|Does the EU cause domestic developments? The problem of case selection in Europeanization research (*)|
|European Integration online Papers (EIoP) Vol. 9
(2005) N° 2;
|Date of Publication in : 14.1.2005|
|| Abstract |
Back to homepage |
| This paper's comments page | Send your comment! to this paper |
Research on the impact of the EU on the political systems of its member states has increased in sophistication over the last two decades. Idiosyncratic descriptive single case studies have given way to theoretically-informed comparative designs. A research programme is evolving that allows for the accumulation of systematic knowledge about the factors and mechanisms shaping national adaptation and non-adaptation to EU pressures, incentives and ideas (Bulmer and Lequesne 2003; Caporaso, Green Cowles et al. 2001; Dyson and Goetz 2003; Featherstone and Radaelli 2003; Goetz and Hix 2000; Haverland 2003a; Knill and Lehmkuhl 2002).
With regard to theory, Europeanization research has successfully utilized theoretical approaches of the political and social sciences, including rational choice, sociological institutionalism and constructivism. In this paper, I argue that Europeanization research would also benefit from a greater awareness of social science methodologies in particular regarding the issue of proper case selection. In the remainder of the paper, I will first elaborate on the specific problem that I identify with regard to case selection: the lack of variation in the independent EU-level variable (Section 2). Then I will discuss two possible solutions: counterfactual reasoning (Section 3) and comparing EU members with non-members or if research is restricted to EU countries cases where the source of the potential EU effect is present with cases where the source is absent (Section 4). In the conclusion, I will suggest some guidelines for case selection (Section 5). Closer attention to these issues, and the use of these potential solutions, will pay dividends to the Europeanization research agenda.
If one follows the emerging consensus to define Europeanization as the effect of the EU on member states polity, politics and policy, the issue of causality is of paramount importance.(1) It must be demonstrated that the EU is indeed causally important for domestic developments in the case(s) under investigation. To be sure, as Anderson has pointed out, in some instances it is very easy, even trivial, to attribute causal importance to the EU (Anderson 2003). Hence, the establishment of new administrative units within ministries to coordinate EU matters can be easily traced back to the existence of the EU. Less obvious, however, are domestic developments that make national policies more compatible with European legal requirements which may in fact be causally unrelated to the EU. The liberalisation of many economic sectors in many member states brings these sectors more in line with EU requirements, but they are arguably caused by international or domestic developments or both rather than by the EU (see below). The problem of causality is compounded if the investigator hypothesizes rather indirect effects of the EU, for instance, if the researcher want to test whether EMU causes cuts in German pension system because of a shift in the opportunity structure (Haverland 2003b), or whether European regional policy resulted into more co-operative government society relations on the sub-national level due to a diffusion of the European model (Kohler-Koch 2002).
One of the reasons why it is difficult, if not impossible, to demonstrate causality lies in the way cases for investigation are typically selected. Quite often only cases are chosen where potential EU pressures, incentives or ideas are present. Hence selection is based on the key independent variable, which depending on the unit of analysis- is operationalized, for instance, as EU membership, EU treaty competencies in a certain area, EU funds, or discourses emanating from the EU. Hence research is typically confined to EU member states; with regard to policy studies, typically those policy sectors are studied where EU competencies exist; when the EU is theorised as agent of new ideas, typically areas are selected where the EU promotes a new idea. This case selection implies that the independent variable we are most interested in, the EU, shows no variation.
It is important to stress that without variation in the independent variable the EU matters hypothesis can be rejected when the hypothesised patterns or outcome is not matched by our observations. For example, if a European policy requires a certain regulatory regime and this regime has not been established in the member state, the EU pressure to adapt was not effective. But without variation in the independent variable, it is difficult to confirm the EU matters hypotheses when the hypothesised pattern or outcome is matched by our observations, as there is no systematic control for other potential explanations, such as globalisation (Collier and Mahoney 1996; King, Keohane, and Verba 1994). Hence if the EU requires a certain regulatory regime and this regime is adopted in the member state, we should wonder whether this regime would not also have been adopted in the absence of the EU requirement. To be sure, the precise timing, speed or sequence of introducing such a regime may be shaped by a EU policy, but it should be more interesting to investigate whether or not a roughly similar regime would have been adopted anyway.
Even orthodox quantitative methodologists concede, however, that the social science text book selection rule random selection is often not desirable in political science as it may negatively affect the substantive relevance of the project (King, Keohane and Verba 1994: 125). Why study Europeanization if one does not include any of the most important EU member states: Germany, France or the UK, or important EU policy fields like monetary integration? Also, random selection may seriously bias conclusions in a situation with a limited number of potential cases (King, Keohane and Verba 1994: 126). Hence the intentional selection of cases is often a wise thing to do. But one should select on the independent variable in a way that ensures variation. Without this variation, patterns that confirm the EU matters hypotheses are too easily attributed to the EU.(2)
I will not suggest here that Europeanization research is totally unaware of the bias towards EU-level explanations. Several authors have warned against this bias (see for example Anderson 2003; Hix and Goetz 2001: 20-3). Moreover, a number of researchers have added global or domestic factors in their explanatory framework from the outset, or as they came across these factors in the course of their empirical process tracing. This awareness that other factors might be at work is indicated by Schmidts comprehensive review of her own and others work on the impact of the EU on policy developments in seven economic sectors in Germany, the UK, and France. Only in seven out of 21 cases, the EU has been the dominant source for adaptation pressure; in seven cases global pressures have been decisive; and in two cases internal pressures. In the remaining cases combinations of two of the three sources of adaptation have been at work (Schmidt 2002). The possibility of arriving at such a result indicates that authors are indeed open for alternative explanations. And this theoretical awareness often leads them to question whether the EU is the sole driving force of domestic change. However, I argue that in order to increase the validity of the findings, theoretical awareness should be reflected in the case selection.
There are a few studies that have included control cases. In a quantitative study of OECD countries, Verdier and Breen investigated the impact of economic globalisation and/or European integration on broad patterns of interest representation in the capital market and other areas while controlling for a number of other variables. The EU impact was conceptualised in both possible ways, either as strengthening economic globalisation or as weakening it. With regard to the capital market, an area quite likely to be affected by the EU, given EMU and Commission and Council activity in this area, no effect of European integration could be found (Verdier and Breen 2001). Levi-Faur also provides evidence that in areas where most scholars would probably expect an EU effect, developments are actually driven by other factors. In his extremely comprehensive qualitative study of telecommunication and energy reforms in 28 countries, he generally found that although the Europeanization of the telecoms and electricity industries led to some liberalisation, it is highly plausible that the major features of the liberalization would have been diffused to most member states even if the Commission and other agents of Europeanization had not existed (Levi-Faur 2004: 25).
The fact that studies which systematically control for other variables through their case selection find a rather weak effect of EU-level variables in cases where strong effects are often expected confirms the importance of increasing the variation of the independent variable. Variation can be increased in two ways. First by including real cases with a zero value on the EU variable: a comparison of EU member states with non-members, policy sectors where EU competencies do exist with policy sectors where EU competencies are lacking, administrative sub-units dealing with EU affairs with sub-units, not involved with the EU etc. There is, however, another alternative: rather than choosing real cases of a lacking potential EU effect, the researcher can mentally construct the situation in which the EU variable would be absent, in other words, he or she can create a counterfactual scenario. It is this alternative to which I turn first.
The statement that European pressures, incentives or models have resulted in a particular outcome implies that if the EU variable had been absent, the particular outcome would not have occurred. Making such a statement explicit allows the researcher to develop the counterfactual scenario to substantiate his or her view that the EU was indeed causally important (for the general logic of counterfactual reasoning, see Fearon 1991; Ned Lebow 2000; Tetlock and Belkin 1996). In the context of Europeanization research, some authors have proposed employing counterfactual reasoning to establish the causal importance of the EU (Anderson 2003; Schmitter 1999). Others have been sceptical about its potential (Hix and Goetz 2000, 12).
Traditionally, counterfactual reasoning has been mostly used by historians and international relation scholars to increase the understanding of particular (important) events. A case in point is the Cuban missile crises, where to many observers the world was very close to a devastating nuclear war. One of the many counterfactuals advanced is that had Kennedy displayed greater resolve prior to the crisis, Khruchev would not have sent missiles to Cuba (Ned Lebow 2000). Other events or situations which have invited counterfactuals are the outbreak of World War I, the British attempt to appease Hitler in 1938, the importance of Stalin for the shape of Soviet Communism, and the importance of nuclear weapons for maintaining peace after World War II (Fearon 1991; Tetlock and Belkin 1996). The rational to apply counterfactual reasoning in these cases is the following:
Key events occur only once, whereas for the purpose of valid causal inference we would like to rerun history many times and to examine the resulting distribution of outcomes that reveal how strongly cause and effects co-vary (Tetlock and Belkin 1996: 9).
Hence, according to its advocates, counterfactual reasoning is a promising research strategy if the universe of potential cases is very low, for instance if entire political systems, or their macro-political institutions or policy sectors are the unit of analysis. In addition, it is recommended when the hypothesised relationships are rather complex: involving a large number of variables with intricate interactions between them (Tetlock and Belkin 1996: 6). Both conditions are often met in Europeanisation research. Many studies are interested in the Europeanization of a member state per se, or crucial institutions such as parliaments or courts, or critical policy sectors. Also, current Europeanization research frequently hypotheses interaction effects among numerous variables, for instance between the goodness of fit (the fit of EU requirements with the domestic status quo), the preferences of key actors in the policy process, institutional veto points and new cognitive inputs (Boerzel and Risse 2003; Haverland 2003; Knill and Lehmkuhl 2002; Liebert 2003; Schmidt 2002).
An important merit of the counterfactual scenarios is that the researcher can to a large extent plausibly control for other potentially important variables. A counterfactual is a thought experiment, in the methodological meaning of the word: it allows for the manipulation of the variable of interest while controlling for other variables. For instance, if the researcher is interested in the impact of the EU telecoms directives on the Netherlands, the researcher hypothesizes the absence of these directives, but other potential explanatory variables such as the political, institutional, technological or economic condition of the Netherlands remain unchanged.
It is important to note that counterfactual reasoning should not lead to unlimited speculation, nor are all counterfactual claims equally compelling. Several authors have suggested criteria for the development of good counterfactuals (see for instance Ned Lebow 2000; Tetlock and Belkin 1996). Three criteria are particularly compelling and are undisputed in the literature: clarity, historical consistency and theoretical consistency. I will discuss these criteria in turn and use an intensive case study on packaging waste policies as illustration. In that study the claim is made that the European Union was causally important for a strong legal codification of the Dutch packaging waste policy: a shift from a voluntary covenant focused on a selective target group to a generally binding ministerial regulation. The argument is that in the absence of the European packaging directive there would not be a strong legal codification of the policy (Haverland 1999).(3)
Clarity, the first criterion, demands that the hypothesised antecedent (independent variable) and consequent (dependent variable) are clearly specified and delimited. In other words, the researcher has to make explicit which variables are changed in his or her mental thought experiment and which remain unchanged. In the packaging waste study it is specified that the relatively informal Dutch packaging covenant would not have been supplemented with a formal ministerial regulation (change in the dependent variable) without the adaptation pressure induced by the EU packaging waste directive (change in the independent variable). Hence, it is explored whether the lack of the EU packaging directive, a hypothetical change of the independent variable, would have changed the value of the dependent variable (degree of legal codification of national policy).
In this example, the first criterion, a precise delimitation of the variables, is relatively easy to be met, as only very limited changes are hypothesized and it is plausible that the changes can be restricted to these two variables. However, precisely because counterfactual reasoning is often employed in situations where variables are interconnected, it may be difficult to determine which antecedent(s) have to be changed in the equation and which antecedents can remain unchanged. Assume that the researcher wants to research the effect of the EU single market on national environmental policies. Given the fact that many EU environmental policies are framed as single market measures, it would not be very plausible to assume the absence of a European single market, without also assuming the absence of EU environmental policy. However, if the researcher changes the value of both antecedents (i.e. to assume their absence): the EU single market and EU environmental policies, to investigate the causal effect of the EU single market on national environmental policy, the whole exercise becomes rather dubious as the EU single market effect would not be isolated.
Historical consistency demands that antecedents be specified in a way that requires few changes to historical facts. In other words, counterfactuals should rewrite history as little as possible. Hence, the packaging waste study merely explores what would have happened in the absence of the European packaging directive. Alternatively, the researcher could have gone much further by exploring what had happened without the existence of the European Union, or at least without the Netherlands being a member of it. Then, however, one would have to rewrite history much more profoundly and the story would become more speculative and therefore less compelling as by implication much more phenomena may be different as well. For instance, one could argue the following: Dutch-based multinationals are a crucial player in Dutch packaging waste politics. Without the European Union and its single market, their threat of exit (i.e. shifting investments to other countries) would be less credible and their political power therefore weaker, which would have substantial consequences on the shape of Dutch packaging waste policies. Generally speaking, the more changes we make the greater the number of consequences and the lower their predictability. However, if the investigator wants to use counterfactual reasoning to explore to what extent a member state, a macro-political institution or an entire policy sector, rather than a specific policy has been Europeanized, he or she has to move some way in the direction of far reaching and tricky scenarios.
Consistency with (well-established) theoretical laws is a criterion of counterfactual reasoning that requires that the hypothesized linkages between antecedent and consequent are consistent with well-established theoretical generalizations. This demands that in addition to historical constrains, we also need to constrain our counterfactual in theoretical terms. As some writers have conceded, one problem that arises is that are only a few, some would even say no, well-established theoretical laws in social sciences (Ned Lebow 2000: 583; Tetlock and Belkin 1996: 25-6). Moreover, the quality of theories is subject to constant debate, so there can be disagreement as to what extent a theory supports the counterfactual. Still, theoretically informed counterfactuals are generally more compelling than a-theoretical accounts. They will at a minimum provide the reader with a more explicit perspective from which to evaluate the plausibility of the counterfactual (Ned Lebow 2000: 583). In the case of packaging waste, the author based the counterfactual in a rational choice approach. He argued that the government did not have sufficient incentives to formalize the packaging waste policy in the absence of the EU. The costs were higher than the benefits. The benefits decreased due to the declining public support for environmental measures while the political costs were high because the Dutch policy would be binding for tens of thousand companies, including the inclusion of many small- and medium-sized companies, rather than only for those 300 companies that participated voluntarily in the covenant (Haverland 1999).
A counterfactual analysis has the advantage that alternative explanations can be mentally controlled for. Moreover, the discussion of the above-mentioned criteria shows that counterfactual reasoning does not need to result in unlimited speculation; there are some rules that must be adhered to if counterfactuals are to be plausible. The discussion also reveals, however, that in situations of causal complexity, where counterfactuals are particularly recommended, it is especially difficult to construct one. Therefore, counterfactual analysis is not a panacea for Europeanization studies and other strategies should be considered. The following section suggests a strategy involving the use of real cases in which the EU influence is completely absent.
The second strategy to address the no variation problem is to include real cases with a value of zero on the independent variable EU. I will discuss two options here: first, to compare EU member states with non-members; and second, if the investigator concentrates only on EU member states, to include policy areas, economic sectors, administrative units, interest groups, etc. where EU factors are lacking.(4)
Comparing EU member states with non-members is the most straightforward strategy to cope with the problem of no variation. It is the strategy chosen in the study by Verdier and Breen and it comprises an important building block in the study by Levi-Faur. In a quasi-experimental fashion, one compares countries where the stimulus, the EU, is present, with a control group where the stimulus is absent (Lijphart 1971). This strategy is not without problems, however, as it demands a critical trade off. Ideally, the researcher needs to choose countries where he or she can be confident that the EU has no effect.
At the same time, he or she wants to choose countries which are comparable in the sense that they are similar with regard to other explanatory variables (Lijphart 1971). Comparability enables the researcher to control for other effects that might confound the hypothesised effect of the EU. The logic of this most similar systems design is based upon Mills method of difference: similarities across countries cannot explain differences between them (Frendreis 1983). The need to ensure variation in the EU variable but similarity with regard to other variables represents the crux of the trade off. Let me elaborate.
Norway and Switzerland are usual suspects as non-EU countries in a comparative case study design. Both countries are included in the studies by Breen and Verdier and Levi-Faur (see above).(5) These countries are similar to the EU member states, for instance in regard to their geographical location and history, macro-political institutions, economic systems and societal aspects: they are European democratic and capitalist welfare states. The problem arising here is twofold. First, the sheer number of suitable countries is small. Therefore the opportunity for literal replication of the findings of the first round of research is severely restricted (Yin 1994). In other words, only weak conclusions can be drawn as they are based on a maximum of two non-EU countries. Idiosyncrasies of Norway, Switzerland or both have a large impact on the outcome.
In addition to the small n problem, there are also theoretical problems involved. One of the main theoretical claims of Europeanization research is that the effect of the EU on its member states is not confined to merely legal compliance pressure. As stated above, the EU may have an indirect effect on its member states. It may change the domestic opportunity structure. It may change the terms of the domestic debate, as domestic actors can use European policies and ideas to legitimize their own policy positions or to de-legitimize the positions of others. Also, European policies and ideas may spur changes in domestic belief systems, hence contributing to policy-oriented learning (Boerzel and Risse 2003; Checkel, 1999; Liebert 2003, Knill and Lehmkuhl 2002, Radaelli 1997). Taking into account that the EU may indeed have these indirect effects on its member states, it may also be the case that these effects reach beyond the member states. Whereas legal obligations to comply with the acquis communitaire are in principle confined to the territory of the EU, learning or other indirect mechanisms do not need to stop at the European Union borders. For example, a new economic policy concept developed by EU institutions may inform political choices in Switzerland or Norway. In other words, the EU may have an impact on non-member states. If this is the case, then evidence that EU members and non-members exhibit similar domestic development cannot be taken as evidence against an impact of the EU. It could even be taken as evidence for the rather strong causal power of the EU!
In a single country study that was part of a larger project examining small states, Norway was portrayed an adaptive non-member (Sverdrup 1998). There are also a number of studies suggesting that the EU had a significant impact on Switzerland (for example, Kux and Sverdrup 2000; Sciarini, Fischer and Nicolet 2004).
Theories about policy learning and diffusion suggest that these mechanisms are more likely to operate when the recipient of the idea shares borders, core beliefs and political and economic conditions with the entity from which the ideas emerge (Berry and Berry 1999; Sabatier 1999). For theoretical reasons therefore, countries which fulfil these properties should not be selected for a comparative and quasi-experimental research design that seeks to isolate the EU effect by selecting for a control group those countries where the hypothesised EU effect must be absent.
The irony here should not go unnoticed: the very same properties that make countries like Norway and Switzerland a good choice for general methodological reasons make them a bad choice for theoretical reasons. A methodological virtue turns into a theoretical vice.
Choosing countries which are dissimilar to the EU member states with regard to historical, political, socio-economic and cultural features mirrors the problem. There might be good theoretical reasons why the EU should not have an effect in Latin American, African or Southeast Asian countries. At the same time these countries vary according to so many other variables as well, that the risk of an overdetermined (or indeterminate) research design looms large. In other words, if the investigator would find a large variation in outcomes between EU member states and non-members, the difference may not be caused by the EU factor, but rather to another factor which also varies between the EU member states and non-members, for instance the level of socio-economic development, or the general nature of the political system (democracy or authoritarian regime).
What conclusions can we draw from this discussion? First of all, case selection depends on the theoretical argument. If the hypothesised effect has a rather direct character, e.g. institutional compliance (Knill and Lehmkuhl 2002), then it makes sense to compare a European Union country with Switzerland, or if the investigator strives for literal replication two EU countries and Switzerland and Norway, at least if the issues under consideration are not dealt with by EEA agreements or bilateral treaties. If the hypothesised effect is rather indirect, the investigator should select countries that are less close to the EU members but that are still not completely dissimilar in regard to political, institutional and socio-economic variables. Such a selection might include Australia, Canada, New Zealand and the United States.
Another strategy to increase the variation in the independent variable confines the research to EU member states but compares individual policy fields, institutions, interest organisations, elites, or other entities subject to EU pressures, incentives or ideas with counterparts not exposed to these EU-level factors.(6) For instance, if the investigator wants to know whether the EU has an impact on the professionalization of interest groups, he or she can compare interest groups which have an incentive to lobby EU institutions, or national institutions about EU policies, with lobby groups who have no incentive to do so.(7)
There are several studies that follow this cross-sector strategy. Levi-Faur, investigating whether the EU is causally important for liberalisation, not only compares member states with non-members, but also facets of electricity and telecommunication regimes that are subject to EU pressures with facets that are not. Similarly, Sciarini, Fischer and Nicolet compare free movement of persons, a sector facing direct EU pressures, telecommunication reform, a sector exposed to indirect pressures, and pension reform, a sector experiencing only negligible pressures, in an effort to investigate the effects of the EU on domestic politics (Sciarini, Fischer, and Nicolet 2004).
Unfortunately, the cross-sector design suffers from essentially the same trade off as the cross country design. Comparing sectors where EU pressure, incentives or models exist with other sectors comes close to the ideal of the most similar systems design, because the investigator can keep all national factors constant. At the same time, because the country is a member of the European Union it might be difficult to argue that the cases are independent from one other. Horizontal effects, from one sector to another may be at work. This horizontal impact, driven for instance by diffusion or learning, may have its origin at the EU level. Thus not only borders between members and non-members become blurred but also boundaries between sectors.
For example, if the investigator wants to research whether partnerships between sub-national governments and business in regional economic policy are causally related to EU regional policy, it might be problematic to compare it with regional vocational training policies, as the norm of partnership may have diffused via regional economic policy to vocational training policy as well. If this is the case, than the EU has caused changes in vocational training policy even though it has almost no formal competencies here. For this example, it would be important to choose sectors that are disconnected from each other. In general, it is important to choose sectors that are not too similar with regard to theoretically relevant properties. For instance, if policy sectors are concerned, it is often wise to choose sectors that are regulated by different ministries.
As in the cross-country design, however, a design that selects most different sectors may suffer from overdetermination. If the hypothesized effect occurs in sectors subject to EU pressures but does not in sectors that are not subject to such pressures, the variation could be caused by sector-specific properties. For instance, if the researcher hypothesizes that the incentive to lobby the EU results in a professionalization of interest groups and he or she compares interest groups in telecommunication with interest groups of teachers of primary education, one could argue that if telecommunication interest groups are indeed more professionalized it is related to economic globalization or technological changes rather than European integration. Therefore in choosing sectors it is important to control for sector-specific variables. Again, it is worth noting that concrete selection criteria depend very much on existing theories and hypotheses. A crucial aspect is whether the relevant theories hypothesize direct or indirect effects.
The paper has addressed an important methodological problem in Europeanization research: the no variation problem. In many studies, case selection is confined to instances where potential EU pressures, incentives or ideas are present. Control cases are lacking. It is difficult, therefore, and some would even argue impossible, to demonstrate that the EU has causal importance in instances where the hypothesized effect of the EU does occur.
Two strategies to cope with the problem have been discussed. The analysis shows that both strategies, counterfactual reasoning and including non-EU cases, have strengths and weaknesses. Counterfactual analysis has the advantage that it can mentally control for alternative explanations. Though the discussion of counterfactuals has shown that it may not be possible to plausibly control the other variables completely, the level of control is certainly better than in designs where the investigator increases variation by including real non-member or sectors where no EU pressures, incentives or ideas are present. Moreover, counterfactual reasoning does not need to result in limitless speculation as there are some methodological rules that help to make counterfactuals more compelling: counterfactuals should be explicitly and clearly delimited, restricted to a minimum, and hypothesised linkages between antecedents and consequents should be supported by theoretical arguments. The discussion does also reveals, however, that precisely in situations of numerous and interrelated variable where according to its advocates counterfactuals reasoning should demonstrate its competitive advantage, it is particularly difficult to construct a counterfactual. Given this rather severe problem, counterfactual analysis should probably not be used as the sole strategy to cope with the no variation problem. Rather it could be used as an additional heuristic device at the outset of the study in order to draw the researchers attention to other potential explanations which then could also inform his or her case selection. In addition, it can be used after the empirical research as a supplementary instrument to evaluate the internal validity of the findings.
Including real cases with a zero value on the EU variable is probably the better strategy for demonstrating the causal effect of the EU. But this strategy is also not without problems due to the trade off between methodological and theoretical concerns. Whereas, for theoretical reasons, most different countries or sectors should be chosen to minimize the potential impact of indirect EU effects beyond member states, methodological reasons require that most similar countries or sectors should be chosen to control for alternative explanations. The precise nature of the trade off depends very much on the character of the theoretical argument advanced. For instance, if an indirect learning or diffusion effect seems possible then countries or sectors far away from the potential EU pressures should be chosen, though a certain comparability should be assured. As far as countries are concerned, Australia, Canada, New Zealand and the US may be adequate. If the hypothesized effect is restricted to legal compliance, countries closer to the EU, like Norway and Switzerland can be chosen at least after the researchers has checked whether these countries are not subject to legal compliance pressures emanating from the EEA agreement, bilateral agreements or other sources of compliance pressure.
Again, it should be emphasised that neither of the strategies discussed is a panacea to the no variation problem. However, this paper demonstrates that Europeanization research would benefit from a greater awareness of the no variation problem, the merits and pitfalls of the discussed methodological strategies, and thus from a better informed selection of cases, if it wants to achieve its major objective, to demonstrate whether or not the EU makes a difference.
Anderson, Jeff. 2003. "Globalization and Europeanization: A conceptual and theoretical overview." in Germany, Europe and the Politics of Constraint, edited by K. Dyson and K. Goetz. Oxford: Oxford UP & British Academy.
Berry, France Stokes and William. D. Berry 1999. Innovation and Diffusion Models in Policy Research. Theories of the Policy Process,.edited by P. A. Sabatier. Boulder, CO, Westview Press: 169-200.
Börzel, Tanja A. and Thomas Risse. 2003. "Conceptualising the domestic impact of Europe." in The Politics of Europeanisation, edited by K. Featherstone and C. Radaelli. Oxford, Oxford UP.
Bulmer, Simon and Christian Lequesne 2003. Member States and the European Union. Oxford, Oxford UP.
Caporaso, James, Maria Green Cowles, and Thomas Risse (eds,). 2001. Transforming Europe. Europeanization and Domestic Political Change, Cornell: Cornell UP.
Checkel, Jeff T. 1999. "Social Construction and Integration." Journal of European Public Policy 6(4): 545-560.
Collier, David and James Mahoney. 1996. "Insights and pitfalls. Selection bias in qualitative research." World Politics 49:56-91.
Dyson, Kenneth and Klaus Goetz 2003. Germany, Europe and the Politics of Constraint. Oxford, British Academy & Oxford UP.
Eising, Rainer 2003. Europäisierung und Integration. Konzepte in der EU-Forschung. Europäische Integration, edited by M. Jachtenfuchs and B. Kohler-Koch. Opladen, Leske + Budrich UTB: 387-416.
Falkner, Gerda 2003. "Comparing Europeanisation Effects: From Metaphor to Operationalisation." European Integration online Papers (EIoP) 7(13) http://eiop.or.at/eiop/texte/2003-013a.htm.
Fearon, James D. 1991. "Counterfactuals and Hypothesis Testing in Political Science." World Politics 43:169-195.
Feartherstone, Kevin and Claudio Radaelli (eds.) 2003. The Politics of Europeanisation. Oxford: Oxford UP.
Frendreis, John. P. 1983. "Explanation of Variation and Detection of Covaration. The Purpose and Logic of Comparative Analysis." Comparative Political Studies 16(2): 255-272.
Geddes, Barbara. 1990. "How the cases you choose affect the answer you get: selection bias in comparative politics." Political Analysis 2:131-150.
Goetz, Klaus and Simon Hix (eds.) 2000. Europeanised Politics? European Integration and National Political Systems. London and Portland, OR: Frank Cass.
Graver, Hans Petter. 2000. Supranationality and National Legal Autonomy in the EEA-agreement, ARENA Working Papers 00/23. Oslo. http://www.arena.uio.no/publications/wp00_23.htm
Haverland, Markus. 1999. National Autonomy, European Integration and the Politics of Packaging Waste. Amsterdam: Thela Thesis.
---. 2003a. "The impact of the European Union on national environmental policies." in The Politics of Europeanisation, edited by K. Featherstone and C. Radaelli. Oxford: Oxford University Press.
---. 2003b. "Social Policy: Transforming Pensions, Challenging Health Care? in Germany, Europe and the Politics of Constraint, edited by K. Dyson and K. Goetz. Oxford: The British Academy & Oxford UP.
Hix, Simon and Klaus H. Goetz. 2000. "Introduction: European Integration and National Political Systems." West European Politics, Special Issue on Europeanised Politics? European Integration and National Political Systems 23:1-26.
Hooghe, Liesbeth. and Gary Marks 2001. Multi-Level Governance and European Integration. Lanham, Boulder, New York, Toronto, Oxford, Rowman & Littlefield.
King, Gary, Robert O. Keohane, and Sidney Verba. 1994. Designing Social Inquiry. Scientific Inference in Qualitative Research. Princeton: Princeton UP.
Knill, Christoph and Dirk Lehmkuhl. 2002. "The national impact of European Union regulatory policy: three Europeanization mechanisms." European Journal of Political Research 41:255-280.
Kohler-Koch, Beate. 2002. "European Networks and Ideas: Changing National Policies?" European Integration online Papers 6. http://eiop.or.at/eiop/texte/2002-006a.htm
Kux, Stephan and Ulf Sverdrup. 2000. "Fuzzy borders and adaptive outsiders, Norway, Switzerland and the EU." Journal of European Integration/Revue d'intégration Européenne XXX.
Levi- Faur, David 2004. "On the Net Impact of Europanization. The EU's Telecoms and Electricity Regimes Between the Global and the National." Comparative Political Studies 37(1): 3-29.
Liebert, Ulrike. 2003. "Explaining Europeanisation: A Mechanisms Based Approach." Paper presented at the Conference "Forschungslogik und Methodik in den Internationalen Beziehungen. Evangelische Akademie Hofgeismar.
Lijphart, Arend 1971. "Comparative Politics and Comparative Method." American Political Science Review 65: 682-693.
Ned Lebow, Richard. 2000. "What's so different about a counterfactual?" World Politics 52:550-85.
Olsen, Johan P. 2002. "The Many Faces of Europeanization." Journal of Common Market Studies 5(5): 921-52.
Pollack, Mark 2000. "The end of creeping competence? The EU Policy-Making since Maastricht." Journal of Common Market Studies 38(3): 519-38.
Radaelli, Claudio. 1997. "How does Europeanization produce policy change? Corporate tax policy in Italy and the UK." Comparative Political Studies 30:553-75.
Radaelli, Claudio. 2000. "Whither Europeanization? Concept stretching and substantive change." European Integration online Papers 4: http://eiop.or.at/eiop/texte/2000-008a.htm.
Sabatier, Paul A. 1999. The Advocacy Coalition Approach: An Assessment. Theories of the Policy Process, edited by P. A. Sabatier. Boulder CO, Westview Press: 117-166.
Schmidt, Manfred G. 1999. "Die Europäisierung der öffentlichen Aufgaben." Pp. 385-394 in 50 Jahre Bundesrepublik Deutschland. Rahmenbedingungen, Entwicklungen, Perspektiven, edited by T. Ellwein and E. Holtmann. Opladen/Wiesbaden: Westdeutscher Verlag.
Schmidt, Vivian A. 2002. "Europeanization and the mechanics of economic policy adjustments." Journal of European Public Policy 9(6): 894-912.
Schmitter, Philip C. 1996. Imagining the future of the Euro-polity with the help of new concepts. Governance in the European Union, edited by G. Marks, F. W. Scharpf, P. C. Schmitter, and W. Streeck. London, Thousand Oaks, New Dehli, Sage: 121-150.
Schmitter, Philip C. 1999. "Reflections on the Impact of the European Union upon 'Domestic' Democracy in its Member States." Pp. 289-298 in Organizing Political Institutions. Essays for Johan P. Olsen, edited by M. Egeberg and P. Laegreid. Oslo: Scandivian University Press.
Sciarini, Pascal, Alex Fischer, et al. 2004. "How Europe Hits Home: Evidence from the Swiss case." Journal of European Public Policy 11(3): 353-378.
Sverdrup, Ulf. 1998. Norway an adaptive non-member. Adapting to European Integration small states and the European Union, edited by K. Hanf and B. Soetendorp. London, Longman Publishers: 149-166.
Tetlock, Philip E. and Aaron Belkin 1996. Counterfactual Though Experiments in World Politics. Logical, Methodological, and Psychological Perspectives. Counterfactual Though Experiments in World Politics. Logical, Methodological, and Psychological Perspectives, edited by P. E. Tetlock and A. Belkin. Princeton UP, Princeton, New Jersey: 1-38.
Verdier, Daniel and Richard Breen 2001. "Europeanization and Globalization. Politics Against Markets in the European Union." Comparative Political Studies 34(3): 227-262.
Vink, Maarten. 2003. "What is Europeanisation? And other questions on a new research agenda." European Political Science 3(1): 63-74.
Yin, Robert K. 1994. Case Study Research and Method. 2nd edition. Thousand Oaks, London, New Dehli, Sage.
(*) Drafts of this paper have been presented at the ECPR General Conference Marburg 18-21 September 2003 and the Second Pan-European Conference on EU Politics, ECPR Standing Group on the European Union, Bologna, 24-26 June 2004. I would like to thank the participants of these meetings as well as David Lowery, Peter Mair, Mark Rhinard, Bertjan Verbeek, and two anonymous reviewers for helpful comments.
(1) The precise definition of the very concept of Europeanization and its operationalization is another relevant methodological issue. As this issue has already attracted scholarly attention I will not cover it in this paper (see Anderson 2003; Eising 2003, Falkner 2003; Olsen 2002; Radaelli 2000, Vink 2003).
(2) A related problem arises from selection bias in the narrow sense of the word, that is selecting on the dependent variable in a way that the outcome, e.g. the hypothesised effect of the EU is always present, or always has a high score. Hence, it would be unwise to study elite orientation towards Europe by only looking at elites that think European, or to study only countries which policies are in line with European requirements. Depending on further properties of the research design such a selection would either bias in favour or against the EU-level explanation (see Geddes 1990, Collier and Mahoney 1996). However, almost no Europeanization study suffers from this bias. Therefore, I will not further elaborate on it.
(3) Note however that the original research does not include a fully-fledged counterfactual analysis
(4) There are other possibilities as well. For instance, if the researcher hypothesises a long term socialisation or emulation effect of the EU on its member states, he or she can include founding member states, member states who joined the EU later, and non member states.
(5) Note, however, that at least as far as single market policies are concerned, Norway, (and Iceland and Liechtenstein) cannot be considered as non-Members as they are part of the European Economic Area. Accordingly, they are legally obliged to comply with (almost all) single market policies (Graver 2000, Sverdrup 1998). Switzerland is not a member of the EEA, but has engaged in bilateral treaties with the EU. As far as issues are concerned that are regulated by those treaties, Switzerland cannot be regarded as non-member for the purpose of this design as well (Sciarini, Fischer, Nicolet 2004). Moreover, countries that seek membership of the EU, i.e. candidate countries, cannot be regarded as non-members, as it is quite likely that the EU has already had an effect there.
(6) In the remainder of the paper I will use the term sector as a general label for all units of comparison except countries (i.e. policy fields, economic sectors, institutions, elites, interest organisations, etc.).
(7) In regard to policy sector comparisons, case selection can be informed by attempts to measure the degree to which policy decisions are taken at the EU level, as a proxy for European integration. Lindberg and Scheingold undertook the most comprehensive early effort in this respect (1970). Schmitter updated that work to 2000 based on expert judgements about the Maastricht Treaty and its implication (1996). Hooghe and Marks (2001), Pollack (2002), and M.G. Schmidt (1999) provide updates of Schmitters results.