradition of EU policy analysis and historical reconstruction, most social scientific analyses of the EU (not least those "Made in the USA") rest on "soft" primary sources, such as public justifications by governments or journalistic commentary and other secondary sources-or indeed on sheer speculation. Yet given the large amount of ex post commentary and justification of any major European decision and the incentives for governments to be dishonest and commentators to be speculative, dozens of "soft" sources can be mustered to support almost any claim about integration, no matter how accurate. Such analysis, Ian Lustick has recently shown, tells us little.
In sum, the normal research design in EU studies remains an isolated, anecdotal, unstructured case study grounded in secondary sources. The result: "confirmed" hypotheses and theories about the EU proliferate without bound. Yet where hypotheses are rarely rejected, they are rarely being truly confirmed. Surely EU studies would benefit greatly if we focused more on testing explicit hypotheses, stating in advance what would constitute disconfirmation, considering multiple cases and collecting more reliable qualitative and quantitative data. The ultimate goal should not be to add yet another theory of integration, but (finally) to remove some that are empirically vulnerable.
An obvious benefit of meeting the theoretical and methodological "n = 1 challenge" in this way would be a more accurate (and less ideological, but that is another issue) understanding of EU politics. A less direct but perhaps even more important benefit would be a set of lessons for scholars and policy-makers interested in other forms of international cooperation. The study of European integration has traditionally been a harbinger of future theoretical trends, such as theories of regimes and interdependence. Today the EU provides the best laboratory for studying theoretical issues only just emerging elsewhere, such as threats of exit and exclusion, binding interstate legislative procedures, multi-level systems and legal dispute resolution. For policy-makers, too, lessons from the EC experience are directly applicable to problems facing the WTO, NAFTA and other international organizations--Kalypso Nicolaïdis' recent OECD reports on regulatory mutual recognition constitute striking proof. Such lessons can be drawn with confidence, however, only if the analysis on which they are based distinguishes generalizable from contingent phenomena. Overcoming the n = 1 challenge in this way offers the best means for scholars and policy-makers to exploit and extend the EU's singular success.
Andrew Moravcsik is associate professor in the Department of Government
and the Center for European Studies at Harvard University.
Mark A. Pollack
THE SUI GENERIS OR n = 1 argument in EU studies is relatively straightforward: the EU, it is argued, is unique in the world as an experiment in political and economic integration, and hence students of European integration have only a single case-the EU itself-to study. If we accept this view, the implications for EU studies are two-fold, and both are disturbing. First, the n = 1 view suggests that theoretical propositions will be difficult or impossible to test in the EU context, because of the well-known problems of testing hypotheses on a single case. Second, it suggests that any propositions or findings to emerge from EU studies will not be generalizable beyond the EU to other international or domestic political systems, because of the EU's supposed uniqueness.
Fortunately for the field of EU studies, the European Union need not represent a single case, but may indeed generate multiple cases, or observations, both within the EU and across the EU and other domestic and international political systems. This, in turn, renders possible both hypothesis-testing in the EU, and generalizability beyond the European Union.
To understand why EU studies is not restricted to a single case, consider briefly the standard literature on the comparative method in political science. In the deductive-nomological tradition of comparative political inquiry, the researcher typically begins with a deductive theory specifying a dependent variable as the object of study, a hypothesized explanatory or independent variable, and a series of control variables. Case selection, including the number of cases, follows in this view from the researcher's question and from the dependent and independent variables specified in the theory. Thus, in order to determine the number and selection of cases in an EU study, we need to begin not with the characteristics of the EU itself, but rather with the research question animating a given study.
Traditionally, the central research question in EU studies was how to explain "integration," defined in terms of either institutional development or policy development, or both. For such studies, the EU was indeed the only case, or at least an extraordinary outlier in comparison with other cases, and the sui generis argument was compelling. However, as Simon Hix has pointed out, integration per se is no longer the only, or even the most common, dependent variable in EU studies. Instead, students of the EU increasingly study variables such as the outcomes of intergovernmental bargains; the adoption (or non-adoption) of advanced social regulations in multi-tiered political systems; the interactions of legislative principals and their executive and judicial agents; the voting behavior of parliamentarians; and the implementation of EU policies in the member states. For each of these research questions, and for many others, students of the EU can generate, and have generated, multiple cases, in two ways.
First, as Keohane, King and Verba point out, a single "case" like the EU may often be disaggregated into a relatively large number of observations at a lower level of analysis. Thus, for example, Andrew Moravcsik finds five major intergovernmental bargains in the history of the EU, Stephan Leibfried and Paul Pierson find a long and complex history of social regulations, Alec Stone finds hundreds of ECJ rulings across various issue-areas, Amie Kreppel finds dozens of roll-call votes by hundreds of MEPs, and Marc Smyrl finds multiple cases of Structural Fund implementation in French and Italian regions. Within the single "case" of the EU, therefore, students of the Union have found material for comparative and even statistical testing of social-scientific hypotheses.
This brings me to a second point, concerning generalizability beyond the EU to other domestic or international "cases." As noted earlier, the supposed incomparability of the EU arose from the fact that early students of European integration took integration itself as their dependent variable, and could not find similar cases of international integration elsewhere. If we shift our focus away from integration to other dependent variables, however, the EU then represents one of a number of cases, and can be profitably compared with other cases--both domestic and international--to test hypotheses about intergovernmental bargaining, regulatory policymaking, principal-agent interactions, parliamentary voting behavior, implementation, and other subjects as well.
There are, of course, methodological problems with such comparisons, which should be acknowledged and dealt with by researchers. Most comparative studies of advanced industrial societies, for example, rely explicitly on "most similar systems" designs, in which cases are selected so as to resemble each other across a range of control variables, while providing variation across the key explanatory variable. Any observed variation in the dependent variable across cases, therefore, is imputed to the hypothesized explanatory variable. The danger with using the EU as a case in such studies is that the EU is likely to differ from other cases along multiple dimensions, making the effect of the hypothesized explanatory variable difficult to ascertain.
This does not, however, mean that comparison of the EU with other political systems is impossible. The problem of variation across control variables, for example, also arises in comparative country studies, and may be alleviated by careful, empirically rich case study analysis which is attentive to the possible effects, if any, of these variables. Alternatively, the same problem can also be overcome by using a "most different systems" research design, in which a particular empirical relationship is studied comparatively across a range of very different cases. Thus, for example, if policy communities behave similarly in centralized Britain and in the decentralized EU, or if activist courts behave similarly in the domestic US context and the "international law" context of the EU, then our hypotheses about these phenomena may in fact be strengthened by testing them across a range of cases including the EU.
In sum, students of the European Union are not doomed to studying a single case. Instead, they may generate multiple observations within the EU, and proceeding with the proper methodological caution, may begin to engage in comparative hypothesis-testing and generalization beyond the EU as well. For most of us, the n = 1 problem in EU studies is not a problem.
Mark A. Pollack is assistant professor of political science at
the University of Wisconsin at Madison.
ECSA members receive the ECSA Review, a triannual periodical, as a benefit of membership. Click here to find out how to join.
European Union Studies Association
415 Bellefield Hall * University of Pittsburgh * Pittsburgh PA 15260 USA