Decline of the Nation State? How the European Union Creates National and Sub-National Identifications

Transcription

1 Decline of the Nation State? How the European Union Creates National and Sub-National Identifications Karl-Dieter Opp, University of Leipzig Abstract A widely held argument is that organizations such as the European Community will only succeed if an identification with these organizations develops; and the EC creates this identification. This, in turn, is supposed to diminish identification with national and subnational regions. Based on this argument, some testable propositions are suggested referring to the conditions under which identifications with hierarchical regions in this paper Europe, nation states and sub-national regions arise and how these identifications are causally related. The propositions are tested with a two-wave panel study carried out in 2000 and 2002 in East Germany. In contrast to the previous argument, we found positive correlations between the three identifications. Furthermore, European identification has positive causal effects on sub-national and national identifications, and sub-national identification has positive causal effects on national and European identification. Introduction After World War II efforts increased to ban armed conflicts between nations forever. A means for achieving this goal was and still is to increase economic interdependence between nations. Free trade spurs economic exchange and makes armed conflict more costly. At the same time, a rise of economic well-being should be achieved. Another strategy to make conflicts less likely is to limit the power of nations by transferring basic rights to super-national organizations such as the European Community, NATO and the United Nations. This paper addresses one central argument about the effects of the emerging EC: it is widely held that the integration of nation states into a unified Europe and the termination of severe conflicts and an increase in economic well-being can only be successful if a European identification develops. It is further held that European unification creates the necessary identification. In addition, this identification reduces identification with the lower-level national and sub-national units. Because identification with the nation state is an aspect of nationalism, the argument implies that the development of the EC reduces nationalism. This argument raises several questions. Why should we expect that an emerging institution such as the EC creates identification with itself, and why does this influence other This paper was written while I was a visiting professor at the Department of Sociology at the University of Washington in Seattle. I wish to thank the department for its generous support and particularly the participants of the Seminar of Institutional Analysis for helpful comments on an earlier version of this paper. I am also grateful to Michael Hechter, Michael Mas and the anonymous reviewers of Social Forces for valuable suggestions. The data used are based on a research project conducted by this author and Professor Kurt Miihler (University of Leipzig). We are grateful to the Deutsche Forshungsgemeinschaft (German National Science Foundation) for support of this project. Direct correspondence to Karl-Dieter Opp, Sullkyweg 22, Hamburg, Germany. opp@sozio.uni-leipzig.de. The University of North Carolina Press Social Forces, Volume 84, Number 2, December 2005

2 654 Social Forces Volume 84, Number 2 December 2005 identifications? The hypothesis that a European identification generates other identifications suggests that there are causal interrelationships between identifications. But what exactly are these interrelationships? If there are causal interdependencies, what other factors are involved that affect the identifications? There is little theory and empirical research that address these questions. The present paper develops hypotheses focusing on these questions. In doing so several theories pertaining to the family of rational choice theory are applied: the Ajzen theory of attitude formation (i.e., the theory of planned behavior), social identity theory (in the version of H. Taifel) and balance theory (based on the work of Fritz Heider). These hypotheses are tested by a two-wave panel study administered in Saxony in 2000 and The geographical regions these data refer to are Europe, Germany and Saxony which is a German federal state. The Theoretical Context The Argument The great importance that is attributed to the formation and change of identifications with Europe, national and sub-national units raises the question why these identifications or "identities" are regarded as so important. It is widely held that "the successful integration of Europe might demand an iconography of identity that would complement, but not necessarily replace, national, regional and local identities." (Graham 1998:42-43) Other writers are less cautious. They consider the emergence of a European identity not only important but necessary for a successful European integration: "People in general need clear and tangible concepts in order to develop a sense of belonging. And that sense of identity is an indispensable factor in achieving and maintaining European unity." (Wistrich 1994:80) This author cites Jean Monnet who held that "most of the issues of monetary, taxation, trade and other matters are overshadowed by political considerations shaped by values which largely spring from distinct cultural attitudes and identities." (Wistrich 1994:80) More specifically, a European identity decreases popular resistance against the forging of Europe (Jacobs and Maier 1998:28). An indicator for the importance of a European identity is that in June 1984, the EC set up a committee charged with suggesting ways to strengthen the identity and improve the image of the community. (Wistrich 1994:90). In a report from representatives of several European countries about problems related to European unification the editors note that some of the accounts "place great emphasis on this issue (of European identity, KDO) as a sine qua non for real European integration." (The National Institute 1993:38) Although these and other authors do not provide any empirical evidence which substantiates these assumptions, we take for granted that it is worthwhile to address the question of how identifications with Europe and with national and sub-national units originate and how these identifications are causally related. Our first assumption is that the forging of the EC generates a European identification. As one author puts it: "Institutions... provide purpose and legitimacy to rules and practices. They equip individuals with an identity and constitutive belonging, cultural affiliations and boundaries, and interpretations and accounts which help individuals make sense of life." (Olsen 1996:251) However, despite the progress of European unification, the nation state will not become unimportant. Many writers emphasize that the European nation states are still the most important decision makers (Sakwa and Stevens 2000: 256; Jenkins 2000: 64). However, the nation state increasingly has "to share its importance as a solidarity-forming entity with both

3 The EU and National Identifications 655 regions and blocks of states." (Allardt 1993:96) This decreasing importance has an effect on the identification with the nation states: "A diminution in the capacity of the nation-state to secure desired policy outcomes promotes dissatisfaction and disillusionment which, in turn, threatens to undermine the basis of national identity." (Mitchell and Russell 1988: ; Wallace 1994; Wilson 2000:139) This argument may be applied to sub-national units such as federal states. If the nation's sovereignty and importance decrease so do those of the subordinate units. There should be a diminution of national as well as sub-national (or regional) identifications. In other words: the implication of the argument is that there will be negative correlations between European, national and sub-national identifications. Although this seems straightforward the slogan of a "Europe of the regions" suggests that the nation state will become redundant and will be replaced by sub-national units; these will then become more important. Thus, identification with the regions should increase and not decrease. However, this development has not taken place (Loughlin 1997). As one writer puts it, "there is no single relationship" between regions and the EC (Jones 1995:289). Thus, the "Europe of regions" with a diminishing importance of the nation state and increasing importance of regions is not yet reality. Although, it may be expected to take place in the future (Bullmann 1997:17). But if there is no overall increasing strength of sub-national regions, the previous argument suggests that the importance of the regions decreases with the loss of power of the nation states. With the progress of the EC then, we would expect that national as well as sub-national identities decrease. Let us summarize the argument so far: (1) Development of a European identification is central for the success of the European Union. (2) The development of the EC creates a European identification. (3) The development of the EC reduces the importance of the nations and of sub-national units. (4) The decreasing importance of the nations and sub-national units lessens identification with these units. The conclusion is that there should be a negative correlation between European, national and sub-national identifications, and there should be a negative causal effect of European identification on national and sub-national identification. Thus, if European identification increases, national as well as sub-national identification is reduced. Furthermore, the previous quotations indicate that national and subnational identifications are not only influenced by European identification, but that other factors are involved as well. This argument is depicted as a causal model in Figure 1. Figure 1. The Negative Effects of European Identification EuropeanEuropean unification identification '^ and other factors!f ^1 National identification Sub-national identification Note: Arrows pointing to sub-national and national identification indicate that other causal factors influence these variables.

4 656 Social Forces Volume 84, Number 2 December 2005 Although this argument is put forward by a large number of authors, we also find alternative hypotheses in the literature. It is sometimes argued, for example, that there is a development of European identity "which is held in tandem with a national identity." (Miall 1994:8). Inglehart suggests that with the decreasing importance of the national state (1977:70-71) a "cosmopolitan identity" (1977:342) develops; the latter could refer to a European identity. These arguments would be tantamount to a positive correlation between European and national identification. At the same time, however, Inglehart sees an increase (and not a decrease) of regional identities (pp. 243 ff.). In other words, European identification correlates positively with sub-national identification. Keating suggests that "territorial identities... do not necessarily displace national identities but are supplementary to them." (Keating 1998:94) This also amounts to a positive correlation between regional and national identification but "not necessarily," as the author adds. Other writers diagnose a "new selfconsciousness" of regions (Loughlin 1997:147). If this is true we would expect a positive correlation between European and sub-national identification (increasing European goes with increasing sub-national identification) and a negative correlation between national and subnational identification (decreasing national identification raises identification with regions). The first assumption of the argument claiming that European identification is important for the success of a unified Europe may be doubted as well. It may be argued that achieving the goals of the EU is based on the kind of institutions that are built and not on elusive phenomena such as identities. However, institution building is an outcome of political decisions which, in turn, depend on the parties that are in power; and election outcomes depend, among other things, on attitudes of the voters. We must leave the question of the validity of assumption 1 open because we cannot test it with our data. In regard to the causal interrelationship of identifications, it thus seems that almost all possible relationships are put forward in the literature. The previous argument, however, seems to be most widely held and is, at least at first sight, not implausible. As was said before, we will focus on this argument. However, we will test alternative hypotheses about relationships between the identifications as well. Some Problems with the Argument The previous argument is burdened with four major problems: important concepts are unclear, the detailed causal interrelationships between identifications are not spelled out, no general theory is applied that explains when identifications change, and authors do mostly not provide any empirical evidence for their hypotheses. Let us discuss these points in order. (1) The major concepts in the previous hypotheses especially identity and nationalism are most of the time not defined clearly or are used in different meanings.' It is beyond the scope of this paper to provide a detailed discussion of the use of these concepts in the literature. We prefer to suggest a clear definition that captures some aspects of the concepts in the literature that is not contradictory to the use of the concept in the debate about European unification. But what aspects should the concept include? We choose a definition that is of theoretical import (i.e., is part of the theories that are to be applied). A rather clear concept is identification in the sense of an attitude' toward Europe, a nation and sub-national units. The previous hypotheses are then understood as referring to an attitude or, equivalently, an attachment or loyalty' to those units. In this sense, the concept of identification has been used for a long time (e.g., Johnston 1969: 7-12). "Identity" is certainly not equivalent with "identification." However, looking at the various formulations of the argument, "identification" is certainly one of the aspects the hypotheses referring to "identity" are about. For example, if it is said that due to the decreasing importance of nations "national identity" decreases, it is

5 The EU and National Identifications 657 certainly meant, among other things, that the attitude or, equivalently, the attachment to a nation decreases. In regard to "nationalism," this term is certainly not identical with "national identification." However, identification with a nation is a feature of nationalism. In what follows, we thus address the identification with Europe, nations and sub-national units. The advantage of focusing our discussion on identification is that we need not go into the extensive debates about the concepts of identity (and nationalism). If our decision to use the concept of identification as outlined above is rejected, we would argue that the critic should replace "identification" with another clearly defined concept, and then test the respective hypotheses empirically. It can then be seen which of the two sets of alternative hypotheses fare better. (2) Proponents of the previous argument usually do not put forward or discuss any detailed hypotheses about the causal interdependencies of the identifications with Europe, nation states and sub-national units. For example, could not a strong national identification also affect the identification with Europe? If individuals strongly identify with a sub-national region such as a federal state, could this not have an effect on national identification? 5 Furthermore, there are no clear hypotheses about other factors that affect those identifications. (3) Another typical feature of contributions to the previous argument is that no general theory is applied that could lend some support to the claims made. We only find singular causal statements of the kind 'A causes B" such as "a decrease of importance of a nation causes a decrease of identification with the nation." Are there any theories that imply the proposed hypotheses? At best, we find orienting statements. To illustrate, it is held that regional identity "may be rooted in historical traditions and myths but, in its contemporary form, it is a social construction, forged in a specific context under the influence of social, economic and political pressures." (Keating 1998: 87) But exactly what the variables are that affect regional "identity" and in what ways remains unclear. (4) The authors who write about the argument typically do not provide any empirical evidence. To be sure, there are several empirical studies that report positive correlations between identifications (Bornewasser and Wakenhut 1999: 58). Examples are the ISSP (International Social Survey Program) from 1995 (Gerhards 2000: 126), and the Allbus (General Social Survey of Germany) of 1990 where items on feeling close to various regions show also positive correlations (Opp, see items v1 24 and v1 28). We find very few reports about negative correlations in the literature (Bornewasser and Wakenhut 1999: 58; Huici et al. 1997: 104). When Do Identifications Change? Some Testable Propositions The previous argument assumes, among other things, that the development of the EC brings about a European identification. Why is this to be expected? Does the emergence of an institution always generate identification with this institution? There are institutions that are not likely to spawn identification such as dictatorships. For other institutions, the reverse is plausible. This may hold for European welfare states such as Germany 20 years ago with their provision of benefits against all sorts of personal predicament. The question then is under what conditions are institutions likely to bring about negative or positive identification. A similar question arises for thesis (3) of the above argument claiming that the strengthening of Europe generates a negative identification with the nation and with sub-national units. Why is this to be expected? Some of the theoretical considerations of proponents of the previous argument give a clue to answer these questions. It is, more or less, implicitly assumed that the progressive unification of Europe is valued positively by the population. These positive features may be economic well-being or "forging a common heritage." (Banos 2002: 163) The nations' declining sovereignty and, thus importance, seems to be a negative feature. In

6 658 Social Forces Volume 84, Number 2 December 2005 general, the implicit hypothesis seems to be that negative features of an attitude object lead to an overall negative evaluation of this object. This is the basic theoretical idea of a wellconfirmed social-psychological theory that we will apply to generate some propositions about the explanation of the relationships between the three identifications: we are referring to the theory of planned behavior (TPB) of I. Ajzen (Ajzen 1988, 1991, 1996). The theory explains how attitudes originate and can be applied to explain attitudes toward territorial units or regions or institutions such as the EC, nation states or sub-national units (Mohler and Opp 2004). The theory holds that a positive attitude toward an attitude object (such as Europe) develops if the number and value of positive features of the object are relatively large and if individuals strongly believe that the object has these features. Thus, a positive identification with a region is a function of the advantages an individual attributes to the region. For example, if the consequences of the advancing European unification such as easier traveling, the common currency or increased economic well-being are regarded as highly advantageous and perceived as highly likely by the citizens of the member states, one can expect that a positive identification with Europe will develop over time. Causal Interdependencies Between Identifications What predictions can be derived from the TPB in regard to the negative effects of an increasing European identification on national and sub-national identifications? If the aspects of Europe, the nation and sub-national regions are the decisive factors that generate identifications we would expect: Hypothesis 1: lithe unification of Europe is associated with positively valued aspects and at the same time leads to negatively valued aspects of the nation and of sub-national regions, the identification with Europe will reduce the identification with the nation and with sub-national units. Are the conditions specified in hypothesis 1 actually realized? Assume, as before, that individuals ascribe increasingly positive attributes to the unified Europe. Why should this change the belief that the existence of national states still has many advantages? If the nation state is still the most important decision maker in the EC, why should the citizens then negatively view the transfer of some rights to the EC? Thus, the decrease of national sovereignty in the course of European unification may not be regarded as important at all. Furthermore, it is argued that only a distinctive culture promotes identification. Even if the cultural distinctiveness of nations and sub-national units decreases in the wake of a unified European culture, there is still cultural distinctiveness in regard to many cultural features, and these may be in the perception of the citizens the most important cultural items. Given these items, the import of new cultural items (Starbucks, McDonalds) may be seen as adding some cultural variety. For everyday life, sub-national units such as federal states and the nation may be seen as the most important units even if European unification progresses. In other words, Europe may be regarded as a new institutional setting that has many advantages, but that may not diminish the roots people have in their nation or their subnational regions. There is a complex interaction between federal states and nations and, furthermore, between nations and the EC. For example, nations contribute to the formation of a unified Europe by transferring rights to the EC. If this is regarded as advantageous by the citizens, then the nations and other subordinate regions such as federal states have contributed to this process, and they contributed voluntarily. If there is such a give and take

7 The EU and National Identifications 659 between regions and if this results in outcomes that are regarded as positive, it is plausible that these outcomes are ascribed to all participating units. We would expect positive correlations between the three identifications. Negative correlations between these identifications are further implausible for the following reason. As was said before, our data refer to Europe, Germany and Saxony (a federal state of Germany). The German federal states are part of Germany and Germany is part of Europe. In terms of balance theory (Heider 1958), these regions stand in a unit-relation (Heider's term). The second type of relation between units is, according to balance theory, a liking relation. Cognitive balance (i.e., psychic well-being) is given if unit and liking relations coincide. For example, if a person P interacts regularly with a person 0 (interaction is a unit relation) and if P likes 0, balance exists. In regard to the objects we are concerned with, cognitive balance exists if the unit relations between the individual/europe/germany/federal states are squared with liking relations. In other words, if I realize that I am part of a federal state and that the federal state is part of a nation, balance is achieved if I like the federal state and the nation. Balance theory further implies that the emergence of a new unit such as Europe that includes the nation and sub-national regions generates liking relations. These relations also refer to the features of the units. For example, if citizens ascribe positive features to their nation and federal state, it is likely that an emerging unit of a higher level such as Europe will be associated with positive features as well. In general, balance theory implies a matching of identifications. In statistical terms, we expect positive correlations between identifications. Political leaders also strengthen the feeling that the European countries are one unit. A typical argument for EU membership is that the specific country belongs to Europe, is part of the European culture, etc. (Brusis 2001: 55). Opponents against the extension of the EU by including Turkey argue that Turkey is in fact not a "real" European state because of its different culture. In terms of balance theory, politicians try to construct a unit-relation between the specific country and Europe or they argue that there is no unit relation. Do TPB and balance theory yield any predictions about the causal order of identifications? We first hypothesize that the process of identification runs from top to bottom (see upper panel of Figure 2) if three conditions are given. (1) A new institutional or, equivalently, regional hierarchy such as the EU develops. (2) One institution, the EU, is regarded as the major agent of forging this hierarchy of institutions. (3) This hierarchy is considered very advantageous by the citizens. If these conditions are given, one can expect that an increasing European identification will have positive effects on the identification with lower units in the hierarchy. In this case, the TPB would predict that identifications will emerge successively with the development of the institutions. Positive features of Europe are ascribed to the nation and lower-level units because what is ascribed to the whole is also a property of its parts. There is thus a top-to-bottom identification. The second proposition is that there is a bottom-to-top identification process. (See the second panel of Figure 2.) Identifications with a region or institutions develop in childhood. The child first experiences his or her immediate environment. TPB suggests: if the child experiences his or her given regional or institutional environment in a rather positive way a positive identification will develop.' Children first develop an identification with their local or lower-level regions. Only later will identification with the higher-level regions such as the nation or Europe originate. There are thus two causal processes: (1) an emerging European identification strengthens national and sub-national Identification, i.e. there is a hierarchical downward identification (top-down identification); (2) the socialization process generates a hierarchical upward identification (bottom-up identification). Both hypotheses assume that there are positive correlations between the three identifications, but the causal processes differ.

8 660 Social Forces Volume 84, Number 2 December 2005 Figure 2. Top-Down and Bottom-Up Identifications European identification Top-Down A Identification + National identification European f unification and other factors Sub-national identification ~ European European identification ^ unification and other factors Bottom-Up Identification + National identification Sub-national identification Note: Bold arrows symbolize strong effects. Arrows pointing to sub-national and national identification indicate that other causal factors influence these variables. Is it plausible that the top-down and bottom-up transmission of identification embraces all units in a hierarchy to the same extent? The effect of an identification on the next unit in the hierarchy (upwards or downwards) will always be strongest. It is therefore assumed that a sub-national identification has a relatively strong effect on the identification with the nation, but a rather weak effect on the identification with Europe; a European identification has a relatively strong effect on the identification with the nation and a weaker effect on the identification with sub-national regions. To summarize: Hypothesis 2: (a) There will be positive correlations between European, national and sub-national identifications. (b) Sub-national identification has a strong positive effect on national identification and a weaker positive effect on European identification. (c) European identification has a strong positive effect on national identification and a weaker effect on sub-national identification. As was said above, research shows that positive as well as negative correlations have been found. This suggests that the previous proposition holds only under certain conditions. We hypothesize that a condition for a positive correlation between identifications with institutions is that there are no con flictual relationships between the units or regions (Huici et al. 1997: ). Assume, for example, that some superpower has conquered Germany, France and Belgium and builds some overarching institution such as the European Union. The citizens will

9 The EU and National Identifications 661 not perceive a "unit relation" (in Heider's sense) between the units in this institutional hierarchy. Accordingly, balance theory would not predict the emergence of liking relations. In terms of TPB, this hierarchy will not be associated with advantages. We would thus expect that there will be negative correlations between the existing units and the new unit. Another situation would be that citizens of a sub-national region attribute many negative features to the nation which they want to be separated from, but positive features to Europe. This would yield a positive correlation between sub-national and European, but a negative correlation between sub-national and national identifications. The signs of the top-down and bottom-up arrows (see Figure 2) will therefore change. The theories applied can thus predict different signs of the relationships between regional (or institutional) units. For the situation in Germany our data refer to, there are no conflictual relationships between the federal states, the nation and Europe. Thus, our data do not allow us to test hypotheses involving conflictual relationships. We will therefore not further address this issue. However, if there are no conflictual relationships between large-scale units such as federal states, nations and Europe, there may be problematic relations between groups within units. For example, East Germans often feel badly treated by West Germans. So far we applied general theoretical propositions to explain identification. A widely applied theory that specifically addresses the origin of identification with groups is social identity theory (SIT).' One of its propositions is that any membership in a salient group creates in-group identification and out-group hostility. In applying this proposition to predict relationships between the three identifications, the question arises what the salient group memberships are. Our data show that respondents identify strongest with the federal state of Saxony, and less so with Germany and Europe. This would yield the following prediction: Hypothesis 3: Identification with sub-national regions leads to negative identification with the other regions.' However, this proposition is implausible for the situation that is at issue in this paper: the identification objects stand in a hierarchical, non-conflictual relationship. To be sure, in the minimal group situation (where people don't know each other) antipathy to a non-membership group might originate. But if we imagine a situation in which group A is part of group B and B is part of C, and if there are interactions between members of these groups whose outcomes improve the welfare of each group, it is highly doubtful whether membership in only group A will generate negative identifications with B and C. Other Causes of Identification We will now extend our model by suggesting some hypotheses about the other factors that might influence identification. (For a summary of the following argument see Figure 4.) In order to determine these other factors we need to specify positive and negative features of the regions, i.e. Europe, Germany and Saxony, which might affect identification. What aspects of the EC could generate a positive or negative identification? We distinguish two kinds of properties. The first refers to the features of the EU's governing institutions: the EC is first of all a bureaucracy of huge dimensions and its decisions have a strong impact on the development of the EU, whereas a democratic element of the EC is the European parliament. Based on the TPB we would submit: the more positively the work of these institutions is valued, the stronger is the identification with the EC. The second property if the EC is the effect of its policies. For example, how do citizens value the new currency of the Euro? (For details, see the section on measurement.)

10 662 Social Forces Volume 84, Number 2 December 2005 Figure 3. Negative Effects of Sub-National Identification European identification -- European unification and other factors National Sub-national identification identification Note: Arrows pointing to sub-national and national identification indicate that other causal factors influence these variables. A hotly debated issue in Germany is immigration, citizenship and integration of foreigners. On the one hand, in times of high unemployment many citizens wish to restrict immigration. On the other hand, many simple jobs are not taken by Germans; it is further well known that due to the skewed age distribution, immigration is sorely needed. A new immigration law that is discussed in the parliament for a long time reflects the controversial issues of how to deal with foreigners in Germany. In general, we submit that the extent to which citizens are satisfied with the treatment of foreigners in Germany has a positive impact on identification with Germany. In regard to Saxony, we suppose that a positive valuation of various features of Saxony such as Saxon culture and Saxon history is important for an identification with Saxony. As was noted before, even if there are no conflictual relationships between regions, there may be conflicts between groups within a region. It is well known that East Germans often feel hostility to West Germans (and vice versa). We assume that strong hostility to a group leads to strong identification with the in group (i.e. Saxony) because perceived discrimination by others raises the salience of the positive properties of the in group. What would be the effect of hostility toward West Germans for identification with the whole Germany? A negative identification with Germany would probably be cognitively dissonant because Saxony is part of Germany and heavily relies on transfers from the German government. It seems plausible that a low positive identification with Germany would be the best way to minimize cognitive dissonance. What about identification with Europe? Saxony is only a tiny part of Europe. An existing conflictual relationship with West Germany should not generate a negative identification with Europe. Saxons do not feel rejected by the EC or Europeans; actually, they are Europeans. It seems plausible that perceived conflict with West Germans may not only increase salience of positive features of the in group but also positive features of other membership groups (such as Europe) where one does not feel discrimination. We suppose that perceived conflict with West Germany has a weak positive effect on European identification. To summarize: perceived conflict with West Germany will have a relatively strong positive effect on identification with Saxony, a weaker positive effect on identification with Europe and a still weaker effect on identification with Germany. The final variable of our model is whether a person is born or not born in Saxony. We assume that being born is Saxony has a strong positive effect on identification with Saxony.

11 The EU and National Identifications 663 Figure 4. Summary of the Hypotheses About "Other Factors" Influencing Identification Positive valuation of the European institutions \\ N, Positive valuation " `.. European of outcomes of EU \ identification Medium "N. `^ ^^ German Small Perceived Satisfaction ` identification 4conflict with with foreigners `^^ West Germans in Germany. Saxon Strong identification Positive valuation of // features of Saxony Born in Saxony Note: All effects are supposed to be positive. For the causal interrelationships between the identifications, see Figure 2. As was said before, most children experience their socialization positively and associate positive features with their immediate environment. The implication is that being born in Saxony should have a positive effect on identification with Saxony. If there are interrelationships between the identifications, at least some of the "other factors" should have effects on several identification variables. More specifically, we should find top-down and bottom-up effects. (See the dashed arrows in Figure 4.) Only "Satisfaction with foreigners in Germany" should have an exclusive effect on identification with Germany because there is no bottom-up or top-down effect. (See hypothesis 2.) We will not develop hypotheses about the strength of these relationships. We only assume that the effect of a variable that directly addresses a feature of Europe, Germany or Saxony should have a significant and a relatively strong effect on the identification with the respective region. Research Design and Measurement We tested our propositions with a two-wave panel survey. The first wave consisted of a representative sample of 3,005 respondents of Leipzig (N=1,500) and of a rural area in Saxony (Mittlerer Erzgebirgskreis, N = 1,505), administered from May through July 2000 by a professional survey institute. The data of the second wave was collected by the same institute April-June 2002 and consists of 1,895 respondents (813 in Leipzig and 1,082 in the rural area). Thus, 63 percent of the original sample could be re-interviewed. We chose a rural and an urban area because we assumed that identifications would differ, and including an urban and rural site would thus increase the value range of the identification variables. Further, if the area has an effect on identification, it would be a good test of the quality of the theoretical propositions whether they can explain effects of the area variable.' A comparison of demographic characteristics of the first wave with official statistics indicates that the survey matched these characteristics very well. We further analyzed the

12 664 Social Forces Volume 84, Number 2 December 2005 extent to which various variables demographic variables and some of our theoretical variables such as the identification scales of participants and non-participants in the second wave differed. Again, we find surprisingly low differences. But even if differences were much worse, we would argue that the major goal of this paper is to test theoretical propositions. If these are sound, they should be confirmed by the present data set. Identification with Europe, Germany and Saxony was measured by two questions for each region. (See Table 1.) It is important to note that the identifications were measured so that a respondent could express her or his identification for a given region independently of the other regions. In principle, thus, all correlations between the three identifications are possible. The remainder of Table 1 summarizes the measurement of all other variables. When scales consist of at least three items, we performed factor analyses to examine dimensionality. The selected items were added and divided by the number of items, we thus constructed additive scales. Some Descriptive Findings An interesting question is to what extent the respondents identify with Europe, Germany and Saxony. This question is of interest because the distribution of the identification variables may be so skewed that statistical analysis may be problematic. As Figure 5 shows, there is great variation in the identifications between the regions. We first present the extent to which respondents "feel" European, German and Saxon. In both waves, identification with Saxony is very strong compared to the maximal scale value 5. Identifications in wave 1 (4.11) and wave 2 (4.07) are very similar. "Feeling" German is clearly lower; the means are 3.31 in wave 1 and 3.33 in wave 2. Still lower is identification with Europe. In contrast to "feeling" Saxon and German, average values for Europe increase from 2.77 (wave 1) to 2.96 (wave 2). The second indicator referring to the extent to which respondents are "proud of" being European, German or Saxon yields a very similar picture. All arithmetic means are a little smaller than those of the first indicator, with the exception of pride in being European. This indicator has the same value as the "feel" indicator, viz The lower means indicate that the pride indicator measures a stronger attachment than simply "feeling" European. Thus, identification is relatively high with lower-level units, which has also been found in other research. The standard deviations and frequencies (that are not presented) show that there are no statistical problems in using the identification variables as dependent or independent variables. The indicators measuring the extent to which individuals value the features of the unified Europe are more problematic. Table 2 lists these indicators in the first column. The first two indicators refer to the European institutions, indicators 3 to 6 to certain outcomes of decisions of the EC, and the indicator 7 measures a general evaluation of the EC. Items 1 through 6 refer to the issues and achievements of the EC that are discussed in the media: there is, for example, a standard critique of the powerlessness of the European parliament and of the excessive bureaucracy in Brussels. The major policy achievements are expressed in items 3 through 6. Three findings in particular are striking: (1) In general, a relatively small percentage of the respondents marked the categories "good" and "very good." The exception is the "open borders" of the EC. Only here does a majority express a positive valuation. The means in the last column confirm the low valuation of features of Europe: only the mean for the "open borders of Germany" and the general evaluation of the EC are a little higher than the middle point 3 of the scale. (2) Most of the time, the middle category received most answers. It may be argued that this category is an outlet for respondents who "don't know" what to say. But,

13 The EU and National Identifications 665 Table 1: Measurement of the Variables Name of the variables Measurement Dependent variables Identification with Europe, Germany, Saxony, Independent variables For each region, respondents were asked: (1) How strongly do you feel as a European, German, or Saxon? Answer categories were: very strong, strong, medium, weak very weak. (2) How proud are you to be European, German, or Saxon? Answer categories were: very proud, proud, partly proud/partly not so proud, less proud, not at all proud. For each region, both indicators were added and divided by 2 because the correlations were relatively high. The items were recoded so that high values mean high identification. Positive valuation of the This is an additive scale consisting of items 1 to 3 of European institutions Table 2. Positive valuation of outcomes This is an additive scale consisting of items 4 to 6 of of EU policies Table 2. Satisfaction with foreigners in Additive scale of two items referring to the extent to Germany which respondents are satisfied with the politics of asylum of Germany and the number of foreigners in Germany: five categories, from "very dissatisfied" (code 1) to "very satisfied" (code 5). Satisfaction with living Respondents were presented with various features of conditions in Saxony Saxony and asked to what extent they value them positively: five categories, from "very bad" (code 1) to "very good" (code 5). The items referred to Saxon history, culture, economy, scientific successes, language, traditions and customs, and successes in sports. Conflict with West Germans A scale consisting of two items to which respondents could more or less agree - five categories: from "fully disagree" (1) to "fully agree" (5): (1) An East German can try as hard as he will, he will never achieve the same as a West German; (2) it is highly frustrating to see how East Germans are treated by West Germans. Born in Saxony Refers to whether a respondent was born in Saxony (1) or not born in Saxony (0).

14 666 Social Forces Volume 84, Number 2 December 2005 Figure 5. Identification with Europe, Germany and Saxony (Waves 1 and 2) 4.5 Wave 1 Wave 2 Wave 1 Wave n oa'i 3.21 k qe^ Median of the scalesa 2.68 ^.7 C,' m Jc CO v, o 2 tie m N 38 W 1.5 r r r Feel as... Indicator for Identification a Proud... this is implausible because those who were unable to answer the questions had the opportunity to express this immediately by marking "cannot judge." (3) The great number of respondents who chose this category is the third striking finding. Even in regard to the common currency that people use every day, 66 respondents said that they cannot judge how good or bad they find the Euro. For the other items, the number of "cannot judge" responses varies between 214 and 808. The Missing Values Problem for the Items About Features of Europe There are several possibilities of handling the "cannot judge" answers. (1) One strategy is to assign missing values. This seems justified: if the hypothesis is that certain beliefs about Europe affect identification with Europe, then those who "cannot judge" an item referring to such beliefs don't have such beliefs and should therefore be excluded. (2) However, it is possible that those who "cannot judge" have nonetheless a strong positive or perhaps a negative attitude toward Europe. If it turns out that the "cannot judge" answers do not affect the dependent variables, it is justifiable to assign a missing value to these answers. In order to test this assumption we included the "cannot judge" respondents in the statistical analyses by constructing for each item in Table 2 dummy variables. For example, for item 1 the first dummy variable gets value 1 for "very bad" and 0 for all other categories; the second dummy variable for item 1 receives value 1 for "bad" and 0 for all other values; the sixth dummy for item 1 gets 1 for "cannot judge" and 0 for all other values. It is now possible to include the dummies for each item in regression analyses by excluding one dummy of each item as a reference category. This procedure is complicated, and, as our analyses showed, multicollinearities are high. In general, however, the "cannot judge" dummies have only low effects on the identification variables. To avoid the problems mentioned we proceeded in another way: we constructed for each of the six items of Table 2 referring to the special features of the EC only one dummy variable with value 1 for having answered the question with "cannot judge" and 0 for any other., W

15 The EU and National Identifications 667 Table 2:Attitudes toward the European Union (Percentage of Respondents for Each Item, N of "cannotjudge"and of "no answer") U) d ^L 0 U 3 U) C AnswerCategories Means Interview question: what of the European Ve<y bad Bad Medium Good V«y'P>J No Union is good and whal is bad? (I) (2) (3) (4) (5) answer (B N 2 Co ') Co Lq M N CV (N (N M (N C') z Co Co N Cl) M N v M N ' f- CO cr C') M Co 1- O ^^ C O O Co Co LO -2, Ur 6") - CO CO LC) 1'- ao CV N O M Co M CO CO N CO m Cu O CO (0 N C') CO C() CO CO M N M M N N Co Co ti CO a M N CO (N CO N-- C6 Cn N O N O a) _ a O U) O C O Cam- CL N 7 C C W..O U) E 0 -(D Cu N fn E o > E w m N > 0C Q 2 O ^ a) w w U) E C.0O CuQ -- L a) W7 o N CO CD a) O O 20 O O C L CD - 0 C m ca i W ; O O -C Cu O E O 0 -O O 7.Q N CT C Cu 3.9 Q Cm O `-' O -O = O 0 2C 2 2 Z3 W y W c Cu CO The European InslJtubons (1 ) WorXing of European parliament , 50' 2.77 (2) sevce machrle<y., Brussels' '.' (3) Inlluence d Brussels ongerman poidx:s Policy 0u1almes d... EU (4) The E..." ' (5) Openbortlets d Germany (6),_d Easl E<ropeanalUfllnes General E...bond... EU {7J JlJ" ai, how has the EUdeveIolJed so fa< " jol6 opinion N: lanswercalegor<s lor... 'em are no<. 'alpstdied (I), no<pstdied (2), partly pstdied (3), pstdied(4), haypstdied (5). O W O O O_ 0 >+ 00 (n cn c? I-- 0 c <. ii Q U) 0 O M a) N _T Q N_ 0 O C

16 668 Social Forces Volume 84, Number 2 December 2005 answer. Those who have refused to answer get, as before, a missing value. The correlations of the six dummies with the dependent variables (identification with Europe, Germany and Saxony) were between -.12 and +.07 and are thus very low. We further constructed the variable "number of 'cannot judge' answers" for each respondent, referring to the first six items of Table 2. From the 1,895 respondents, 35.2 percent have not given any "cannot judge" answer; 19.1 percent have given 1, 19.2 percent have given 2, 14.4 percent 3; a total of 12.3 percent of the respondents could not judge four to six items. In other words, there is only a relatively small percentage of respondents who are not able to answer more than three of six items. The correlation of this measure with the dependent variables is.05 (identification with Europe),.06 (identification with Germany), and.04 (identification with Saxony). We further included these variables as an additional variable in several multiple regression analyses. There was no statistically significant effect. In addition, the correlations with the independent variables are low. A consequence is that the regression coefficients remain almost the same when the variable "number of cannot-judge responses" is included in the analyses. To ease interpretation of the findings we did not include this variable in the analyses to be presented later. We further explored who the cannot-judge respondents are. We first correlated the variable "number of cannot-judge answers" with the following variables: political interest, leftright orientation, education, age, gender, not married and household income. The data show that a large number of "cannot judge" answers goes together with low political interest (this is the highest correlation with.44), high age, low education and income, and being female. For the present analysis, these correlations are unproblematic because these variables show neither substantive correlations with the dependent nor with the independent variables. These findings suggest that it is advisable to assign missing values to "cannot-judge" answers. But excluding these respondents from all analyses (i.e. applying listwise deletion of missing values) would mean to forgo information on all the other variables. We therefore used pairwise deletion of missing values. However, various regression analyses with listwise and pairwise deletion show almost identical results. The Procedure of Testing the Propositions The model exhibited in Figure 4 does not include a time dimension. When we test this model with a panel design it has to be specified which variable is taken from wave 1 and wave 2. The dependent identification variables were taken from wave 2. In all analyses, we always include the lagged dependent variable, i.e. the respective identification variable from wave 1 as an independent variable. This means that we determine the effects of the independent variables (such as being born in Saxony) on the identification variables given the value of the respective dependent identification variable in the previous time period. We thus ascertain the extent to which an independent variable leads to a change of the dependent variable. (For details, see Finkel 1995.) The two independent variables referring to the European institutions and the outcomes of EC policies were only measured in the second wave and not in the first wave. The other independent variables were taken from wave 2. Based on this procedure the model to be tested is exhibited in Figure 6. Let us first look at the lagged effects (i.e., the effects from the first to the second wave). We see the bottomup causal relationships from Saxon to national and to European identification. The figure further exhibits the top-down effects (the effects of European on national and on Saxon identification). We further include the stabilities (the effects of the lagged dependent on the

17 The EU and National Identifications 669 dependent variables). In order to test our model we will, among other things, examine whether the causal relationships which do not correspond to our hypotheses exist. For example, we test whether the effect of national on European identification is actually zero, as the model in Figure 6 assumes. Figure 6. The Model to be Tested Wave 1 Wave 2 European identificatio National identificatio Saxon iri.ntifiratin European 4, X identification " e, i -- / N tional ^^ X; i ntific ion =r--- ez ' Saxon Xk identification --- e 3 X; = Features of the EU X. = Satisfaction with Foreigners in Germany Xk = Positive valuation of features of Saxony, Born in Saxony The hypothesized causal relationships between the identifications may not only exist between waves 1 and 2 (as lagged effects), but also within wave 2 (as simultaneous effects). The latter are symbolized by the arrows between the identifications of wave 2. We further symbolize the effects of the other independent variables of wave 2 which we call Xi, Xj and Xk Finally, the figure includes the error terms e1, e2 and e3. We will test whether these error terms correlate. Strong correlations point to problems of the model. Why did we include the independent variables Xj and Xk (see Figure 6) from wave 2 and not from wave 1? We performed various multivariate analyses with those wave 1 variables as independent variables. The effects were in general very weak. The reason might be that the time period between the measurements of about two years was too long. Lagged models "will be appropriate only whenever the causal lags are approximately equal to the time period between measurements." (Finkel 1995:24) For example, assume that the causal effect of an independent variable Xt1 that is measured at wave 1 occurs after a short time and generates Yt1 which we do not measure. After some time, the values of Xt1 change to Xt1' so that both (Xt1 and Xtl') correlate only weakly (or even negatively). Now assume that we measure the effects of Xtl', i.e. Yt2, at wave 2. We thus measure Xt1 and Yt2 (the effect of the independent variable Xt1 and not of Xtl'). The causal impact of Xt1 on the dependent variable of wave 2 (Yt2) may be zero or even negative although a clear positive causal effect obtains. If we find that given causal variables do not have a cross-lagged but only a simultaneous effect, then this suggests that the causal lags are not captured by the time difference between wave 1 and wave We therefore include the causal independent variables (except the identifications) from wave 2. (See Figure 6.)

18 670 Social Forces Volume 84, Number 2 December 2005 How can we interpret effects of wave 2 independent variables on wave 2 dependent variables? Remember that the lagged dependent variable is included in each regression equation. The analyses thus show the extent to which the independent variables of wave 2 affect the change of the dependent variables of wave 2 (because, as said before, the lagged dependent variable is always included in the analyses). This procedure is not a perfect test of causality, but better than a pure cross-sectional analysis. In the latter type of analysis, a lagged dependent variable does not exist. Therefore, in cross-sectional analyses, it cannot be ascertained to what extent a causal independent variable leads to a change in the dependent variable (given the value of the dependent variable at some earlier time). A Test of the Propositions Two Identification Models In regard to the bivariate correlations between the three identifications, there are two major findings (Table 3): (1) There are only positive correlations, and each correlation is highly significant. This clearly falsifies any hypotheses claiming a negative correlation between identifications. (2) When we look at the size of the correlations we observe the following pattern: Correlations between identifications with Germany and Europe are always higher than correlations between identifications with Germany and Saxony and with Europe and Saxony. For example, in wave 1 the identification with Europe and Germany correlates at.61, whereas the correlation with Europe and Saxony is only.26, and the correlation between Germany and Saxony is.41. A similar pattern holds for wave 2 (lower right hand panel of Table 3). In regard to correlations between waves (upper right hand panel of Table 3), this pattern is also present but the differences between the correlations are lower. In testing our other propositions about the causal order of the identifications we will proceed in two steps. First, we tested only the causal relationships between the identifications. The other independent variables, symbolized by Xi, Xj and Xk in Figure 6 are thus not considered at this point. In a first analysis of the causal interdependencies of the identification variables, we tested a model that consists of all possible lagged relationships of the three identification variables; we then eliminated the non-significant effects and included error correlations. Model 1 of Figure 7 shows the results." The hypothesis that a strong European identification has a positive effect on national identification is confirmed (Beta =.13). There is no effect on the identification with Saxony. The topdown proposition is thus only partially confirmed. Furthermore, there are effects from "Saxon" to "national" and to "European" identification, which is in line with the bottom-up proposition. The quality of this model is rather good: with a chi-square of 5.83 and 3 degrees of freedom the p-value is.12. The RMSEA (Root Mean Square Error of Approximation) of.02 which is also highly satisfactory. Less positive are the rather strong intercorrelations of the error terms. These correlations refer to the covariations between the wave 2 identification variables that are not explained by the independent variables. This means that additional variables should be included or that there are simultaneous effects (i.e., feedback-effects) between the wave 2 identification variables. In a next step we modified model 1 of Figure 7 by including simultaneous effects that correspond to the cross-lagged effects. It turned out that the cross-lagged effects became either insignificant or, due to multicollinearity, negative. When we delete these effects from the model we get model 2 of Figure 7. This model consists of the lagged dependent variables and the simultaneous effects. The three simultaneous effects are highly significant. They correspond to the lagged effects. The quality of the model in regard to chi-square, p-value and

19 The EDand National Identifications 671 Table 3: Correlations Between Identification with Europe, Germany and Saxony (Bivariate Pearson Correlations) Identification with: Wave 1 Identification with: Europe Germany Saxony Wave 2 Europe Germany Saxony Wave 1 Europe Germany Saxony Wave 2 Europe Germany Saxony Note: All correlations are significa nt at least at the.0 1 level. one-ta iled tests. I = RMSE is the same as model 1. However. the error correlations are strongly reduced. Model 2 is thus better than model 1. We will not discuss the results of these two models in greater detail at this point. The models may be misspecified because the other independent variables have not yet been included. The next step then will be to test the full model. TheFull Model Based on model 2 of Figure 7 we first estimated a model by adding the causal relationships of the other causal variables. (See the Xi. Xj and Xk variables in Figure 6.) We then eliminated non-significant effects as well as non-significant variables. and tested for other possible effects. The causal diagram of Figure 8 shows the causal model that has been confirmed. Let us first look at the causal interrelationships of the identification variables. It was assumed in hypotheses 2b and 2c: (2b) Sub-national identification hasa strongpositiveeffect on national identification anda weakerpositive effect on European identification. This is the bottom-upproposition. (2c) European identification hasa strong positive effect on national identification anda weakereffecton sub-national identification. This is the top-down proposition.

20 672 Social Forces Volume 84, Number 2 December 2005 Figure 7. Causal Relationships Between Identifications: Two Models (Standardized Coefficients) Wave 1 Wave 2 European '43** European identification,13** identification R2=.24 Model 1.50** National 29** National Chi-square = 5.83 identification identification 33** Df = 3.15**. 13** Rz.20 P-value =.12 39** RMSEA =.02 Saxon.46** Saxon identification Identification R2 =.22 European 43** European identification identification R2 =.37 Model 2.30** ^ Chi-square = 5.83 National.29** National Of = 3 identification identifica on )108** P-value =.12 R2 = RMSEA=.02 17***.32 **.46** Saxon Saxon identification Identification R2 =.22 In regard to hypothesis (2b), we find a simultaneous effect of Saxon identification on national identification of.11 and a lagged effect on European identification of.08. Although the differences of the betas are low, they are in line with bottom-up proposition. Hypotheses (2c) is also in line with the data. The effect of European identification on national identification is.36, whereas the effect on Saxon identification is smaller, as expected (beta =.20). These are all simultaneous effects. Let us turn to the effects of the other variables that affect the identifications. We assumed that "positive valuation of European institutions" and "positive valuation of outcomes of EC policies" (see Figure 4) have a major effect on European identification. Only the former variable had the expected effect, but this effect is weak. Thus, the more positively that individuals value the working of the European parliament, the effects of decisions of the EC on German politics and the civil service machinery of the EC, the more they identify with Europe. Surprisingly, the scale including possible achievements of the EC such as the introduction of the Euro, the open borders of Germany to the adjacent states and the planned extension of the EC by incorporating new member states from Eastern Europe does not have an effect on European identification. Furthermore, there is no effect on the other identifications. i )

21 The EU and National Identifications 673 Figure 8. The Confirmed Model (Standardized Regression Coefficients) Wave 1 Wave 2 Wave 2 European identificat' National identificat Saxon identification.41 European 13 identification R2 =.31 I^ p 17\.36 National 09 R2 =.53 Identifications 24 R2 = Error Correlations: European - National identification:.17 European - Saxon identification:.12 National - Saxon identification:.12 All coefficients are at least significant at the.01 level (one-tailed tests) Model fit: Chi-square = 8.08; df = 7; p =.33; RMSEA =.009 Positive valuation of European institutions Satisfaction with foreigners in Germany Satisfaction with living conditions in Saxony Conflict with West Germans Born in Saxony In regard to satisfaction with foreigners in Germany there are weak effects on all three identifications. "Conflict with West Germans" has a strong effect on Saxon identification and a weaker but still strong effect on European identification. There is no effect on national identification, which was supposed to be lower than the two other effects. "Born in Saxony" only affects identification with Saxony. The hypothesis that there is some spread of effects of the independent variables on the identifications is correct, but the kind of spread could not well be predicted. The prediction that the feature that refers directly to a kind of identification such as the positive valuation of features of Saxony refers to identification with Saxony has an effect on this identification has been confirmed. The hypothesis that if there is some spread then the "focal" effect should be strongest does not hold for the satisfaction with foreigners in Germany, but it holds for positive valuation of features of Saxony." In regard to the effects of the identifications of the first wave, it is first important to note that the lagged dependent variables have relatively strong effects, in particular the lagged European identification for which the standardized regression coefficient is.41. The bivariate lagged correlations are still higher: r for European identification is.47, for German

22 674 Social Forces Volume 84, Number 2 December 2005 identifications.41 and for Saxon identifications.47. National identification can be explained best with an explained variance of.53. Discussion Although the previous propositions have been confirmed it must be noted that the data we used to test the propositions are rather special: they are about one urban and one rural area of one German federal state (Saxony) located in East Germany. The question arises whether this sample has features that favor the confirmation of the propositions. One distinctive feature of Saxony is that it has a long history that goes back to the early middle ages (e.g., the University of Leipzig was founded in 1409) and has preserved some cultural distinctiveness such as the very special Saxon dialect. As our data indicate, such features are a cause for the identification with Saxony. It is not implausible that a bottom-up identification is set in motion especially if the identification with the lowest unit is, on the average, relatively strong. In this case dissonance would be particularly high if identifications with the higher-order units do not match. The data may thus favor a confirmation of the bottom-up identification hypothesis. Another distinctive feature of East Germany and Leipzig in particular (where part of the sample comes from) is that it was the venue of a revolution in Due to the isolation from the western world before unification it seems plausible that in terms of balance theory unit relations with the whole Germany and Europe originated or became more salient after unification. These salient unit relations may generate particularly high imbalance if the pertinent liking relations (identifications) would not develop. This situation may favor a confirmation of the bottom-up identification hypothesis. However, it may also be argued that identifications with a region develop in childhood and do not depend on the history or the cultural tradition of a region. Furthermore, it may be held that the development of an overarching institution like the EU is always salient because it is felt in everyday life. Thus, the exposure to the western world after the revolution in 1989 could be irrelevant. All this is pure speculation. But this speculation is important for setting up the design of future research. Based on the previous possible limitations of the data, a useful design would be to select a region without a long history. That might be a federal state of a European country that was recently formed after an administrative restructuring. It would be ideal if the propositions would be tested in another context. The EU is not the only association of states. Two other examples are ASEAN (Association of Southeast Asian Nations) and NEPAD (New Partnership for Africa's Development). 12 Both associations of states are rather loose groupings of states, compared to the EU, with some overarching organizations or committees. Cooperation between the member states aims at securing peace and improving economic development. These aims are pursued on a contractual basis. There is apparently no delegation of power to an organization such as the EC. For such loose associations of states we would predict that a top-down identification will hardly develop because the three conditions mentioned above are not given. The development of a bottomup identification process depends on the extent to which unit relations of the regions and the highest order associations (in this case ASEAN and NEPAD) are perceived. It seems plausible that such unit relations are compared to the EU relatively weak for ASEAN and probable absent for NEPAD because the ASEAN achieved more contractual agreements (especially in regard to economic cooperation) than NEPAD. It would be ideal if a research project could be implemented in these regions to test the previous propositions. The descriptive analysis (Figure 5) showed that there is a clear increase of European identification between 2000 and 2002 whereas identifications with Saxony and Germany were

23 The EU and National Identifications 675 largely stable. How can this be explained? Based on the model of Figure 8, the question is which factors of the model changed between 2000 (first wave) and 2002 (second wave) and which events between 2000 and 2002 have influenced these changes? Unfortunately, the "other factors" of the model are only measured in the second wave. We therefore don't know which of these factors changed, but we can suggest some plausible speculations. The major events during 2000 and 2002 were related to positively valued decisions of the EU and, we surmise, were ascribed to the working of the EU institutions. The TPB will therefore predict that these events will increase identification with the EU. Perhaps the most important event between the two waves was the introduction of the new currency, the Euro, in January Another major event was the admission of 10 new East European states to the EU (May 2004). This means that, among other things, import of cheap East European goods becomes much easier. This is especially an advantage for the rural area where part of our sample comes from because this area is closer to Poland. This hypothesis is in line with our data: the increase of European identification from 2000 to 2002 is larger among respondents in this rural area than in Leipzig. A third factor that may be of importance is a decision of the EU that beginning with the year 2000 Saxony would receive a total of 4.9 billion Euros of subsidies over six years. We did not expect that the achievements of the EC had relatively weak effects. A possible explanation is that the aspects of the EC were not valued very positively by our respondents, as Table 2 shows. Perhaps there is a threshold effect in the sense that a transmission of attitudes from the aspects of an attitude object to this object only occurs if the valuation of the aspects is rather intense. It is difficult to determine what aspects of a region affect identification. We examined the effects of various features of Saxony and Germany on the respective identifications and on other identifications. In particular, various political discontents referring to Saxony as well as Germany or involvement in social networks which were ascertained in our questionnaire were not related to identification with Saxony or Germany. The TPB would predict that those aspects are important when they are salient. "Salience" could mean that the distribution of a valuation covers all possible values of the value range of a variable. This was the case for the variables whose effects were tested with the exception of the EC features. But it could be argued that even if an aspect is valued very positively it may not be salient in the sense of being important to an individual. For example, somebody may feel as a German and be proud of being German, but being German may not be a salient (or important) "identity." Our data do not confirm such a claim. We asked respondents how important it is for them to be Saxon. When we correlate this item with the "feel" and "pride" items, we find correlations higher than.70. The correlations of the identification scales in waves 1 and 2 (which are additive scales of the feel and pride items, see Table 1) with the importance item are even higher than.80. It thus seems plausible that the value of the identification scales measures salient identifications. It is not implausible that TPB fails in this context. When the socialization process is important for the development of an identification the theory may work early in the socialization process where the positive aspects of an attitude object matter. But then when identification has been developed in the formative years of a person it may be stable and only dramatic changes of aspects of an attitude object may change identification. This seems reasonable from the point of view of dissonance theory: it is a cognitive cost to change attitudes because they are linked to other attitudes. This speculation may explain why the independent variables for the identification with Saxony have relatively strong effects: the scale "positive valuation of features of Saxony" that has a relatively strong effect refers to features rooted in the history of this federal state such as achievements of history. Further research should ascertain those basic features of the nation and Europe that seem relatively stable over time.

24 676 Social Forces Volume 84, Number 2 December 2005 Another question is whether identifications not only generate attitudes but also have behavioral effects. It is beyond the scope of this paper to provide a detailed discussion of this question. It must suffice to refer to the existing literature on identification or, as it is sometimes put, identities. This literature indicates that identification with a group has behavioral effects under certain conditions. Social Identity Theory (SIT) supports this view as well as A. Hirschman's "Exit, Voice, and Loyalty" (1970). Empirically, we may again mention the theory of Ajzen that is well tested and specifies conditions for the effects of attitudes and beliefs on behaviors. The attitude objects our propositions refer to are Europe, the nation states (in this research Germany) and sub-national units (in this research Saxony, a federal state of Germany). These units are territories. They thus have a spatial aspect. We therefore address spatial identifications in this paper. However, these regions are also institutions in the sense of sets of rules. Accordingly, the theories applied in this paper may be used to extend existing institutional analysis. The standard assumption is that institutions are behavioral constraints (e.g., Furubotn and Richter 1997; Nye and Drobak 1997; Brinton and Nee 1998). There can be no question that Europe, the nation states and sub-national units impose such constraints as well as create behavioral opportunities. For example, the increasing permeability of the national borders lowers constraints for trade and mobility. But this behavioral-effects model is incomplete. As this research shows, developing institutions may also have effects on attitudes or preferences. The question of when institutions generate such changes is rarely posed. The previous theoretical propositions specify conditions under which institutions bring about identifications. These identifications may strengthen (or weaken) the behavioral effects. For example, the greater permeability of the borders of the national states increases mobility. A low identification with a nation may be another cause for mobility: people who are not very attached to their nations will be more willing to move than people who have developed a strong emotional bond to their country. But institutions may also generate attitudes that are not related to the behavior an institution brings about. For example, it is argued that a strong identification with a nation increases negative attitudes toward foreigners. These examples indicate that the theoretical analysis of this paper may be applied to generate propositions about the extent to which institutions change preferences or attitudes. Based on this paper, a set of propositions could be generated that are similar to those suggested before and that refer to institutional units that stand in hierarchical relationships. It is certainly correct that the theoretical argument in this paper does not do justice to all the complexities of the relationships between the European states, their regions and the institutions of the EU, and does not address the historical development of the EU either. Instead, the paper focuses on a specific argument that is put forward in the literature. Nonetheless, it is an interesting question how the complexities mentioned and the history of the EU could be addressed, based on the theoretical approach advanced in this paper. According to the theories applied, identification originates on the basis of perceptions of the features of the relevant institutions or regions. We saw that the perception of Saxon history is an important variable for explaining identification with Saxony. Future empirical studies may try to devise similar measures for Europe and for nations. Maybe the valuation of the history of Europe or the nation is another factor that influences identification. In regard to the complicated interrelationships between units and sub-units, perceptions are relevant because they have an effect on the extent to which unit relationships are perceived. Perceptions are further of importance for ascribing properties to regions. Future research may explore in more detail the perceptions of the relationship between the EU, the membership states and the regions as well as the properties ascribed. It would also be interesting to explain different perceptions. The question asked in the title of this paper about the decline of the nation state can now be answered in the following way, based on the results of the research reported in this paper.